EPA-600/2-75-045
September 1975
Environmental Protection Technology Series
A METHOD FOR
EVALUATING S02 ABATEMENT STRATEGIES
Industrial Environmental Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, N.C. 27711
-------
EPA-600/2-75-045
A METHOD
FOR EVALUATING
SO2 ABATEMENT STRATEGIES
by
C.T. Chi, E.G. Eimutis, W. H. Hedley,
M. V. Jones, R. Jones, and L. B. Mote
Monsanto Research Corporation
1515 Nicholas Road
Dayton, Ohio 45407
Contract No. 68-02-1320, Task 11
ROAPNo. 21ADE-008
Program Element No. 1AB013
EPA Project Officer: Gary L. Johnson
Industrial Environmental Research Laboratory
Office of Energy, Minerals, and Industry
Research Triangle Park, NC 27711
Prepared for
U.S. ENVIRONMENTAL PROTECTION AGENCY
Office of Research and Development
Washington, DC 20460
September 1975
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development,
U.S. Environmental Protection Agency, have been grouped into
five series. These five broad categories were established to
facilitate further development and application of environmental
technology. Elimination of traditional grouping was consciously
planned to foster technology transfer and a maximum interface in
related fields. The five series are:
1. Environmental Health Effects Research
2. Environmental Protection Technology
3. Ecological Research
4. Environmental Monitoring
5. Socioeconomic Environmental Studies
This report has been assigned to the ENVIRONMENTAL PROTECTION
TECHNOLOGY STUDIES series. This series describes research
performed to develop and demonstrate instrumentation, equipment
and methodology to repair or prevent environmental degradation from
point and non-point sources of pollution. This work provides the
new or improved technology required for the control and treatment
of pollution sources to meet environmental quality standards.
EPA REVIEW NOTICE
This report has been reviewed by the U.S. Environmental Protection Agency,
and approved for publication. Approval does not signify that the contents neces-
sarily reflect the views and policies of the Agency, nor does mention of trade
names or commercial products constitute endorsement or recommendation
for use.
This document is available to the public through the National
Technical Information Service, Springfield, Virginia 22161.
ii
-------
ABSTRACT
A non-deterministic, probabilistic methodology has been
developed as an aid to the EPA decision maker when evaluating
sulfur oxides abatement alternatives. This methodology is
not restricted to gas scrubbing models and can handle clean
fuel substitution processes. A Monte Carlo Simulation
approach is described which can treat fully developed cost
models and also those which encompass a high level of uncer-
tainty either in structure or process parameters. Impact
Assessment techniques are presented which develop the
methodology of alternative future scenarios. A specific
example, the Wellman-Allied Process, is used to demonstrate
the utility of this simulation methodology.
iii
-------
CONTENTS
Sections Page
I Summary 1
II Introduction 5
A. Previous Work 5
III Probabilistic Approach 11
IV Simulation Methodology 15
V Decision Analytic Approach 17
A. The Deterministic Phase 20
1. Description of Alternatives 21
2. Description of Process and Cost 21
Models
VI Simulation Analysis 2?
VII Impact Assessment Methodology 39
A. State Variable Analysis 39
B. Discussion 5^
VIII Efforts Required for a Realistic Analysis 57
IX Example of Methodology 6l
X References 77
v
-------
FIGURES
Figure Page
1 The SO2 Systems Analysis Program in 18
a Decision Analysis Framework
2 Relationship Between Factors Included in 23
the Total Capital Investment
3 Relationships Between Factors Included in 24
the Total Operating Costs
4 Sample Log-Normal Probability Distribution 28
5 Sample Normal Probability Distribution 29
6 Normal Cumulative Density Distribution 30
7 Log-Normal Cumulative Density Distribution 31
8 Example of Using Cumulative Density 32
Function
9 Sample Frequency Histogram 34
10 A Comparison of the Normal and Weibull 37
Distributions
11 The Price of Coal in 1984: A Projected 45
Index with Explanatory Factors
12 Consumer Price Index 46
13 Regional Differences in the Costs of Coal 52
for Steam Electric Power Plants
14 Underground Mining Productivity 53
15 Wellman-Allied Process Flowsheet 62
16 General Procedure for Calculation of 65
Common Cost Factor for Add-On Scrubbing
Systems
17 Calculation of Annual Production Cost of 67
Wellman-Allied Process
vi
-------
TABLES
Table Page
1 Illustrative State Variables Applying 41
to the Choice of SO Control Technologies
X
2 Increases in the Price of Fuels Compared 47
General Price Increases
3 Employment in Mining 48
4 Employment in Coal Mining Compared to All 49
Types of Non-Agricultural Employment
5 Typical Costs of Reclaiming Land That 50
Has Been Strip-Mined
6 Average Per Ton Costs of Transporting 51
Bituminous Coal to Market
7 Projected Increases in Labor Productivity 53
8 An Index Measuring Regional Differences 56
in the Increase in Union Wage Rates in
the Construction Field in the 1969-74
Period
9 Input Variables, Wellman-Allied Process 69
10 Simulation Results - Utility Boilers, 71
Wellman-Allied Process (1974 Prices)
vii
-------
ACKNOWLEDGEMENTS
Two of the authors, M. V. Jones and R. Jones, are
associates of the Impact Assessment Institute,
Bethesda, Maryland, who worked on this project as
consultants to Monsanto Research Corporation. We
also wish to acknowledge the helpful advice of EPA
personnel, Charles J. Chatlynne, Gary J. Poley,
John 0. Smith, and W. Gene Tucker.
viii
-------
SECTION I
SUMMARY
We have developed a methodology for evaluating pollution
abatement and clean fuel alternatives in a well-traceable9
systematic fashion. Employing a combination of classical
engineering economics and process analysis together with
the disciplines of decision analysis and impact assessment,
we present a practical, flexible technique as a utilitarian
aid to the decision maker.
Several novel approaches were used to overcome previous
difficulties encountered when using a deterministic
approach. When dealing with future projections on any
variable, we are faced with uncertainty. To deal with
the many complexities inherent in this situation, we first
applied the impact assessment approach - a method that
presents, in a readily traceable fashion, techniques
useful in developing future scenarios.
Having developed a rationale for treating the future,
we then developed quantitative methods for treating
uncertainty. These techniques include the use of statis-
tical inputs and the use of simulation to produce statis-
tical results.
-------
Comparison of clean fuel alternatives with flue gas
scrubbing has presented conceptual as well as computational
problems. An innovative approach was used to allow a
common basis o.f comparison. This required the inclusion
of the cost of coal and the transportation distance from
the mine mouth to.the using site.
In addition, we have shown how this methodology can take
into account the uncertainties in any step of an ill-
defined (because of the lack of experimental data) process,
both in the cost of equipment and in the operating cost,
and to allow for projections to a fully commercialized
process.
The job of allocating R&D funds for S0x pollution abatement
in an optimum manner, so as to minimize the future total
cost for SO abatement in the U.S., is made difficult by
A
the rapid changes in energy availability, national
regulations, and economic growth. This versatile method-
ology was designed to assist the Control Systems Laboratory
in allocating these funds in light of these changes.
This system can, over a period of time, answer an in-
creasingly widening series of questions such as:
1. At what level of sludge disposal costs do
regenerable processes become preferable?
2. How will fuel availability and cost affect the
allocation process?
3. How will the future growth of utilities and
industries affect the relative merits of candi-
r
date alternatives?
-------
4. How should low sulfur coal be compared to flue
gas cleaning and how do we model the attitudes
of utility and industry representatives?
5. How do transportation costs affect the relative
merits of the alternatives?
This program and resulting models are modular and quick
in response. Because of their flexible nature the
programs will be able to treat unexpected rapid develop-
ments in a very orderly framework and a convenient well-
defined manner.
-------
SECTION II
INTRODUCTION
A. PREVIOUS WORK
The objective of this work VMS to develop a methodology
for evaluating alternatives for sulfur oxides abatement.
This methodology has been developed, and should be
helpful in assessing the economics of various competitive
methods. A great deal of valuable information was
gathered by M. W. Kellogg personnel (EPA Contract 68-
02-1303)1 both on emission sources and on cost models for
four specific processes for SO abatement. A problem
a
arose, however, when attempts were made to use the
equations to calculate the costs of control for large
groups of SO sources, such as all coal-fired power
.A.
plants or all coal-fired industrial boilers in the U.S.
Further problems arise when one considers the future.
It is probable that more power plants will be needed
to fulfill the ever-rising power demand in the U.S.
It is hard to be specific, however, about where plants
will be built in the year 1980; for example, how much
will the demand increase in southern Ohio? Will the
plants that are needed to supply this additional
capacity be built in southern Ohio, or will they be
built in Kentucky and the power transmitted to Ohio
over high tension, long distance transmission wires?
5
-------
What will be the availability of sludge disposal areas
near these plants, whose number, size, and location
are at present unknown?
The amount of specific information needed to apply this
"deterministic" approach and the impossibility of
obtaining this information for future plants make this
approach almost impossible to apply even to one small
region, much less, the whole U.S. Even if some method
could be found to obtain the necessary information, the
volume of data needed for each specific power plant
multiplied by the total number of coal-fired power
plants would require extensive data processing facilities
and complex data management procedures.
As inputs to this study, attempts were made to define
the future demand for electric power in the U.S. and
future expected development costs and operating costs
for a number of SO abatement processes by means of the
A
Delphi technique. Panels of experts were convened to
comment on detailed questionnaires which had been filled
out by these consultants. A large amount of information
was compiled from these questionnaires and meetings;
however, the difficulties in applying this information
to a plant-by-plant calculation still remained.
Pacing the uncertainties as to how to further proceed,
Monsanto Research Corporation (MRC) was assigned the
task (under EPA Contract 68-02-1320, Task 5) of developing
an approach which could be used to evaluate alternative
costs for SO abatement. During this work a new approach
A
was generated, in which plants would be considered in
-------
large groups rather than individually. Since it is
known that all plants have different cost factors, the
differences in wage rates at different plants were taken
into account by using a distribution of wage rates rather
than one specific value. Difficult calculations, such
as costs to haul sludge from the limestone scrubbing
process to available landfill sites, are avoided by
summarizing sludge disposal costs as a distribution
function. As long as the distribution function represents
the costs of sludge disposal reasonably well over the
average of many plants, manipulation of these data by
simulation techniques leads to reasonable overall
estimates.
There is a large number of variables that are used to
calculate operating costs for SO abatement processes.
A.
Some of these variables have a much larger effect on the
operating costs than others. We performed sensitivity
analyses on the input variables to determine which ones
had the most significant effect. Variables, such as
the percent of ash in coal, which would not be expected
to have a large effect on the calculated results, were
entered as specific values. Variables which would
significantly affect the costs were treated as distribu-
tion functions.
Having developed this new approach to this problem,
MRC was then asked to develop this concept into a prac-
tical application for evaluating national SO abatement
j\,
alternatives. Drawing heaviliy upon previous work by
M. W. Kellogg, MRC personnel assembled cost models for
the Limestone scrubbing, Wellman-Allied, Lurgi Synthetic
-------
Natural Gas, and Solvent Refined Coal processes into
several computer programs. Input data on sizes of power
plants, and their load factors were also drawn from
previous work. Using these equations we performed
sensitivity analyses on the variables and determined
which one needed to be used as a distribution function.
Distribution functions have been compiled for key vari-
ables. Based upon numerous specific assumed values for
variables, costs for national SO abatement using these
Jv
four specific processes were calculated as examples of
how this method might be used in the future.
The sample results calculated using the methodology
developed in this work are strongly a function of assumed
input variables. This should prove helpful since it
means the possible economic effects of making changes
in numerous variables can be quickly investigated. This
also means, however, that one should beware of merely
abstracting values from this work and using them in an
unquestioning fashion as being the actual costs which
will prevail for a specific process in 1980, since many of
the values were for test purposes only and may not be
correct. One must verify that the values for input variables
are in accord with best estimates currently available before
using the reported values. This work was done to establish
the method, not to produce guaranteed cost figures.
In this work, the probabilistic method for analyzing
future SO abatement costs has been developed. Equations
X
for specific processes have been combined with values of
input variables for coal-fired electric utility boilers,
8
-------
industrial boilers, sulfuric acid plants, and sulfur
recovery plants. Classes of SO sources and annual SO
x x
abatement costs for these plants have been simulated.
This system should be useful in testing the effect of
different values of input variables on expected SO
J\.
abatement costs. It should also be useful in helping
to determine what effect process improvements or develop-
ment of a new SO abatement process would have on future
A
expected costs.
-------
SECTION III
PROBABILISTIC APPROACH
One approach that would answer questions on R&D fund
allocation would be to assemble a set of equations
that could be solved to give one unique answer for a
set of basic assumptions. By assuming rational, best
estimates on variables such as fuel cost, labor cost,
by-product cost, transportation cost and interest rates,
one would then be able to calculate a unique value of
national cost for SO abatement for the processes avail-
A
able. A model of this type is called deterministic
because it enables a specific, unique solution to be
determined.
A significant problem arises, however, in using this
type of approach. Because the future is always uncertain,
the appropriate values to use for given variables are
unknown. What will be the price of coal in 1980? What
interest rates will prevail at the time someone wishes
to build a new plant for SO abatement or the production
A
of clean fuel? We know these costs will change with
time.
The deterministic approach when applied to processes
strongly influenced by future variables, as is the case
of R&D allocation, encounters severe uncertainties. To
intelligently assess events in 1980, a range of values
11
-------
for futuristic variables such as the price of coal
must be considered. A minimum of two values would
have to be selected for each deterministic future
variable in order to determine its effects and inter-
actions with other variables. With a deterministic
model containing 20 variables at 2 values each, we
would calculate 220 or over 1 million results for
national SO abatement costs.
.A.
These results would be difficult to analyze, and the
relative probability that any one of these values was
better than another would be unspecified. Hence, a
selection of the most probably value of these 1 million
results would not be possible. We present an alternative
approach that avoids this situation.
If we describe the inputs as statistical distributions
then by performing a probabilistic simulation, any
desired output variable can be plotted to form on
probability distribution curve. This curve may consist
of cost/unit energy for a given alternative on the abscissa
and the probabilities from 0 to 1.0 on the ordinate.
At the 0.5 probability point on the curve, a unique
value of cost can be read. This procedure also has
another advantage - it describes the range of uncer-
tainty on the obtained results.
The approach of assigning probability values to
futuristic variables rather than giving them predeter-
mined values will be referred to as the probabilistic
approach. It has the advantage of reducing the complexity
12
-------
of the problem to manageable bounds and it also defines
the uncertainty in the values calculated. This method
is well founded in the science of decision analysis and
probabilistic simulation.
Our final thought on the probabilistic approach is best
expressed by quoting from a letter dated July 26, 1967,
written by Robert E. Pfenning, Vice President and Comp-
troller of the General Electric Company. The letter
was addressed to all of General Electric's department
general managers and managers of finance, and it marked
an unprecedented change in policy. It read in part as
follows: "... Conventional budgets and forecasts
present single value data as expressions of business
plans, yet the very nature of the planning process
involves conditions of uncertainty which are better
communicated using probabilities, ranges of values, and
expected values..."
13
-------
SECTION IV
SIMULATION METHODOLOGY
The deterministic approach is straight-forward, using
single valued information and computing a single result.
When statistical distributions are used as inputs,
statistical mathematics must be applied. Many of these
mathematical approaches are extremely complex even when
dealing with the simple Gaussian or Normal distribution.
When the inputs are not Normally distributed there are
no known analytical techniques that can be applied. We
must turn then to a process called probabilistic simu-
lation or "Monte Carlo analysis."
The expression "Monte Carlo analysis" was coined by two
mathematicians, John von Neumann and S. Ulam, in the
late 1940Ts to describe a mathematical technique. They
developed this technique while working on problems
concerning the shielding of nuclear devices which they
found to be too complicated to treat analytically and
too expensive and risky to solve by physical experimen-
tation. The name they chose seems appropriate enough
since the basic principle is the same as one finds in
the operation of casinos at Monaco. Devices are used
there which produce random samples from well defined
populations.
-------
In the present case, however, the device that we use
is the electronic digital computer. The Monte Carlo
method really refers to a collection of techniques
invented to solve a number of different types of
problems. In this collection can be found a common
idea which serves to bind the techniques together.
This is the notion of approximating the solution to
a problem by sampling from various processes. Thus,
instead of using one value for the amount of sulfur
in coal, we use a distribution of values for sulfur
in coal. Or, as in the case of the location cost
factor, we have a distribution of cost factors for
various cities in the United States. The same holds
true for other critical variables, such as the price
of coal. This is not an experimentally derived
distribution but one that is projected on the basis
impact analytic techniques and encoding of uncertainty
methodology.2
16
-------
SECTION V
DECISION ANALYTIC APPROACH
The questions that might be asked will mainly deal
with future events. Because of this, a degree of
uncertainty is inherent. Because of this very impor-
tant fact, we have chosen elements from a methodology
that is ideally suited for cases where probabilistic
statements will be the rule rather than the exception.
This methodology is decision analysis. Decision analy-
sis can be described as a merger of the two fields of
decision theory and systems analysis. Decision theory
provides the philosophy for logical behavior in decision
situations constrained by uncertainty. Systems analysis
as used here represents systems and modeling methodology
that captures the interactions and dynamics of complex
problems. Combining the two results in a theory and
methodology that allows analysis of complex, dynamic,
and uncertain decision situations.2 We will describe
how decision analysis can be used to attack this
particular decision problem.
The decision analysis cycle can be summarized as
follows (see Figure 1). It is made up of three dis-
tinct phases:3
17
-------
Prior
Information
oo
Deterministic Phase
Bound Decision
Identify Alternatives
Establish Outcomes
Select Variables
Create Structural
Model
Create Value Model
Measure Sensitivity
-To Variables
-To Alternatives
Probabilistic Phase
Encode Uncertainty
on Critical Variables
Develop Cost Distributions
for Alternatives
New Information
Design and Execute
Information Gathering
Program
Informational Phase
Measure Economic
Sensitivity
(Determine Value of
Eliminating Uncertainty
in Crucial Variables)
Explore Feasibility
of Information Gathering
Budget
Gather New Information
Figure 1. The S02 systems analysis program in a decision analysis framework
-------
(1) deterministic
(2) probabilistic
(3) informational
The deterministic phase is concerned with the basic
structuring of the problem. The structuring entails
defining relevant variables, characterizing their
relationships in formal models, and assigning values
to possible outcomes. The importance of the different
variables is measured through sensitivity analysis.
In the probabilistic phase uncertainty is explicitly
incorporated by assigning probability distributions
to the critical variables. These distributions are
transformed in the models that exhibit the uncertainty
in the final outcome, which again is represented by a
probability distribution. We can then quantitatively
rank the alternatives in the face of uncertainty.
The informational phase determines the economic value
of information by calculating the worth of reducing
uncertainty in each of the important variables in the
problem. The value of additional information can then
be compared with the cost of obtaining it.
If the gathering of information is profitable, the three
phases are repeated again. The analysis is completed
when further analysis or information gathering is no
longer profitable.
19
-------
A. THE DETERMINISTIC PHASE
The basic steps of the deterministic phase are as
follows:
(1) Bound the decision
(2) Identify the alternatives
(3) Establish the outcome
(4) Select the decision variables and the state
variables
(5) Create a structural model
(6) Create a value model
(7) Measure the sensitivity to identify the
crucial state variables
(8) Measure the sensitivity to the elimination
of alternatives
In this project the first step was to define and bound
the decision problem. In this particular case, the
various abatement technologies (i.e., whether to use
clean fuels as an alternative, etc.), were identified.
The basic decision problem was thus how to fund various
abatement or clean fuel technologies in order to mini-
mize S02 levels at minimum cost to industry.
The next step was to identify outcomes that would be
sufficient to describe the results of various alterna-
tives. This was total annual cost and mills/kwh. In
relating outcomes to the alternatives we defined the
factors that were relevant to the decision. These
factors were separated into decision variables (factors
that can be controlled) and state variables (factors
20
-------
that cannot be controlled). The decision variables
had been identified by Kellogg as consisting of various
process parameters, coefficients, flow rates for certain
processes, energy requirements for certain processes
for a certain size plant, etc.
The state variables, which cannot be controlled but for
which a distribution of values can be predicted, were
energy demand, industrial growth, utility growth, price
of coal, cost of labor, etc.
1. Description of Alternatives
The Kellogg Company has prepared process and cost
models for the following specific processes for four
SO control technology groups:
X
(1) Throwaway flue gas cleaning process: the
wet limestone scrubbing process
(2) Regenerable flue gas cleaning process: the
Wellman-Lord process
(3) High-Btu gasification process: production
of pipeline quality synthetic natural gas
by Lurgi coal gasification process
(4) Chemical coal cleaning process: the solvent-
refined coal process
2. Description of Process and Cost Models
Kellogg's process and cost models were prepared by
relating the key process variables with equipment
sizes, utility consumption (cooling water, process
21
-------
water, power, natural gas, and steam), raw material
requirements, and by-product, if any. For example,
the Lurgi coal gasification process (for pipeline-
quality synthetic natural gas) contains two key inde-
pendent variables: 1) the substitute natural gas
production rate, and 2) the oxygen content of the dry,
ash-free coal. Two minor independent variables are
the plant stream factor and the sulfur content of the
coal. The process and cost model for the Lurgi high
Btu process were based upon these process variables.
The Kellogg cost models estimate the total capital
investment and the annual operating costs. The total
capital investment is a function of:
(1) key process variables
(2) the retrofit difficulty factor
(3) the location
The relationship between factors included in the total
capital investment and the total operating costs are
shown in Figures 2 and 39 respectively.
In Kellogg's cost model, fixed costs (Figure 3) are
based on 18.25% of the total capital investment (TCI)
for 15 years plant life and 16.75$ of TCI for 20 years
plant life. The TCI includes cost of site, working
capital, startup costs, and interest on construction
capital (Figure 2). The fixed costs are not a function
of these expenses. Therefore, the accuracy of the annual
22
-------
Major Equipment Costs
Other Material Costs
Direct Field Construction Labor Costs
Fringe Benefits & Payroll Burden
Field Administration, Supervision
Construction Equipment & Tools
Engineering
Contingency
Direct Plant
Construction Costs
Indirect Costs
Cost of Site
Working Capital
1
Fixed Capital
Investment IFCI)
»«»<»" ConSoXta,
1
Total Capital Investment
ITCI)
Figure 2. Relationship between factors included in
the total capital investment
23
-------
ro
-Cr
- Operating Labor & Supervision
- Maintenance Labor & Materials
Plant Supplies
Raw Materials
Utilities
Waste Disposal
Direct Costs
Payroll Overheads
L- Plant Overheads
Indirect Costs
Total Operating Costs-
Depreciation
Interim Replacements
Federal & Local Taxes
Insurance
Cost of Money
Fixed Costs
Figure 3. Relationships between factors included in the total operating costs
-------
operating cost estimate would increase if the fixed
costs were based on a certain percentage of the fixed
capital investment (FCI) rather than the TCI.
The MRC cost models identify the most sensitive para-
meters of cost, including process variables and non-
process variables. Process variables such as plant
size, flue gas volume, etc., are derived from the
process model. Non-process variables consist of items
such as capital charge rate, load factor, utility costs,
credits for by-products, etc. The most sensitive
parameters were identified by varying process and
non-process variables over a range of values. The
effect of each parameter on both plant investment and
operating costs was then determined.
-------
SECTION VI
SIMULATION ANALYSIS
As was discussed in the introduction, a simulation
approach was used, since the inputs were not normally
distributed and since we are dealing with uncertainty.
As an example, the distribution of sizes for coal-
fired electric utilities is log-normally distributed.
A sample log-normal distribution is shown in Figure 4.
Figure 5 shows what would be an example of a normal
distribution. These are true frequency distributions.
The frequency distribution cannot be used for sampling
purposes. We used instead the cumulative density func-
tion (or distribution), which is an integral of a
function such as shown in Figure 6. The cumulative
density, functions of both the normal and the log-normal
distributions are shown in Figures 6 and 7, respectively
An example of the way a cumulative distribution is used
is illustrated in Figure 8. The dashed line going from
the 0.60 point on the ordinate to the curve selects the
value for that particular level and indicates that 60%
of the values for x will be less than 3-5.
During a simulation, several distributions such as that
shown in Figure 8 were sampled and a range of values
was computed. To perform this sampling, a digital com-
puter which generates uniform random numbers between
27
-------
1.0
0.8
&
CD
13
CT
CD
0.6
0.4
0.2
0 0.5 1.0 1.5 2.0 2.5 3.0
Figure 4. Sample log-normal probability distribution
28
-------
0.5
0.4
0.3
o
c
CD
13
ar
0.2
O.I
o
-3-2-1 0 I 23
Figure 5. Sample normal probability distribution
(Units in standard deviations from mean)
29
-------
1.0
0.8
o
c.
o>
ZJ
*-
JO
Z3
E
13
O
0.6
0.4
0.2
0
-3 -2-101
Figure 6. Normal cumulative density distribution
30
-------
1.0
0.8
o
c
o>
13
r 0.6
0)
E
3
O
0.4
0.2
0
Figure 7. Log-normal cumulative density distribution
31
-------
1.0
0.8
o>
0>
0.6
5 0.4
_to
ZJ
E
o
0.2
0
Figure 8. Example of using cumulative density function
32
-------
0 and 1 was used. Essentially, the process is the same
as selecting a random number, finding its intersection
with the cumulative distribution curve, and determining
a value for the variable of interest. Using that value
throughout the calculation, a result is obtained and
saved. The process is repeated several thousand times,
and then the calculated values are used to generate a
frequency histogram such as that shown in Figure 9
If the input distributions are normally distributed,
then one can use statistical mathematics to calculate
the resulting distributions. In the present case, the
distributions were not normal. We therefore used another
distribution that had an analytical solution to an
integral as defined by the following equation:
x
F(x) = I f (z) dz (1)
o
The Weibull distribution4 can approximate the log-normal
distribution, the normal distribution, the Bernoulli
distribution, and many others merely by a change in
parameters. There are also analytical functions that
describe both the probability density function and the
emulative density function. The following equations
describe the Weibull cumulative density function, the
probability density function, and a way of calculating
independent variables from uniform random numbers (R ),
respectively:
33
-------
0.5
0.4
o
o>
£ 0.3
0.2
0
I 23456789
Figure 9. Sample frequency histogram
-------
P(x) = 1 - exp (-axb) (2)
different iat ing,
f(x) = abx13"1 exp (-axb) (3)
F(x) = Ru
where a, b, and x > 0.
The analysis then proceeds as follows. Let the distri-
bution be as shown in Figure 8. We define it as the
Weibull cumulative density function and calculate the
parameters a and b. Using equation 2, we transpose,
invert and take double natural logarithms to obtain:
In In U _ F(X) = b In x + In a (5)
Having transformed the cumulative density function
into a linearized expression, we can now plot it in
the slope-intercept form. We perform a least squares
analysis on the computer and obtain the parameters
a and b from the slope and the intercept:
define Y = In In L _ F(x) (6)
X = In x (7)
K = In a (8)
35
-------
The least squares analysis is performed on the following:
Y = bX + K (9)
One of the advantages of a Weibull function is the
flexibility of its expression. This is true because
b, a shpae parameter, really makes the function a family
of curves. When b equals 1, the probability density
function is the familiar log-normal distribution. The
Rayleigh distribution, which deals with the distribution
of errors in a two-dimensional grid, is obtained when b
is equal to 2. When b equals 3.25, the Weibull distri-
bution closely approximates the normal distribution.
This is shown in Figure 10. Having obtained values for
a and b, we now sample from a distribution by using
equation 4. Since the Weibull function is continuous,
it will have values for all ranges of x from zero to
infinity, and since the random number generator could
conceivably select a random number such as 0.99999»
we would get an extremely large magnitude for a value of
x. Therefore, in order to restrict the sampling range,
we restrict the range of the uniform random number
generator and calculate a transformed number R, by per-
forming the following:
Rt ~ ClRu +
Thus, if we want a lower random number R and a high
J6
limit number R, , we would use the uniformly generated
numbers 0.0 and 1.0 and solve the following simultaneous
equations for GJ and c2:
36
-------
1.4
1.2
1.0
0.&
0.6
0.4!
0.2i
0'-
0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8
Figure 10. A comparison of the normal and Weibull
distributions
37
-------
Ez = G! (0.0) + c2 (11)
Rh = GI (1.0) + c2 (12)
Thus, if we want to limit the sampling to be between
0.99 and 0.01, we would make the following transformation
on the random number generator:
p.01 = G! (0.0) + c2 (13)
0.99 = G! (1.0) + c2 (14)
then G! = 0.98
c2 = 0.01
Therefore,
R, = 0.98 R + 0.01 (15)
i» u
This type of transformation can be performed on any
of the input distributions.
38
-------
SECTION VII
IMPACT ASSESSMENT METHODOLOGY
A. STATE VARIABLE ANALYSIS
Defined generically, state variables are those that
can affect the success or failure of a decision-maker's
operations but which are essentially beyond his ability
to control. For instance, during the last year major
international airlines have experienced severe financial
problems because of large, arbitrary increases in the
price of aviation fuel which is an important item of
their costs but over which they have no control.
A manager of an electric power plant or an industrial
plant, such as a smelter, also has to cope with state
variables when he decides which type of technical
process he should install to reduce the volume of SO
X
emissions that fuel combustion in his plant generates.
Table 1 lists some of these state variables.
It would be costly, time-consuming, and methodologically
difficult to incorporate a large number of state variables
into an analytical model that a Government R&D agency,
39
-------
like EPA CSL, might employ. Fortunately, screening
criteria can be employed to reduce a large list, such
as that provided in Table 1, to manageable proportions.
Several of these criteria are defined below.
Defineability and Measurability
Unless a clear consensus can be reached relative
to the meaning of a variable and a quantitative
method developed to measure it, it is very diffi-
cult to handle that type variable in a formal
analytical model. For instance, political
strategies of both the oil importing and exporting
nations are relevant state variables, but such
strategies are at this point not susceptible to
systematic analysis.
Data Availability
Unless a reasonable quantity of historical data
is available, there is no basis from which to
project the future status of a state variable.
For instance, potential equipment shortages, such
as new stack scrubbers now under development,
could be a relevant state variable that would
affect the decisions of companies relative to the
choice of SO control technologies. However,
Jv
there is very little statistical information from
which to project such potential shortages.
-------
Table 1. ILLUSTRATIVE STATE VARIABLES APPLYING TO
THE CHOICE OF SO CONTROL TECHNOLOGIES
A.
Market Costs of Production Factors
Labor
Land
Construction
Equipment
Materials and supplies
Transmission and distribution
Financing (interest rates)
Waste material disposal
etc.
Competitive Fuel Relationships
Coal
Oil
Gas
Hydro
Nuclear
-Supply
&
-Price
Consumer Elements
Per capita requirements for energy
Choice of fuel for heating, etc.
etc.
Government Regulations
Effecting supply
Strip-mining controls
Off-shore oil policy
Licensing (e.g., AEC, pipe lines, transmission
lines)
Other public land leasing
etc.
Effecting costs
Strip-mining land restoration requirements
Oil depletion allowances, taxes, etc.
etc.
-------
Table 1 (Cont.) ILLUSTRATIVE STATE VARIABLES
APPLYING TO THE CHOICE OF SO
CONTROL TECHNOLOGIES
Effecting both supply and costs
Mine safety & health regulations
etc.
Effecting fuel usage
Clean-air regulations governing use of
polluting fuels
etc.
Effecting consumer demand
Various rationing measures
etc.
Effecting prices and rates
ICC rates for transportation
Natural gas price controls
International Considerations
Political Policies: Fuel Exporting Nations (OPEC)
Consumption Patterns: Fuel Importing Nations
(Japan, Common Markets etc.)
etc.
Logistical Constraints
Shortages of labor (e.g., coal miners)
Shortages of transportation (e.g., pipe lines)
Shortages of equipment (e.g., stack scrubbers)
Shortages of desirable fuels (low sulfur, coal,
natural gas, etc.)
Other resource shortages (e.g., water)
etc.
Technological Parameters
Amount of R&D funding
Supply of scientists and research facilities
(time lags)
etc.
-------
A Substantial Item
If an item of cost or credit is relatively small,
even if it is measurable and affects different
SO technologies to a different degree, it would
A.
not be worthwhile to include it in the analysis.
The revenues that certain processes might make
possible through the sale of some byproducts is
an example.
Taking all of these factors into considerations and
using the results of the sensitivity analyses, we have
selected the price of coal as an illustration of state
variable analysis. Besides the sensitivity analyses,
we have consulted approximately a dozen different
government and private sources in reaching the conclu-
sions that follow:
For two reasons, it appears that the trends in
the price of coal will work toward the competitive
advantage of the Solvent Refined Coal (SRC) process
as opposed to the Lurgi-Synthetic Natural Gas (SNG)
process. The average price for coal is expected
to double in the next decade whereas the general
cost of living and most industrial prices are
expected to increase by only 50$. These two factors
mean that a process, such as SNG, which is more
sensitive to the price of coal would be at a
competitive disadvantage compared to another process,
such as SRC, which is less sensitive to the price
of coal.
-------
It is very difficult to project rationally what level the
price of coal will actually reach a decade from now. In
the last year, coal sold under contract doubled in price.
Non-contract coal (spot purchases) increased by as much
as 1,000 percent to $120 per ton in some instances.
Most observers, perhaps as an article of faith and
hope, anticipate that future percentage increases in
the price of coal will be much less annually over the
next decade than they were in the last year. Our own
projection that the price of coal will double in the
next decade (see Figure 11) is based more on certain
fundamental long-term trends than in terms of last
year's experience. First, for about five years the
price of coal increased much more sharply than did the
average of other prices. Second, there is evidence
that future increases in the price of coal will con-
siderably exceed increases in the general price level.
The projected 100% increase in the price of coal in
the next decade will result from the combined action
of many factors:
a. Fifty percent of this increase will result from
a general inflation in all prices, domestic and
foreign. This continued inflation means that all
of the costs - wages, equipment, materials, etc. -
involved in producing a ton of coal will, on the
average, increase at this 50% rate. This projected
annual average incrase for the next decade of 5%
roughly equals the average annual increase in the
1964-1973 period as shown in Figure 12.
-------
10CH
0>
in
to
a?
&_
o
c
o>
u
o>
Q_
50-
c
b
1974
1979
1984
Q iin
b -
c -
d -
e -
f -
g -
Figure 11.
Code:
general inflation
market competition for all fuels
shortages of mining labor
safety and health regulations
strip mining compliance
additional transportation costs
declining labor productivity
The price of coal in 198^: A projected
index with explanatory factors
1974 = 100
-------
1967=100
155
150
140
130
120
110
100
90
85
FOOD'
1964 1965 1966 1967 1968 1969 1970 1971 1972
1973
Figure 12. Consumer price index (all items)5
b. Fifteen percent of this increase will be due to
market place competition that forces up the prices of
all fuels at a faster rate than the increase in prices
generally, this relative increase in fuel prices
will be due partly to a growing domestic and world
demand for energy, coupled with a tight supply.
The nation's drive to satisfy a larger percentage
of its total energy needs from domestic sources
will contribute to the tightness in energy supply.
-------
This greater-than-average increase in the prices
of fuel, taken as a group, as compared to prices
generally, is reflected by the data reported in
Table 2.
Table 2. INCREASES IN THE PRICE OP FUELS*
COMPARED TO GENERAL PRICE INCREASES6
1967 = 100
Fuels
All prices
Annual
Average
1973
134.3
13^.7
September
1974
225.0
167.2
*Includes coal, coke, gas fuels, electric power,
crude petroleum, refined petroleum products
c. Ten percent of this increase will be due to
a shortage of coal miners that will push their
hourly rate up faster than the average of wages
in other industries. This labor shortage will
be due partly to the fact that older miners will
be retiring at an earlier age, and their sons
will be migrating to the cities to seek other
types of more attractive work rather than working
in the mines. The declining labor force in all
types of mining has been a very long run trend
extending over a period of at least 35 years,
as the data in Table 3 reveal.
-------
Table 3. EMPLOYMENT IN MINING7
Year
19^7
1952
1957
1962
1967
1972
Employment
955,000
898,000
828,000
650,000
613,000
607,000
This declining labor force in mining, especially
coal mining, is at variance with trends in other
sectors of the economy as indicated in Table 4.
These data show that whereas in other sectors of
the economy employment has increased by an average
annual rate of 2.J%, in coal mining it has decreased
4.4$ annually. Even with all of the pressures to
increase coal production in the balance of this
decade, it is anticipated that for the twelve-year
period 1968-80, coal mining employment will increase
at less than one-half the rate of the increase in
employment generally. Finally, it is estimated
that after 1980 there will be no increase at all
in coal mining employment.
48
-------
Table 4. EMPLOYMENT IN COAL MINING COMPARED TO
ALL TYPES OP NON-AGRICULTURAL EMPLOYMENT8
( average annual percentage change)
Total non-agricul-
tural employment
Mining
Average Annual
Percent Change
1959-68
2.7
-4.4
1968-80
2.4
1.1
1980-85
1.4
0.0
d. Six percent of the increase will be due to new
mine safety and health regulations, and a stricter
enforcement of existing regulations. The estimated
impact of safety and health regulations is primarily
an assessment of the effect of the Federal Coal
Mine Health and Safety Act that became effective
in April 1970. This Act, which is designed primarily
to make the physical mine safe, has been in effect
too short a time and the evidence accumulated to
date is insufficiently complete, to permit conclu-
sive judgements. However, some observers have
contended that the impact of this Act, has been to
reduce production by 15 to 25$ and to increase the
per ton cost of coal 20 to 30%.9
e. Six percent of the increase will be due to the
costs associated with compliance with new stricter
strip-mining regulations. Here again, judgements
must be formed on very incomplete information. As
this report is being written, (December 1974), the
U.S. Congress is involved in a heated debate relative
49
-------
to passing new strip-mining legislation. Whether
or not such legislation passes at this session of
the Congress, stricter strip-mining legislation is
highly likely. For one thing, the more liberally-
oriented Congress that took office early in 1975
might pass such legislation. Perhaps more important
is the fact that individual states, such as
Pennsylvania, have already passed much stricter
strip-mining laws than the pending Federal legis-
lation. The cost of reclaiming land that has been
strip-mined varies considerably depending upon such
factors as the terrain of the land and the depth
of the coal deposit under the strip-mined surface.
One set of estimates that generalizes on these
costs is shown in Table 5.
Table 5. TYPICAL COSTS OF RECLAIMING LAND
THAT HAS BEEN STRIP-MINED10
Mining Area
Eastern States
Western States
Dollars
per Acre
$3,000-5,000
$1,000-4,000
Dollars
per Ton
$1.00-3.00
$0.20-2.00
Cents
per kWh
0.3-0.10
0.001-0.01
f. Eight percent of the increase will be due to
the spiraling costs of transporting coal to market.
Except for a slight dip in 1972, the per-ton costs
of transporting coal to market have steadily in-
creased since 1968, as shown in Table 6. Although
published data were not available to us at the time
this report was being written, it is our under-
standing that the rise in transportation costs has
50
-------
continued in the two years since the last date
shown in Table 6. Of course, the farther away
that users are from the coal mines, the greater
will they feel the impact of increases in the costs
of transporting coal to market. The difference in
the transportation costs is one factor that explains
why it costs so much more for electric power plants
in some regions of the country to use coal as a
fuel than in. others. For instance, the costs of
steam electric power plants using coal in New England
are more than twice as much as in the west south-
central states. Other regions fall between these
two extremes as shown in Figure 13.
Table 6. AVERAGE PER TON COSTS OF TRANSPORTING
BITUMINOUS COAL TO MARKET11
Year
1968
1969
1970
1971
1972
Costs per Ton
$ 3.02
3.12
3.43
3.72
3.68
Price Index
100.00
103.31
113.58
123.18
121.85
g. Five percent of the increase will be due to
declining labor productivity in the coal mines.
There is considerable evidence to substantiate
this productivity problem. For instance, a Federal
report shows that coal mining was the only one of
approximately 50 manufacturing, mining, transpor-
tation, and other industry sectors that registered
51
-------
60 -
--280
45
210
3 304-
CD
O
140
70
00
B
00
H
Code:
A - West, South Central States
B - Mountain States
C - East, South Central States
D - West, North Central States
E - East, North Central States
F - Middle Atlantic States
G - South Atlantic States
H - New England States
Cents
21.0
22.7
32.5
34.0
38.9
42.1
42.6
49.7
100
108
154
162
178
200
203
237
Figure 13.
Regional differences in the costs of coal
for steam electric power plants
(Unit of measure: "As Burned" cents per
million Btu, 1972)
52
-------
a decline in output per manhour in the 1968-1973
period.13 Figure 14, based on data from the U.S.
Bureau of Mines, shows the extent of this decline.
10.64
12.03 11.9
11.0
1960 61 62 63 64 65 66 67 68 69 70 7] 72
73
Figure 14. Underground mining productivity'1'4
Taking all mining as a class, government analysts
have projected that over the next decade the improve-
ment in productivity will be much less than in other
sectors of the private economy. These data are
shown in Table 7.
Table 7- PROJECTED INCREASES IN LABOR PRODUCTIVITY15
(Annual Average Percentage Increases)
Total private
economy
Mining
1972-80
3.2
0.9
1980-85
2.8
0.8
53
-------
Several factors account for the fact that produc-
tivity will continue to be a problem in the mining
industry. The shortage of miners will bring into
the labor force marginal, less efficient workers
who would not be employed if it were not for the
labor shortage. A contributing factor will be the
restlessness and low morale of men who work in the
mines. Although a decade from now mining wages
will compare favorably with many other industries,
and working conditions in the mines will be much
improved over those today, still, the amenities
and fringe benefits provided to miners will be
fewer than those provided to other trades, occupa-
tions, and professions.
B. DISCUSSION
The primary purpose of this report has been to illustrate
how one would approach a state variable analysis and
present the findings in a relatively simple fashion. The
specific projection that the price of coal will double
in the next decade is based on several optimistic
assumptions:
a. The 50$ increase in general inflation over a
period of a decade, assumed above, is highly opti-
mistic by recent standards. In other words, we
have assumed that general prices will increase by
an average of only 4$ per year in the next decade
as compared to the 12% experienced in the last year.
-------
b. We have also assumed that a persistent energy
shortage can be kept from becoming a major crisis.
This optimism is based on an expectation that the
nation will be able to bring many more nuclear
plants into operation and will be able to tap the
oil of the Atlantic coast, the Alaskan slopes, and
that recently discovered in Mexico.
c. We have also assumed that the petroleum-using
countries and the OPEC nations will be able to
come to terms to maintain a high supply of mid-east
petroleum at relatively stable prices.
If one or more of these optimistic assumptions fail to
materialize, the rise in general prices will certainly
be much higher and the rise in coal prices proportionally
larger than our projections indicate.
The foregoing relatively optimistic scenario is, of
course, only one of many possible scenarios. One alter-
native would be to assume that the nightmarish inflation
that has plagued the United States and other nations in
the last year will become the new "wave of the future."
A comprehensive, probabilistic analytical model would
generate not only one scenario for the price of coal,
but a whole series of them based on different assumptions
relative to the values of major parameters.
Finally, a comprehensive, probabilistic, analytical
model would incorporate an analysis of many of the other
state variables listed in Table 1. One other such variable
might be the price of construction labor in various parts
55
-------
of the country. For instance, over the last five years
construction labor wage rates have increased considerably
more on the east coast than in central or western states.
This is shown in Table 8. Various factors explain this
fact. One of these has been the relatively heavy indus-
trialization of the south. A related one has been the
building boom in Florida and other states that has
accompanied this industrialization plus the migration
of retired people -to that area. Finally, the greater
unionization of labor in southern states has tended to
reduce historical differences in construction industry
wage rates. In any event, projected regional differences
in percentage changes in construction labor costs could
very well influence the choice that plant managers will
make in the coming decade as to which of the various
SO control technologies they would decide to adopt.
Table 8. AN INDEX MEASURING REGIONAL DIFFERENCES IN
THE INCREASE IN UNION WAGE RATES IN THE
CONSTRUCTION FIELD* IN THE 1969-74 PERIOD
Region
Index
Western States
South Central States
East, North Central States
West, North Central States
North Eastern States
South Eastern States
100
115
116
119
128
131
*This index is a weighted average using sheet metal
workers, electricians, and laborers. Basic data were
taken from: CMA News, Contractors Mutual Association,
Washington, D.C., March 1, 1971*.
56
-------
SECTION VIII
EFFORTS REQUIRED FOR A REALISTIC ANALYSIS
One basic assumption in this analysis is that the clean
fuel facilities will be located at or near the coal
mine mouth. If this is an unwarranted assumption, there
is nothing in the method that prevents us from including
a transportation cost distribution to these facilities.
However, viewing this aspect realistically, we feel that
quite the reverse might well be true, i.e., not only
would the clean fuel facilities be located at the mine
mouth and thus not incur any transportation cost, but
also these facilities might well be owned or jointly
owned by the mining company supplying the coal, thereby
resulting in a lower mine mouth cost of coal to this
facility. Since the cost of coal is an important vari-
able, clean fuel facilities provide coal with different
heat contents and thus resulting transportation costs,
this aspect warrants additional attention.
On the other hand, this analysis used a somewhat arbi-
trary distribution for transportation distances for
coal to the using utilities. Again, since the same
distance distributions were used for clean or dirty
coal, the only effect this has is that the utility vs.
industrial boiler cost comparisons may not be valid.
The relative cost comparison of the four alternatives
for utilities or industrial boilers is valid with
57
-------
respect to transportation distance. Naturally, the
absolute value for mills/kWh is a function of many
things and it was not the objective or scope of this
work to arrive at absolute values of cost. We have
raised questions about the validity of the clean fuel
cost models. These questions and objections can be
treated with our approach. Any portion of the process
or cost models that is deterministic can be treated
probabilistically to take into account undemonstrated
process reliability. This in turn can be reflected in
uncertainty of equipment size, reaction times, reactant
or catalyst amounts required, etc. Process complexity
and thus unforeseen increased manpower requirements for
any segment of the process can be treated with our metho-
dology. Deterministically this is easy to follow; e.g.,
there are problems with filtration in the SRC process
due to unforeseen scale-up agglomeration. If we know
this, we simply triple the equipment costs to indicate
that more equipment will be necessary to successfully
accomplish that particular process step. Or, the number
of workers required for that step can be increased, thus
increasing both capital and operating costs. But since
the process has not been demonstrated commercially and
we do not know if the price of that piece of equipment
should quadruple, triple, double, we now have a method
of treating this uncertainty- This approach should have
utility in allowing cost estimates to be made for undemon-
strated processes.
Another very important aspect of a realistic analysis is
the realistic projection of by-product disposal costs and
by-product credits. Using SRC as an example, would the
-------
increased availability of cresylic acid be an increased
incentive for its utilization, thereby making its price
relatively inelastic, or would its increased availability
result in a glut on the market and contribute to a lower
price? Again, the methodology we have developed would
allow a treatment of this question.
We have data to show that the price of coal and its
heating value are functionally related; the higher the
heating value the higher the price. Percent sulfur and
the price may be additionally related, however, we have
no data to support this.
Transportation cost per unit distance was considered a
constant, with the transportation distance being the statis-
tical variable. We know that unit transportation cost is
a variable as well, and a realistic analysis should take
this into account.
In the flue gas scrubbing models we also have some un-
realistic aspects. The Kellogg Company model for the
Wellman/Allied process uses a separate train for each
boiler- In actual practice, a common train would likely
be used for all the boilers in a power plant. The percent
sulfur removed is a constant for both scrubbing processes.
In actual practice, it would be a function of the percent
sulfur in the coal burned.
As a final comment, we feel that the unrecoverable release
of sulfur into the atmosphere in any form will be highly
undesirable due to the projected increasing worldwide demand
for sulfuric acid. In the future, the economics of either
59
-------
sulfur recoverable flue-gas desulfurization or clean
fuel preparation are likely to become competitive with
throwaway processes.
The methodology we have developed can be used for any
number of alternatives. Realistic results will be
obtained only from a well documented, thorough probabi-
listic analysis of process and cost uncertainties and
all relevant input parameters affecting these alternatives
The objectives of this program were limited in the scope
of results developed, but this program has shown the
fundamentals of a way uncertainty can be treated in a
quantitative manner. A realistic assessment of alterna-
tives would require the modeling not only of the processes
but also of the availability of construciton, transpor-
tation, capital, manpower and the seemingly intractable
and intangible social, economic and political perceptions
of utility viewed from highly subjective points of view.
The methodology for doing these things is called Decision
Analysis. We have taken a small subset from that science
to show how it might be useful to the decision maker when
he is operating under conditions that he faces every hour
of every day, operating under conditions of uncertainty.
60
-------
SECTION IX
EXAMPLE OP METHODOLOGY
Wellman-Allied Process - This process is based on a sodium
sulfite/sodium bisulfite cycle to remove sulfur dioxide from
the glue gas. The reactions that take place in the process
are abbreviated as follows:
Absorption
S02 + SOj + H20 -» 2HSOg (13)
Regeneration
2Na+ + 2HS03 ^ Na2S03+ + S02t + H20 (14)
S02Reduction
2S02 -> C02 + 2H20 + S2 (15a)
+ 4C02 + 4H20 + 4H2S + S2 (15b)
2H2S + S02 -> 3/2 S2 + 2H20 (15c)
The process flow scheme, shown in Figure 15, can be divided
into four main sections:
61
-------
I ARI
AREA I
FLUE GAS
REHEAT
FLUE GAS
H20-
ON
ro
FLUE GAS
COMPRESSION
FLUE GAS
TO STACK
AREA III
1
MAKE-UP
SYSTEM
PRESCRUBBING
AND
S02 REMOVAL
SULFITE' SOL'N
EVAPORATION
AND
CRYSTALLIZATION
FLY ASH
SLURRY
AREA I
AREA II
AREA III
AREA IV
ABSORBER
-S02 REGENERATION
PURGE/MAKE-UP
S02 REDUCTION
AREA III
_J L
VENT GAS
TO ABSORBER
S02
S02 PURIFICATION
(CONDENSATION/
STRIPPING)
CONDENSATE
| ARI
AREA IV
PURGE SYSTEM
(CRYSTALLIZATION
AND DRYING)
T
AREA
SO2
1
S02
REDUCTION
NATURAL
GAS
TAIL GAS I
TO ABSORBER I
PURGE
SOLIDS
SULFUR
Figure 15- Wellman-Allied Process flowsheet
-------
(1) The absorber section, including gas reheat and
discharge;
(2) The S02 regeneration section;
(3) The purge/make-up section; and
The S02 reduction section.
The S02-rich gas is contacted countercurrently in the
absorber by the sodium sulfite solution and exits from the
absorber top stripped of S02- The solution leaving the bottom
of the absorber is rich in bisulfite and is pumped to an
evaporator/crystallizer in the regeneration section. Low
pressure steam is used to heat the evaporator and drive off
S02 and water vapor. The gas stream leaving the evaporator
is partially condensed to remove the majority of the water
vapor and is then discharged to the 862 reduction unit where
natural gas is used to convert S02 to product sulfur.
To operate this scrubbing process, more coal feed to the
boiler would be necessary to provide the electricity and
steam required.
Cost Comparison Procedures
The major problem in comparing SO abatement methods of
J\.
different generic types on a common cost basis is the finding
of a common cost factor. The common cost factor used in this
study is a summation of the annual costs of fuel, fuel
handling, and flue gas desulfurization for each of the four
methods as applied to utility power plants or industrial
boilers.
63
-------
Electric Utility Power Plants - In this industry, the following
assumptions (based on M. W. Kellogg's work) were made for
each power plant to calculate the common cost factor:
The number of boilers is 4 if the plant capacity
is higher than 50 megawatts. This number becomes
3 is plant capacity is between 20 and 50 megawatts,
and 2 if the plant capacity is between 10 and 20
megawatts. It is assumed that there is only one
boiler in the plant if its capacity is less than
10 megawatts.
When the number of boilers is 4, it is assumed that
.the distribution of the plant generation capacity
among these boilers is 51.8$, 28.4$, 12.6$, and
7.2$ of the plant capacity. The capacity distri-
bution becomes 65$, 26$, and 9$ of the plant capacity
when the number of boilers is 3. This distribution
becomes 75$ and 25$ when the number of boilers is 2.
If the load factor for the first boiler is unity,
the load factor for the second boiler is 0.75.
The fraction is 0.55 for the third boiler, and
0.1 for the smallest one.
All the sulfur content in the coal feed to the
boiler becomes S02 in the flue gas.
Each power plant has a load factor of 0.5.
Wellman-Allied Process - A general procedure is shown in
Figure 16. The calculations can be grouped into three
areas. Area I contains the calculations associated with the
operation of the power plant Itself. No cost is involved in
-------
en
ui
Plant Capacity
in
Megawatts
Plant Capacity
in
Btu/day Required
Capacity and Load
Factor
for Each Boiler
Coal Feed Rate
Sulfur and Flue
at Full Capacity
for Each Boiler
for Each Boiler
Total Annual
Production Cost for
Scrubbing System
Coal Storage
and
Preparation
Unit Cost
of Delivered Coal
of Coal Handling
Transportation
Annual
Cost of Coal
and Coal Handling
Annual
Cost of Fuel,
Fuel Handling and
Flue Gas Scrubbing
Variables Included in the Block are Sulfur Content, Heating Value, Water Content,
and Ash Conent.
Area I - Power Plant
Area 11 - Flue Gas Scrubbing System (Detail is shown in Figs. 2 and 3)
Area 111 - Coal-Related Operations
Figure 16. General procedure for calculation of common
cost factor for add-on scrubbing systems
-------
this area. The input variables are: plant capacity in
megawatts, type of coal, and the modified load factor for
each boiler from Area II. The input variables included in
the type of coal are coal sulfur content, heating value,
water content, and ash content. The outputs from this area
are sulfur and flue gas flow rates for each boiler at full
capacity, load factor of each boiler before the scrubbers
are installed, and annual consumption of coal.
Area II contains the calculations to obtain the annual
production cost of the scrubbing system. Detailed procedures
involved in this portion of the Wellman-Allied process are
shown in Figures 16 and 17, respectively.
The cost models for these two processes as derived by the
M. W. Kellogg Company have been modified to change the costs
of electricity and steam to the cost of additional coal
needed and the additional cost of desulfurizing the
increased flue gas. Several simultaneous algebraic equations
were derived and solved to obtain the additional heat input
needed. The increased annual production cost is calculated
by using a modified load factor for each boiler.
The outputs from Area II are the modified load factors for
each boiler and the total annual production cost for the
scrubbing system.
Area III includes cost of. coal and its handling. The input
variables are mine-mouth costs of coal, coal transportation
costs, costs of coal storage and preparation, and costs of
ash disposal. The output is the annual cost of coal and coal
handling.
The total annual production cost for the scrubbing system
obtained from Area II and the annual cost of coal and coal
66
-------
Coal Feed Rate
for a Boiler*
Cost Of:
Average Hourly Labor
Ammonia
Fuel Oil
Sodiun Carbonate
Natural Gas
Process Water
Cooling Water
Filter Aid
Annual Cost of
Materials
Less By-Product
Credit
Only the procedure for one boiler is shown
here. This procedure is the same for other
hoilers.
Total Annual
Production Cost
for
Scrubbing System
Summation for
All the Boilers
in the Plant
Total Annual
Production Cost
for the Boiler
Figure 17.
Calculation of annual production cost
of Wellman-Allied Process
-------
handling obtained from Area III are then added together to
get the annual cost of fuel, fuel handling and flue gas
desulfurization.
Discussion - In all the cost calculations for utility power
plants, a heat rate of 10,400 Btu/kwh was used to calculate
the fuel demand for each boiler regardless of the size and
age of the boiler. This heat rate represents an overall
cycle efficiency of 32.7%, and is about the average for all
utility boilers. If one is looking at the cost for S02
abatement for a new and large plant or an old and small plant,
this heat rate may not be accurate. However, if the same
heat rate is used for comparing all abatement methods, it is
believed that the comparison is still valid even at these
extreme cases.
Sample Results
Table 9 lists the deterministic input values used in the
Wellman-Allied simulation. The state variables were fitted
to Weibull distributions; their one and ninety-nine percentile
values are tabulated below.
Item Low value (1%) High .value (99$)
Location factor 0.929 2.10
Percent sulfur in coal O.l4l 3.94
Utility plant size (mega- 19.1 2366
watts)
Cost of coal $9.23 $22.23
Transportation distance 453 miles 1607 miles
Industrial boiler size 10.0 2710
(million Btu)
68
-------
Table 9. INPUT VARIABLES, WELLMAN-ALLIED PROCESS
Name
Item
Value
CO
CWATER
CDPOSE
CSULP
HV
FWATER
PASHRC
CNH3
CPOIL
CELEC
CONTIN
CS
CN
CH
CC
CFA
Average Hourly Labor Rate
Cost of Water
Cost of Waste Disposal
Credit for Sulfur By-Product
Heating Value of Coal
Fraction of Water in Coal
as Received
Fraction of Ash in Raw Coal
as Received
Cost of Ammonia
Cost of Fuel Oil
Cost of Electricity
Contingency
Purchased Price of Sodium
Carbonate
Purchased Price of Natural Gas
Purchased Price of Steam
Purchased Price of Cooling
Water
Purchased Price of Filter Aid
$ 7.00 Per Hour
$ C.20 Per K Gal
$ 3-00 Per Ton
$ 50.00 Per Ton
12000. Btu/Lb
0.20
0.08
$ 200.00 Per Ton
$ 0.80 Per MBtu
8.00 Mils/Kwhr
0.10
$ 90.00 Per Ton
$ 0.50 MSCF
$ 0.50 Per K Lb
$ 0.02 Per K Gal
$ 50.00 Per Ton
69
-------
Table 10 lists sample results for the Wellman-Allied process,
The baseline calculation is a simulation performed with all
probabilistic inputs being sampled. The remaining rows show
the sensitivity of the model to the state variable inputs.
Thus "Location Factor High" means that that input was fixed
at the value shown in parentheses and all remaining state
variables were sampled probabilistically.
Cost Model for Wellman-Allied Process
Nomenclature
GP
GT
NA
SF
S7
S28
N7
N28
E
M
Total flue gas to control
plant
Total flue gas to each
absorber train (maximum
value of GT = 500)
Number of absorber trains
Total sulfur flow in flue gas
to control plant
Total sulfur flow in flue gas
to control unit per train
of sulfur-related equipment
in absorber and S02 regen-
eration areas (maximum
value of S7 = 7)
Total sulfur flow in flue gas
control unit per equipment
train in the purge/make-up
and SO2 reduction areas
(maximum value of S28 = 28)
Number of trains of sulfur-
related equipment in the
absorber and S02 regen-
eration areas
Number of equipment trains
in the purge/make-up and
S02 reduction areas
Major equipment cost (direct
material and subcontracts)
Field materials costs
KCFM (actual)
KCFM (actual)
K Ibs/hr
K Ibs/hr
K Ibs/hr
$K
$K
70
-------
Table 10. SIMULATION RESULTS - UTILITY BOILERS
WELLMAN-ALLIED PROCESS (1974 PRICES)
Baseline
Location
factor
high (2.10)
Location
factor
low (.929)
% Sulfur in
coal
high (3.9*0
% Sulfur in
coal
low (0.14)
Power plant
size low
(19.D
Cost of coal
high (22.23/ton)
Cost of coal
low (9.23/ton)
Transportation
distance low
(453 miles)
Average
TAG ( x 10 )
$ 32.2
$ 32.5
$ 32.3
$ 35.4
$ 30.1
$ 2.20
$ 37.8
$ 27.4
$ 25.3
Average cost
per megawatt
plant capacity
$ 88,300
$ 89,000
$ 87,700
$ 97,800
$ 80,900
$115,000
$102,000
$ 75,900
$ 70,900
Mils
per
Kwh
19-2
19.4
19.1
20.1
18.1
25.0
22.2
16.5
15.4
VALUES FOR 10,000 SIMULATIONS
71
-------
Nomenclature
L
A, S, P
IP
RB
RP
BARC
TPI
TCR
CONTIN
AS
AAO
AN
AFA
AE
AH
ACW
AW
Field labor costs
Letters following E, M, L
A refers to absorber area
S refers to S02 regenera-
tion area
P refers to purge/make-up
area
R refers to 862 reduction
area
No letter following refers
to total for all area
Particulate index (IF - 1 if
particulates are present in
flue gas. IF = 0 if par-
ticulates are absent).
Retrofit difficulty factor of
each boiler
Retrofit difficulty factor of
gas-related equipment in
the absorber area which is
not in parallel trains, i.e.,
the fuel oil system; assumed
to be equal to the highest
RFB
Bare cost of the control unit
Total plant investment
Toal capital required
Contingency
Annual cost of sodium car-
bonate
Annual cost of anti-oxidant
(amine)
Annual cost of natural gas
Annual cost of filter aid
Annual cost of electric
power
Annual cost of steam
Annual cost of cooling water
Annual .cost of process water
$K
$K
$K
$K
$K/yr
$K/yr
$K/yr
$K/yr
$K/yr
$K/yr
$K/yr
$K/yr
72
-------
Nomenclature
AP
ASC
APS
CS
CN
CPA
CE
CH
CCW
CP
VSC
VPS
TO
CO
LP
AOC
TAG
P
Annual cost of fuel oil
Annual sulfur credit
Annual purge solids credit
or debit
Purchase price of sodium
carbonate
Purchase price of natural gas
Purchase price of filter aid
Purchase (or transfer) price
of electricity
Purchase (or transfer) price
of steam
Cost of cooling water
Purchase price of fuel oil
Unit value of sulfur (negative
if credit)
Unit value of purge solids
(negative if credit)
Total number of operators
Unit cost of operating
labor
Load factor of the power
plant
Annual net operating cost
Total annual production cost
Location factor
$K/yr
$K/yr
$K/yr
$/ton
$/KSCP
$/ton
mills/kwh
$/K Ibs
$/K gal
$/M Btu
$/long ton
$/ton
$/hr
$K/yr
$K/yr
73
-------
Capital Cost Model Summary of Equations
NA
EA
ES
EP
ER
r -,
= £ RB 726 (GT/550)°-5+ 639 (GT/550)0-9 + 119 RP (GP/3300)0-5
n=l L Jn .
+ [l33 (S7/7)0-5 + 127 IP (S7/7)°'6]N7 $K
= [209 (S7/7)0'5 + 618 (S7/7)°'6+ 157 (S7/7)°-9]N7 $K
= [525 (S28/28)°-5+ 380 (S28/28)0-6 + 86 (S28/28)0-7
+ 306 (S28/28)°-8+ 519 (S28/28)°-9~JN28
'
= 998 (SF28)~"~ + 287 (SF/28)0-6 + 683 (SF/28)0
M = 0.429 EA + 0.742 ES + 0.827 EP + 0.772 ER
L = 0.224 EA + 0.310 ES + 0.433 EP + 0.623 ER
BARC = 1.15 (E+M) + 1.43 L-F
TPI = 1.12 (1.0 + CONTIN) BARC
TCR =1.15 TPI + 0.8 TO-CO (1+F) +0.4 ANR
Operating Cost Model
AS
AAO
AN
AFA
AE
AH
ACW
AW
AF
= 28.2 CS-LF (SF/28)
= 0.04 CAO-LF (SF/28)
= 1460 CN«LF (SF/28)
= 1.24 CFA-LF-IF (GP/3300)
= [l54 (GP/3300) -I- 79 (SF/28)] CE-LF
= 5430 CH-LF (SF/28)
= [856 (GP/3300) + 19,900 (SF/28)~| CCW-LF
= 64 (SF/28) CW-LF
= 1,800 (GP/3300) CF-LF
$K
$K
$K
$K
$K
$K
$K
$K/yr
$K/yr
$K/yr
$K/yr
$K/yr
$K/yr
$K/yr
$K/yr
$K/yr
74
-------
ASC = 95.11 (SP/28) VSC-LF $K/yr
APS = 37-3 (SP/28) VPS-LF , $K/yr
ANR = AS + AAO + AN + APA + AE + AH + ACW + AW $K/yr
+ AF + ASC + APS
AOC = 0.078 TPI + 2-TO-CO (1+F) + ANR $K/yr
TAG = 0.237 TPI + 2.1-TO-CO (1+P) + 1.04 ANR $K/yr
75
-------
SECTION X
REFERENCES
1. Shore, D., J.. J. O'Donnell, and P. K. Chan. Evalua-
tion of R&D Investment Alternatives for SO Air
Pollution Control Processes. EPA 650/2-74-098, 1974.
2. North, D. W. Systems Science and Cybernetics.
SSC-4 (3), 1968. p. 200.
3. Howard, R. A. Ibid., p. 211.
4. Weibull, W. Fatigue Testing and Analysis of Results.
New York, Pergamon Press, 1961.
5. U.S. Department of Labor. Monthly Labor Review.
November 1974.
6. U.S. Department of Commerce. Survey of Current
Business. November 1974. p. 109
7. U.S. Department of Labor. Monthly Labor Review.
September 1974. p. 85.
8. U.S. Department of Labor. The U.S. Economy in 1985:
A Summary of BLS Projections. 1974. p. 54.
9. Continental Oil Company. Coal and Energy Shortage.
(Presented to Security Analysts. December 1973) p. 19-
10. Ibid., p. 24.
11. MoodyTs Transportation Manual. 1974. p. 20.
77
-------
12. National Coal Association. Steam Electric Plants
Factors. 1973. p. 55.
13. U.S. Department of Labor. Monthly Labor Review.
August 1974. p. 70.
14. Coal and Energy Shortage. Op. Cit., p. 18.
15. The U.S. Economy in 1985; A Summary of BLS Projections
Op. Cit., p. 37.
78
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1 REPORT NO
EPA-600/2-75-045
2.
3. RECIPIENT'S ACCESSION-NO.
4. TITLE AND SUBTITLE
A Method for Evaluating SO2 Abatement Strategies
5. REPORT DATE
September 1975
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
C.T.Chi, E.C.Eimutis, W.H.Hedley,
M. V. Jones, R. Jones, and L. B. Mote
8. PERFORMING ORGANIZATION REPORT NO.
MRC-DA-^92
9. PERFORMING OR9ANIZATION NAME AND ADDRESS
Monsanto Research Corporation
1515 Nicholas Rd
Dayton, Ohio 45407
10. PROGRAM ELEMENT NO.
1AB013: ROAP 21ADE-008
11. CONTRACT/GRANT NO.
68-02-1320, Task 11
12. SPONSORING AGENCY NAME AND ADDRESS
EPA, Office of Research and Development
Industrial Environmental Research Laboratory
Research Triangle Park, NC 27711
13. TYPE OF REPORT AND.PERIOD COVERED
Task Final; 2-12/74
14. SPONSORING AGENCY CODE
15. SUPPLEMENTARY NOTES
16. ABSTRACT
The report describes a non-deterministic probabilistic method that can be
used to evaluate sulfur oxides abatement alternatives. The method is not restricted
to gas scrubbing models, but can also handle clean fuel substitution processes. The
method utilizes a Monte Carlo simulation approach which can treat fully developed
cost models and also those which encompass a high level of uncertainty either in
structure or in process parameters. The report presents impact assessment tech-
niques which can be used to develop methodology for alternative future scenarios.
A specific example, the Wellman-Allied process, is used to demonstrate the
utility of simulation methodology.
7.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.lDENTIFIERS/OPEN ENDED TERMS
c. COSATI Field/Group
Air Pollution
Sulfur Oxides
Abatement
Strategies
Evaluation
Monte Carlo Method
Cost Engineering
Mathematical Models
Air Pollution Control
Stationary Sources
Wellman-Allied Process
13B
07B
14A
12A
8. DISTRIBUTION STATEMENT
Unlimited
19. SECURITY CLASS (ThisReport)
Unclassified
21. NO. OF PAGES
85
20. SECURITY CLASS (Thispage)
Unclassified
22. PRICE
EPA Form 2220-1 (9-73)
79
------- |