EPA's Response to IPM v6 Peer Review Report
March 2022
Summary of Contents
Section 1: Introduction
Describes the peer review process (goals, charge, panelists), commendations, and the structure
of the rest of the response document in addressing recommendations.
Section 2: Addressing Main Recommendations
2.1 Updates to model to improve the model's ability to represent the ongoing evolution of
the industry: demand/supply, new technologies, transmission, evolving state and
regional policies, and ISO/RTO market rules
2.2 Types of uncertainty that the model handles
2.3 Coal plant turndowns, operating reserves, continued penetration of renewables,
dispatch
2.4 Incorporating upstream emissions
2.5 Investment decision-making of utility and merchant power plants and capacity markets
2.6 Gas markets and natural gas pricing
2.7 Alternative load duration curves changes in load shapes from new forms of demand
(such as electric vehicles); regional and temporal resolution
2.8 Improving representation of behind-the-meter generation
2.9 Increasing transparency of retail pricing model
2.10 Representation of various policy mechanisms and publishing alternative/side cases
2.11 Documentation improvements
Appendix: Table for Detailed Accounting of Peer Review Recommendations and Narrated
Responses
1
-------
Response to the Peer Review Report
EPA Reference Case Version 6 Using IPM
U.S. EPA, Clean Air Markets Division
SECTION 1
INTRODUCTION
Background on Peer Review Process
In May 2018, the U.S. Environmental Protection Agency (EPA) released a new version of EPA's
power sector modeling platform (designated Integrated Planning Model (IPM) version 6)1. This
new EPA modeling platform incorporated important structural improvements and data updates
with respect to EPA's previous version (version 5). EPA published several updates to EPA
modeling platform version 6 Reference Case between May 2018 and September 2021.
IPM is a multiregional, dynamic, deterministic model of the U.S. power sector that provides
projections of least-cost capacity expansion, electricity dispatch and emissions. The EPA uses
the platform to project and evaluate the cost and emissions impacts of various policies to limit
emissions of sulfur dioxide, nitrogen oxides, particulate matter, mercury, hydrogen chloride,
and carbon dioxide.
In September 2019, EPA commissioned a peer review of EPA's v6 Reference Case using the
Integrated Planning Model (IPM). Industrial Economics Inc., an independent contractor,
facilitated the peer review of the EPA Version 6 Reference Case in compliance with EPA's Peer
Review Handbook (U.S. EPA, 2006) and produced a report from that peer review.2 Industrial
Economics Inc. selected five peer reviewers (Dr. Dallas Burtraw, Dr. Seth Blumsack, Dr. James
Bushnell, Dr. Frank Felder, And Frances Wood) who have extensive expertise in energy policy,
power sector modeling and economics to review the EPA Version 6 Reference Case and provide
feedback. The panel focused on the latest available Reference Case version and its
documentation at that time (May 2019 Reference Case).
Peer review panel has been asked to:
• Evaluate the suitability and scientific basis of the methods (model formulation), model
assumptions, model outputs, and conclusions derived from the model;
1 EPA periodically publishes updated projections and their documentation. Documentation, input and output files for
the latest EPA v6 Reference Case using IPM and links to the previous versions are located at
https://www.epa.gov/airmarkets/power-sector-modeling
2 https://www.epa.gov/airmarkets/ipm-peer-reviews-and-responses
2
-------
Identify specific strengths, weaknesses, limitations, and errors in the model
formulation, model assumptions, model outputs, and conclusions derived;
• Propose specific options for correcting errors and fixing or mitigating weaknesses and
limitations in the model formulation, model assumptions, model outputs, and
conclusions derived;
• Check the appropriateness of the set of model-scenarios for addressing uncertainty in
potential future power-sector trends and of particular relevance to future power sector
emissions.
The peer reviewers evaluated the adequacy of the framework, assumptions, and supporting
data used in the EPA Version 6 Reference Case using IPM, and they suggested potential
improvements. Overall, the panel found much to commend EPA; stating that the modeling
platform:
• lends itself well to EPA analyses of air policy focused on the power sector
• includes significant detail related to electricity supply and demand
• includes data-rich representation both across different geographic areas and across
time
• provides a reasonable representation of power sector operations, generating
technologies, emissions performance and controls, and markets for fuels used by the
power sector
• is well suited to assess the costs and emissions impacts
• documentation is well written, clearly organized, and detailed in its presentation of
most model characteristics
The independent peer review panel provided expert feedback on whether the analytical
framework, assumptions and applications of data in the Version 6 Reference Case using IPM are
sufficient for the EPA's needs in estimating the economic and emissions impacts associated
with the power sector due to emissions policy alternatives. The panel made recommendations
to improve the model's ability to represent the ongoing evolution of the industry; in particular:
• Continued penetration of renewables
• Increasing developments in energy storage technologies and markets
• Changes in load shapes from new forms of demand, like electric vehicles
• Evolving state and regional policies
• Evolving ISO/RTO market rules
• Increasing need for and advances in modeling capabilities of temporal resolution
Executive summary recommendations included:
1. Clarify types of uncertainty that the model is capable of handling
3
-------
2. Reconsider coal plant turndowns and addition of operating reserves
3. Consider incorporating upstream emissions
4. Distinguish investment decisions between utility and merchant power plants
5. Address the evolving gas markets regionalization and emerging sectors
6. Consider alternatives to the current load duration curves
7. Improve representation of behind-the-meter generation
8. Increase transparency of retail pricing results
9. Consider improvements in the representation of various policy mechanisms
10. More thorough citing of sources and expanded explanations in documentation
Body of the Peer Review Report included over 100 recommendations (of which most of them
tied back to the Executive Summary Recommendations) and about 50 edits to documentation.
For quick and easy reference, all of the Peer Review Report recommendations and EPA's
responses to those are tabulated in the Appendix and also referenced to the Section 2 of this
document (EPA's response to Peer Review Report) for narrated responses.
Section 2 of this document provides a high-level response to the Executive Summary
recommendations of the Peer Review Report, where we also grouped, incorporated and
addressed many of the recommendations included elsewhere in the Peer Review Report.
Before and after Peer Review Panel completed their work, EPA published five updated v6
Reference Cases; namely May 2018, November 2018, May 2019, January 2020, and Summer
2021 Reference Cases. Vast majority of the Peer Review Panel recommendations, both in terms
of capability improvements and documentation, have been addressed in the last public release
with the Summer 2021 Reference Case (published in September 2021). EPA anticipates that
future updates will continue to improve some existing features and will introduce new
capabilities, as well as more detailed documentation as needed EPA is also working on
publishing a number of side cases with alternative set of assumptions.
4
-------
SECTION 2
ADDRESSING MAIN RECOMMENDATIONS
2.1 Updates to model to improve the model's ability to represent the ongoing evolution of
the industry: demand/supply, new technologies, transmission, evolving state and regional
policies, and ISO/RTO market rules
EPA continuously evaluates and makes updates or improvements to the model capabilities,
parametrization heuristics, input data, and assumptions. Some of these are routine updates
that are updated with every new reference case (such as the fleet information), some are
integrated as new data becomes available (such as demand, generation cost and performance
assumptions), and some categories are specifically evaluated as they become more prominent
and potentially impacting projections through emerging future power sector dynamics and
policy. EPA's reference case reflects on the books state and regional policies, and relevant
ISO/RTO market rules. In addition, model has existing and potential capabilities for various
possible policy mechanisms. Documentation of these capabilities, are usually not part of the
Reference Cases but are routine part of the incremental documentation or Technical Support
Documentation that accompanies policy or scenario analysis. Appendix of this document gives a
detailed account of such capabilities mentioned in the Peer Review Report.
2.2 Types of uncertainty that the model handles
Since IPM is a deterministic model, a single run is not able to capture the range of uncertainty
of the types highlighted by the reviewers. To date EPA has focused on a central case, which
highlights conditions that can be reasonably expected. However, in order to evaluate how key
uncertainties impact model projections, EPA has previously released3 and plans to continue to
release a range of scenarios that outline a representative cone of outcomes. These scenarios
will estimate the impact of changing natural gas prices, renewable technology costs, and
demand.
2.3 Coal plant turndowns, operating reserves, continued penetration of renewables,
dispatch
EPA models the turndown rate for coal plants at the unit level. The unit level turndown
percentages for coal units were estimated based on a review of recent hourly Air Markets
Program Data where most of the coal capacity has a turndown rate between 40% and 60%. EPA
believes having unit-specific turndown rate is beneficial to our model projections because it
accounts for the variation in performance of coal plants rather than representing it as a single
value for the entire fleet, which is the approach employed by many other power sector models.
3 https://www.epa.gov/airmarkets/results-using-epas-power-sector-modeling-platform-v6
5
-------
EPA's turndown approach is an aspect of the model that we have revisited regularly and expect
to continue to do so in the future, given that load following behavior has recently become more
common in the coal-fired fleet.
EPA has evaluated the inclusion of operating reserve constraints in IPM through a variety of test
runs and determined them to be beneficial for the projections, particularly for scenarios with a
high deployment of renewable resources. At this time, EPA does not believe additional reserve
products beyond operating reserve are necessary but will continue to evaluate this moving
forward.
2.4 Incorporating upstream emissions
It is important for IPM to maintain the ability to show emissions from the combustion of the
fuel and emissions at the power plant stack. This is central to both air quality modeling efforts
and the majority of EPA EGU rulemakings that regulate stack emissions. Incorporating upstream
emissions would be done in a "feature" way that could be activated depending on the elements
of a particular policy sensitivity analysis. It would unlikely be activated in EPA's base case.
To facilitate this ability, EPA is coordinating with ORD and other divisions within OAP to derive
possible "upstream coefficient" for coal, oil, and gas. This coefficient would take into account
upstream methane emissions and C02 emissions associated with the extraction, production,
and distribution of the fuel source. These upstream coefficient source could be described in the
documentation of the reference case or any corresponding sensitivity that incorporated these
impacts, and utilized in policy sensitivities (such as CES analysis) that may revolve around
standards that synthesize the upstream emissions with the point-of-combustion emissions.
EPA's standard output and reporting of emissions would still present the sector-specific
emissions from grid-connected EGUs.
2.5 Investment decision-making of utility and merchant power plants and capacity markets
While new build financing assumptions are not differentiated based on utility/merchant
categorization, retrofits do include this differentiation. This in turn results in more realistic
retrofit/retirement decisions for the existing fleet.
Within a cost minimizing framework assuming differentiated financing for new builds would
result in possible over-builds and under-builds as a result of effective differences in levelized
costs. Based on prior runs, these builds may be unrealistic in their concentrations. Instead, IPM
assumes a weighted average financing charge for all new builds of a given technology type.
Based on prior testing we believe the current approach, i.e. differentiated financing for retrofits
and weighted average financing for new builds is the most reasonable modeling convention.
6
-------
2.6 Gas markets and natural gas price
Comments in this area tended to focus on two deficiencies: 1) more transparency in
documentation, 2) more definition given to under what scenarios EPA would reconstitute its
natural gas supply curves.
In regard to the former, EPA has supplemented the documentation with additional language
regarding the LNG export volume, non-power sector demand assumptions (particularly how it is
accounting for significant changes in the petro-chemical industry), and the relationship
between GMM and IPM oil price assumptions. This is somewhat similar to the basin-specific
discussion included in the coal supply section (except that gas supply curves are national in
scope and the gas supply implementation is different from the coal supply implementation)
where EPA provides detail on mining techniques, market conditions, and geological factors that
are basin specific and experiencing change.
In regard to identifying when EPA would rerun GMM and update gas supply curves during the
course of analysis, EPA is pursing methods to develop quantitative metrics for doing so. There
are several challenges and there are no clear-cut and distinct levels to trigger such a re-running
of GMM. EPA could perform sensitivity analysis to determine how responsive the gas supply
curves are relevant to changes in gas demand from the power sector. There may be certain
levels where the impacts on GMM are insignificant, and therefore minimal or no updates ot the
supply curve are necessary. EPA will explore thresholds for the level of change in power sector
gas consumption that would trigger the need for EPA to revisit the need to run GMM again.
Such work may accompany the data that EPA publicly releases as part of routine updates, or
also be part of some policy or technical work that the EPA might conduct in the future (which
will also be made public).
2.7 Alternative load duration curves changes in load shapes from new forms of demand (such
as electric vehicles); regional and temporal resolution
As an input to the model, the impact of alternative load shapes has been tested in a number of
scenarios and applications. For example, EPA is working on an analysis to support the
evaluation of impacts of warming temperatures on the power sector in the USA using IPM and
IPCC scenarios. This side case will demonstrate and quantify incremental impacts relative to the
EPA's reference case, taking into consideration impacts on electricity demand, power plant
capacity, power plant heat rates, transmission capacity and hydropower impacts, in addition to
identifying additional areas and improvements needed for further study. EPA has also
completed a number of internal analyses evaluating the impact of electric vehicle charging load,
varying both its magnitude and timing. Since the model's input structure allows to modify load
(both its shape and magnitude) as needed, we have evaluated various Energy Efficiency cases in
the past and will continue to do so.
7
-------
IPM can be configured with varying number of seasons. For example, in v6, a winter shoulder
season was added to better capture seasonality in wind generation. The load segments can also
be customized to account for time of day to better capture solar generation. The seasonal
structure and segmental configuration is reviewed with each update and might need to be
revised in the future to capture electric vehicle load.
2.8 Improving representation of behind-the-meter generation
To improve the representation of behind-the-meter generation, EPA has recently updated its
approach so that non-dispatchable distributed generation affects the shape of the load
duration curve, instead of simply reducing the net energy for demand used in the projection.
2.9 Increasing transparency of retail pricing model
EPA is improving the documentation for the Retail Price Model by providing further clarification
on and discussion of key components of the model. Additional improvements to the
documentation will also include an enhanced discussion of the purpose of the model, and
explain how that relates to the different methodologies for estimating retail price in
competitive and regulated regions.
2.10 Representation the of various policy mechanisms and publishing alternative/side cases
EPA has the capability to run a wide array of scenarios in IPM to inform and shed light on
important power sector projections. Previous iterations of IPM that have been released have
included alternative scenarios, for public dissemination and review. These scenarios have
included alternative assumptions for electric demand (high and low), renewable energy costs
(high and low), and natural gas price.EPA continues to consider, develop, and perform
alternative scenarios to inform its efforts to address pollution from the power sector, and will
continue such efforts. Where appropriate, EPA will release and disseminate scenarios to
accompany future IPM updates. In addition, EPA will consider such scenarios in other contexts
where IPM is being used, such as regulatory development.
2.11 Documentation improvements including results viewer
A number of documentation improvements were reflected in the Summer 2021 Reference Case
full-fledged documentation providing additional detail and clarity. These are tracked in the
Appendix table. Documentation updates will continue with each update as needed in light of
both formal reviews and comments received from stakeholder and user community.
8
-------
EPA has refined the Results Viewer to make it more intuitive and easier to use. The controls
were modified to automatically match between primary and comparison cases to make use
easier. The units displayed above charts were updated to clearly indicate the cases being
compared. And finally, the "Read Me" guide was edited and updated for clarity.
-------
Appendix: Table for Detailed Accounting of Peer Review Recommendations and Narrated Responses
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.1
ES
ii
1
Consider changes to
the model formulation
that would improve
the model's ability to
represent the ongoing
evolution of the
industry
represent the ongoing
evolution of renewable
industry
increases in the penetration of
renewables
2.1
ES
ii
1
Consider changes to
the model formulation
that would improve
the model's ability to
represent the ongoing
evolution of the
industry
represent the ongoing
evolution of EF
adoption
changes in load shapes... [from] electric
vehicles
2.1
ES
ii
1
Consider changes to
the model formulation
that would improve
the model's ability to
represent the ongoing
evolution of the
industry
represent the ongoing
evolution of storage
industry
changes in load shapes... [from] energy
storage
2.1
ES
ii
1
Consider changes to
the model formulation
that would improve
the model's ability to
represent the ongoing
evolution of the
industry
represent the ongoing
evolution of state and
regional policies
state and regional policies
10
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.1
ES
ii
1
Consider changes to
the model formulation
that would improve
the model's ability to
represent the ongoing
evolution of the
industry
represent the ongoing
evolution of changes in
LDC
revising the intra-annual load segments
2.1
ES
ii
1
Consider changes to
the model formulation
that would improve
the model's ability to
represent the ongoing
evolution of the
industry
represent the ongoing
evolution of modeling
for the power sector
solving the model chronologically
2.1
ES
ii
1
Consider changes to
the model formulation
that would improve
the model's ability to
represent the ongoing
evolution of the
industry
represent the ongoing
evolution of modeling
for the power sector
solving a companion model that
describes chronological demand and
system operation using capacity
assumptions from IPM
2.1
ES
ii
1
Consider changes to
the model formulation
that would improve
the model's ability to
represent the ongoing
evolution of the
industry
represent the ongoing
evolution of storage
industry
richer representation of energy storage
11
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.1
ES
ii
1
Consider changes to
the model formulation
that would improve
the model's ability to
represent the ongoing
evolution of the
industry
represent the ongoing
evolution of market
rules
incorporating changes in capacity
market rules into the model
(particularly as they relate to variable
renewable energy)
2.2
ES
ii
2
Clarify the types of
uncertainty that EPA's
Platform v6 is capable
of handling
Clarify the types of
uncertainty that are
not captured by the
model
documentation should provide guidance
to model users that more clearly
articulates the types of uncertainties
captured and not captured by the
model
2.2
ES
ii
2
Clarify the types of
uncertainty that EPA's
Platform v6 is capable
of handling
Clarify the types of
uncertainty and
address uncertainty in
a broader manner
consider evolution in the model
structure to address uncertainty in a
broader manner
2.3
ES
ii
3
Reconsider coal plant
turndown constraints
and possible addition
of operating reserves
Reconsider coal plant
turndown constraints
to determine if it
creates bias in coal
operations
EPA examine the turndown constraints
more closely to determine if they create
bias in coal plant operations, especially
in scenarios with low gas prices or high
renewable generation
2.3
ES
ii
3
Reconsider coal plant
turndown constraints
and possible addition
of operating reserves
Reconsider coal plant
turndown constraints
and consider operating
reserves as an
alternative solution
consider whether adding explicit
operating reserve requirements in the
dispatch would provide a better
representation of the impact of high
levels of renewable generators on the
grid
12
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.4
ES
iii
1
Consider
incorporating
upstream emissions in
addition to source-
level (power plant)
emissions
consider including
upstream emissions
consider including upstream emissions
in its reference case as a separately-
reported item (so upstream and stack
emissions are not combined together)
2.5
ES
iii
2
Distinguish between
investment decision-
making of utility and
merchant power
plants
Distinguish between
investment decision-
making of utility and
merchant power plants
Distinguish between investment
decision-making of utility and merchant
power plants
2.5
ES
iii
2
Distinguish between
investment decision-
making of utility and
merchant power
plants
Distinguish between
investment decision-
making of utility and
merchant power plants
evaluate whether a weighted average of
existing firms within a power region or
some other rule is a reasonable
representation of which type of firm is
more likely to make an incremental
investment
2.6
ES
iii
3
Address evolving gas
markets where Henry
Hub is less central to
pricing and where
emerging
petrochemical
production has
greater influence
Address evolving gas
market by describing in
the documentation the
model process for using
GMM
[Describe in the documentation the
model process for] iterating with the
Gas Market Model that generates the
natural gas supply curves and basis
differentials
13
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.6
ES
iii
3
Address evolving gas
markets where Henry
Hub is less central to
pricing and where
emerging
petrochemical
production has
greater influence
Address evolving gas
market by tracking
emerging
petrochemical sector
emerging petrochemical sector in the
Appalachian production region is likely
to affect regional natural gas pricing in
ways that may not be well represented
in the gas market model that EPA's
Platform v6 relies upon
2.7
ES
iii
4
Consider alternatives
to the current load
duration curves (LDCs)
Consider alternatives to
the LDC to better
account for inter-
regional trade
[How EPA] aggregates time into LDCs in
a way that... creates biases related to
the opportunities for inter-regional
trade
2.7
ES
iii
4
Consider alternatives
to the current load
duration curves (LDCs)
Consider alternatives to
the LDC
assess the trade-offs between different
approaches to aggregating load into
LDCs
2.8
ES
iv
1
Improve
representation of
behind-the-meter
generation
Improve representation
of behind-the-meter
generation
capture policies that encourage behind-
the-meter generation... [beyond]
represented as a change in demand
2.9
ES
iv
2
Increase transparency
of retail pricing results
Increase transparency
of retail pricing results
When EPA uses the RPM, we
recommend that the reporting of retail
rates be broken into component parts
so that the user can understand which
elements are endogenous to the model
and which are dominated by external
sources and assumptions
14
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.10
ES
iv
3
Consider
improvements in the
representation of
various policy
mechanisms
Improve representation
of policy mechanisms
such as dynamic
allocation of emission
credits
dynamic allocations within various
forms of emissions trading programs
such as output-based allocation under
cap and trade and a clean energy
standard
2.10
ES
iv
3
Consider
improvements in the
representation of
various policy
mechanisms
Improve representation
of policy mechanisms
such as EE
expenditures and
carbon pricing
expenditures on energy efficiency that
are linked to revenue from carbon
pricing
2.10
ES
iv
3
Consider
improvements in the
representation of
various policy
mechanisms
Improve representation
of policy mechanisms
such as flexible demand
ability to represent flexible demand that
may be encouraged at the retail level to
promote the integration of variable
renewable energy
2.11
ES
iv
4
More thorough citing
of sources and
expanded
explanations
throughout the EPA
Reference Case v6
documentation
Update the
documentation to
include the
development of the
load segments
development of load segments
2.11
ES
iv
4
More thorough citing
of sources and
expanded
explanations
throughout the EPA
Reference Case v6
documentation
Update the
documentation to
include treatment of
interregional trade
treatment of interregional trading
15
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.11
ES
iv
4
More thorough citing
of sources and
expanded
explanations
throughout the EPA
Reference Case v6
documentation
Update the
documentation to
include aggregation of
model plants
aggregation of individual plants to
model plants
2.11
ES
iv
4
More thorough citing
of sources and
expanded
explanations
throughout the EPA
Reference Case v6
documentation
Update the
documentation to
include more detail for
the retail price model
retail pricing model
2.1
2
2
3
Uncertainty
periodically review the
model to determine
whether model
structure should be
modified or
complemented with
other modeling
capabilities
EPA should periodically review the
model to determine whether EPA's
application of IPM model structure
should be modified or complemented
with other modeling capabilities
This is part of routine model development
process.
2.1
2
3
3
Chronological
modeling
restructure IPM as a
chronological model
restructuring IPM as a chronological
model
Possibility of making IPM a chronological
model is a significant task and may be
investigated. However, there is segmental
output information that can be used. In
addition, a production costing model such
as PROMOD can be used in conjunction
with IPM when required.
16
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.1
2
3
3
Chronological
modeling
develop a companion
short-run chronological
model of system
operation
develop a companion short-run
chronological model of system
operation that would enable comparing
the outcomes of the model's load
duration curves with a more realistic
characterization of the temporal nature
of demand
ICF has run GE MAPS for EPA, while
performing analyses in support of the
MATS rulemaking, for example. ICF runs
PROMOD production costing model
routinely and could setup such a
framework if so desired by EPA. PROMOD
is a chronological model that can be run
annually and does not make investment
decisions. ICF runs PROMOD either at the
interconnect level or at a subset of an
interconnect level.
2.1/2.10
2
3
4
Demand response
incorporate demand
response
consider incorporating additional
factors into the model's formulation of
demand response [including]... changes
in total electricity consumption in
response to changes in price
2.1/2.10
2
3
4
Demand response
incorporate demand
alternatives/substitutio
ns to electricity
consider incorporating additional
factors into the model's formulation of
demand response [including]...
substitution between electricity and
other forms of energy consumption
This is not done through model
formulation. Gas power plants are a form
of substitution between electricity and
other forms of energy consumption. A
similar approach can be evaluated to
estimate a kWh to Btu relationship and
can be used in IPM.
2.1/2.10
2
3
4
Demand response
incorporate changes in
the load shapes
consider incorporating additional
factors into the model's formulation of
demand response [including]... changes
in the load shapes that will be observed
and projected under different scenarios
This is not done through model
formulation but load shapes are adjusted
based on scenarios evaluated.
17
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.1/2.10
2
3
4
Demand response
incorporate variations
in supply-side short run
marginal costs (due to
increased VRE)
consider incorporating additional
factors into the model's formulation of
demand response [including]...
variations in supply-side short run
marginal costs (due to increased
penetration of variable renewable
energy)
In addition to the battery approach, we
could also utilize the DSM/EE option
functionality. NREL simulates flexible
demand/DR with a 100% efficient battery
that is time constrained. We can adopt a
similar approach in IPM to model DR
impacts under changing pricing patterns.
2.1/2.10
2
3
4
Demand response
incorporate retail TOD
pricing or retail pricing
linked to RE/clean
energy
consider incorporating additional
factors into the model's formulation of
demand response [including]...
potentially demand side retail prices
that vary by time of day or are linked to
resource availability directly require
cross-time-period analysis of electricity
demand
The endogenous demand response
capability allows us to estimate demand
response by load segment. In v6, we use
TOD based load segments and hence
demand response can indeed be linked to
TOD. Due to the TOD based load segment
structure, the generation from solar units,
for example, accounts for TOD. IPM is a
wholesale price model, which makes
linking to retail pricing very challenging.
We can make a simplification and allow
demand to move in response to wholesale
pricing.
2.7
2
3
5
Climate change
considerations
periodically evaluate
the model with respect
to weather
normalization of key
data inputs
recommend that EPA periodically
evaluate the model with respect to
weather normalization of key data
inputs
This has been evaluated in the past and
we will continue to do so.
2.7
2
3
5
Climate change
considerations
represent climate
change impacts in
generation
consider a more explicit representation
of climate change in the model's
specification of generation
On-going as scenario study.
18
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.7
2
3
5
Climate change
considerations
represent climate
change impacts in
transmission
consider a more explicit representation
of climate change in the model's
specification of... transmission
On-going as scenario study
2.7
2
3
5
Climate change
considerations
represent climate
change impacts in load
consider a more explicit representation
of climate change in the model's
specification of... load assumptions
On-going as scenario study
2.1
2
4
1
Transmission capacity
regularly revisit
implementation of
transmission
regularly revisit and, as appropriate,
revise EPA's implementation of
transmission outcomes and the
assumptions that shape anticipated
future transmission siting decisions
Transmission assumptions are regularly
updated in v6.
2.1
2
4
2
Storage
model energy storage,
including end-use
storage
rigorous treatment of energy storage
within the formulation of the model,
particularly with respect to the
opportunity to schedule demand and
achieve thermal and battery storage for
end-uses
Storage assumptions and parametrization
are regularly visited and updated as
needed in v6.
2.1
2
4
3
Capacity markets
represent resource
adequacy as they are
structured
representations of resource adequacy
requirements, as opposed to modeling a
generation reserve requirement
We will continue to monitor relevant
market developments and make
appropriate changes. The introduction of
the operating reserve constraint begins to
approximate this constraint.
2.11
2
4
4
Additional operational
constraint
include operating
reserve requirements
include operating reserve requirements
This is implemented. See Section 3.7
19
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.2
2
5
1
decision-maker
uncertainty
evaluate model results
in an option value
framework
improve the way they use model results
by explicitly considering them in an
option value framework
EPA has previously released and plans to
continue to release a range of scenarios
that outline a representative cone of
outcomes. These scenarios estimate the
impact of changing natural gas prices,
renewable technology costs, and demand.
2.11
2
5
2
runtime restrictions
publish runtime
restriction
requirements
any runtime restrictions required by the
EPA should be made explicit and used to
appropriately structure EPA's
application of IPM to the task at hand
This is a pragmatic preference rather than
restrictions.
2.10
3a
6
4
Demand
include electricity price
response in policy and
sensitivity cases where
prices vary significantly
We view the use of fixed electricity
demands without response to electricity
prices as problematic in policy and
sensitivity cases where prices vary
significantly from the Reference Case.
We recommend that EPA use this
feature when analyzing policy scenarios
that have significant price impacts
(perhaps roughly greater than 20%
variation in wholesale prices).
The capacity to perform demand response
already exists. EPA has used IPM's
demand response functionality while
conducting carbon policy analyses in the
past.
2.10
3a
6
4
Demand
publish in more detail
how the elasticity is
applied when used
recommend that the EPA Reference
Case v6 documentation describe in
more detail how the elasticity is applied
when used
When a certain parameter/capability is
used, they are always documented in the
corresponding side/alternative case or
policy case.
20
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.1
3a
7
2
Demand
develop better load
shapes over time
a more systematic way is needed to
develop load shapes over time rather
than just using a single metric of load
factors from ES&D or the AEO to shift
the curves
Usually, load factors are the only piece of
data that is available from AEO and NERC
projections. If future year load shapes that
underlie AEO or NERC demand projections
are available (for the use in side cases
possibly), then we can develop a
methodology to use those load shapes.
2.7
3a
7
4
Demand
evaluate using data for
a single year vs. a
multi-year average or
weather normalization
would be more
appropriate
recommend that EPA consider whether
using data for a single year creates any
biases and whether a multi-year
average or weather normalization
would be more appropriate
Both approaches could have pros and
cons for the various EPA applications
(including AQM). Using a multi-year
average could result in load shapes that
are very different from the original load
shapes and is not considered at this time.
2.11
3a
7
4
Overall
have consistency
among AEO vintages
used for data
assumptions
have consistency among AEO vintages
used for data assumptions
This is a goal but is hard to implement in
practice as not all parameters are
available or updated in any given year.
AEO or other sources, we strive to
incorporate most recent data available
with significance with every update.
Inevitably, not all data categories will
reflect the same calendar year or vintage
in any given IPM version.
2.3
3a
8
3
Dispatch
allow steam plants to
shut down for lowest
load time segments
when they run at full
capacity during the
peak segment
it appears that steam plants would not
be able to shut down for any time
segments (such as segments with
lowest load) if they are expected to run
at full capacity during the peak
segment.
Our turndown approach is an aspect of
the model that we have revisited regularly
and expect to continue to do so in the
future. Turndown constraints can be
reconfigured to allow coal plants to shut
down at time of lowest load. However,
this change should be considered with
care as to disallow overoptimization
through cycling.
21
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.3
3a
8
3
Dispatch, turndown
rate
publish information on
dispatch by time
segment
output files from EPA's Platform v6 do
not include information on dispatch by
time segment
Dispatch by time segment is not available
to either the public or EPA
2.3
3a
8
4
Dispatch, turndown
rate
develop turndown
rates based on
technical operational
considerations rather
than historical
economic
circumstances
the turndown constraints vary
considerably by unit, and some are as
high as 80% with most of them between
40% and 60%. Because these values are
based on historical operations rather
than current or projected engineering
considerations, they may reflect
historical economic circumstances that
may not apply in the future
The turndown assumptions can be
updated based on current data to reflect
the current operating behavior of coal
plants. Our turndown approach is an
aspect of the model that we have revisited
regularly and expect to continue to do so
in the future. In addition, if the low gas
price environment persists, then the
assumptions could also be relaxed
(turndown targets lowered) to reflect
increased cycling.
2.3
3a
8
5
Dispatch, turndown
rate
examine the turndown
constraints for bias in
coal scenarios with low
gas prices or high
renewables
recommend that EPA examine the
turndown constraints more closely to
determine if they create bias in coal
plant operations, especially in scenarios
with low gas prices or high renewable
generation.
We are currently working on this.
2.1
3a
9
1
Dispatch
add operating reserve
requirements
consider whether adding explicit
operating reserve requirements in the
dispatch would provide a better
representation of the impact of high
levels of variable renewable energy
This feature is implemented in the current
platform. See Section 3.7
22
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.1?
3a
9
2
Dispatch
analyze historical
generation versus the
model patterns by time
period for hydro
consider analyzing historical generation
patterns versus the model patterns by
time period to assess whether EPA's
application of IPM is significantly
overoptimizing hydro generation
The concern about overoptimization is
primarily related to run-of-river hydro
units that do not have storage. In the
current update, we have aggregated run-
of-river hydro units separately and then
model their generation through a
generation profile based on recent
history. See Section 3.5.2
2.1
3a
9
4
Transmission
update transmission
loss assumptions
The application of a 2.4% transmission
loss to each interregional transfer
strikes us as high for the Eastern
Interconnect, especially given the size of
the model regions and hence relatively
short distances for many of these
transfers. For example, in NEMS a 2%
loss factor is assumed for transfers
between regions and there are fewer
regions.
We are currently evaluating this
recommendation and can easily
update/implement.
2.1
3a
9
5
Transmission
perform sensitivity
cases where
transmission capacity is
added
performing sensitivity cases in which
additional transmission capacity is
added exogenously
The recommendation is to perform
sensitivity analyses where we exogenously
add transmission capacity. We have
incorporated endogenous transmission
builds, this sensitivity analysis may be
unnecessary.
23
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.3
3a
10
3
Capacity Expansion -
Setting the Capacity
Targets
add reserve ramping
constraint
application of IPM may need to be
modified to reflect other capacity
requirements beyond a simple reserve
requirement and declining capacity
values for variable renewable capacity.
For example, it may be appropriate to
add an additional reserve constraint
requiring a percentage of capacity is
capable of meeting a certain ramping
capability
We monitor the electricity markets
continuously and make updates to the
model to be consistent with changes in
the markets. As the resource mix changes
and requirements like California's or other
mechanisms become more common, the
model will be updated accordingly to
account for the changing dynamics. A step
in the direction is the incorporation of the
operating reserves constraints in v6. We
do not believe additional reserve products
are necessary for the scenarios we are
currently pursuing but will continue to
evaluate this moving forward.
2.3
3a
11
2
Capacity Expansion -
Setting the Capacity
Targets
modify the short-term
supply cost adders for
capacity expansion
The cost adders to capacity expansion
costs when expansion is rapid... are
quite steep with roughly a 45% cost
penalty on the second step. It might be
better to have smaller initial steps with
smaller cost penalties for the second
step.
These constraints are applied to all new
plants for the 2021-2035 run years.
However, they usually get activated for
solar and wind builds in runs having
stringent RPS/CES standards. In EPA v6,
the short-term capital cost adders step
widths are from AEO. However, the
approaches differ from AEO in the sense
that in IPM we are not updating the step
widths to account for the IPM builds.
2.3
3a
12
2
Capacity Expansion -
Rating the Capacity of
Alternative Resources
update the solar
capacity credits to
latest AEO
if the solar capacity credits are still
benchmarked to those of the AEO2017,
as indicated in the documentation, this
should be revisited because the AEO
methodology and resulting credits for
solar have changed considerably since
the AEO2017 was published
Solar capacity credits are no longer being
benchmarked with the AEO version. See
Section 4.4.5
24
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.5
3a
12
2
Capacity Expansion -
Rating the Capacity of
Alternative Resources
examine capacity
market rules and
consider the
implications of non-
performance risk
we recommend that EPA examine the
capacity credit methodology as system
operators change their capacity market
rules and consider the implications of
non-performance risk
This is standard practice, although there is
a balance between chasing today's rules
versus the 'true' value, as understood by
the model. We will continue to monitor
the changing rules.
2.5
3a
12
3
Capacity Expansion -
Rating the Capacity of
Alternative Resources
account for demand
response and energy
efficiency in capacity
markets
worth noting that demand response and
energy efficiency are providing non-
trivial shares of total capacity and even
larger shares of new capacity in many
capacity markets
We are continuing to plan for how to
incorporate these resources into our
modeling projections.
2.1
3b
13
1
Storage
incorporate additional
storage technologies
into the model
recommends that EPA consider
incorporating additional storage
technologies into the model
Work is ongoing.
2.1
3b
13
1
Storage
regularly revisit energy
storage cost,
performance, and
market assumptions
because the technologies, cost
structure, performance, operating
strategies, market rules, and regulations
related to storage are rapidly changing,
EPA may need to regularly revisit the
model's representation of storage
Ongoing work. We routinely consider this
for all technologies including storage.
2.1
3b
13
2
Storage
update energy storage
technologies, costs,
and operational
assumptions regionally
consider regional variations in energy
storage technology, costs, and
operations
In the current version our implementation
has regional variations of cost and
capacity credit. See section 4.4.5
25
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.1
3b
13
3
Nuclear
consider more flexible
nuclear dispatch
consider more flexible nuclear dispatch
in EPA's application of IPM
Nuclear dispatch is already flexible and is
not hardwired. The reviewers appear to
suggest that we model nuclear O&M costs
as a function of the level of dispatch. EPA
is participating in inter-agency workgroups
to research and implement updates in
modeling as needed.
2.1
3b
13
4
Heat Rates
vary heat rates by
season and over time
recommend that the heat rates of
generating units in EPA's application of
IPM vary by season and perhaps over
time
Heat rates are less impacted than capacity
by change in temperature. This issue is
less important as compared with the
impact on capacity. We considered this in
the past but have not found value for our
applications so far. It might be interesting
to consider grid reliability / reserves in
light of units that might not be able to get
the cooling water they need and therefore
have to limit generation. But that would
be either more episodic and hard to
reflect in a long term capacity expansion
model or would be considered as a side
case evaluating warming impacts.
2.1
3b
14
2
Heat Rates
vary the available
capacity of a given unit
by season
recommend that EPA's application of
IPM vary the generation capacity of a
given unit by season, or add text to the
documentation explaining why seasonal
variation is not necessary
This primarily impacts CT and CC units.
The primary impact is we might be
underestimating generation potential in
the winter season. We will evaluate the
LOE required to implement this feature in
IPM.
2.7
3b
14
3
Generation
Assumptions
update generation over
time to account for
climate change
consider adjusting the EPA's Reference
Case generation assumptions over time
to account for climate change
This is a scenario case.
26
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.1
3b
14
4
Generation
Assumptions
vary generation
assumptions by
market/regulatory
environment
consider varying some generation
assumptions by market/regulatory
environment
For existing units, EPA uses unit specific
heat rates, emission rates, and technology
assumptions. In addition, where possible,
unit and state level emission regulations
are also modeled in detail. Some of the
other assumptions such as unit level
availabilities can be used if such data is
available.
2.4
3c
15
1
Assignment and Scope
of Emissions Factors
document upstream air
emissions in fuels
prices or in generator
marginal costs
consider documenting how upstream
air emissions are reflected in fuels
prices or in generator marginal costs
within its Power Sector Modeling
Platform
Upstream air emissions can be estimated
through post processing. However, the
challenge will be in defining the scope of
what constitutes upstream and then
developing the associated emission
factors.
2.11
3c
15
3
Emission Control
Options
periodically review the
technology options for
emissions control
suggest that EPA periodically review the
technology options for emissions
control in EPA's application of IPM to
determine if this portion of the model
could be made simpler with the
reduction of emissions control
technologies from which modeled
plants can choose
This is always considered with major
updates.
27
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.11
3c
15
4
Emission Control Costs
publish the raw
engineering data used
to develop the unit cost
values
Chapter 5 of the EPA platform v6
documentation includes unit cost
estimates derived from the Sargent and
Lundy study but does not provide a
formal citation for the study or the raw
engineering data used to develop the
unit cost values. Publication of these
data would make the cost figures used
by EPA's Platform v6 more transparent
than they are currently. We recommend
that EPA consider the costs and benefits
of this additional data transparency as
weighed against the benefits of being
able to access and use proprietary data,
which in some cases may be more
granular or up-to-date than data
existing in the public domain.
As of 2022, we are working with S&L and
in the process of updating reports.
2.11
3c
16
1
Emission Control Costs
periodically compare
emissions control cost
data with publicly
available data
EPA should also periodically compare its
emissions control cost data with
relevant information that exists in the
public domain, such as the Integrated
Environmental Control Model (IECM)
developed by Carnegie-Mellon
University.
We will consider this.
28
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.11
3c
16
2
Emission Control Costs
include natural gas
combined cycle plants
with carbon capture as
a capacity option
the option to choose natural gas
combined cycle plants with carbon
capture appears to be turned off within
the model... we recommend that EPA
restore this technology option under
relevant analyses.
These options are currently back in v6.
2.11
3c
16
3
Emission Control Costs
document interactions
between IPM and
GeoCAT
recommend that EPA incorporate
additional specificity... in the
documentation... [relating to] any
interactions between IPM and GeoCAT.
IPM and GeoCAT are never iterated
together, nor there are any interactions
between the two tools. Updated
documentation provides additional detail.
Please see Section 6.2
2.11
3c
16
4
Emission Control Costs
re-evaluate the oil price
assumption related to
EOR and the C02
storage cost curves
should re-evaluate the oil price
assumption related to EOR... [and] re-
evaluate the C02 storage cost curves
On-going work, we update oil prices
regularly.
2.11
3c
16
5
Emission Control Costs
remove C02 transport
pipeline economies of
scale and document
model approach
Some elements of the C02 transport
model are also not clear, particularly
related to the economies of scale in
pipeline transportation. The method
described in Section 6.3 of the
documentation appears to assume that
C02 sources that are transporting C02
over longer distances for long-term
geologic sequestration are taking
advantage of some undescribed scale
economies in the form of capacity
sharing in C02 pipelines.
In the latest reference case, EPA is no
longer accounting for scale economies
while estimating the cost of C02
transportation. Please see Section 6.3
29
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.5
3d
17
5
Power Sector Finances
and Economics
add equation and
reference for the
capital charge rate
The description of the calculation of the
capital charge rate would be made
substantially more clear with an
equation. In particular, whether EPA's
Platform v6 uses the common "short
cut" version of the capital charge rate
(Stauffer, 2006) could be made more
clear
The reviewers indicated a desire to have
more information on the capital charge
rate and to ensure that the concerns of
Stauffer (2006) are not affecting the
capital charge rate. Stauffer, a founder of
ICF, indicates that there could be
confusion between real and nominal
capital charge rates and input parameters,
which is not a problem. He may have
indicated other concerns, but we have not
reviewed his 2006 article in detail. We
could review the article and determine
next steps, if any.
2.5
3d
17
6
Power Sector Finances
and Economics
describe the debt life
versus the asset life
recommend an explicit statement in the
documentation describing the debt life
versus the asset life.
In general, the debt life is shorter than
the book life. This is based on the tenure
of debt, especially in the IPP sector. See
Section 10.10.2
2.5
3d
17
7
Power Sector Finances
and Economics
update assumptions on
debt-to-equity ratios
and the cost of
merchant debt
assumptions on debt-to-equity ratios
and the cost of merchant debt, which in
the market environment at the time of
this writing may be high. EPA's Platform
v6 uses a value of 7.2%, but one of the
stated data sources for debt-to-equity
ratios currently suggests that the cost of
debt may be substantially lower.This is a
data point that we suggest be updated
in future revisions of EPA's Platform
Financial assumptions are regularly
updated, and D:E ratios are one of the
metrics we closely track.
30
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.5
3d
18
2
Power Sector Finances
and Economics
use a different WACC
formula with a constant
leverage ratio
[Consider using a different WACC
formula other than Hamada] Brealy and
Myers (2011) point out that a constant
leverage ratio is a more realistic
assumption
The Hamada equation is used to adjust for
differences in the reported debt to equity
structure and the targeted structure. The
reviewers point to a source that they
assert favors a constant leverage
assumption. In the case of IPPs, there has
been very high debt shares, large amounts
of financial distress, especially in some
periods. Accordingly, we believe this
unusual situation warranted adjustments
to more sustainable debt levels.
31
-------
Power Sector Finances
and Economics
account for
differentiated risk
appetite of utility
versus merchant
investment costs of
capital
consider addressing this differentiated
risk appetite [of utility versus merchant
investment costs of capital] in future
versions of EPA's modeling platform.
One possibility would be to introduce
different hurdle rates for different
investor decision-makers and effectively
split investment decisions within EPA's
application of IPM.
While new build financing assumptions
are not differentiated based on
utility/merchant categorization, retrofits
do include this differentiation. This in turn
results in more realistic
retrofit/retirement decisions for the
existing fleet. Within a cost minimizing
framework assuming differentiated
financing for new builds would result in
possible over-builds and under-builds
because of effective differences in
levelized costs. Based on prior runs, these
builds may be unrealistic in their
concentrations. Instead, IPM assumes a
weighted average financing charge for all
new builds of a given technology type.
In the peer review, the issue that was
identified as the most important is the use
of a weighted average cost of capital of
regulated utilities and merchant
powerplants. In nominal terms, the
WACCs of utilities are 4.9% versus 6.7%
for IPPs; the weighted average is 5.6%.
The latest modeling bases its financial
assumptions on a 60:40 utility: merchant
weighting. The 60:40 weighting
approximately equals the 2015-2019
average for renewable and thermal
additions in the US. The concern is that
the use of the average may not
adequately characterize the financing
costs. The peer review suggests
designating some regions as regulated
utility and others as merchant IPP. The
decision to use an average was based in
part on the uncertainty about the
structure in the long term. The approach
32
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
also reflects the concern that there could
be an unrealistic skewing of regional
results. Namely, low capital cost regions
would disproportionately make capital
investments including disproportionately
investing for export to high capital cost
regions.
33
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.11
3d
19
4
Power Sector Finances
and Economics
update documentation
on tax credits for wind
energy
The documentation could also explain
more clearly how tax credits for wind
energy are treated.
The tax credits for both PTC and ITC
are modeled as a reduction in the
levelized capital costs of those resources.
We provided better documentation.
2.11
3e
20
7
Coal
update documentation
on coal mine closures
there is not enough information
provided in the documentation to
discern whether coal mine closures are
exogenous or endogenous within the
model and the degree to which closures
in the model reflect recent changes in
regional fuel supplies
2.11
3e
21
2
Coal
evaluate differences
between ElA's and
EPA's coal
prices/supply to ensure
consistence across
other sector demands
Because the AEO2017 view of coal
prices and supplies reflected in export
and non-electric sector demand may
not match EPA's view of coal prices and
supply, the projections of other sector
demands and exports may be
inconsistent with power plant demand.
We acknowledge that there may be
inconsistencies in our current approach.
There will always be seams between IPM
and AEO, we will continue to investigate
to limit them and their impact.
2.11
3e
21
2
Coal
keep the base
projections for coal up
to date
consider keeping the base projections
up to date (using AEO2018 (or AE02020
if an update is done) versus AEO2017)
This is always considered and usually
implemented at every update.
34
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.6
3e
21
4
Natural Gas
run the GMM and IPM
iteratively when
necessary
One disadvantage of this static curve
approach is that it treats prices in
different years as independent in EPA's
Platform v6 context, rather than as a
function of cumulative production that
may vary by EPA scenario, even though
the underlying curves were developed
with that consideration by GMM. This
can be addressed by re-estimating the
curves by running the models
iteratively, as in the Reference Case set-
up, when it seems necessary due to
significant changes in gas demand.
While IPM's endogenous gas model can
address this issue, an alternate approach
is to regenerate the gas supply curves
whenever there is significant divergence
in the gas demand relative to that in the
reference case. An initial Ref Case is set up
by iterating between GMM and IPM. But
we do check when/if the static curves are
no longer appropriate for a given case.
35
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.6
3e
21
4
Natural Gas
publish methodology
for deriving gas supply
curves
recommend that EPA publish more
information about the methodology for
deriving the curves
The slopes of the gas supply curves are
derived based on ICF's assessment of
change in natural gas prices historically
based on several parameters like rig
count, production, etc. and certain
assumptions of the natural gas resource
base moving forward to determine the
short-term and long-term supply elasticity
that feed into the supply curves. The way
the supply curves are built is that they are
more elastic over time compared to the
short-term elasticity as the resource base
can respond to price changes. In other
words, the short-term elasticity is higher
than the long-term elasticity. More
elaborate documentation is provided. See
Section 8.2.1
2.6
3e
22
1
Natural Gas
describe how LNG
exports are determined
It is also not clear the degree to which
LNG exports, both export capacity
expansion and utilization, are
determined endogenously versus
predetermined.
ICF assumption of LNG exports for EPA
base case is exogenous; however, GMM
has the capability to change the LNG
exports over time in response to change in
natural gas prices. More elaborate
documentation is provided. See Section
8.3.5
36
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.6
3e
22
2
Natural Gas
update seasonal gas
price differentials
across scenarios
The GMM also serves as the basis for
seasonal price differentials that capture
the difference between the Henry Hub
price and gas prices in model regions.
While these differentials are
endogenously projected by GMM with
variable costs as a function of pipeline
throughput and pipeline capacity
expansions, they are fixed in a given
scenario context.
Under scenarios with major changes in
natural gas demand regionally, the change
in basis can be captured by GMM based
on the pipeline infrastructure build-out
necessary to support the demand growth
under that particular scenario. However,
this requires iterations between GMM and
IPM.
2.6
3e
22
3
Natural Gas
include petrochemical
sector demand in the
GMM
Within the GMM, econometric
equations project other sectoral
regional gas demands. The elasticity of
these demands presumably impacts the
overall supply elasticity of gas to the
power sector. We note, however, that
an emerging petrochemical sector in the
Appalachian production region is likely
to affect regional natural gas pricing in
ways that may not be well represented
in the gas market model.
GMM base case forecast for EPA base case
projects significant growth in natural gas
production from the Marcellus and Utica
region from 2019 through 2050 (about 25
Billion Cubic Feet per Day) which does
account for growth in NGL demand and
exports from the Appalachia region
exogenously.
37
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.6
3e
22
5
Other Fuels
ensure consistency
across oil price
assumptions between
IPM and GMM
oil prices are treated inconsistently
across EPA's Platform v6 and ICF's GMM
platforms. While oil prices for power
generation are based on the AEO2017,
diesel fuel prices used in developing rail
rates for coal are from the AEO2016. At
the same time, oil prices used in the
GMM, which are used to determine fuel
switching in the industrial sector, are
quite different from those from the AEO
that are used in the rest of EPA's
Platform v6
In the future reference cases, DFO and
RFO fuel prices will be made consistent
with the crude oil price projections used in
the GMM.
2.1
3e
23
4
Renewable Resources
apply generic
transmission costs to all
units including wind
and solar
It would seem more consistent for the
generic transmission network costs to
be applied to all units rather than
exempting wind and solar. Otherwise
this provides a bias towards wind and
solar PV development
These costs are currently applied in the
v6. Documentation incorporated better
explanation of distance to transmission vs.
generic transmission network costs,
aligning NRELand AEO approach as much
as possible. Please see Section 4.4.2
2.7
3f
24
5
Regional and
Temporal Resolution
evaluate differences in
peak load and peak
net-load
Load aggregation can dilute outcomes
that are concentrated into a small
number of hours... EPA's Platform v6
addresses this well by specifying a very
high peak load segment, representing
only 1% of all hours. However, key
transient outcomes in the system may
not be limited to only peak hours,
particularly with extensive adoption of
renewable energy resources.
We will investigate this in the near future.
38
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.7
3f
24
6
Regional and
Temporal Resolution
evaluate load
aggregation
implications regarding
inter-regional trade
Load aggregation necessitates difficult
modeling choices regarding inter-
regional trade... diagnose the full
impacts of this implementation. Our
intuition is that it constitutes a hidden
penalty on trade between regions; in
order to export during hours in which
trade is beneficial, the model may be
forcing additional trade in hours in
which trade is not beneficial. If true, this
means the model will bias downward
trade between regions.
The intuition is correct. There could be
hours when power might be exported
during hours when it might not be
needed. Additional analysis is not required
to confirm this assessment.
2.7
3f
26
4
Regional and
Temporal Resolution
document interregional
trade
One additional comment on this point is
that the documentation does not
describe this aspect of interregional
trade.A description with an
accompanying example would help
promote understanding of this feature
of the model
2.7
3f
26
5
Regional and
Temporal Resolution
evaluate aggregating
load over a larger
geography
Geographic aggregation involves trade-
offs between accuracy over time vs.
space... One way to reduce the
problems identified above [related to
inter-regional trade] is to aggregate
over larger geography.
We will investigate this in the near future.
39
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.7
3f
27
1
Regional and
Temporal Resolution
evaluate the model's
ability to represent
regulations focused on
peak or episodic
emissions
Load aggregation limits modeling of
inter-temporal constraints... For
regulations that concern total output or
emissions from a power plant during a
season or year, the aggregation is likely
relatively benign. However, for the
purposes of assessing any
environmental regulations focused on
peak emissions, episodic emissions, or
emissions intensity, the aggregation
could be more problematic.
This will be considered in the future if
such policy design is necessary.
2.7
3f
27
4
Regional and
Temporal Resolution
Publish model outputs
by load-segment
Publish more output details: Currently
model outputs are not broken out by
load-segment. This additional output
detail may allow stakeholders to better
judge the relative impacts of the various
aggregation assumptions in a given
policy context
Dispatch by time segment is not available
to either the public or EPA.
Duplicate with row 55
2.7
3f
27
5
Regional and
Temporal Resolution
evaluate tradeoffs
between regional and
temporal aggregations
Investigate the Time vs. Geography
Trade-off: It is possible that the goals of
the model may be better implemented
with more temporal resolution and that
this could be aided by less geographic
resolution
2.7
3f
28
2
Regional and
Temporal Resolution
evaluate grouping
hours first by time of
day and then by load
segment
Consider grouping hours first by time of
day and then by load segment, instead
of the other way around.
The current approach was implemented
for simplicity. The alternate approach can
also be implemented. Such an approach
will eliminate the possibility of load
segments having zero hours.
40
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.7
3f
28
3
Regional and
Temporal Resolution
evaluate grouping
hours first by load
segment for a whole
interconnection and
then by region
Investigate the implications of grouping
hours first by load segment for a whole
interconnection and then by region...
group hours by their interconnection-
wide load level and then subdivide into
regions. For example, the top 37
summer hours would be chosen from
the hours with the highest total load
across the WECC.
We will investigate this in the near future.
2.7
3f
28
4
Regional and
Temporal Resolution
evaluate grouping
hours into time-of-day
blocks (e.g., 4 hours)
and model them
sequentially
Represent time as a sequence of "model
hours" or "model days." ... One
alternative would be to group hours
into time-of-day blocks (e.g., 4 hours)
and model them sequentially, allowing
for better representation of some inter-
temporal constraints, inter-regional
trade, and probably renewable energy
and storage output. One such "model
week" per season could capture a peak
day and other important load
characteristics.
A sequential hour approach may not work
in models that are based on a load
duration curve. Needs further evaluation.
2.7
3f
28
5
Regional and
Temporal Resolution
evaluate grouping
hours into time-of-day
blocks for typical
weekdays and weekend
days with a
preservation of peak
loads through a peak-
day or other method
Represent time as a sequence of "model
hours" or "model days." ... Another
alternative would be to group hours
into time-of-day blocks for typical
weekdays and weekend days with a
preservation of peak loads through a
peak-day or other method.
IPM has the capability to separate load
segments based on weekday and
weekend days in addition to TOD. We
have performed some test runs. This
functionality can double the number of
load segments.
2.7
3f
28
6
Regional and
Temporal Resolution
link the model results
with a dispatch model
Run the output of a model scenario
through a more detailed dispatch
model.
41
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.7
3f
29
2
Regional and
Temporal Resolution
model fewer future
years to compensate
for more detail within a
given model year
Model fewer future years. One way to
compensate for more detail within a
given model year would be to run fewer
model years. Seven, instead of eight,
explicit model years would reduce the
number of demand segments (region +
hours).
Outputs serve multiple purposes of EPA
applications and we do consider output
years carefully. The future run years in
five-year increments are also important
due to significant reductions in new unit
costs and the start of several state level
clean energy standards.
2.5
3f
29
3
Regional and
Temporal Resolution
evaluate using a higher
discount rate in the
objective function
Consider the impact of the discount rate
in the objective function... An even
higher discount rate may be appropriate
to minimize the impact of out-year
decision making on model outcomes
We use discount rates based on our
financial analyses. Further thought needs
to be given in regards to use of discount
rates whose primary function is to reduce
the impact of out years on model
outcomes. Previously we used a post-
processing tool to change policy cost NPV
with different discount rate, not in the
objective function, but it is a tool for
evaluation.
2.10
3g
29
5
emerging policy and
industry issues to
consider
model dynamic
allocation of emissions
allowances in cap and
trade programs
Although trading programs are
represented well in EPA's Platform v6,
the documentation indicates that the
model does not include any explicit
assumptions on the allocation of
emission allowances among model
plants under any of the programs. An
element of cap and trade that may be
challenging to model is dynamic
allocation of emissions allowances that
maintains the emissions cap.
IPM has the capability to model output-
based allowance allocations methods and
has performed such analyses in the past.
However, there may not be an immediate
need for this functionality. The cap and
trade programs promulgated by EPA do
not account for this policy lever; it is not
one of the central policy parameters
under discussion, and is not utilized
anywhere at the moment (including
RGGI).
42
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.10
3g
30
2
emerging policy and
industry issues to
consider
model clean energy
standards with
effective emissions
targets that adjust over
time in response to the
quantity of production
challenges arise in representing clean
energy standards, which are emissions
intensity standards with effective
emissions targets that adjust over time
in response to the quantity of
production.
These constraints can be approximately
modeled through a set of two constraints
in IPM. Constraint 1 can model the
emission intensity standards through
Ibs/MWh constraints and Constraint 2 can
model the effective emissions targets
through cap based constraints. We are
currently exploring issues related to CES
policy design.
2.10
3g
30
3
emerging policy and
industry issues to
consider
model dynamic
allocation of emissions
allowances in cap and
trade programs
Other challenging elements in
representing power sector
environmental policies include dynamic
adjustments to emissions budgets
based on the prevailing price in an
auction, as illustrated by the emissions
containment reserve in the Regional
Greenhouse Gas Initiative
2.10
3g
30
5
emerging policy and
industry issues to
consider
document the
interactions between
electricity
sales/transmission and
renewable energy
credit markets
there is an interaction between
electricity sales and transmission, and
renewable energy credit markets.
Although we understand that this
interaction is embodied in the model,
we have not seen it represented in
previous exercises of the model or
described in the documentation
IPM solves for the power, fuels, and
environmental markets simultaneously.
The interaction among these markets and
within these markets are modeled
endogenously in an integrated manner.
We provided further detail in
documentation. Please see Section 2.3.10
43
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.10
3g
31
1
emerging policy and
industry issues to
consider
evaluate leakage risk
between RPS policies
and carbon pricing
policies
Policies such as carbon pricing that
promote an increase in renewable
generation in one region could
precipitate a decrease in renewable
generation in another region if
renewable energy credits become
available in the region introducing
carbon pricing that can be used for
compliance with renewable portfolio
standards in other jurisdictions.
Such an analysis may only be relevant
when EPA is designing or promulgating
any of these policies. When necessary,
leakage issues will be addressed in policy
contexts in new analyses or rulemakings
where this is relevant.
2.10
3g
31
2
emerging policy and
industry issues to
consider
evaluate impact of the
New Source Review in
constraining existing
generation and limiting
new investments
Some prescriptive policies such as New
Source Review constrain the utilization
of an existing generating unit and limit
investments in new units in a
geographic area. The Agency should pay
special attention to this in evaluating its
modeling
As part of the flat file generation process,
emissions from new units are not assigned
to areas which are non-compliant. This
approach can work in instances where
only a subset of an IPM region is affected
by these requirements. However, if an
entire region is affected by these
restrictions, then we may need to disallow
the build of such units in those regions.
2.10
3g
31
3
emerging policy and
industry issues to
consider
model state level
policies that encourage
behind-the-meter
generation
The model seems to capture policies
that affect the bulk power system - but
does not seem to capture state level
policies that encourage behind-the-
meter generation except represented as
a change in demand.
The model could be developed and
enhanced to do this kind of analyses,
however this is not our priority currently.
We can address this in other ways using a
more simplified approach. This is
something we would model separately
and provide it as an input into IPM.
44
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.10
3g
31
4
emerging policy and
industry issues to
consider
account for MOPR rules
impact on nuclear and
capacity market
compliance
if EPA models nuclear generation
incentives from state-level zero-
emission-credit (ZEC) policies, it would
also need to evaluate whether those
nuclear plans should count towards
satisfying a regional capacity constraint.
Under final MOPR rules, which are as of
yet to be determined, such nuclear
capacity may not be part of capacity
market compliance
Once the final rules are known, such units
can be provided with zero capacity credit.
2.10
3g
32
2
emerging policy and
industry issues to
consider
improve the
representation of the
C02 emissions rate for
imports to California
one last finding pertains to the
representation of the C02 emissions
rate for imports to California, at 0.428
MT/MWh... A careful solution to this
could be found through iteration,
solving the model twice varying the
level of demand in California in order to
identify the marginal resource providing
power to the state and region, but this
may require additional Agency
resources.
The approach implemented was a tradeoff
to minimize complexity. Potential
approaches can be investigated to better
represent AB 32. One potential solution is
to focus more on the recent CA clean
energy and RE requirements. However, it's
not clear to identify a good path towards
modeling imports into CA. We will
continue to work on the methods for how
to improve upon our current approach.
2.9
3h
33
1
Retail price estimates
document the purpose
of the NUG adder in
RPM
While it is not obvious that the capital
costs of a merchant NUG would be
directly passed on to retail rates, we
assume the NUG adder is included
because these costs are captured in
long-term contracts between the
generator and the local load-serving
entity (but this is not explained in the
documentation)
We will include this in the documentation.
45
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.9
3h
35
1
Retail price estimates
document the purpose
of the NUG adder in
RPM
The general structure of the generation
pricing formulas seems appropriate, but
the purpose and magnitude of
elements, such as the NUG adder,
should be better explained.
We will provide better documentation
2.9
3h
35
2
Retail price estimates
define the difference
between competitive
vs. regulated regions
The logic behind the definition of
competitive vs. regulated regions is
unclear. There are many possible
definitions of "regulated" and
"competitive" and the RPM utilizes
definitions from ElA's Annual Energy
Outlook for its assignments of regions
to these categories. This may not be the
best definition for this application.
We will provide better documentation
2.9
3h
35
4
Retail price estimates
publish more of RPM
results and their
components
Based on the above comments, we
recommend that EPA improve
transparency of the RPM results and
their components... One suggestion
would either be a table or stacked bar
chart detailing not just the total rate (or
change in rate) but also the components
that make up that total. Most of these
details are available but take
considerable effort to put together.
We consider RPM as a first order price
generation tool. It is also used more for
estimating the change in prices rather the
absolute level of prices. Providing retail
price components might be confusing and
a digression. Provided better
documentation and disaggregated
impacts.
2.9
3h
73
1
Retail price estimates
define the difference
between competitive
vs. regulated regions
EPA should evaluate and articulate the
purpose of distinguishing between
competitive and regulated regions
46
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.9
3h
37
2
Retail price estimates
consider regression
analysis for estimating
the retail rate
Consider a simpler retail price formula
based upon regression analysis of the
relationship between generation costs
and retail rates over time.
The purpose of the RPM is to measure the
impact of a policy on prices.
2.2
4
38
4
non-parametric
uncertainty
show how behavior
may change or may
depart from expected
net present value
maximization in the
presence of uncertainty
one limitation of the focus on
parametric uncertainty is that sensitivity
analysis does not show how behavior
may change or may depart from
expected net present value
maximization in the presence of
uncertainty
Investment decisions can be impacted
under uncertainty. If the intent is to create
a strategy that is robust across a range of
futures, then we can evaluate an
approach such as a stochastic LP that can
generate robust results under a range of
scenarios. Additionally, this does not have
to be analyzed through IPM development
but a conceptual or weight-of-evidence
response. We have been evaluating this
for retirements.
2.2
4
39
1
non-parametric
uncertainty
adjust the hurdle rate
for investment and
retirement options to
account for option
theory
option theory suggests rational decision
makers will delay irreversible
investments (and retirements) in the
face of uncertainty to gain more
information about the uncertain aspects
of the scenario. This behavior will not
be evident in an inter-temporal
optimization linear program such as
IPM. However, this element of decision-
making under uncertainty might be
represented by adjusting the hurdle
rate for investment and retirement
options, perhaps implemented as a
shadow cost of capital for investments
that would be vulnerable to specific
parametric uncertainty.
This does not have to be evaluated
through IPM development but a
conceptual or weight-of-evidence
response.
47
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.2
4
39
3
parametric
uncertainty
publish which
parameters or
combinations of
parameters most
heavily influence model
outputs
Beyond the scenario runs made public
by EPA, it is not clear what sensitivity
analyses EPA conducts to determine
which parameters are the most
important in determining variation in
model outputs. EPA's application of IPM
is so complex that it may be the case
that no single parameter is driving the
model outputs all by itself. Some
attempt at investigating and publishing
which parameters or combinations of
parameters most heavily influence
model outputs in the Reference Case
would be very useful.
Having methodical and documented
sensitivity runs when we go through the
development phase would be a very costly
and time-consuming undertaking. We do
this on an ad-hoc basis (dozens of
sensitivities are run/tested); and many
times we do not obtain the full outputs of
test runs (as they are not necessarily fully
QA'd) but assess what the results
directionally suggest. Also, not all of those
sensitivities can be planned beforehand
but as the need emerges (in conjunction
to other updates being made). Best
method would be to envisage/group these
runs retrospectively after we have arrived
at a reference case. This would still render
significant additional effort but it would
be a more concise and targeted work
serving documentation purposes and
justifying various assumptions/updates
made. We will consider this when we have
resources.
2.2
4
40
1
parametric
uncertainty,
loadshapes
model scenario that
includes changes in the
shape of the LDC from
vehicle electrification
Changes in the shape of the load
duration curve: (1) vehicle
electrification
This is ongoing work.
2.2
4
40
1
parametric
uncertainty,
loadshapes
model scenario that
includes changes in the
shape of the LDC from
TOD pricing
Changes in the shape of the load
duration curve: (2) time-varying retail
rates that encourage load shifting and
peak-time demand response,
48
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.2
4
40
1
parametric
uncertainty,
loadshapes
model scenario that
includes changes in the
shape of the LDC from
demand response
Changes in the shape of the load
duration curve: (3) wholesale
(aggregated or individual customer)
demand response that is generally
dispatched during summer peaks to
ameliorate very high market clearing
prices or reduce peak system loadings
for reliability reasons
Ongoing work. In the past we added
demand response in for meeting capacity
markets.
2.2
4
40
1
parametric
uncertainty,
loadshapes
model scenario that
includes changes in the
shape of the LDC from
behind-the-meter
generation and energy
storage
Changes in the shape of the load
duration curve: (4) the penetration of
behind-the-meter generation and
energy storage.
We made adjustments to the LDC for
distributed solar PV and implemented.
When there is an EPA outlook/expectation
we can consider this for other distributed
technologies.
2.2
4
40
3
parametric
uncertainty,
loadshapes
model scenario with
negative demand
growth from EE
Even without changing the load
duration curve, we also suggest
including a scenario in EPA's Platform,
along with the Reference Case, that
involves negative demand growth
arising through greater energy
efficiency measures for buildings and
appliances
We can execute a first order estimate run,
assuming the load shape doesn't change
overtime. However, a more realistic
approach would require more work.
49
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.2/2.10
4
41
1
parametric
uncertainty, fuel
supply
model scenario
involving negative
shocks to fuel supplies
encourage EPA to publish scenarios
alongside the Reference Case involving
negative shocks to fuel supplies,
particularly in the northeastern U.S.
where resistance to additional fuel
delivery infrastructure has been high.
These negative shocks could be
modeled as outages or de-rates to
certain types of generating units in
certain regions within EPA's application
of IPM, or (perhaps preferably) using
high fuel prices to indicate shortage (see
an example for natural gas in Bent, et
al., 2018).
These can be modelled either by changing
fuel prices exogenously or through a full-
scale iteration with GMM.
2.2/2.10
4
41
3
multiple parametric
changes
model combo scenarios
that interact shifts in
LDC with existing
parametric scenarios
(such as low/high gas
prices and renewable
costs)
Scenarios that interact shifts in load
duration curves with existing parametric
scenarios (such as low/high gas prices
and renewable energy costs)
We can do this if needed, deemed
valuable and priority. LDC runs could be
tested when we have resources.
2.10
4
41
4
multiple parametric
changes
model combo scenarios
involving very low gas
prices and low
renewables costs
Scenarios involving very low gas prices
and rapidly declining capital costs for
renewable power generation
We can do this if needed and if deemed
valuable and priority. We are posting two
alternative reference case runs.
50
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.10
4
41
5
multiple parametric
changes
model combo scenarios
involving fuel supply
shocks and low
renewable costs
Scenarios involving fuel supply shocks
and low capital costs for renewable
power generation (implying a larger
dependence on renewable energy
during supply shocks, and the response
of the system to that known
dependence)
We can do this if needed, deemed
valuable and priority. We are posting two
alternative reference case runs.
2.10
4
41
6
parametric
uncertainty,
unexpected events
model parametric
surprise events
we observe that EPA's Platform v6 as
currently configured is ill-equipped to
handle unexpected events that might
arise over the multi-decadal time frame
that it models... however, we do see a
straightforward way for EPA to be able
to model specific scenarios that involve
parametric surprise events, and
encourage EPA to publish the results of
such scenarios alongside the Reference
Case
EPA has performed such analyses in the
past. We can do this if needed, deemed
valuable and priority.
2.10
5
43
4
policy analysis
model policies or
technologies that
endogenously shift load
across time
policies or technologies that
endogenously shift load across time
would introduce challenges and may
not be achievable given the current
model configuration, as we understand
it, except through an iteration
procedure.
51
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.10
5
44
1
policy analysis
model energy efficiency
and improve demand
representation
Because of the various federal proposals
to promote energy efficiency, EPA may
need to revisit its representation of
demand in order to be useful to analysis
of these policies.
Energy efficiency can be modeled
explicitly in IPM and has been done in the
past. The level of detail can be at the
measure level.
2.10
5
44
2
policy analysis
model state policies
governing retail tariffs,
including payments for
DG, electrification, and
shifting demand to
align with VREs
One of the largest challenges for EPA
going forward may be the
representation of policies at the state
level governing retail tariffs, including
payments for distributed generation,
and incentives to promote
electrification that may intentionally
align demand growth with the
availability of variable renewable energy
resources
Demand side policy representation will
have to follow an updated approach to
representing demand that doesn't rely
solely on EIA data. This would be a phase
two of any demand work we would
execute. We are developing in-house
capabilities to run NEMS.
2.10
5
44
3
policy analysis
model retail TOD prices
or retail RE prices
a possibly important policy mechanism
in the next decade is the determination
of retail prices that are differentiated by
time or type of electricity use
52
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.10
5
44
3
policy analysis
model time varying
prices applied to new
sources of
electrification
However, potentially more important
are time varying prices applied to new
sources of electricity demand such as
electric vehicles, water heating, and
building heating that embody
technologies with inherent storage
capability. These types of electricity
uses do not require all the attributes of
typical "instant on" electricity use.
Consequently, they may not be priced
at the same level and they may not be
burdened with the sunk costs
associated with the reliability aspects of
the existing grid, and retail prices may
be adjusted accordingly
Managed charging essentially addresses
this as we have been working on. For a
more comprehensive approach, we would
need to develop an EPA approach to
modeling demand before we can start
modeling cases like this in IPM.
2.10
5
44
4
policy analysis
model time-varying
prices including cross-
time-period elasticities
of electricity use and a
demand side model
To represent the meaningful aspects of
time-varying prices requires cross-time-
period elasticities of electricity use
within a fully functioning demand side
model.
IPM's DSM/EE modeling capability can be
exercised to model some of the demand-
side optionalities available in the market.
However, we do not plan to have a fully
functional demand-side model in the
near-term.
2.10
5
44
5
policy analysis
account for the effects
of uncertainty on
economic behavior
Another potentially important limitation
of EPA's policy analyses (that we also
raise in the context of EPA's Platform v6
representation of baseline uncertainty)
is the model's ability to account for the
effects of uncertainty on economic
behavior
We have been doing analytical work on
retirements outside of IPM to address
this.
53
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.11
6
46
3
model documentation
update documentation
on developing load
segments
Development of load segments: The
process used for developing load
segments as described in the
documentation is unclear.
The documentation is already clear.
2.11
6
46
4
model documentation
update documentation
on treatment of
interregional trading
Treatment of interregional trading: The
documentation's description of inter-
regional trade, especially related to the
load segments, is not very clear. The
documentation indicates that trade is
modeled on a seasonal basis, yet it is
our understanding after discussions
with EPA that trade is modeled by load
segment.
2.11
6
46
5
model documentation
update documentation
on aggregation of
model plants
Aggregation of model plants: The
documentation's description of the
aggregation of model plants also
requires clarification..., it is our
understanding that fossil units are
aggregated no further than at the plant
level
Section 4.2.6 documentation is already
clear.
2.11
6
46
6
model documentation
improve the
publication of data
tables on the EPA
website
Publication of data tables on the EPA
website: The use of tables uploaded
directly to the web is understandably
necessary given the large size of many
of the data inputs. However, a few
improvements are suggested.
The list of tables posted separately are
listed at the end of each chapter.
54
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.11
6
47
2
model documentation
include in the
documentation a more
complete description of
which AEO case for
what
Guide to EPA's Platform v6 Output Files:
It would be helpful to include a
reference in section 2.5.2 of the
documentation to the output file guide
that is on EPA's website. In addition,
when ElA's AEO cases are used to set up
alternative sensitivity cases, a more
complete description of which AEO case
is being used and what inputs are being
used from the case would be helpful.
So Far EPA has always used AEO reference
cases. If and when a different AEO case or
alternative demand cases are used, these
will be appropriately documented.
2.11
6
47
5
results viewer
insert a few
clarifications in the
READ ME instructions
for the Results Viewer
To avoid user confusion, we would
recommend that EPA insert a few
clarifications in the READ ME
instructions.
Addressed.
2.11
6
47
6
results viewer
update Results
Viewer's distinction
between "plant type"
and "plant category"
We found the Results Viewer's
distinction between "plant type" and
"plant category" confusing. For
example, it is unclear what a user
should choose for nuclear plant type.
The readme tab indicates that plant
type and plant category may be merged
in the future, and we agree this would
be clearer.
Revised.
55
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.11
6
47
7
results viewer
update Results Viewer
so that the displayed
results indicate the
cases being compared
The displayed results should indicate
the cases being compared (i.e., the
difference between what to what) and
the units of measure reflected in the
results. The readme tab indicates that
the results represent "changes from the
comparison model," but it would be
helpful to include this on the graphics
page accompanying the map.
Addressed.
2.11
6
47
8
results viewer
Make more intuitive
the "comparison case"
for other metrics, such
as capacity factors and
emissions rates
The use of the "comparison case" for
other metrics, such as capacity factors
and emissions rates, is clever but not
very intuitive.
Already limited the dropdowns to be
active when absolutely necessary and to
automatically match the Primary Case
selections where it makes sense to do so.
The Results Viewer is squeezed for screen
real estate, so the popup box seems like
the best way to give users a handy cheat
sheet (rather than permanently displaying
it)
2.11
6
48
2
results viewer
allow map sheet to
display two sets of
absolute values
In the map sheet, the comparison
functionality is confusing and only
works for displaying differences, rather
than two sets of absolute values.
While the display could be altered to show
this, the challenge is that displaying two
numbers per state would become illegible
(either too small a font or overlapping
values). The whisker chart is an alternate
graphing method that can fill a user's
need.
56
-------
Response
Document
Section
where
addressed
PR
Sect.
#
PR
Page
#
PR
Para.
#
PR Recommendation
Section/Category
PR Recommendation
Summary
Detailed Recommendation Text
Additional EPA Response Notes
(detailed narratives start after Executive
Summary rows)
EPA's IPM Summer 2021 Reference Case
Documentation is available here
2.9
6
48
4
retail price model
documentation
update the discussion
of utility depreciation
costs
In the discussion of utility depreciation
costs, the units are mills/kWh but these
are not defined by year. In addition, the
"directly from" is not explained
sufficiently as to whether the reader can
find these in a published document or
table or whether this was provided by
EIA
This will be addressed when an updated
RPM documentation published.
2.9
6
48
5
retail price model
documentation
update documentation
with additional detail
on the NUG adder and
the regional tax rates
The documentation would benefit from
additional detail for the non-utility
generators (NUG) adder and the
regional tax rates used in the RPM.
This will be addressed when an updated
RPM documentation published.
2.9
6
48
6
retail price model
documentation
define regional tax
dollars
Also related to regional tax rates, it is
not clear what is included in "regional
tax dollars" referenced in the
documentation.
This will be addressed when an updated
RPM documentation published.
2.9
6
48
7
retail price model
documentation
describe how the
percentage of each
region that is
deregulated or
regulated were derived
Attachment 1 of the documentation
includes a table showing the percentage
of each region that is deregulated or
regulated. We recommend that EPA
describe how the percentages were
derived, rather than simply citing the
AEO.
This will be addressed when an updated
RPM documentation published.
57
-------
Peer Review of EPA's Power Sector
Modeling Platform v6 using
Integrated Planning Model
Final Report
By:
Dallas Burtraw +
Seth Blumsack, James Bushnell, Frank A. Felder, Frances Wood
Work performed under contract to the U.S. Environmental Protection Agency
through
Industrial Economics, Incorporated
Jason Price, Project Manager
+ Review Panel Chair
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
EXECUTIVE SUMMARY
The U.S. Environmental Protection Agency (EPA) contracted with Industrial
Economics, Incorporated (lEc) to manage an expert review of the EPA's Power
Sector Modeling Platform version 6 (v6) using Integrated Planning Model (IPM).
IPM is a multi-regional, dynamic, deterministic linear programming model of the
U.S. electric power sector developed by ICF International. It provides projections
of least-cost capacity expansion, electricity dispatch, and emission control
strategies while meeting energy demand and environmental, transmission,
dispatch, and reliability constraints. EPA uses IPM to evaluate the cost and
emissions impacts of alternative policies to limit emissions of sulfur dioxide (SO2),
nitrogen oxides (NOx), carbon dioxide (CO2), and air toxics such as mercury (Hg)
and hydrochloric acid (HCI) from the electric power sector's operations. EPA has
applied IPM in the regulatory impact assessments for several rulemakings. For
example, EPA used IPM in its analysis of the costs and emissions impacts
associated with the Clean Power Plan and the Affordable Clean Energy Rule.
This peer review was guided by a specific set of charge questions developed by
EPA. The review presented in this document focuses on the specific issues
raised by these questions. Furthermore, EPA provided the panel with detailed
documentation of the EPA's Platform v6 and several supplementary documents
and data files. The panel reviewed this material and also participated in two
informational teleconferences and one in-person meeting (all organized by lEc)
with EPA, ICF, and lEc staff during the review process to seek clarification on a
variety of questions related to the model's design and functioning. The input
provided by the panel in this document reflects the contents of the EPA
Reference Case v6 documentation and related material provided to the panel as
well as the input provided through these interactions.
Overall, we found much to commend EPA's Platform v6. The model formulation
and structure lends itself well to EPA analyses of air policy focused on the
electric power sector. In addition, it includes significant detail related to electricity
supply and demand, with a data-rich representation both across different
geographic areas and across time. Based on the current structure of the
industry, EPA's application of IPM provides a reasonable representation of power
sector operations, generating technologies, emissions performance and controls,
and markets for fuels used by the power sector. The model is also well suited to
assess the costs and emissions impacts of the types of power sector policies that
EPA and other federal agencies have considered over the past several years.
The panel also found the EPA Reference Case v6 documentation to be well
written, clearly organized, and detailed in its presentation of most model
characteristics.
We also recommend that EPA consider several improvements and refinements
to EPA's application of IPM and the associated documentation. These
recommendations are presented in detail in the main body of this document, but
our highest priority recommendations are as follows:
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
• Consider changes to the model formulation that would improve the
model's ability to represent the ongoing evolution of the industry:
The electric power industry is undergoing fundamental changes with
potentially expansive scope and at an uncertain rate of evolution. Such
changes include significant increases in the penetration of renewables,
significant changes in load shapes (related to the accelerated introduction
of electric vehicles and energy storage), and significant changes in state
and regional policies affecting the industry. In response to these
changes, we recommend that EPA consider modifications to the model
formulation that would enable it to better represent these changes. Such
modifications could include revising the intra-annual load segments,
solving the model chronologically (as opposed to by load segment), or
alternatively solving a companion model that describes chronological
demand and system operation using capacity assumptions from IPM.
Other potential modifications include incorporating a richer representation
of energy storage into the model, and incorporating changes in capacity
market rules into the model (particularly as they relate to variable
renewable energy).
• Clarify the types of uncertainty that EPA's Platform v6 is capable of
handling: Within the current structure of EPA's Platform v6, the model is
capable of capturing uncertainty related to the value of key model
parameters, but the model is not capable of quantifying uncertainty in
model structure, decision rules, or processes. We recommend that the
documentation should provide guidance to model users that more clearly
articulates the types of uncertainties captured and not captured by the
model, and that EPA consider evolution in the model structure to address
uncertainty in a broader manner.
• Reconsider coal plant turndown constraints and possible addition of
operating reserves: To prevent the dispatch algorithm from setting
capacity utilization to unrealistically low levels, especially during specific
timeblocks given its utilization in other timeblocks, EPA's application of
IPM assigns minimum capacity factors that vary by plant type (and by
plant in some cases). While it is appropriate for EPA's Platform v6 to
constrain modeled dispatch to account for the lack of chronological load
segments and explicit unit ramping, we are concerned that the turndown
constraints are overly restrictive in some circumstances and perhaps not
quite restrictive enough in others. We recommend that EPA examine the
turndown constraints more closely to determine if they create bias in coal
plant operations, especially in scenarios with low gas prices or high
renewable generation. EPA might also consider whether adding explicit
operating reserve requirements in the dispatch would provide a better
representation of the impact of high levels of renewable generators on the
grid. The new constraint would require sufficient flexible capacity in each
time period to supply operating reserves that can be met by holding
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
capacity back from generating in the time period or supplied by quick start
capacity.
• Consider incorporating upstream emissions in addition to source-
level (power plant) emissions: As designed, EPA's application of IPM
explicitly considers only stack emissions at the point of fuel use. It does
not consider the "life cycle" emissions associated with upstream fuel
production, processing, and transportation. Since EPA may need to
evaluate regulations related to fuels production that are relevant for the
power sector, the panel recommends that EPA consider including
upstream emissions in its reference case as a separately-reported item
(so upstream and stack emissions are not combined together).
• Distinguish between investment decision-making of utility and
merchant power plants: Because utility-owned and merchant plants
face different risks and have different options at their disposal for
managing risk, decision-making is likely to differ significantly between
utility-owned and merchant plants. They may also behave differently with
respect to specific types of investment. Further, we suggest EPA evaluate
whether a weighted average of existing firms within a power region or
some other rule is a reasonable representation of which type of firm is
more likely to make an incremental investment.
• Address evolving gas markets where Henry Hub is less central to
pricing and where emerging petrochemical production has greater
influence: Natural gas pricing within EPA's Platform v6 is specified based
on static differences between the Henry Hub price and the price in
individual model regions. Under scenarios with different gas demand
patterns (quantities and locations) than the Reference Case, these basis
differentials could be quite different. It is the panel's understanding that
EPA addresses this issue by iterating with the Gas Market Model that
generates the natural gas supply curves and basis differentials, but this
process is not described in the model documentation. Relatedly, the
emerging petrochemical sector in the Appalachian production region is
likely to affect regional natural gas pricing in ways that may not be well
represented in the gas market model that EPA's Platform v6 relies upon.
Review of the gas market model, however, was outside the panel's
charge.
• Consider alternatives to the current load duration curves (LDCs): As
currently designed, EPA's Platform v6 aggregates time into LDCs in a
way that can assign the same calendar hour to different load segments in
different regions. This creates biases related to the opportunities for inter-
regional trade. It is difficult to assess the quantitative impact of these
assumptions, but we recommend that EPA assess the trade-offs between
different approaches to aggregating load into LDCs. Some approaches
may bias inter-regional trade less, but we recognize they may provide
less detailed resolution with respect to load within a given region.
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
• Improve representation of behind-the-meter generation: EPA's
Platform v6 captures policies that affect the bulk power system but does
not seem to capture policies that encourage behind-the-meter generation,
except as represented as a change in demand. Because future federal
policy may introduce new requirements that encourage increased
generation behind-the-meter, EPA's modeling may, in the near future,
need to capture these policies in greater detail than is currently possible
in the model.
• Increase transparency of retail pricing results: EPA uses the Retail
Price Model (RPM) as a post-processor to IPM results to estimate the
change in retail electricity prices associated with a given policy scenario.
Many elements of retail prices as reflected in the model remain constant
across scenarios and policies. These elements rely upon external
sources whose quality is difficult to assess. When EPA uses the RPM,
we recommend that the reporting of retail rates be broken into component
parts so that the user can understand which elements are endogenous to
the model and which are dominated by external sources and
assumptions.
• Consider improvements in the representation of various policy
mechanisms. Such improvements include dynamic allocations within
various forms of emissions trading programs such as output-based
allocation under cap and trade and a clean energy standard, expenditures
on energy efficiency that are linked to revenue from carbon pricing, and
the ability to represent flexible demand that may be encouraged at the
retail level to promote the integration of variable renewable energy.
• More thorough citing of sources and expanded explanations
throughout the EPA Reference Case v6 documentation: This
additional detail is particularly needed in portions of EPA Reference Case
v6 documentation pertaining to the development of load segments, the
treatment of interregional trading, and the aggregation of individual plants
to model plants, and the retail pricing model.
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
iv
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
TABLE OF CONTENTS
I. INTRODUCTION 1
II. MODEL FORMULATION 2
III. MODEL ASSUMPTIONS AND OUTPUTS 6
a. Power Sector Operation 6
b. Generating Technologies 12
c. Emission Factors and Control Alternatives 14
d. Power Sector Finances and Economics 17
e. Fuels and Renewable Resources 20
f. Regional and Temporal Resolution 23
g. Power Sector Policies 29
h. Retail Price Estimates 32
IV. BASE SET OF MODEL SCENARIOS 38
V. IMPROVEMENTS TO SUPPLORT POLICY ANALYSIS 43
VI. EPA's PLATFORM V6 DOCUMENTATION 46
VII. REFERENCES 57
V
Burtraw, Blumsack, Bushnell, Felder, and Wood March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
I. INTRODUCTION
The U.S. Environmental Protection Agency (EPA) contracted with Industrial
Economics, Incorporated (lEc) to manage an expert review of the EPA's Power
Sector Modeling Platform version 6 (v6) using Integrated Planning Model (IPM),
Version 6. IPM is a multi-regional, dynamic, deterministic linear programming
model of the U.S. electric power sector developed by ICF International. It
provides projections of least-cost capacity expansion, electricity dispatch, and
emission control strategies while meeting energy demand and environmental,
transmission, dispatch, and reliability constraints. EPA uses IPM to evaluate the
cost and emissions impacts of alternative policies to limit emissions of sulfur
dioxide (SO2), nitrogen oxides (NOx), carbon dioxide (CO2), and air toxics such as
mercury (Hg) and hydrochloric acid (HCI) from the electric power sector's
operations. EPA has applied IPM in the regulatory impact assessments for
several rulemakings. For example, EPA used IPM in its analysis of the costs and
emissions impacts associated with the Clean Power Plan and the Affordable
Clean Energy Rule.
This peer review was guided by a specific set of charge questions developed by
EPA. The review presented in this document focuses on the specific issues
raised by these questions. Furthermore, EPA provided the panel with detailed
documentation of the EPA's application of the model and several supplementary
documents and data files. The panel reviewed this material and also participated
in two informational teleconferences and one in-person meeting (all organized by
lEc) with EPA, ICF, and lEc staff during the review process to seek clarification
on a variety of questions related to the model's design and functioning. The input
provided by the panel in this document reflects the contents of the EPA
Reference Case v6 documentation and related material provided to the panel, as
well as the input provided through these interactions.
This document is largely organized according to the order of questions as they
appear in the charge. The one exception is the panel's input on the EPA
Reference Case v6 documentation, which is presented in the final section of this
document.
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
1
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
II. MODEL FORMULATION
Identify strengths, weaknesses, limitations, and errors in the structure of the model
formulation (e.g. objective function, constraints, and decision variables and their indices).
Propose options as needed. Specifically, are all the necessary elements included in order
to meet EPA's analytical needs? Are there any extraneous elements? Could
simplifications be made?
The Integrated Planning Model (IPM) is a long-term capacity expansion and
production-cost model of the U.S. electric power sector. It is a linear program,
which enables quickly solving a large and detailed model using off-the-shelf
solvers. It covers input fuels, air emission, and electricity markets and is
designed to analyze the impacts of alternative regulatory policies on the power
sector over the long term (i.e., investment horizon).
In considering the formulation of the model, we are cognizant of the fact that one
model cannot address all questions or serve all purposes. Consequently, the
structure of EPA's modeling should be linked to the modeling objective.
Examination of the strengths, weaknesses, limitations, and errors in the structure
of the model is constrained by our limited access to the coding of the objective
function, constraints, and decision variables and their indices. We are able to
focus, however, on the importance of aligning generation aggregation, temporal
aggregation, regional definitions, transmission expansion, trading capabilities,
and dynamic and elastic load with the objective of the particular modeling
assignment at hand. In this context, we find that for long-term (20-40 years)
modeling of overall trends, differences between the outputs of scenarios, and
general emission levels changes, EPA's application of IPM has a great deal of
appropriate structure (with caveats we address below with respect to demand
price elasticity and transmission expansion) trading off computational time,
assumption details, and long-range planning horizons. For shorter periods of time
and where hourly dispatch is critical to assess emissions for more detailed air
pollution modeling, where changes in near term wholesale and retail prices are
important, or where the feasibility and reliability of large-scale renewable
penetration is being considered, EPA may consider reconfiguring the model or
complementing it with a chronological hour capability.
We also considered this charge question in the context of both the current state
of the electric power industry and likely changes in the industry in the coming
years, some of which have already started to occur. The electric power industry
is undergoing fundamental changes with potentially expansive scope and at an
uncertain rate of evolution (U.S. DOE, 2015). Based on this ongoing evolution,
particularly industry changes related to the penetration of renewables, load
shapes (including accelerated introduction of electric vehicles and energy
storage), state and regional policies, etc., the model formulation may need to be
adjusted according to suggestions offered here. However, as some of these
uncertainties unfold, EPA should periodically review the model to determine
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
whether EPA's application of IPM model structure should be modified or
complemented with other modeling capabilities.
The suggestions that we provide here regarding model formulation, as well as
suggestions that we present in later sections of this review, are informed by
uncertainty as an overarching consideration that affects the usefulness of EPA's
application of IPM in projecting alternative futures for the industry. Uncertainty
affects the choice of parameters in the model, structural relationships in the
model, and the way that model results should be interpreted. Specifically related
to the structure of the model, as the relative costs of renewable resources decline
and more state and regional policies directly or indirectly support renewables and
energy efficiency, we recommend that EPA consider the following to ensure that
the model's formulation and structure endow it with capabilities critical to
providing insights into the industry's response to changes in policy:1
a. Chronological modeling: We expect the role of chronological operation
of the power system to become increasingly important with expanded
availability of variable renewable energy, electricity demand growth, and
demand flexibility. EPA may therefore consider restructuring IPM as a
chronological model or develop a companion short-run chronological
model of system operation that would enable comparing the outcomes of
the model's load duration curves with a more realistic characterization of
the temporal nature of demand with chronological load profiles, perhaps
using capacity assumptions taken from the long-term model.
b. Demand responses: We recommend that EPA consider incorporating
additional factors into the model's formulation of demand response. This
would include changes in total electricity consumption in response to
changes in price, substitution between electricity and other forms of
energy consumption, and changes in the load shapes that will be
observed and projected under different scenarios. In addition, variations
in supply-side short run marginal costs (due to increased penetration of
variable renewable energy) and potentially demand side retail prices that
vary by time of day or are linked to resource availability directly require
cross-time-period analysis of electricity demand.
c. Climate change considerations: We recommend that EPA periodically
evaluate the model with respect to weather normalization of key data
inputs and consider a more explicit representation of climate change in
the model's specification of generation, transmission, and load
assumptions.
1 Some of the recommendations pertaining to the formulation of the model offered here overlap
with recommendations in later sections of this review. This reflects the fact that the model
formulation and structure are closely intertwined with the various issues related to IPM
assumptions, inputs, calculations, and outputs identified in later sections.
3
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
d. Transmission capacity: Transmission expansion is one of the ways that
increasing amounts of variable renewable energy may be integrated into
the power system. Transmission system decisions, however, are tightly
connected with regional and local land use decisions that cannot be
precisely forecast. We therefore recommend that EPA regularly revisit
and, as appropriate, revise EPA's implementation of transmission
outcomes and the assumptions that shape anticipated future transmission
siting decisions.
e. Storage: EPA's application of IPM would benefit from a more rigorous
treatment of energy storage within the formulation of the model,
particularly with respect to the opportunity to schedule demand and
achieve thermal and battery storage for end-uses, and for storage
technologies that enable sending power back to the to grid.
f. Capacity markets: The treatment of capacity markets and other
representations of resource adequacy requirements, as opposed to
modeling a generation reserve requirement, may be increasingly
important going forward. In several large regions of the country (e.g., in
PJM), capacity markets have started to change (PJM, 2019). Potential
changes to capacity market rules include availability requirements (which
may result in generation units pursuing dual fuel options, firm gas, or
including a risk premium in their capacity offers) and changes in the
conditions under which demand can participate in capacity markets.
Such changes may have important effects on the level and technology of
capacity investment and accompanying transmission investments.
g. Additional operational constraints: A more explicit representation of
such constraints may be necessary in a capacity expansion model such
as IPM because of the increased penetration of variable renewable
energy. As the penetration of renewable resources increases, it may be
necessary to also include operating reserve requirements, in addition to
adjusting renewable capacity credits toward meeting capacity reserve
margins, to account for the impact renewable resources have on
economic dispatch.
We also highlight that projections with a deterministic model such as IPM, or
even a set of projections, are not necessarily representative of the decision-
making under uncertainty that the model is attempting to explore. Decision
makers may hedge investments in various ways that do not align with the cost
minimization outcome identified under a given scenario.2 EPA may also improve
2 Burtraw et al. (2010) illustrate one approach using Taylor series approximation to represent
hedging behavior in the context of a deterministic and highly parameterized model..
4
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
the way they use model results by explicitly considering them in an option value
framework.3
Finally, any runtime restrictions required by the EPA should be made explicit and
used to appropriately structure EPA's application of IPM to the task at hand.
There seems to be an implicit requirement that IPM run time should be
"overnight" so that EPA can have scenarios turned around within a day. Whether
increasing this turnaround time would enable the model to accommodate more of
the complexities listed above should be investigated.
3 Echeverri et al. (2013) use stochastic dynamic optimization in an engineering-economic model
to describe technology choice in response to technology-forcing regulations in an uncertain
environment.
5
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
III. MODEL ASSUMPTIONS AND OUTPUTS
Identify strengths, weaknesses, limitations, and errors in the model assumptions, model
outputs, and conclusions derived. Propose options as needed. Specifically consider how
well the representations of the following items suit EPA's analytical needs: power sector
operations, generating technologies, emission factors and control alternatives, power
sector finances and economics, fuels and renewable resources, regional and temporal
resolution, power sector policies, and retail price estimates.
A. POWER SECTOR OPERATION
IPM uses a linear program inter-temporal optimization approach to power sector
capacity expansion and economic dispatch in which costs are minimized over the
projection period subject to multiple constraints. Major components the model's
representation of power sector operation include demand and load growth,
dispatch of existing assets, trade, representation of transmission constraints, and
generating capacity expansion.
Demand
For most applications of EPA's Platform v6, electricity demand is exogenously
specified. Projected annual demand is taken from the 2018 Annual Energy
Outlook (AEO) Net Energy for Load and mapped from the National Energy
Modeling System's Electricity Market Model (EMM) regions to EPA's model
regions. The mapping currently uses relatively old data (2007 and 2011), but
EPA has informed the panel that it is performing an update to 2016. Although
IPM includes the capability of using price elasticity to impact demand, EPA
generally does not use this feature. In addition, distributed generation (DG) is not
explicitly included in the model's specification of electricity demand, with only
own-use DG implicitly reflected within the demand projection.
EPA's application of IPM converts annual demands to demand by time segment
by using seasonal load duration curves (load sorted by level) and then by time of
day, as described more in Section lll-F below. Based on this specification, the
load segments in EPA's application of IPM are not chronological. Currently 2011
load data are used in all regions, except for ERCOT where 2016 is used, to
develop the hourly load curves. Changes in load shapes over time are driven by
assumption about load factors, with NERC electricity supply & demand (ES&D)
load factors used to project peaks for 2021-2027 and changes in AEO2018 load
factors for years post-2027.
We view the use of fixed electricity demands without response to electricity prices
as problematic in policy and sensitivity cases where prices vary significantly from
the Reference Case. We recommend that EPA use this feature when analyzing
policy scenarios that have significant price impacts (perhaps roughly greater than
20% variation in wholesale prices). We also recommend that the EPA Reference
Case v6 documentation describe in more detail how the elasticity is applied when
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
used. Based on input provided by EPA staff, it is our understanding that the
baseline demands and prices are stored and that the elasticity is applied to the
percent change in the scenario prices from the baseline set. In addition, the
wholesale price is grossed up to approximate a retail price. However, a better
approach for sensitivity cases, such as high/low natural gas price cases, in which
electricity prices are expected to vary from the AEO Reference Case is to
develop alternative baselines, which we understand is EPA's practice, rather than
using elasticities for modifying demands from the Reference Case.
As described in Section V, the demand side of the power system is likely to
become more important as the grid evolves with greater shares of intermittent or
variable renewable energy sources and possible expanded uses of electricity,
such as electric vehicles. At a minimum, a more systematic way is needed to
develop load shapes over time rather than just using a single metric of load
factors from ES&D or the AEO to shift the curves. While peak demand is
important for reserve margin requirements and capacity needs, the shape/time of
demand is also important for dispatch and use of variable renewable energy.
Although the 24 load segments per season within EPA's Platform v6 are at a
fairly high level of resolution for a long-term planning model, the lack of
chronology creates some challenges for representing trade among regions. As
described in Section lll-F, the importance of trade is greater as the number of
regions increases. Section lll-F provides specific suggestions regarding other
segmentation methods that might be considered by EPA, recognizing that there
is a trade-off between capturing correlations of load and variable renewable
energy availability and trade among regions, while maintaining reasonable model
run time performance.
As EPA updates load shapes to 2016 values, we would recommend that EPA
consider whether using data for a single year creates any biases and whether a
multi-year average or weather normalization would be more appropriate. Also, it
may be useful to have consistency among AEO vintages used for data
assumptions. For example, the model currently uses AEO2018 loads and peak
load factors but Mexican trade from the AEO2017.
Dispatch
EPA's Platform v6 performs an optimal economic dispatch subject to several
constraints. For most plant types, availability (using forced and scheduled
outages based on the Generating Availability Data System) determines maximum
generation; availability is defined by season where no planned maintenance is
assumed to occur in peak demand season.
EPA assigns oil/gas steam units minimum annual capacity factors to avoid over
optimization that might result in them not operating despite their historical use.
These minimum capacity factors account for considerations that cannot
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
practicably be reflected in EPA's application of IPM, such as local transmission
constraints or grid reliability concerns. These minimums terminate over time
based on units' individual historical capacity factors (minimums for units with low
historical capacity factors are removed sooner), but all minimums terminate by
age 60. In addition, to prevent an unrealistic dispatch of fossil steam units
(oil/gas and coal) within the existing non-chronological time segments, turndown
constraints apply to all 23 non-peak time segments in a season if the plant runs
at 100% in the peak load time segment. Oil/gas units are assumed to dispatch
no less than 25% of the unit capacity for these segments. For coal, unit level
turndown percentages apply based on historical rates that vary from 20% to 78%,
with most units in the 40-60% range.
Wind and solar generation are determined by hourly generation profiles prepared
by NREL. Hydropower dispatch is governed by seasonal capacity factors
specified by model region based on a 9-year historical period (2007 to 2016) but
otherwise is assumed to be fully dispatchable within those constraints.
While it is appropriate for EPA's application of IPM to constrain modeled dispatch
to account for the lack of chronological load segments and explicit unit ramping,
the model's turndown constraints strike us as overly restrictive in some
circumstances and perhaps not quite restrictive enough in others. For example,
it appears that steam plants would not be able to shut down for any time
segments (such as segments with lowest load) if they are expected to run at full
capacity during the peak segment. In practice, some plants likely can shut down
at night and ramp up to full capacity by the middle of the following day. It was not
as clear from the documentation what operations are allowed in other time blocks
if units run at partial load on peak. This is not very likely to occur in practice, but it
might occur in the model as a way around the minimum in the other load
segments. Because output files from EPA's Platform v6 do not include
information on dispatch by time segment, it is not possible to determine if this
occurs.
To the degree that the model's turndown constraints result in unrealistic dispatch,
they may introduce bias into the model. Specifically, they may impact the cost-
effectiveness and retirements of steam plants and, by extension, annual
emissions. The degree of flexibility of fossil plant dispatch in EPA's application of
IPM can also impact the attractiveness of variable renewable energy. For coal
plants, the turndown constraints vary considerably by unit, and some are as high
as 80% with most of them between 40% and 60%. Because these values are
based on historical operations rather than current or projected engineering
considerations, they may reflect historical economic circumstances that may not
apply in the future.
We recommend that EPA examine the turndown constraints more closely to
determine if they create bias in coal plant operations, especially in scenarios with
low gas prices or high renewable generation. EPA might also consider whether
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
adding explicit operating reserve requirements in the dispatch would provide a
better representation of the impact of high levels of variable renewable energy on
the grid. The new constraint would require sufficient flexible capacity in each
time period to supply operating reserves that can be met by holding capacity
back from generating in the time period or supplied by quick start capacity.
EPA's assumption that hydroelectric generation is fully dispatchable within
seasonal capacity factor constraints may be too generous. Run-of-river
conditions, location of multiple dams on a single river, and environmental
considerations may make a portion of hydrogeneration somewhat inflexible. EPA
might consider analyzing historical generation patterns versus the model patterns
by time period to assess whether EPA's application of IPM is significantly
overoptimizing hydro generation.
Transmission
EPA's Platform v6 defines transmission capacity limits for firm (capacity) and
non-firm (energy) trading between model regions. The model also specifies joint
limits between groups of regions to account for reliability considerations.
Wheeling charges are assessed to move power between regions, except for
trading within the same regional transmission organization (RTO) region. Line
losses are assessed for inter-regional transmission at 2.8% for WECC and 2.4%
for ERCOT and the Eastern Interconnect.
The application of a 2.4% transmission loss to each interregional transfer strikes
us as high for the Eastern Interconnect, especially given the size of the model
regions and hence relatively short distances for many of these transfers. For
example, in NEMS a 2% loss factor is assumed for transfers between regions
and there are fewer regions. In EPA's Platform v6, the large number of regions,
especially in the Eastern Interconnect, means that the distances between them
are relatively short so a smaller loss factor would be expected.
EPA's Platform v6 also includes the capability to add transmission capacity,
though EPA rarely uses this feature. Not using this feature may be overly
restrictive in some scenarios, especially given the large number of regions
represented in the model. In general, one would expect new generation capacity
to be added somewhat near growing loads and to be distributed across all
regions. However, in scenarios that lead to an economic propensity for high
levels of renewable capacity additions, there may be an economic rationale for
transmission expansion to move power from regions with high levels of wind, and
perhaps solar, resources to other regions. EPA could alleviate this concern by
performing sensitivity cases in which additional transmission capacity is added
exogenously by user assumption.
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
Capacity Expansion
Setting the Capacity Targets
EPA's Platform v6 projects capacity additions and retirements as part of the
optimization of meeting future electricity demands. Total capacity requirements
are determined by reserve margins and peak loads. Reserve margins are set for
each model region based on the requirement of the NERC region to which it
belongs.
Overall, we find the current method for setting capacity requirements using
reserve margins by region to be reasonable, even though not all regions explicitly
follow that model. The most significant deviation is ERCOT, which does not have
a capacity market and hence has no way to enforce or incentivize achievement of
a specific reserve requirement. In theory, capacity and "energy-only" markets
such as ERCOT should achieve similar outcomes over the long run if held to
similar standards. The biggest difference is that the ERCOT standard is based
upon short-term operating reserves and relies upon higher energy and ancillary
services (AS) prices when short-term reserves fall below the reliability standard.
Therefore, the main modeling question is how the model's annual peak planning
standard translates to an ongoing operating reserve margin. In the absence of
uncertainty, the two concepts should be reconcilable.
California maintains a "flexible" capacity requirement that is intended to ensure a
percentage of installed capacity is both fast ramping and not energy limited. The
flexible capacity standard continues to evolve but may become a model for other
States in the near future. If so, EPA's application of IPM may need to be modified
to reflect other capacity requirements beyond a simple reserve requirement and
declining capacity values for variable renewable capacity. For example, it may
be appropriate to add an additional reserve constraint requiring a percentage of
capacity is capable of meeting a certain ramping capability.
To meet capacity needs, EPA's application of IPM selects new capacity additions
from a slate of multiple technology options, each with its own characteristics that
vary by vintage to reflect technological improvements over time. Short-term cost
adders are applied when annual capacity additions would exceed upper bounds.
We view EPA's application of cost adders to capacity expansion costs when
expansion is rapid as a reasonable approach for limiting modeled deployment of
a given generating technology on a large scale. We would expect that sharp
increases in capacity expansion would increase the costs of a technology due to
factors such as manufacturing constraints and would also expect that these
increases in costs would limit deployment of the technology in question. EPA
has adopted capital cost penalties like those in NEMS, although the
implementation is quite different. As mentioned above, in EPA's Platform v6, the
amount of capacity that can be built in a period without invoking a cost penalty is
specified by time period and is currently assumed to be constant over time (same
10
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
amount per year) through 2035, after which no penalty is applied. In contrast,
NEMS defines the annual capacity addition step size before a cost penalty as a
function of the level of previous capacity additions. This difference is likely to
accommodate the model's intertemporal optimization where the model might
otherwise select a technology when not economic simply to allow lower cost
future builds. However, it means that the capacity additions allowed in EPA's
Platform v6 without a cost penalty are higher than in NEMS for technologies that
have had low deployments, and the penalty threshold does not diminish as
greater deployments occur. For example, the steps are quite large for nuclear
capacity at roughly 10 GW year. Figure 1 below illustrates an example of the
potential annual nuclear capacity additions that are allowed within the first step
(i.e. base cost without a cost adder) where for NEMS 2200 MW are assumed to
already have been recently built (planned units currently under construction).
Note that 10 GW cannot be added in a single year at the base cost until many
years of steady capacity additions have occurred unless a cost penalty is paid
and capacity additions exceed the first step at least once. Also, after 2035,
EPA's Platform v6 allows unlimited additions at the base cost which could be
problematic in scenarios in which a technology cost declines over time and
becomes cost-effective only late in the projection.
As in NEMS, the cost penalties are quite steep with roughly a 45% cost penalty
on the second step. It might be better to have smaller initial steps with smaller
cost penalties for the second step. These cost adders are not likely to be
incurred in a Reference Case but could become important in scenarios with
either different market conditions or technology assumptions or significant
policies that impact the relative attractiveness of new capacity types.
Figure 1. Comparison of EPA's Platform v6 and NEMS AE02020 Nuclear
Capacity Additions in First Step at Base Cost
Nuclear Capacity Additions
in First Step at Base Cost
25,DOO
0
—•—NEMS
IPM
0
5,000 10,000 15,000 20,000
Greatest PriorYear Capacity Additions (MW)
25,000
11
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
Rating the Capacity of Alternative Resources
All existing capacity in EPA's Platform v6 counts 100 percent toward planning
reserve requirements except those units that depend on variable renewable
energy resources. Capacity credits for wind and solar capacity are based on a
supply curve approach where the credit declines as more capacity is developed.
The current method used in EPA's Platform v6 of assessing capacity credits for
variable renewable energy seems to be a reasonable approximation assuming
average performance. The sorting of resources in order of likely builds and
assessing the renewable energy capacity contribution, taking into account
generation and load profiles, is not a fully endogenous process and cannot
account for the joint impact of wind and solar but is likely adequate. One caution
is that if the solar capacity credits are still benchmarked to those of the AEO2017,
as indicated in the documentation, this should be revisited because the AEO
methodology and resulting credits for solar have changed considerably since the
AEO2017 was published. In addition, we recommend that EPA examine the
capacity credit methodology as system operators change their capacity market
rules and consider the implications of non-performance risk. For example, the
performance penalties could discourage many variable renewable energy
generators from participating in capacity markets at all, due to the financial risk of
the penalties. They may also result in more "dual fuel" fossil capability than
would otherwise be economic.
It is also worth noting that demand response and energy efficiency are providing
non-trivial shares of total capacity and even larger shares of new capacity in
many capacity markets. While performance criteria are still controversial and
may evolve, demand side resources are quite likely to play an increasing role in
meeting reliability requirements.
B. GENERATING TECHNOLOGIES
EPA's Platform v6 models existing, planned-committed, and new generation
technologies to determine the least-cost power system over the modeling
horizon. Key assumptions are the types of generation units that are modeled,
capital, fixed and variable costs, and performance (e.g., heat rates, capacity
factors, availability, emissions, etc.). Existing and planned-committed units are
based upon the National Electric Energy Data System (NEEDS) v6 database.
Assumptions for new generation technologies are based upon a variety of
sources.
As described below, our recommendations pertaining to generating technologies
relate to energy storage, nuclear dispatch, heat rates, the specification of a unit's
generating capacity, and changes in generation assumptions over time.
12
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
Storage Technologies
The panel recommends that EPA consider incorporating additional storage
technologies into the model. U.S. energy storage installations are increasing
significantly (EIA, 2018), and multiple states have mandates for energy storage.
Currently, EPA's Platform v6 models only two types of energy storage facilities:
pumped storage and lithium-ion batteries (with a four-hour charge capacity). As
energy storage technologies mature and as the market responds to cost
reductions, performance improvements, and regulatory changes, energy storage
will likely become an increasing part of the power sector, requiring more attention
and focus by the platform. In addition, because the technologies, cost structure,
performance, operating strategies, market rules, and regulations related to
storage are rapidly changing, EPA may need to regularly revisit the model's
representation of storage.
In addition to including more storage technologies, EPA should also consider
regional variations in energy storage technology, costs, and operations. These
variations may include if and how energy storage counts in capacity markets.
Capacity market rules may also affect the length of time energy storage facilities
are discharged, which affects costs and dispatch levels. The current modeling
platform does not account for these regional variations.
Nuclear Dispatch
We also suggest that EPA consider more flexible nuclear dispatch in EPA's
application of IPM. As the quantities of renewables are likely to increase over
time, nuclear units may be dispatched to respond to changes in net load, i.e.,
demand minus non-dispatchable supply. Currently, EPA's Platform v6 models
nuclear units with low fuel and variable operations and maintenance costs, which
results in them being run at the maximum possible available output. Although
EPA's Platform v6 appropriately does not model very short-term fluctuations in
net load, if in the future some nuclear units are used as flexible resources, their
specific modeling assumptions may need to be revised to account for lower
capacity factors of nuclear units due to increased ramping. For example, nuclear
units in France dispatch at different levels based on variable changes in
renewable generation.
Heat Rates
We recommend that the heat rates of generating units in EPA's application of
IPM vary by season and perhaps over time. EPA's Platform v6 assumes a fixed,
single heat rate for each generation unit that does not vary by season, although it
does have a heat rate improvement option for coal units, which is not currently
activated. Furthermore, EPA assumes that the Reference Case heat rates
remain constant over time due to increased maintenance and component
replacement over time. Empirical evidence suggests that such expenditures are
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
necessary to maintain heat rate performance, which otherwise tends to degrade
over time (Linn et al., 2014).
Related to seasonal heat rates, we also recommend that EPA's application of
IPM vary the generation capacity of a given unit by season, or add text to the
documentation explaining why seasonal variation is not necessary. Seasonal
temperature changes affect the generation capacity (MW) of many types of
generation units. Currently EPA's Platform v6 uses "net summer dependable
capacity".
Generation Assumptions
We also suggest that EPA consider adjusting the EPA's Reference Case
generation assumptions over time to account for climate change. As average
climate temperatures rise along with associated increases in cooling water, many
electricity modeling assumptions, including assumptions related to generation,
may need to be adjusted such as summer and winter capacity, heat rates,
electrical losses, etc. (Chandramowli and Felder, 2013).
EPA may also consider varying some generation assumptions by
market/regulatory environment. Currently, EPA's Platform v6 does not make
different assumptions for generation units in wholesale markets versus regulated
regions (except for the cost-of-capital). Generation costs (including both fixed
and variable operating costs) and unit availability, however, may vary based on
the market/regulatory environment in which they operate (as well as ownership)
(Fabrizio, et. al, 2007). Some of these differences may be captured in the
regional cost variation factors shown in Table 4-15 of the documentation (at least
historically), but a clear distinction of costs between regulated and market regions
is needed.
C. EMISSION FACTORS AND CONTROL ALTERNATIVES
Accounting for air emissions from electric generation units and representing the
decisions to invest in air emissions control technologies are fundamentally
important requirements of the model for EPA's analytical purposes. The current
version of EPA's Platform v6 represents emissions rates by plant type (the
"types" being aggregations of actual plants as defined in the documentation) and
features a highly granular set of air emission control technologies with which
plants can be outfitted. These air emissions factors and control technologies
cover multiple relevant pollutants, including oxides of sulfur and nitrogen,
particulate matter, mercury, and carbon dioxide.
Assignment and Scope of Emissions Factors
The method by which EPA's Platform v6 assigns emission factors to specific
plant types within the model appears to be appropriate to meet EPA's analytical
needs. EPA uses appropriate data sets to determine stack emissions from
different power generation technology types. The model, however, explicitly
14
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
considers only stack emissions at the point of fuel use. It does not consider the
"life cycle" emissions associated with upstream fuel production, processing, and
transportation. There can be considerable differences between stack emissions
and life-cycle emissions (Jaramillo, et al., 2007), and the life-cycle air emissions
may vary considerably by location since primary energy extraction, processing,
and transportation technologies (and fuels used in those phases of the life cycle)
can be location-specific. In some cases, EPA may need to evaluate regulations
related to fuels production that are relevant for the power sector. A recent
example of this would be proposed regulations addressing fugitive methane
emissions from unconventional natural gas production (Brandt, et al., 2016).
Control costs to address fugitive methane emissions would add some cost to
delivered natural gas (ICF, 2014; Osofsky et al., 2018), which in some cases
would render natural gas a more expensive fuel than coal at the margin. EPA
should consider documenting how upstream air emissions are reflected in fuels
prices or in generator marginal costs within its Power Sector Modeling Platform
for analysis of regulations where such upstream emissions would be relevant to
power system investment and operations.
Emissions Control Options
EPA's Platform v6 contains a highly detailed representation of emissions controls
options from which modeled plants can choose. The level of detail in terms of the
air emissions addressed by these technologies and the different technology
options for each is impressive. The extensive library of control options is a
strength of EPA's Platform v6, given the analyses that EPA has needed to
conduct in the past.
As the U.S. power sector continues to move away from the heavy use of coal to
the heavy use of natural gas for power generation, it may be the case that some
of the emissions control options currently included within EPA's Platform v6
become less relevant to the kinds of analyses that EPA is asked to perform. We
suggest that EPA periodically review the technology options for emissions control
in EPA's application of IPM to determine if this portion of the model could be
made simpler with the reduction of emissions control technologies from which
modeled plants can choose.
Emissions Control Costs
The emissions control cost data used in EPA's Platform v6 appear to come from
a study by Sargent and Lundy (2017) that relies on proprietary data. Chapter 5 of
the EPA platform v6 documentation includes unit cost estimates derived from the
Sargent and Lundy study but does not provide a formal citation for the study or
the raw engineering data used to develop the unit cost values. Publication of
these data would make the cost figures used by EPA's Platform v6 more
transparent than they are currently. We recommend that EPA consider the costs
and benefits of this additional data transparency as weighed against the benefits
of being able to access and use proprietary data, which in some cases may be
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
more granular or up-to-date than data existing in the public domain. EPA should
also periodically compare its emissions control cost data with relevant information
that exists in the public domain, such as the Integrated Environmental Control
Model (IECM) developed by Carnegie-Mellon University.
For the capture of carbon dioxide specifically, we note that EPA's Platform v6
considers the costs of CO2 capture, transport, and long-term geologic
sequestration. While the model scenario runs published by EPA generally do not
choose carbon capture as an emissions control option, this capability is likely to
become more important in the future. In particular, the option to choose natural
gas combined cycle plants with carbon capture appears to be turned off within
the model in the EPA Reference Case. This technology option may become more
relevant in policy scenarios where carbon capture is a more economic choice, so
we recommend that EPA restore this technology option under relevant analyses.
As part of the cost estimation algorithm, EPA's application of IPM uses a
proprietary model (GeoCAT) to provide information on CO2 sequestration
potential in different storage areas. While this is not described specifically in
Section 6 of the documentation, it appears that IPM solves an optimization
problem to find a least-cost solution for CO2 transportation (mode and
sequestration location) for each model region in which power plants are capturing
CO2. We would recommend that EPA incorporate additional specificity on this in
the documentation, particularly any interactions between IPM and GeoCAT.
The CO2 storage cost curves include opportunities for use of CO2 for enhanced
oil recovery (EOR) that yield negative storage costs assuming an oil price of
$75/barrel. Given more recent developments in unconventional oil recovery, the
assumed oil price may be too high. EPA should re-evaluate the oil price
assumption related to EOR. Based on our reading of the documentation (Section
6.2), it appears that CO2 sourced from industrial facilities at a positive storage
cost of $50/ton could displace some CO2 for EOR (sourced either from industrial
facilities or power plants) with a negative net storage cost. Although it is
necessary to consider competition between power and industrial sources, this
appears to give the industrial sources preferential treatment that may not be
appropriate in all scenarios, such as under power sector policies that provide an
economic advantage to power compared to industrial CO2 sources. In these
circumstances or when oil prices (and hence EOR demand for CO2) varies from
the Reference Case, EPA should re-evaluate the CO2 storage cost curves.
Some elements of the CO2 transport model are also not clear, particularly related
to the economies of scale in pipeline transportation. The method described in
Section 6.3 of the documentation appears to assume that CO2 sources that are
transporting CO2 over longer distances for long-term geologic sequestration are
taking advantage of some undescribed scale economies in the form of capacity
sharing in CO2 pipelines. The documentation justifies this by saying that "the
longer the distance from the source of the CO2 to the sink for the CO2, the greater
the chance for other sources to share in transportation costs...." although there
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
does not appear to be any such probability calculated within EPA's application of
IPM or used as an input. We interpret this statement as implying that EPA's
Platform v6 assumes that CO2 transportation over long distances will involve
lower average transport costs because long-distance transportation implies
multiple users sharing the cost of a larger pipeline. A CO2 source in EPA's
Platform v6 would thus face a lower cost per mile of transportation by moving
captured CO2 over longer distances. There is little real-world data to back up or
refute this assumption, but if our interpretation is correct, we suspect that it may
be biasing EPA's Platform v6 towards very long-distance CO2 transportation in
some cases.
D. POWER SECTOR FINANCES AND ECONOMICS
The most important economic factors governing power plant dispatch - fuel costs
(where applicable) and O&M costs - are represented in EPA's Platform v6
through the use of relevant O&M cost data, fuel price data, and heat rate data
where applicable. The most important economic factors governing power plant
investment including new capital, retirements, and retrofits are equipment costs
and the weighted average cost of capital (WACC), each of which is represented
in EPA's Platform v6. The model also contains features meant to capture
different investment incentives and risks in restructured versus traditionally-
regulated jurisdictions.
The methods that EPA uses to determine variable dispatch cost (fuel costs plus
variable O&M) seem appropriate for EPA's purposes and are based upon
credible data sources.
Section 10 of the documentation has a highly detailed description for calculation
of the WACC. Several elements of the determination of WACC are not clear:
• The description of the calculation of the capital charge rate would be
made substantially more clear with an equation. In particular, whether
EPA's Platform v6 uses the common "short cut" version of the capital
charge rate (Stauffer, 2006) could be made more clear.
• The documented distinction between the book life of debt and the asset
life for making investment decisions could be more clear. We recommend
an explicit statement in the documentation describing the debt life versus
the asset life.
• This module of EPA's Platform v6 is necessarily replete with assumptions
because little financial data is in the public domain. Some of these
assumptions may be questionable, although it is not clear how important
they are in determining the overall WACC. Examples include the
assumptions on debt-to-equity ratios and the cost of merchant debt, which
in the market environment at the time of this writing may be high. EPA's
Platform v6 uses a value of 7.2%, but one of the stated data sources for
17
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
debt-to-equity ratios currently suggests that the cost of debt may be
substantially lower.4 This is a data point that we suggest be updated in
future revisions of EPA's Platform.
• The use of the Hamada equation (Hamada 1972) is a technical
assumption that is made for convenience but the conditions under which
the Hamada equation is valid include an assumption about a constant
dollar value of debt. This is seldom true if firms are continuously
refinancing their debts. Brealy and Myers (2011) point out that a constant
leverage ratio is a more realistic assumption.
The most important issue on power sector finances that arose during our review
concerned the treatment of utility versus merchant investment costs of capital in
EPA's Platform v6. The documentation describes differentiated costs of capital
for producers in deregulated versus traditionally-related markets, and also
differentiates between capital charge rates for utility and merchant investments.
This is intended to reflect differences in risk between these two types of
environments (regulated/deregulated and merchant/utility) and is summarized in
Table 10-2. Table 10-3 shows a single aggregate cost of capital that appears to
be derived by taking the weighted average of the WACC figures for utility and
merchant investment, based on the ratio of utility to merchant investment that
prevailed during the period 2012-2016. This WACC is then applied to all new
investments within the model. Thus, EPA's Platform v6 appears to be developing
differentiated costs of capital for different market actors in the power sector, but
in the model formulation is using a single cost of capital that represents a
weighted average of these differentiated costs of capital.
If this is indeed how EPA's Platform v6 treats the cost of capital, we are
concerned that the finance module does not sufficiently differentiate between risk
attitudes by different investor types. Large integrated utilities in the U.S., for
example, have had a much greater appetite for big capital projects such as
nuclear power plants and carbon capture and sequestration pilot projects than
merchant generators. These utilities can socialize risk across their customer
base, and can also amortize costs over longer time horizons. Merchant
generators cannot manage risk in this way - they may try to use financial
markets to shift risk but the exposure for a merchant generator is generally higher
than for a rate-regulated utility.
Additionally, the literature suggests that there are other cost differences in utility
versus merchant generation firms, beyond the cost of capital. Specifically,
Fabrizio, et al. (2007) suggest that independent power producers are lower-cost
operators as compared to utilities. We therefore recommend that EPA consider
4 See http://people.stern.nyu.edu/adamodar/New Home Page/datafile/wacc.htm.
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
specifying the costs of new units differently for cost-of-service regions and
competitive regions.
We recommend that EPA consider addressing this differentiated risk appetite in
future versions of EPA's modeling platform. One possibility would be to introduce
different hurdle rates for different investor decision-makers and effectively split
investment decisions within EPA's application of IPM. This would make EPA's
application of IPM more complex in adding a number of new decision variables
but could potentially be simplified by limiting investments of each type on a
regional basis. For example, a fully "deregulated" state would have no utility
investment and would have only merchant investment. The payback period for a
utility (in terms of time horizon) would be much closer to life-of-plant, whereas the
payback period required for new investment by a merchant generator would be
shorter than the payback period for a regulated utility investment. Section 10 of
the documentation refers to 10-K filings indicating long useful lifespans for
generation equipment, but these are not the same as necessary payback periods
for investment decisions.
It is possible that differentiating between utility and merchant investment in this
way would yield model solutions where utility investors in "traditionally regulated"
regions would engage in substantial new generation builds for export to
"deregulated" regions. Despite this possibility, we still recommend that EPA
consider separating utility versus merchant investment as separate decision
variables with distinct discount rates. There are modeling steps that could be
taken to mitigate the possibility of some odd simulation outcomes. Utility
decision-makers within EPA's model platform, for example, could be constrained
in the amount of new generation investment so that the installed capacity in those
traditionally regulated regions does not exceed peak demand plus an appropriate
reserve margin for that region.
The documentation could also explain more clearly how tax credits for wind
energy are treated. Based on the documentation, it is not entirely clear whether
wind is treated as an investment tax credit (ITC) rather than as a production tax
credit (PTC). Chapter 4 of the documentation indicates that the tax credit
extensions for new wind units as prescribed in H.R. 2029, the Consolidated
Appropriations Act of 2016, are implemented through reductions in capital costs.
As the credits are based on construction start date, the 2019 production tax credit
(40% of initial value) is assigned to the 2021 run-year builds for wind units. The
tax credit extensions for new solar units as prescribed in H.R. 2029 are
implemented through reductions in capital costs. As the credits are based on
construction start date, the 2020 investment tax credit (ITC) of 26% is assigned
to the 2021 run-year builds for solar PV units. There is no discussion of why the
wind tax credit is modeled as an ITC rather than a PTC - perhaps drawing
parallels with other large-scale power sector planning models would be helpful in
the documentation. Although the tax credit is currently set to phase out and so
19
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
may not significantly affect model results, there could be implications for
modeling the credit as a PTC vs. ITC if the tax credits were extended. For
example, PTCs can lead to negative prices, which has implications for the cost-
effectiveness of storage.
The documentation under review did not include representation of the amended
Section 45Q credit for carbon capture projects. The final version of rules
implementing this tax credit appears in 2020 and thus were not included in this
review.
E. FUELS AND RENEWABLE RESOURCES
Coal
EPA's application of IPM uses a set of exogenous coal supply curves for each
coal grade in each supply region to solve for the price and quantities of coal used
for power generation. For the purposes of modeling the coal market, EPA's
Platform v6 treats each model power plant as its own demand region. These
demand regions are linked through a transportation network to 36 supply regions.
EPA's Platform v6 assigns coal grades to each model plant, with generally more
than one coal grade per plant. Multiple transportation modes are specified based
on existing infrastructure with relative competitiveness assessed for each mode.
New plants use generic transport costs for different coal types.
The coal supply curves are constructed by sorting new and existing mines by
cash cost per ton and plotting cumulative production. Costs include operating
costs without capital for existing mines and with amortized capital for mine
expansions. Considerable detail is used to build up transportation costs as well.
Projections of coal exports, imports, and non-electric sector coal demand are
based on the AEO 2017 projections by region and coal grade. Because the
model has more grade types and regions than the AEO, the model solves the
selection of specific coal regions and grades that meet those imposed by
assumption from the AEO.
Based on our review, the use of coal supply curves and a transportation network
in EPA's Platform v6 appear to provide the dynamic trade-offs of coal grade
selections with environmental control options that are necessary to project the
effects of regulatory policy on the cost and emissions of coal generation.
Consideration has been given to coal mine expansion costs and rail and other
transportation costs in detail.
What is less clear is how costs and prices might change with significant
reductions in coal demands that might occur under some scenarios. For
example, there is not enough information provided in the documentation to
discern whether coal mine closures are exogenous or endogenous within the
model and the degree to which closures in the model reflect recent changes in
20
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
regional fuel supplies. The rate of mine closures and bankruptcies in the Powder
River Basin (PRB) in particular seems to have been faster than expected in the
past year, and at this point the only region adding coal production capacity may
be Kentucky. Declining coal demand might impact unit production costs, and
potentially transport costs, as volumes are reduced.
EPA's use of exogenously specified export and non-electric sector demand
allows these demands to be considered in competition with electric demands, but
only partially. Because the AEO2017 view of coal prices and supplies reflected
in export and non-electric sector demand may not match EPA's view of coal
prices and supply, the projections of other sector demands and exports may be
inconsistent with power plant demand. In addition, if coal use for electricity
generation varies significantly within a scenario, there will be no response in
exports or other demand levels. Because these other demands are not of
interest, the lack of feedback only matters if modified demands or exports would
in turn impact coal prices to U.S. power generators. Overall, this is not likely to
be a significant issue, but EPA should consider keeping the base projections up
to date (using AEO2018 (or AE02020 if an update is done) versus AEO2017),
checking that the coal price and demand by electric sector projections between
the AEO and EPA's Platform v6 are in close alignment, and assessing whether
alternative projections are necessary in scenarios where electric sector coal
demand varies materially from the Reference Case.
Natural Gas
EPA's Platform v6 uses natural gas supply curves derived from the proprietary
ICF Gas Market Model (GMM). GMM is a gas demand and transportation
network model that relies on characteristics of supply from ICF's Hydrocarbon
Supply Model. To create the gas supply curves, IPM and GMM are run iteratively
to find equilibrium prices and quantities. Subsequently, EPA's Platform v6 is run
with the resulting set of curves that indicate how Henry Hub prices change with
changes in the power sector quantity demanded, based on the supply curve for
each solution year.
One disadvantage of this static curve approach is that it treats prices in different
years as independent in EPA's Platform v6 context, rather than as a function of
cumulative production that may vary by EPA scenario, even though the
underlying curves were developed with that consideration by GMM. This can be
addressed by re-estimating the curves by running the models iteratively, as in the
Reference Case set-up, when it seems necessary due to significant changes in
gas demand. In addition, while the slopes of the gas supply curves are clearly
important to the modeling of electricity supply, it is difficult to determine how
these slopes are derived based on the available documentation. Therefore, it is
also difficult to assess their reasonableness. We would recommend that EPA
publish more information about the methodology for deriving the curves.
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
It is also not clear the degree to which LNG exports, both export capacity
expansion and utilization, are determined endogenously versus predetermined.
If U.S. demand for natural gas changes significantly in some scenarios, one
would expect that LNG exports might shift as well, which might have price
implications for domestic gas buyers.
The GMM also serves as the basis for seasonal price differentials that capture
the difference between the Henry Hub price and gas prices in model regions.
While these differentials are endogenously projected by GMM with variable costs
as a function of pipeline throughput and pipeline capacity expansions, they are
fixed in a given scenario context. Under scenarios with different gas demand
patterns (quantities and locations) than the Reference Case, these basis
differentials could be quite different. Again, this may require additional runs
iterating with GMM for some scenarios. Even now, the Appalachian trading hubs
have increasingly separated from the Henry Hub in terms of pricing. Until there is
significant pipeline build-out in the Appalachian region, this separation is going to
be structural rather than transient (i.e. not an adjustment process to a new
equilibrium).
Within the GMM, econometric equations project other sectoral regional gas
demands. The elasticity of these demands presumably impacts the overall
supply elasticity of gas to the power sector. We note, however, that an emerging
petrochemical sector in the Appalachian production region is likely to affect
regional natural gas pricing in ways that may not be well represented in the gas
market model. Petrochemical facilities demand natural gas liquids (NGLs), not
dry gas, and these NGLs are a co-product of natural gas production in some
portions of the Marcellus and Utica deposits. Increased NGL demand from
petrochemical facilities will require additional natural gas production, but without
substantial regional storage or pipeline additions, there is likely to be additional
stranded dry gas in Appalachia. The industry's future trajectory is uncertain, but it
represents a non-power sector demand for gas that will be important for regional
supply and pricing.
Other Fuels
Residual and distillate oil prices to electric generators are specified exogenously
in EPA's Platform v6 and are taken from the AEO 2017. Nuclear fuel prices are
from the AEO 2018.
In our view, using an exogenous oil price is appropriate because oil prices are
not likely to change materially with changes in power usage. However, as a
minor point, oil prices are treated inconsistently across EPA's Platform v6 and
ICF's GMM platforms. While oil prices for power generation are based on the
AEO2017, diesel fuel prices used in developing rail rates for coal are from the
AEO2016. At the same time, oil prices used in the GMM, which are used to
determine fuel switching in the industrial sector, are quite different from those
22
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
from the AEO that are used in the rest of EPA's Platform v6. In the GMM, oil
prices increase and then flatten at $75/barrel. There is no indication in the
documentation what oil prices are assumed in the Hydrocarbon Supply Model in
developing associated oil and gas production projections, although we
understand from EPA that they are the same as used in GMM.
Biomass is represented in EPA's Platform v6 by annual regional supply curves
that are derived from those in the Department of Energy's 2016 Billion Ton
report. The curves are built up from county level data and aggregated to each
model/state region. The curves reflect transportation costs, as well as storage
costs that are added to the steps of the agricultural residue supply curves. No
interregional trade is allowed, which seems reasonable given the generally high
cost of long-distance biomass transport.
Renewable Resources
Solar and wind resources are represented regionally in EPA application of IPM
by resource class. For wind, resources are defined by techno-resource groups
(TRG) as developed by NREL that are based on estimated levelized costs
reflecting expected capacity factors, capital costs, and O&M costs. EPA's
Platform v6 also divides resources further into 3 or 6 capital costs steps (3 for
offshore wind and 6 for onshore wind and solar) that represent increasingly
higher transmission costs. This is similar to the methodology used in many other
capacity expansion models.
EPA's capital cost adders for wind and solar, which are based on an estimated
distance to transmission infrastructure, do not appear consistent with those for
non-wind and non-solar units that represent the cost of maintaining and
expanding the transmission network. These latter costs are based on AEO 2017
values and are equal to 97 $/kW outside of the WECC and NY regions and 145
$/kW within those regions. The wind and solar PV adders start as low as $1/kW,
with especially low values for PV even though it is assumed that these are utility
scale rather than rooftop systems. It would seem more consistent for the generic
transmission network costs to be applied to all units rather than exempting wind
and solar. Otherwise this provides a bias towards wind and solar PV
development.
F. REGIONAL AND TEMPORAL RESOLUTION
For planning models such as IPM that consider decisions related to costs and
benefits over decades, a high level of aggregation is often necessary for the
model to function. Consistent with this objective, EPA's application of IPM
aggregates its representation of the electricity system in the three following ways.
• Model Run Years: EPA's Platform v6 contains a roughly 30-year
planning horizon, compressed into eight representative model run years.
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
The results from a representative year are assumed to be representative
of a series of years associated with the model year. Model years are
more spaced further apart later in the horizon (e.g. 5 years between
model years rather than 2 or 3 years as in the nearer term).
• Load Segments: The 8,760 hours in a given model year are aggregated
into 72 representative hours by first dividing seasonal (3 seasons) load
duration curves (LDC), in which hours are sorted by loads into 6 load
levels and then dividing each of these into 4 groupings by time of day
(night, morning, mid-day, and early evening). There are 72 such load
segments for each model region although some may contain no load, for
example the top 1% load hours in the summer may have no nighttime
hours, in which case the summer/peak/nighttime segment would have no
representation.
• Model Regions: EPA's Platform v6 divides the continental U.S. and
Canada into 78 modeling regions. Each region has its own load and
supply characteristics, capacity requirements, and transmission capability
to other regions. The model disaggregates some regions such as NYISO
much more extensively than others, such as the Southeastern U.S.
regions.
While in several ways, EPA's Platform v6 is "state-of-the-art" for a linear-
programming model that utilizes an LDC-style aggregation of load, it is important
to recognize the trade-offs inherent in aggregating space vs. time in an LDC
approach, as well as in using the LDC approach at all. Key considerations
related to these tradeoffs include the following:
• Load aggregation can dilute outcomes that are concentrated into a
small number of hours. If certain key outcomes, such as capacity
requirements, transmission utilization, and episodic peak emissions are
driven by conditions in a small set of hours, such as peak hours, then
aggregation even up to 1% of all hours may average outcomes in the
lower end of the load grouping with the higher end. EPA's Platform v6
addresses this well by specifying a very high peak load segment,
representing only 1% of all hours. However, key transient outcomes in
the system may not be limited to only peak hours, particularly with
extensive adoption of renewable energy resources.
• Load aggregation necessitates difficult modeling choices regarding
inter-regional trade. With a large number of regions aggregated up to
individual load-shapes, the model must represent how exports in a given
load segment (e.g. segment 1) in region A map to imports in another
region's load. The standard practice for models like IPM is to simply
assume that all regions are in a given segment at the same calendar time,
so that trade between regions happens at the load-segment by load-
24
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
segment level. Such an assumption is problematic, as Figure 2 below
illustrates.
The top panel in Figure 2 plots concurrent load in the CAISO region
relative to load in the Bonneville Power Administration (BPA) region for
mid-day spring and fall hours during 2014-2016. The bottom panel plots
the load in Georgia Power against that in Florida Power & Light (FPL) for
the same time periods. If load were perfectly correlated, all observations
would fall along the 45 degree lines in each panel. The spread of
observations in the top panel indicates that peak load in CAISO and BPA
are very much non-concurrent. Florida and Georgia are much more
correlated but there are clearly peak hours in both that fall in lower load
segments in the other.
Figure 2: Load Correlations between Neighboring Control Areas
<
O
CAISO vs. BPA concurrent 714 Load
Spring and Fall mid-day hours
.0001 .00015
Fraction of BPA Annual Load
.0002
O _
o°
Georgia Power vs. FPL concurrent 714 Load
Spring and Fall mid-day hours
-.00005
.0001 .00015
Fraction of FPL Annual Load
.0002
Data shown here from FERC Form No. 714 - Annual Electric Balancing
Authority Area and Planning Area Report.
25
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
To its credit, EPA's Platform v6 goes beyond the simple assumption of
concurrent load and requires that exports from one region's load-segment
"arrive" in importing regions during the same calendar hour in which they
were exported. These calendar hours likely would fall across a range of
different load segments in the importing region because calendar hours
are assigned to different load segments in different regions. For example,
exports from the Pacific Northwest during its 1% daytime summer peak
hours would be assumed to arrive in California during those same
calendar hours. Some of those hours may fall within the second or third
daytime summer load segment in northern California because the 1%
peak demand occurs on different days or hours in the two locations.
While this is likely an improvement over the naive assumption of perfect
correlation of hours and load segments across regions, it still makes
strong assumptions about the timing of trade. It is difficult to diagnose the
full impacts of this implementation. Our intuition is that it constitutes a
hidden penalty on trade between regions; in order to export during hours
in which trade is beneficial, the model may be forcing additional trade in
hours in which trade is not beneficial. If true, this means the model will
bias downward trade between regions.
Again, this is in contrast to other LDC-based models that ignore these
problems. Those models are not built to describe trade between identical
calendar hours and allow trade between hours that do not actually
coincide, simply because of their position in an LDC.
One additional comment on this point is that the documentation does not
describe this aspect of interregional trade. A description with an
accompanying example would help promote understanding of this feature
of the model.
• Geographic aggregation involves trade-offs between accuracy over
time vs. space. The aggregation of geographic regions implicitly
assumes that all plants and loads within a region share common LDC
load segments and face no transmission congestion. Our understanding
is that the increase in the number of model regions in version 5 is
motivated by a desire to faithfully capture important inter-regional
transmission constraints. However, this benefit comes at the cost of
dividing up shared calendar hours into potentially different load segments
- as described above. One way to reduce the problems identified above
is to aggregate over larger geography. Further, since trade between
regions is subject to the distortions of which "time" the trade is occurring,
the representation of transmission flows would also be skewed in the
same way, limiting at least somewhat the benefits of being able to model
a given transmission interface.
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
• Load aggregation limits modeling of inter-temporal constraints.
While electricity markets work to operate generation units in a strict "merit-
order" based upon costs, many units feature operating limits such as
ramping rates, start-times, and minimum down and up times. In addition,
hydro and storage units feature energy limitations or charge/discharge
cycles. Fully representing these considerations requires modeling a
sequencing of contiguous hours that is impossible when aggregating to
an LDC. In many cases, these considerations "average out" over time.
Simpler modeling understates output some hours and overstates it in
other hours.
For regulations that concern total output or emissions from a power plant
during a season or year, the aggregation is likely relatively benign.
However, for the purposes of assessing any environmental regulations
focused on peak emissions, episodic emissions, or emissions intensity,
the aggregation could be more problematic. The aggregation would likely
bias downward the output of high heat-rate "peaking" plants. The
turndown constraint could at least partially offset this bias, but it is difficult
to understand what the effect of this constraint is without running the
model with this constraint disabled.
Given the costs and benefits of the aggregation choices described above, we
would urge EPA to consider alternatives that might be compatible with a linear-
programming implementation. These are not necessarily recommendations, as it
is difficult to know the magnitudes of the issues described above. We note,
however, that accurately representing inter-temporal constraints and modeling
the correct timing of "peak" and "off-peak" net loads could be more important in
the future than is the case today due to evolving electricity supply conditions.
Publish more output details: Currently model outputs are not broken out
by load-segment. This additional output detail may allow stakeholders to
better judge the relative impacts of the various aggregation assumptions
in a given policy context. If this information is not generally shared with
the public, EPA may still consider modifying IPM to generate these
outputs for purposes of model and scenario evaluation before publishing
results in more aggregated form.
• Investigate the Time vs. Geography Trade-off: It is possible that the
goals of the model may be better implemented with more temporal
resolution and that this could be aided by less geographic resolution.
With fewer model regions, LDCs better represent timing within a region
(but could exacerbate dis-alignment between regions). There could also
be some computational savings through geographic aggregation,
although these may be limited by the need to model each power plant
individually, no matter what region it is in. As discussed below, additional
time-based modeling could take the form of additional segments or
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
seasons within the current framework or a more substantial
reconfiguration of timing.
• Consider grouping hours first by time of day and then by load
segment, instead of the other way around. We do not understand the
motivation for dividing hours into only 3 seasonal groups before sorting
into load shapes instead of grouping into 3 seasons x 4 time-of-day
segments and then aggregating into 6 segments each. The latter
approach would provide a more balanced number of hours in each load
segment within each season/time-of-day block.
• Investigate the implications of grouping hours first by load segment
for a whole interconnection and then by region. Another way in which
the grouping order affects the trade-off between regional vs. local
accuracy in load conditions is the practice of grouping load by model
region first, and then sorting by load level within regions. An alternative
would be to group hours by their interconnection-wide load level and then
subdivide into regions. For example, the top 37 summer hours would be
chosen from the hours with the highest total load across the WECC.
Those hours would then be assigned to the different model regions within
the WECC. This approach would more faithfully recreate the conditions
available for trade between model regions, but would lose accuracy in
capturing, for example, peak conditions within a given model region.
• Represent time as a sequence of "model hours" or "model days."
Computational limits require limiting the number of demand segments
(region + hours) represented in the model. LDC aggregation allows for all
hours of a year to be approximately represented but at the cost of not
knowing when each hour falls relative to others. One alternative would be
to group hours into time-of-day blocks (e.g., 4 hours) and model them
sequentially, allowing for better representation of some inter-temporal
constraints, inter-regional trade, and probably renewable energy and
storage output. One such "model week" per season could capture a peak
day and other important load characteristics. However even this
aggregation would require 126 hours of modeling, which may be
infeasible.
Another alternative would be to group hours into time-of-day blocks for
typical weekdays and weekend days with a preservation of peak loads
through a peak-day or other method. Once the hours are more
contiguous, it may also be worth adjusting time zones across regions so
that the hours represent the same actual time.
• Run the output of a model scenario through a more detailed
dispatch model. Another way to investigate the relative importance of
the inter-temporal constraints would be to take the output of a given
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
model scenario and run the resulting unit configurations through a
different model designed to capture short-term operating constraints in
more detail. This second model could be used to test for "peak"
emissions impacts of intertemporal dispatch constraints, for example.
• Model fewer future years. One way to compensate for more detail
within a given model year would be to run fewer model years. Seven,
instead of eight, explicit model years would reduce the number of demand
segments (region + hours). There is so much uncertainty about future
conditions, particularly in the 20- to 30-year time horizon, that it is not
clear how valuable more model years are that far into the future.
However, care should be given to avoid "end-year" effects in years of
interest.
• Consider the impact of the discount rate in the objective function.
EPA's Platform v6 discounts model objective values by 4%, meaning that
costs 20 years in the future count for roughly half of costs in the first year.
Such discounting is appropriate given the high degree of uncertainty over
future conditions. Important model outcomes should not hinge upon
decisions the model makes 25 to 30 years in the future. An even higher
discount rate may be appropriate to minimize the impact of out-year
decision making on model outcomes.
G. POWER SECTOR POLICIES
Federal environmental policies aimed at the electricity sector are represented in
detail in EPA's Platform v6. These policies include many prescriptive technology
policies, tradable performance (emissions rate) standards, and cap and trade. In
general, we find that EPA models these policies well. Power sector policies
pertaining to the market structure, transmission pricing, reliability standards, and
other features of economic regulation can also have an important influence on
environmental outcomes; these are represented in less detail in EPA's Platform
v6 than the other types of policies listed above.
Although trading programs are represented well in EPA's Platform v6, the
documentation indicates that the model does not include any explicit
assumptions on the allocation of emission allowances among model plants under
any of the programs. An element of cap and trade that may be challenging to
model is dynamic allocation of emissions allowances that maintains the
emissions cap. An example of this approach appeared in Virginia's final 2019
regulation for the introduction of cap and trade for carbon emissions, in which
allowances were to be distributed to generators based on their share of
29
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
generation in a recent period.5 This approach couples a carbon price with a
production incentive that stems from the ability to affect one's allocation by
changing one's production. The model might represent such a policy by
describing the production incentive that is embodied in this type of allocation.
One difference between dynamic allocation and auctioning is the observed price
of emissions allowances and the effect on the price of electricity (Rosendahl and
Stomasten (2011).
Similar challenges arise in representing clean energy standards, which are
emissions intensity standards with effective emissions targets that adjust over
time in response to the quantity of production. This differs from dynamic
allocation of emissions allowances among facilities with a specific emissions
target in place, but it is similar from an analytical perspective in that it embodies a
production incentive. These issues may be challenging for EPA to represent, and
may require model development. EPA's Platform v6 does not currently represent
these policies, except potentially through an iteration procedure.
Other challenging elements in representing power sector environmental policies
include dynamic adjustments to emissions budgets based on the prevailing price
in an auction, as illustrated by the emissions containment reserve in the Regional
Greenhouse Gas Initiative. Another example is program-related spending that
may be tied to allowance proceeds, such as the direction of auction proceeds
among states in the Regional Greenhouse Gas Initiative to investments to
promote energy efficiency. In principle, these policies also can be addressed
through iteration, but this may require more time and multiple runs of the model.
Dynamic policy features may be increasingly relevant in the future, at both the
federal and state level. The Agency needs to anticipate this in considering the
evaluation of power sector environmental policies.
With the expansion of state-based environmental and clean energy policies,
there is increasing interest in leakage of generation and emissions from
jurisdictions introducing regulations to unregulated jurisdictions. Conversely,
unregulated jurisdictions could see an increase in wholesale power prices if they
increase generation to serve demand elsewhere. IPM has been used to model
policies similar to these, and IPM produces projections of the implications of
these market dynamics and is a useful tool for the EPA to evaluate and
understand them. For example, IPM was used to assess impacts of the Clean
Power Plan.
Moreover, there is an interaction between electricity sales and transmission, and
renewable energy credit markets. Although we understand that this interaction is
embodied in the model, we have not seen it represented in previous exercises of
the model or described in the documentation. Policies such as carbon pricing that
5 That final regulation was never implemented. Legislation in 2020 established Virginia's
participation in RGGI and the allocation was replaced with a revenue-raising auction.
30
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
promote an increase in renewable generation in one region could precipitate a
decrease in renewable generation in another region if renewable energy credits
become available in the region introducing carbon pricing that can be used for
compliance with renewable portfolio standards in other jurisdictions. These
issues are the subject of an ongoing study in PJM and various types of border
carbon adjustments have been proposed. These are emerging policy and
industry issues that the panel wants to be sure that EPA considers in its updates
to its modeling platform. EPA may want to begin to consider these issues and
direct development of EPA's application of IPM to track and report the effects of
policies implemented within narrow geographic regions.
Some prescriptive policies such as New Source Review constrain the utilization
of an existing generating unit and limit investments in new units in a geographic
area. The Agency should pay special attention to this in evaluating its modeling,
because it is not obvious in the documentation that generation from units that are
in nonattainment areas or subject to New Source Review are necessarily
constrained from expanding generation, or may be required to introduce pollution
controls that may raise their costs. Relatedly, nonattainment areas are not
congruent with the power regions in EPA's Platform v6. The panel understands
that no explicit generation limits are modeled; however, unit level emission rates
and permit rates are explicitly modeled in the EPA Reference Case. To capture
the influence of New Source Review-type regulations, the model might be
adjusted to preclude increases in generation at a plant, or to condition such
increases on the installation of post-combustion controls.
The model seems to capture policies that affect the bulk power system - but
does not seem to capture state level policies that encourage behind-the-meter
generation except represented as a change in demand. This could be
increasingly important if federal policies introduce new requirements on states to
achieve emissions goals, as represented for example in the recent outline of
potential climate legislation developed by the House Committee on Energy and
Commerce.
Several other possible energy policies may become increasingly important, such
as reliability constraints defined by on-site fuel storage and the minimum offer
pricing rule (MOPR). These policies may have important environmental
implications. EPA's current approach to modeling capacity requirements does not
fully capture the impact of any of these policies, although EPA's Platform v6 may
have the capability to reflect elements of these policies. For example, if EPA
models nuclear generation incentives from state-level zero-emission-credit (ZEC)
policies, it would also need to evaluate whether those nuclear plans should count
towards satisfying a regional capacity constraint. Under final MOPR rules, which
are as of yet to be determined, such nuclear capacity may not be part of capacity
market compliance. This may not require new model development, but it might
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
require the attention of the Agency in specifying the way policies in the model are
represented.
Finally, one last finding pertains to the representation of the CO2 emissions rate
for imports to California, at 0.428 MT/MWh. This is the default rate for power that
is not assigned a specific emissions rate, but the major portion of power coming
into the state is assigned a specific rate. Moreover, it is unclear that the
incremental unit that is affected by a policy under evaluation will have an
emissions rate that is proximate to the average rate. A careful solution to this
could be found through iteration, solving the model twice varying the level of
demand in California in order to identify the marginal resource providing power to
the state and region, but this may require additional Agency resources. With the
expansion of the Western Energy Imbalance Market to reach as far as Colorado,
accounting for emissions intensity of transmitted power will be increasingly
important.
H. RETAIL PRICE ESTIMATES
EPA's Retail Price Model (RPM) is a post-processing model that estimates the
relative retail electricity price impacts of different regulations or scenarios. To
estimate retail price impacts, the RPM relies, in part, on scenario outputs from
EPA's Platform v6 as its inputs. Retail rates in the model are comprised of three
components: (1) transmission costs, (2) distribution costs, and (3) generation
costs. Currently generation costs in actual rates comprise roughly 30 to 50% of
the total costs included in retail rates.
As long as transmission and distribution are not modeled as endogenous choices
in the model for EPA's purposes, these components should not vary between
scenarios or under different regulations. Therefore, the crux of the RPM lies in its
translation of wholesale energy costs into retail prices. The main challenge faced
by the model in fulfilling this objective is the fact that changes to some generation
costs will be transmitted into retail rates differently in fully regulated states than in
states that have undergone regulatory restructuring/deregulation. The model
therefore has two different generation cost pricing formulas to match the different
regulatory environments. For each model region, the final retail price reflects a
weighted average of the regulated and restructured generation price.
a) Cost of Service Generation Cost Pricing.
Final Cost of Power = (Average Generation Cost + Utility
Generation Depreciation Cost + Return on Equity and
Debt + NUG Adder) *(1+Tax Rate)
where the first three terms represent the variable and capital costs of a
cost-of-service generator and the last term captures the cost of pre-
32
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
existing non-utility generation (NUG) present in the service territory.
While it is not obvious that the capital costs of a merchant NUG would be
directly passed on to retail rates, we assume the NUG adder is included
because these costs are captured in long-term contracts between the
generator and the local load-serving entity (but this is not explained in the
documentation).
The Average Generation Cost represents the full average cost (including
capital) of all units built or retrofit within the model run, as well as the
variable cost of generation from existing units. The Utility Depreciation
Cost represents the capital costs of existing units owned by regulated
utilities. The Return on Equity and Debt is the return earned on those
existing assets.
b) Deregulated Generation Cost Pricing
Competitive = (Marginal Energy Price + Reliability Cost +
Generation Cost Renewable Energy Credit Cost)*( 1+Tax Rate)
Based on our review, the RPM includes much more than the rate impacts of
going-forward costs (i.e., the incremental cost impacts of a policy). This is
consistent with some applications of the RPM. The formulas in the RPM in
theory include all components of a retail rate. A key trade-off to consider in the
RPM is whether to attempt to measure the levels of rates or only to focus on the
changes in rates going forward. For example, the formula for cost of service
generation includes only one term, Average Generation Costs, that captures
newly incurred costs. All the other elements in that formula capture components
of the existing rate and should not necessarily be expected to vary by scenario
going forward.
One important reason to capture the full level of rates would be to model the
impact of those prices on the level and shape of demand. Currently EPA does
not regularly use demand elasticity to estimate the demand response to changes
in electricity prices. If that were to change and demand responses to prices
became endogenously modeled, then the ability of the RPM to reflect both the
changes to and levels of retail prices will influence most outputs of the model due
to the feedback between wholesale outcomes and retail demand.
Going-forward impacts of policies can also be divided into a) impacts of policies
on going forward costs and b) impacts of policies on the pass-through of existing
costs to retail rates. These distinctions are important because the differences
between regulated and competitive markets matter less for capturing going-
forward costs than for the pass-through of existing costs. A model such as IPM
effectively represents perfect competition and perfect regulation - resulting in
least-cost investment and operations. In theory, both perfectly regulated and
perfectly competitive markets should produce the same outcome, and total costs
33
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
going forward. Therefore, if the only purpose of the RPM were to measure the
rate impacts of going-forward costs, it is not clear the model needs to make a
distinction between competitive and regulated markets. However, our
understanding is that EPA intends the RPM to measure rate impacts of both
current and going-forward costs.
The model does attempt to capture the differential impact of policies on the
pass-through of existing costs. Since the purpose of the RPM is to capture
the full rate impact of policies, including their effect on the pass-through of
existing capital costs, then there is justification for treating regulated and
competitive markets differently. Two examples of such differences are 1) the
impact of policies on marginal generation costs (energy prices) and 2) the
possibility that policies might "strand" certain existing capital assets.
Given that EPA does intend the model to shed light on transitional rate impacts,
such as the impact of policies on the pass-through of existing costs, then the
general modeling approach taken by the RPM is appropriate. The model should
capture the fact that energy prices will fluctuate more in response to certain
regulations than others, and that rates in competitive states will likely reflect
these fluctuations. To illustrate this point, Figure 3 shows the disparate retail rate
response to natural gas price volatility since 2000. Rates in competitive states
rose more rapidly when gas prices rose in the early 2000s, and fell more quickly
when gas prices declined post 2008. One would expect a similar differential
pattern from a regulation that effects the marginal cost of marginal generation,
such as a carbon tax or the Clean Power Plan. As the figure illustrates, however,
these differential effects tend to average out over long periods of time.
Figure 3. Rate Responses to Natural Gas Price Changes
Year
- Non-restructured States Restructured States
- Elec. Price Difference Natural Gas Price ($/MCF)
Source: Borenstein and Bushnell (2015).
Burtraw, Blumsack, Bushnell, Felder, and Wood
34
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
The general structure of the generation pricing formulas seems
appropriate, but the purpose and magnitude of elements, such as the NUG
adder, should be better explained. The formulas described above capture the
key elements of pricing in regulated and competitive markets in the sense that
regulated prices are based upon IPM's calculations of total average costs and
competitive prices are based upon IPM's calculations of the marginal cost of
serving load (including compliance with state Renewable Portfolio Standards,
etc). However, there are other elements that are somewhat opaque and appear
to contribute significantly to the total rate. For example, the NUG adder is
substantial in many regions in early years. Our understanding is that this is
meant to capture the cost of existing NUG generation in regulated states selling
power through power purchase agreements (PPAs) to regulated utilities.
However, even in deregulated states, there are substantial NUG transactions
through PPAs, and there is no NUG adder to rates in those states.
The logic behind the definition of competitive vs. regulated regions is
unclear. There are many possible definitions of "regulated" and "competitive"
and the RPM utilizes definitions from ElA's Annual Energy Outlook for its
assignments of regions to these categories. This may not be the best definition
for this application.
The RPM is ultimately trying to capture the pass-through of generation costs to
retail prices, and therefore it is the regulatory treatment of the generation that is
most relevant to this calculation. The weights for competitive and regulated
regions in the RPM appear to be too blunt compared to generation ownership
patterns defined by the EIA. Figure 4 compares the percent of generation that is
regulated for each state (top panel, aggregated to states from model regions) in
the RPM against the 2018 share of generation coming from non-utility (bottom
panel, Independent Power Producer and Combined Heat and Power) sources
according to EIA. The RPM tends to assign either a zero or 100%
regulated/competitive designation in many states, while the generation shares
according to EIA are much more blended. One reason for this difference is likely
the NUG generation operating under power purchase agreements selling to the
local utility in regulated states.
Based on the above comments, we recommend that EPA improve
transparency of the RPM results and their components. The components
included in the competitive and regulated pricing formulas seem appropriate,
provided that exogenous components such as distribution charges and taxes are
calibrated correctly. However, some components such as the NUG adder seem
subjective and it would be useful for a consumer of the results to understand how
important different components are to the rate impacts coming out of the RPM.
One suggestion would either be a table or stacked bar chart detailing not just the
total rate (or change in rate) but also the components that make up that total.
Most of these details are available but take considerable effort to put together.
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
Figure 4. Regional Competitive Generation Shares
EIADeregulation Share by State
2018 IPP & CHP Generation Share by State
36
Burtraw, Blumsack. Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
EPA should evaluate and articulate the purpose of distinguishing between
competitive and regulated regions. The choice of whether to model
competitive and regulation regions differently in the RPM depends largely upon
the intended usages of the model. If it is important to capture near-term rate
impacts of policies that change how existing costs are passed through to rates,
then maintaining a regulation/competition distinction is important. If the model is
intended more to capture the rate impacts of going forward costs created by
policies, rather than the impacts of those policies on the pass through of existing
costs, then there may not be much difference between the regulated/competitive
results. One other point to consider is the role of PPAs and other long-term
contracts with NUGs in competitive markets. Even if short-term energy prices
fluctuate in response to policies, to the extent that generator payments are locked
in through PPAs and other contracts, those short-term energy price changes will
not all be passed through to retail prices, even in competitive markets.
Consider a simpler retail price formula based upon regression analysis of
the relationship between generation costs and retail rates overtime.
Predicting the movements in retail prices is a difficult challenge and although the
RPM captures the theoretical relationship between generation costs and retail
prices, it may be more transparent (without being much less accurate) to apply a
simpler formula based upon the historic relationship between costs and prices.
A regression-based approach would allow for the estimation of confidence
intervals or other measures capturing uncertainty as well.
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
37
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
IV. BASE SET OF MODEL SCENARIOS
Check the appropriateness of the base set of model-scenarios for addressing uncertainty
in potential future power-sector trends, focused on answering these questions:
a. Are the base set of model-scenarios (which include a reference case, low demand
case, high demand case, low renewable cost case, high renewable cost case, and
a high gas cost case) appropriately characterized? How well do these scenarios
suit EPA's analytical needs ?
b. Do the model scenarios reflect the most robust sources of uncertainty for the
power sector? Are any of the model scenarios extraneous? Outside of a federal
regulatory context, are there significant areas of uncertainty in the power sector
that are not covered by these scenarios? How well does the range of scenarios
suit EPA's analytical needs?
EPA has made several model scenarios available in the public domain. These
scenarios reflect many of the most important uncertainties driving operations and
investment in the electric power sector, including the price of natural gas and the
cost of renewable energy (particularly wind and solar). EPA has also made public
a scenario involving changes to the U.S. tax code, which affects financial
incentives for investment in the electricity sector.
NON-PARAMETRIC UNCERTAINTY
While we do not question EPA's choice of using a deterministic linear
programming model, it is important to recognize that this structure limits the kinds
of uncertainty that IPM can reasonably handle. The uncertainty reflected in the
model run scenarios is limited to "parametric uncertainty", which can address
uncertainty over the values of one or more input variables without changing the
model's overall decision structure. EPA's Platform v6 as currently configured,
however, cannot handle "intrinsic uncertainty", which is characterized by
uncertainty over decision rules or processes. Any decision process that deviates
from present discounted cost minimization, for example, cannot be captured
directly by IPM.
Because EPA's Platform v6 as currently structured is limited to handling
parametric type uncertainty, there may be some biases reflected in the model
outputs. These biases are probably very situationally dependent. EPA's Platform
v6 as currently structured, for example, may exhibit a bias towards decisions that
reflect current power-sector conditions and incentives even though these
conditions are rapidly changing.
Another limitation of the focus on parametric uncertainty is that sensitivity
analysis does not show how behavior may change or may depart from expected
net present value maximization in the presence of uncertainty. For instance,
option theory suggests rational decision makers will delay irreversible
38
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
investments (and retirements) in the face of uncertainty to gain more information
about the uncertain aspects of the scenario. This behavior will not be evident in
an inter-temporal optimization linear program such as IPM. However, this
element of decision-making under uncertainty might be represented by adjusting
the hurdle rate for investment and retirement options, perhaps implemented as a
shadow cost of capital for investments that would be vulnerable to specific
parametric uncertainty, including those mentioned above and potential policy
uncertainty.
Also of note is that EPA's Platform v6 is not currently structured to capture "deep
uncertainty" (Walker et al., 2013), which reflects a situation where parties cannot
agree on the nature of the uncertainties that the system faces or on how to rank
or compare potential solutions.
PARAMETRIC UNCERTAINTY
The parameter uncertainty represented by the set of model run scenarios made
public by EPA capture many of the most important factors that have driven power
sector investment and operations over the past ten years, and are likely to
continue to influence the power sector in the coming decade (at least). An
important aspect of incorporating parametric uncertainty into models of this type,
in which a large number of assumptions must be made due to limitations in data
or model tractability, is sensitivity analyses to understand the robustness of
model outputs to uncertainty in model input parameters. Beyond the scenario
runs made public by EPA, it is not clear what sensitivity analyses EPA conducts
to determine which parameters are the most important in determining variation in
model outputs. EPA's application of IPM is so complex that it may be the case
that no single parameter is driving the model outputs all by itself. Some attempt
at investigating and publishing which parameters or combinations of parameters
most heavily influence model outputs in the Reference Case would be very
useful.
There are additional factors not represented in the set of model run scenarios
published alongside the Reference Case as part of EPA's Platform v6 that are
likely to be important drivers of power system operations and investment in the
coming decades. We believe that model runs representing these factors are
useful in understanding the influence of fundamental changes in the electricity
industry on the results that EPA's Platform v6 produces, and therefore would
have substantial value as additional sets of model run outputs published along
with the Reference Case. EPA should therefore consider ways to represent these
factors in additional sets of model runs published alongside the Reference Case
in EPA's Platform. Some of these factors include the following:
Changes in the shape of the load duration curve. EPA's Platform v6 currently
features low and high demand scenarios in addition to the Reference Case, that
are taken from AEO projections. These low and high demand scenarios do not
39
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
explicitly modify the shape of the load duration curve, only the rate of overall
annual growth in electricity demand. The panel understands the change in load
shape can be explicitly captured in EPA's application of IPM when such
information is available. Load shapes are adjusted to account for the load factor
embedded in the AEO high and low demand scenarios. However, even under
such scenarios, the model does not capture fundamental changes in the the way
load may be shaped dynamically in the future, reflecting interactions between
demand and the bulk power grid that are likely to be a defining feature of the
evolving electricity system over the next two decades. Scenarios driving these
changes in the nature of demand patterns (not just the amount of aggregate
kilowatt hour consumption or the ex ante assignment of levels of demand over
times of day) are likely to include (1) vehicle electrification, (2) time-varying retail
rates that encourage load shifting and peak-time demand response, (3)
wholesale (aggregated or individual customer) demand response that is generally
dispatched during summer peaks to ameliorate very high market clearing prices
or reduce peak system loadings for reliability reasons, and (4) the penetration of
behind-the-meter generation and energy storage.
These changes in the nature of electricity demand, in isolation or taken together,
can be represented in EPA's Platform v6 through adjustments to the load
duration curve. In some cases, these adjustments may be fairly straightforward
(for example the same solar data used to model location-specific solar production
could be used to model offsets to different demand segments, by correlating
time-varying solar production with the demand segments in each region and
netting regional behind-the-meter solar production against regional demand
segments). In others, the nature of the demand adjustment will itself be scenario-
dependent. The timing and nature of electric vehicle charging, for example, will
influence the impacts on diurnal load curves that may translate to changes in the
demand segments used within EPA's Platform v6. Wide adoption of electric
vehicles combined with primarily nighttime charging will increase the level of
demand in what are currently lower-demand segments in a way that represents
overall nighttime load growth, potentially without corresponding demand
reduction in other load segments.
Even without changing the load duration curve, we also suggest including a
scenario in EPA's Platform, along with the Reference Case, that involves
negative demand growth arising through greater energy efficiency measures for
buildings and appliances.
Scenarios that reflect uncertainty regarding fuel availability: Resilience of
the power grid to fuel supply disruptions has gained some policy attention,
particularly as power generation shifts towards the use of natural gas and away
from coal. There are potential interactions between environmental policy and this
kind of fuel substitution, but even cutting-edge planning models in the literature
(e.g. Bent, et al., 2018) do not adequately capture the multi-decadal implications
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
of shortages in infrastructure to deliver fuel to power plants. We would encourage
EPA to publish scenarios alongside the Reference Case involving negative
shocks to fuel supplies, particularly in the northeastern U.S. where resistance to
additional fuel delivery infrastructure has been high. These negative shocks could
be modeled as outages or de-rates to certain types of generating units in certain
regions within EPA's application of IPM, or (perhaps preferably) using high fuel
prices to indicate shortage (see an example for natural gas in Bent, et al., 2018).
The model scenarios should be compared to empirical experiences, of which
there are some past examples due to weather events.
Scenarios that capture multiple parametric changes: We view capturing
interaction effects between parametric scenarios as being valuable for EPA's
purposes as well as for the broader analytical community. The number of
possible combinations is large, but we would prioritize assessing and making
publicly available the following model scenarios:
• Scenarios that interact shifts in load duration curves with existing
parametric scenarios (such as low/high gas prices and renewable energy
costs);
• Scenarios involving very low gas prices and rapidly declining capital costs
for renewable power generation;
• Scenarios involving fuel supply shocks and low capital costs for
renewable power generation (implying a larger dependence on renewable
energy during supply shocks, and the response of the system to that
known dependence).
Unexpected Events: Finally, we observe that EPA's Platform v6 as currently
configured is ill-equipped to handle unexpected events that might arise over the
multi-decadal time frame that it models. Yet, these kinds of surprise events can
often be critical drivers of energy system change. The rise of unconventional
natural gas as a major domestic fuel source is one recent example that could not
have been foreseen two decades ago. IPM has difficulty handling these kinds of
events because of its implicit assumption regarding perfect foresight. Without
deviation from the linear programming and deterministic structure, however, we
do see a straightforward way for EPA to be able to model specific scenarios that
involve parametric surprise events, and encourage EPA to publish the results of
such scenarios alongside the Reference Case. Such a procedure might progress
as follows.
1. Define a parametric shock that would occur during a defined time interval
over which IPM is run.
2. Run IPM without the parametric shock to obtain a base case of what the
model's outputs would be in the absence of the shock.
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
3. Run IPM a second time, starting at the time of the shock and initialized
with information from the base-case run just before the shock happens.
The outputs from this second run (as compared to the period after the
surprise in the first run) should reflect how decisions change in response
to unexpected information.
4. Choose metrics to compare the base-case and second IPM runs. Aside
from the metrics that EPA often uses now to describe model outputs,
such as investment choices, rate outcomes, and air emissions outcomes,
one particularly interesting feature of this approach could be the ability to
determine which asset types in which regions would wind up "stranded"
by the surprise event. Such "stranded assets" in this context would
include those made under the base case model run but which would retire
or be financially unviable under the scenario with the shock.
To the degree that EPA has applied this approach in past applications of its
modeling platform, this is not described in the model documentation.
The potential universe of shocks that EPA might consider modeling in IPM using
this framework is large, as there are a number of conditions within the power
sector that could change rapidly within the time horizon considered by EPA's
Platform v6. One example of potential policy importance would be a shock to
natural gas supplies that arises because of policy interventions affecting the
utilization of modern hydraulic fracturing techniques, stringent technology
requirements to control methane emissions, or other conditions that rapidly affect
the cost and availability of delivered natural gas supplies to the power generation
sector.
Burtraw, Blumsack, Bushnell, Felder, and Wood
42
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
V. IMPROVEMENTS TO SUPPORT POLICY ANALYSIS
What improvements, if any, could be made to support the analysis of the full range of
policy mechanisms that may be applied to limit power sector emissions? How well does
the model scenario capability of IPM version 6 suit EPA's analytical needs?
EPA's application of IPM has been developed over many years to serve as the
central platform for evaluating federal environmental policies as they affect the
electricity sector. The model is also used widely by state governments and non-
governmental organizations. The model is dynamically consistent with perfect
foresight, identifying scenario results that reflect the cost-minimizing strategies
that would be expected to unfold over many years in the face of the various
identified and anticipated constraints. The model embodies tradeoffs in
granularity along dimensions of space and time, which is discussed elsewhere in
this review.
We find EPA's Platform v6 to be versatile and capable of addressing almost any
well-specified regulation including most prescriptive regulations and flexible
incentive-based regulations at the federal level. Standard environmental policy
mechanisms that have been built into EPA's Platform v6 include tradable
emissions rate performance standards, cap and trade, inflexible emissions rate
performance standards, and technology standards. Based on the evidence
available for this review, the model appears to perform strongly. Nonetheless,
there may be ways for EPA to more fully evaluate its performance that are
discussed elsewhere in this review.
We know from the model structure that policies with a high degree of spatial or
temporal resolution will not be represented perfectly in the model. An example
might be the operation of resources in nonattainment areas, or the ability to site
and build new resources or change the utilization of existing resources that are
subject to New Source Review. Nonetheless the model makes an attempt to
represent these granular constraints where it is important to do so. The panel
also recognizes that the performance and capabilities of the model are co-
dependent on the data configuration underlying the EPA Reference Case.
One policy area that EPA's Platform v6 may not adequately address is energy
efficiency. Demand is taken as parametric in EPA's Platform v6, and demand-
side policies are described as reductions in demand estimated on the basis of
elasticities without changes in the load profile. Alternative load profiles can be
implemented in the model. However, policies or technologies that endogenously
shift load across time would introduce challenges and may not be achievable
given the current model configuration, as we understand it, except through an
iteration procedure. Further, investments in energy efficiency have various rates
of decay, for example, due to various rates of lifetime for appliances. In addition,
changes in prices trigger a partial adjustment process in which effects are
43
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
compounded as behavioral adjustments accumulate. The response to a
sustained change in prices is greater than the response in the first year.
Because of the various federal proposals to promote energy efficiency, EPA may
need to revisit its representation of demand in order to be useful to analysis of
these policies.
One of the largest challenges for EPA going forward may be the representation
of policies at the state level governing retail tariffs, including payments for
distributed generation, and incentives to promote electrification that may
intentionally align demand growth with the availability of variable renewable
energy resources. The potential expansion of flexible demand could be driven by
state or federal policy to promote electrification, federal policy that gives states
flexibility in how to meet emissions goals, and technological factors and may be
important to the evaluation of future environmental policies.
Policies aimed at the demand side and at retail price setting are relevant to the
operation of the electricity system as represented in EPA's Platform v6 because
they are likely to be a key component of strategies to integrate large quantities of
variable renewable energy. Specifically, a possibly important policy mechanism
in the next decade is the determination of retail prices that are differentiated by
time or type of electricity use. Economists have anticipated time-varying retail
prices for more than four decades and although they have yet to emerge widely,
in the last couple of years a number of utilities have begun to introduce time-of-
day prices. If this were to expand with respect to conventional uses of electricity,
it could be important for EPA's application of-IPM. However, potentially more
important are time varying prices applied to new sources of electricity demand
such as electric vehicles, water heating, and building heating that embody
technologies with inherent storage capability. These types of electricity uses do
not require all the attributes of typical "instant on" electricity use. Consequently,
they may not be priced at the same level and they may not be burdened with the
sunk costs associated with the reliability aspects of the existing grid, and retail
prices may be adjusted accordingly.
To represent the meaningful aspects of time-varying prices requires cross-time-
period elasticities of electricity use within a fully functioning demand side model.
One type of modeling approach that would come close is an Almost Ideal
Demand System.6 The key feature in developing a demand side of the model is
that demand should respond over time not only to changes in prices in a given
time block but also to changes in relative prices in different time blocks.
Another potentially important limitation of EPA's policy analyses (that we also
raise in the context of EPA's Platform v6 representation of baseline uncertainty)
is the model's ability to account for the effects of uncertainty on economic
6 An Almost Ideal Demand System (AIDS) is a relatively simple system to estimate and preserves
important properties of consumer theory. See Deaton and Muellbauer (1980).
44
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
behavior. This is relevant in consideration of the analysis of the full-range of
policy options because agents will view future policies and the continuation of
existing policies as inherently uncertain, yet in EPA's Platform v6 the construct of
perfect foresight means agents react based a known future. When using the
model to anticipate the broadest possible range of future policies, the associated
uncertainty of outcomes is also broad. For example, the response of investors to
a new environmental policy will be shaped by legal or political challenges that
might reverse it, and the public consideration of a policy in the near-term will
affect behavior in the Reference Case even if such a policy is not enacted.
Burtraw, Blumsack, Bushnell, Felder, and Wood
45
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
VI. EPA's PLATFORM V6 DOCUMENTATION
Identify strengths, weaknesses, limitations, and errors in the model documentation and
the Results Viewer. Is the documentation dear and well-written?Propose options as
needed. Specifically, are all the necessary elements included? Are there any extraneous
elements? Could simplifications be made? How well does the Results Viewer effectively
communicate model run results? What additional documentation or model results, if any,
would further improve transparency?
Overall the documentation is well organized and well written. We recognize that
it is a challenge to document complex models in a comprehensive way and keep
the documentation up to date. EPA is commended for its achievements in both
regards. However, there were several aspects of the model that we felt were not
as well described as could be in the documentation. There are many instances
where providing model equations would add clarity, such as previously described
for the objective function and capital charge rates. In addition, there are several
places where we would recommend that EPA provide more complete information
about data sources. We also make some specific suggestions regarding data
displays and text, as described below.
The main areas in the EPA Reference Case v6 documentation that require
additional explanation include the following:
Development of load segments. The process used for developing load
segments as described in the documentation is unclear. However, we found
slide 30 in the briefing that EPA presented to the panel on October 16 to be quite
helpful in understanding the distinction between load slices and times of day and
recommend that this graphic be added to the documentation.
Treatment of interregional trading. The documentation's description of inter-
regional trade, especially related to the load segments, is not very clear. The
documentation indicates that trade is modeled on a seasonal basis, yet it is our
understanding after discussions with EPA that trade is modeled by load segment.
Because the load segments are defined uniquely for each region, a mapping of
segments by hour is performed in order to capture simultaneity between regions.
A description of this process with an accompanying example would help promote
understanding of this model feature.
Aggregation of model plants: The documentation's description of the
aggregation of model plants also requires clarification. Section 4.2.6 of the
documentation describes the aggregation of model plants as occurring for plants
that share several key characteristics and that are located within the same state.
Based on our communication with EPA during the review, it is our understanding
that fossil units are aggregated no further than at the plant level.
Publication of data tables on the EPA website: The use of tables uploaded
directly to the web is understandably necessary given the large size of many of
the data inputs. However, a few improvements are suggested. The first time one
of these appears (Table 2-2), a footnote indicating that large tables are on the
46
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
web and that the list of these can be found at the end of each chapter would be
helpful. This footnote could also provide a link to the site where the tables are
posted. In addition, because some users may go directly from a search engine to
the page with the complete document, it would be useful to have a link from here
to the tables as well.
Guide to EPA's Platform v6 Output Files: It would be helpful to include a
reference in section 2.5.2 of the documentation to the output file guide that is on
EPA's website. In addition, when ElA's AEO cases are used to set up alternative
sensitivity cases, a more complete description of which AEO case is being used
and what inputs are being used from the case would be helpful. For example, in
the AEO2018 there are no cases called High or Low Demand, but rather there
are High and Low Economic Growth scenarios and two alternative efficiency
cases.
Table 1 below includes specific comments and suggestions, organized by
chapter of the EPA's Reference Case v6 documentation.
RESULTS VIEWER
In general, the results viewer is a great tool, as the apparent design and intent for
functionality is very good. However, there are limits to it the tool in its current
form that we would recommend that EPA address, because even small
inconsistencies or incompleteness impart uncertainty about the use of the viewer.
To avoid user confusion, we would recommend that EPA insert a few
clarifications in the READ ME instructions. The first is to define the RPE
acronym at the outset of the readme tab. Second, the instructions include a
reference to the "Profile worksheet" in the discussion on Workbook Navigation,
but there does not seem to be such a sheet with that name in the Results Viewer.
It is unclear if that reference should be changed or if a worksheet is missing from
the file. Third, the description of Fuel Type is confusing because it focuses on
the selection of years for displaying data rather than fuel type.
We found the Results Viewer's distinction between "plant type" and "plant
category" confusing. For example, it is unclear what a user should choose for
nuclear plant type. The readme tab indicates that plant type and plant category
may be merged in the future, and we agree this would be clearer. For example,
the merged list could include "all renewables" as well as solar, wind, etc.
The displayed results should indicate the cases being compared (i.e., the
difference between what to what) and the units of measure reflected in the
results. The readme tab indicates that the results represent "changes from the
comparison model," but it would be helpful to include this on the graphics page
accompanying the map. The names of the two scenarios should also appear in a
text field. This would make the charts more useful since they cannot be edited
once exported.
The use of the "comparison case" for other metrics, such as capacity factors and
emissions rates, is clever but not very intuitive. The comment box that guides the
47
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
user on how to set the base and comparison cases is helpful but would perhaps
be more effective as a full display (i.e. always visible). It is also a bit
cumbersome that true comparisons require the user to specify twice what data to
display (for the primary and comparison cases). Another approach would be to
have a single place where users can select data type for comparison graphs.
Only when other types of graphs are selected, such as capacity factors or
emissions rates, would the user specify a second set of data types.
In the map sheet, the comparison functionality is confusing and only works for
displaying differences, rather than two sets of absolute values. Additional
instructions about the map feature would help users better understand this
functionality.
RETAIL PRICE MODEL DOCUMENTATION
The Retail Price Model (RPM) relies heavily on assumptions from the AEO2018
and in many instances the documentation refers only generically to the AEO2018
without sufficient information about what the assumptions are and how they were
derived from either published AEO2018 outputs or data provided by EIA at EPA's
request. Additional specific comments on the RPM documentation include the
following:
• In the discussion of utility depreciation costs, the units are mills/kWh but
these are not defined by year. In addition, the "directly from" is not
explained sufficiently as to whether the reader can find these in a
published document or table or whether this was provided by EIA.
• The documentation would benefit from additional detail for the non-utility
generators (NUG) adder and the regional tax rates used in the RPM. In
both cases, the reader is referred to the ElA's AEO 2018. We
recommend that the documentation present these values and describe
their source (public or specially requested from EIA).
• Also related to regional tax rates, it is not clear what is included in
"regional tax dollars" referenced in the documentation. Are these
revenues collected only from electricity bills or do they reflect other
sources of revenue? We recommend that EPA define these revenues
more precisely based on input from EIA.
• Attachment 1 of the documentation includes a table showing the
percentage of each region that is deregulated or regulated. We
recommend that EPA describe how the percentages were derived, rather
than simply citing the AEO. The version of the RPM provided to the panel
includes a map of model regions that does not match the regions listed in
the Attachment 1. For clarity, these should be updated to be consistent
with EPA's Platform v6.
48
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
Table 1. Specific Comments and Recommendations Related to the EPA Platform v6 Documentation
Page
Number
Comment/Recommendation
Chapter 1. Introduction
Page 1-4
The technology list in Table 1-2 does not include NGCC with CCS although it is included in Chapter 6.
Perhaps it would be best to include with a footnote that it is disabled in the Reference Case (see
recommendation in the CCS section of the review that it be activated).
Chapter 2. Modeling Framework
Page 2-1
". ..used IPM extensively for various ..."
For some of the IPM applications mentioned, it may have been the case that IPM provided input into
those analyses (e.g., economic impact assessment) as opposed to actually conducting such an
assessment. The distinction between applications that IPM can address entirely vs. ones that it provides
important input to may be worth making.
Page 2-1
"... a globally optimal solution".
It may be possible, although highly unlikely for the LP algorithm in IPM to obtain multiple optimal
solutions.
Page 2-1
. . reasonable solution time for LP model..."
See also p. 2-11, Section 2.4. More information on the time to run EPA's application of IPM and any
formal or informal EPA requirement that it run within a particular amount of time would be helpful.
Page 2-1
"IPM is a dynamic linear program..."
See also p. 2-6, Section 2.3.3. The use of the term "dynamic" should be clarified. It may suggest
something beyond what IPM is doing. IPM can have different assumptions that vary by time, but those
assumptions are not dynamic (i.e., change based upon IPM calculations). A suggested revision would
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
Page
Number
Comment/Recommendation
be "IPM is a linear programming model that generates optimal decisions over the projection time horizon
under the assumption of perfect foresight."
Page 2-2
"IPM provides estimates of air emission changes, ..."
The documentation should clarify what EPA estimates these changes relative to (i.e., relative to different
scenarios, previous year, or both).
Page 2-2
The discussion of integrated resource planning states that IPM can optimize demand-side options but a
search on the documentation does not provide any further information or assumptions. Moreover, the
demand side options that are available in EPA's application of IPM appear limited. For example, it is not
obvious that demand could not respond to time varying prices, and demand could not shift between time
blocks. It is important that the documentation not overstate the capability of the model as it is
implemented by EPA so that users understand what is accomplished in the modeling, and what users
must look elsewhere to address. It may be useful for the documentation to indicate where there are
capabilities of the model that EPA chooses not to exercise but might find useful in future analyses.
Page 2-3
"Many of these costs components are captured in the objective function...."
Why the use of the word "many"? Are not all costs captured as described in the sentence? If not, the
documentation should explain.
Page 2-3
"The applicable discount rates are applied to derive the net present value for the entire planning
horizon...."
Rates is plural, whereas Chapter 10 indicates that a single discount rate is used for intertemporal
decisions. Another sentence indicating the the EPA Reference Case uses a single discount rate would
reduce confusion.
Page 2-3
In the discussion of transmission decision variables, the documentation states "...the total cost of
transmission across each link." The text could be made clearer that this refers to the transmission tariff,
not the capital costs.
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
Page
Number
Comment/Recommendation
Page 2-3
As the paragraph on emission allowance decision variables reads, it sounds as though the formulation is
non-linear with the multiplication of the market price of allowances times the allowance decision
variables. Because this is obviously not the case, we recommend that EPA clarify.
Pages 2-5
and 2-6
Section 2.3.1 (Model Plants) should mention "planned-committed", which is one of the three categories of
generation units that EPA's Platform v6 uses.
Page 2-6
The parsing discussion should provide references as well as additional documentation. The discussion
identifies post-processing parsing tools to translate model plant level data into generating unit-specific
results and for deriving inputs for air quality modeling. Neither references nor a detailed description are
currently provided.
Page 2-6
More clarity should be provided that 2050 is both the final model run year and the last reported year.
The documentation leads one to perhaps presume that the model is run farther to handle end-of-horizon
end effects.
Page 2-7
The sentence of the last bullet in Section 2.3.3 reads,"This permits the model to capture more
accurately...."
This sentence and associated paragraph would benefit from more details and discussion. The reader is
unable to ascertain whether this paragraph is saying that by including costs for every year, not just the
individual model run years, the model is more accurate than if it only included costs in the individual
model run years.
Page 2-11
Should the assumption of perfect regulation, that is regulators can determine the actual costs of the
utilities that they regulate and do not allow for gold plating, be incorporated into this section? IPM seems
to also be making this assumption. Also, uncertainty is not a market imperfection (and applies to
regulated portions of the power system as well as to market-based portions).
Page 2-11
The discussion of hardware and programming features would benefit from several additions/clarifications
including:
(1) "MPS" should be spelled out or defined;
(2) More information on how long it takes to run EPA's application of IPM would be helpful. Also, how
quickly can performance be improved with advances in computational power? This is important in
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
Page
Number
Comment/Recommendation
determining potential enhancements to EPA's Platform v6.
(3) The discussion related to the benchmarking tests performed by EPA's National Environmental
Scientific Computing Center warrants more detail on what the unacceptable results were, why did they
occur, and what are the implications for EPA's Platform v6 current and potential configurations.
Pages 2-12
and 2-13
Section 2.5 is short on detail and simply includes a high-level listing of categories of inputs and outputs.
A reference to the more detailed output file guide would be helpful here. In addition, more detail,
including equations or equivalent details, should be included throughout the documentation.
Page 2-13
The header "List of tables that are uploaded ..." might be modified to include the Chapter number. For
Chapter 2, this would be especially helpful because this is the first reference to these tables. Another
option would be to make this a subsection of its own in each chapter.
Chapter 3. Power System Operational Assumptions
Page 3-3
The text and Table 3-1 column heading cite different AEO vintages. Although it does not matter since the
regional definitions are the same, it might be confusing to some readers.
Page 3-7
A more complete description is needed of how demand elasticities may be applied in EPA's application
of IPM. For example, are the elasticities applied for annual demands or by load segment? Also, the
documentation would benefit from discussion of how electricity prices are scaled up to approximate retail
prices.
Page 3-14
In the section about minimum capacity factors for oil/gas plants, it might be clearer in step 3 to insert the
phrase "annual historical average" to describe the capacity factor threshold for removal. It should also
specify that the minimum capacity factors are applied to units as annual averages rather than by
load/time segment.
Page 3-14
Incorporate more detail about how the coal turndown constraints were developed from the recent hourly
Air Markets Program Data (AMPD) data. For example, how many years of data were used and what was
the criteria for setting the minimums (single hour, multiple hour averages, etc.)?
Page 3-15
In Table 3-9, include a citation for the planning reserve margins listed.
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
Page
Number
Comment/Recommendation
Page 3-26
The treatment of 316(b) costs in EPA's Reference Case is unclear. The documentation states that "EPA
Platform v6 includes cost of complying with this rule" and points to another document where the cost
assumptions and analysis for 316(b) can be found but does not provide any information about how these
costs are incorporated. Are these investment (sunk) costs, operating costs (fixed or variable)? Do they
impact plant operations (heat rates or other performance characteristics)?
Chapter 4. Generation Technologies
Throughout
Many assumptions and associated calculations and algorithms to calculate those assumptions are not
documented. For instance, the capacity parsing algorithm (p. 4-4), the level of aggregation of generation
units (p. 4-6), coal switching (p. 4-6), how availability accounts for planned and unplanned maintenance
(p. 4-21), the basis for the upper bound on new power plants in Table 4-14 (discussion on p. 4-22), and
existing nuclear unit assumptions (Section 4.5.1, p. 4-47). All assumptions should be precisely and
specifically documented (as opposed to having incomplete or only high-level names of references)
including providing references for all tables.
Page 4-2
In Table 4-2, the assumptions regarding two new nuclear units to come online may need to be updated.
Page 4-5
In Section 4.2.4, a reference to the U.S. Nuclear Regulatory Commission 80-year life should be
incorporated into the documentation
(https://www.nrc.gov/reactors/operating/licensing/renewal/subsequent-license-renewal.html).
Page 4-5
The last sentence of Section 4.2.4 reads: "The unit, however, continues to make annualized capital cost
payment on any previously incurred capital cost for model-installed retrofits projected prior to retirement."
We recommend that the documentation state if, and how, regulated vs. competitive retired generation
units are treated differently with respect to future annualized capital costs payments.
Page 4-7
and 4-20
The type of energy storage should be specified in Table 4-7 (p. 4-7) and Table 4-12 (p. 4-20).
Page 4-7
Add a definition clarifying what IMPORT represents in Table 4-7.
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
Page
Number
Comment/Recommendation
Page 4-11
The discussion of the data sources for gas-turbine based prime movers references ICF's experience and
expertise related to O&M costs. We recommend that documentation and clarification of the referenced
ICF expertise and experience be provided.
Pages 4-11
and 4-12
Table 4-8 provides ranges of variable O&M costs for some generation technologies. We recommend
that the documentation clarify how how ranges of assumptions are implemented (e.g., use of median
point estimate).
Page 4-18
The basis and reference for the lifespan without life extension expenditures assumptions in Table 4-10
are not provided. In particular, the lifespan of combustion turbine of 30 years may be too long. See
Newell et al. (2014), which assumes a 20-year economic life for combustion turbines.
Page 4-18
The Non-conventional and Conventional labels in Table 4-11 may be unnecessary. They could be
modified to Renewable/Storage and Fossil/Nuclear.
Page 4-19
Table 4-12 should clarify that the capacity information presented represents Summer Capacity (MW).
Page 4-22
In the discussion of regional cost adjustments, the documentation should define what is meant by the
term "ambient conditions." Also, this discussion indicates that regional cost multipliers from the
University of Texas are applied. The documentation is unclear why those are used instead of data from
EIA.
Page 4-29
The values in the second half of Table 4-16 (vintage #4 and later) appear to be shifted over a column
(i.e. the PV values are in the fuel cell column etc.).
Page 4-30
A more complete definition of wind techno-resource groups (TRGs) would be useful for readers who are
not familiar with the National Renewable Energy Laboratory (NREL) Annual Technology Baseline (ATB).
Page 4-30
Assumptions of wind potential by resource and cost class is shown Table 4-17 before the concept of cost
classes is introduced and defined. Adding one sentence about them at the end of the paragraph before
the table would be helpful.
Pages 4-34
to 4-36
The citations for wind resource, capacity factors and generation profiles from NREL should be more
specific and summarize the methodology used by NREL for their derivation.
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
Page
Number
Comment/Recommendation
Page 4-34
The documentation needs to make clear how solar and wind capacity factors are adjusted, if at all, for
planned and unplanned maintenance. As written, the documentation suggests that no adjustment is
made because the capacity factor is multiplied by the installed capacity to obtain the amount of energy
produced for a given season.
Pages 4-
35, 36, and
41
The ranges of wind and solar PV reserve margin contributions in Tables 4-21,4-23,4-25, 4-27 and 4-32
are too large to be meaningful. It would be more useful to show the initial year values by region and
resource class with an indication of how they may change by 2050 in the Reference case. An alternative
would be to show the stacked sequence of cumulative capacity and reserve margin contributions by
region.
Page 4-36
The discussion of wind tax credits does not indicate why EPA chose to model the credit as a reduction in
capital cost (investment tax credit) rather than a production tax credit (PTC). The wording suggests that
this is done as a modeling convenience rather than due to an assumption that wind generators will select
an ITC rather than a PTC, but it is not clear.
Page 4-40
The wind calculation example refers to Table 4-20b as the source of the reserve margin contribution, but
there is no such table.
Page 4-45
The acronym IDC in Table 4-35 should be defined.
Page 4-47
Section 4.5 includes a long discussion of capacity factors by vintage for nuclear power plants, but most
of the existing capacity was built before 1982 and is older than 25 years so have constant capacity
factors anyway. In addition, it is not clear from the Reference Case outputs whether the new planned
units exhibit a lower capacity factor at start.
Chapter 8. Development of Natural Gas Supply Curves for EPA Platform v6
Pages 8-10
to 8-12
In section 8.3.4, more information is needed about LNG exports. It is not clear to what degree these are
an exogenous assumption vs. model outcome. Particularly ambiguous is the phrase "ICF assumes."
Does this mean an ICF assumption that goes into the Gas Market Model (GMM) or an outcome of GMM?
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
Page
Number
Comment/Recommendation
Page 8-17
The discussion on this page outlines the four main drivers of natural gas demand. This seems like it
would be better placed before the discussion of gas demand projections - perhaps at the start of Section
8.5.
Burtraw, Blumsack, Bushnell, Felder, and Wood
56
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
REFERENCES
Bent, R., S. Blumsack, P. Van Hentenryck, C. Borraz-Sanchez and M.
Shahriari. 2018. "Joint Electricity and Natural Gas Transmission
Planning With Endogenous Market Feedbacks." IEEE Transactions
on Power Systems, 33(6): 6397-6409.
Borenstein, Severin and James Bushnell. 2015. "The US electricity
industry after 20 years of restructuring." Annu. Rev. Econ. 7(1): 437-
463.
Brandt, A. R., Heath, G. A., & Cooley, D. 2016. "Methane leaks from
natural gas systems follow extreme distributions." Environmental
Science & Technology. 50(22): 12512-12520.
Brealy, Richard, Stewart Myers, and Franklin Allen. 2011. Principles of
Corporate Finance (10th Ed.), McGraw Hill, New York NY.
Burtraw, Dallas, Karen Palmer, and Danny Kahn. 2010. "A symmetric
safety valve." Energy Policy. 38:4921-4932.
Chandramowli, Shankar N. and Frank A. Felder. 2014. "Impact of
climate change on electricity systems and markets-a review of
models and forecasts." Sustainable Energy Technologies and
Assessments. 5: 62-74.
Couzo, Evan, James McCann, William Vizuete, J. Jason West, and Seth
Blumsack. 2016. "Modeled Response of Ozone to Electricity
Generation Emissions in the Northeastern United States Using
Three Sensitivity Techniques." Journal of the Air and Waste
Management Association. 66(5): 456-469.
Deaton, Angus and John Muellbauer. 1980. "An Almost Ideal Demand
System." The American Economic Review. 70(3): 312-326.
Echeverri, Dalia Patino, Dallas Burtraw, and Karen Palmer. 2013.
"Flexible Mandates for Investment in New Technology." Journal of
Regulatory Economics. 44 (2): 121-155.
Fabrizio, Kira R., Nancy L. Rose, and Catherine D. Wolfram. 2007. "Do
markets reduce costs? Assessing the impact of regulatory
restructuring on US electric generation efficiency." American
Economic Review. 97(4): 1250-1277.
Hamada, Robert. 1972. "The effect of the firm's capital structure on the
systematic risk of common stocks." The Journal of Finance. 27(2):
435-452.
ICF International. 2014. Economic Analysis of Methane Emission
Reduction Opportunities in the U. S. Onshore Oil and Natural Gas
Industries. Available at:
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
.
Jaramillo, P., Griffin, W. M., & Matthews, H. S. 2007. "Comparative life-
cycle air emissions of coal, domestic natural gas, LNG, and SNG for
electricity generation." Environmental Science & Technology.
41(17): 6290-6296.
Linn, Joshua, Erin Mastrangelo, and Dallas Burtraw. 2014. "Regulating
Greenhouse Gases from Coal Power Plants under the Clean Air
Act." Journal of the Association of Environmental and Resource
Economists. 1(1):97-134.
Newell, Samuel A., J. Michael Hagerty, Kathleen Spees, Johannes P.
Pfeifenberger, Quincy Liao, Christopher D. Ungate, and John
Wroble. 2014. Cost of New Entry Estimates for Combustion
Turbine and Combined Cyucle Plants in PJM. Prepared for PJM
Interconnection, LLC. Available at:
.
Organisation for Economic Cooperation and Development, Nuclear
Energy Agency. 2011. Technical and Economic Aspects of Load
Following with Nuclear Power Plants. Available at:
.
Osofsky, Hari M., Zachary R. Barkley, Seth Blumsack, Kenneth J. Davis,
Lara B., Fowler, Thomas Lauvaux, Ross H. Pifer, Ekrem Korkut &
Chloe Marie. 2018. Regulation of Methane Emissions from
Unconventional Oil and Gas: Current Approaches and Possibilities
for Innovation Based on Emerging Science. Penn State University
Center for Energy Law and Policy. Available at:
.
PJM. 2019. Inside Lines, December 30. Available at:
.
Rosendahl, Knut Einar and Halvor Briseid Stomasten. 2011. "Emissions
Trading with Updated Allocation: Effects on Entry/Exit and
Distribution," Environ Resource Econ. 49: 243-261.
Sargent & Lundy. 2017. "IPM Model - Updates to Cost and
Performance for APC Technologies - CO2 Reduction Cost
Development Methodology." Project 13527-001.
Stauffer, Hoff. 2006. "Beware Capital Charge Rates." Electricity Journal
19(3): 81-86.
58
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
-------
Peer Review of U.S. EPA's Power Sector Modeling Platform v6
U.S. Department of Energy (DOE). 2015. Quadrennial Technology
Review: An Assessment of Energy Technologies and Research
Opportunities.
U.S. Energy Information Administration (EIA), U.S. 2018. Battery
Storage Market Trends, May 21. Available at:
.
Walker W.E., Lempert R.J., Kwakkel J.H. 2013. "Deep Uncertainty". In:
Gass S.I., Fu M.C. (eds) Encyclopedia of Operations Research and
Management Science. Springer, Boston, MA
Burtraw, Blumsack, Bushnell, Felder, and Wood
March 2020
59
------- |