&EPA
  United States
  Environmental Protection
  Agency
Office of Transportation                     EPA420-D-04-002
and Air Quality                        March 2004
             Advanced Technology Vehicle
             Modeling in PERE

-------
                                                            EPA420-D-04-002
                                                                 March 2004
  Advanced Technology Vehicle Modeling in PERE
                               Edward Nam

                      Assessment and Standards Division
                    Office of Transportation and Air Quality
                     U.S. Environmental Protection Agency
                                 NOTICE
  This Technical Report does not necessarily represent final EPA decisions or positions.
It is intended to present technical analysis of issues using data that are currently available.
        The purpose in the release of such reports is to facilitate an exchange of
       technical information and to inform the public of technical developments.

-------
                                   Table of Contents

Executive Summary	2
Introduction	4
Conventional Vehicles	5
       Vehicle Model  	5
       Gasoline Spark Ignited Internal Combustion Engine	6
       Transmission	  10
       Diesel Internal Combustion Engine	  12
Advanced Internal Combustion Engine	  14
Hybrid Vehicles	  16
       Strategy	  17
       Motor/Generator/Inverter	  18
       Energy Storage Devices - Batteries, Ultracapacitors and Hydraulics	20
       Vehicle Weight and other Specifications	21
       Accessories	22
       Model Calibration	22
       Validation Results	23
Fuel Cell Vehicles	27
       Validation	29
Sensitivity	31
       Road Load and Track Coefficients	33
Application to MOVES	35
Cold Start	41
Acknowledgments	42
References	43
Appendix A	46
Appendix B	48

-------
Executive Summary

This study proposes a modeling methodology for light duty advanced technology vehicles
including those powered by:
•  Advanced gasoline internal combustion engine
•  Advanced diesel internal combustion engines
•  Hybrid electric and gasoline/diesel powertrains
•  Hydrogen fuel cell hybrid

The model called PERE (Physical Emission Rate Estimator) is designed to support the new EPA
energy and emissions inventory model, MOVES (MOtor Vehicle Emissions Simulator). PERE is
a stand-alone spreadsheet, which can be run by an informed user. It outputs Pump-to-Wheel
(PTW) fuel consumption rates. The purpose of PERE is to fill data gaps in MOVES and to help
it extrapolate to future projections of energy and emissions. The current version of PERE will
model many of the advanced technologies in MOVES and is capable of capturing more. If the
user is knowledgeable of the vehicle parameters required in PERE, other technologies can be
modeled that are not described in this report (e.g. series hybrids etc.). However, since PERE
requires significant development time for each technology, the current version of MOVES
(2004) may directly model some vehicles using modifications to existing rates (e.g. hydraulic
hybrids and some alternative fuels).

MOVES is currently being integrated with GREET (developed at Argonne Laboratory), which
produces upstream Well-to-Pump (WTP) energy and emissions rates. The combination of PERE,
MOVES, and GREET will provide a powerful comprehensive modeling suite for anyone
requiring emissions inventory projections or life cycle (energy/emissions) studies for mobile
sources into the future.

As the name implies, PERE uses physical principles to model propulsion systems in the vehicle.
It is therefore built on a relatively  strong foundation. The model is based on a sound, yet
elegantly simple model for the internal combustion engine. Simulation of hybridization is
achieved by inserting  a secondary power source (usually a battery/motor combination). Fuel cell
vehicles are simulated by replacing the engine with an aggregate fuel cell system. PERE takes
vehicle input parameters, then runs the vehicle through driving cycles defined by the user and
outputs second-by-second fuel consumption rates. All  of the processes are transparant to the
user, which allows for easy updates.

The model is validated to four production conventional gasoline, and five advanced (and hybrid)
vehicles by comparing the PERE output to the rated fuel economy figures. The advanced
vehicles include advanced engine technologies, such as lean burn gasoline. The model performs
well, in most cases the predictions are within  10%, as the figure below shows.

-------
                                                                  DPERE
                                                                    EPA unadjusted
   0.000
          PERE  Toyota   VW     VW    Honda   Honda  Honda
           MIT   Camry   Jetta    Jetta  Civic DX Civic HX   Civic
                         gas    Diesel                 Hybrid
Honda  Toyota  Toyota
Insight Prius '01 Prius '04
Figure ESI. Combined fuel economy (city/highway) estimates from PERE compared to vehicle
rated values.

The fuel cell model describes a direct hydrogen PEM fuel cell hybrid. It is validated to the fuel
economy results from the Honda PCX vehicle. The model predicts fuel economy within 5% of
the measured value.

The report also includes a sensitivity analysis, which will help guide users to determine which
parameters require more accuracy.

Finally, the report describes how PERE might be used to support MOVES. It provides  guidance
for how parameters may be estimated in order to perform future projections. It also describes a
methodology by which the PERE output can be integrated into MOVES.

-------
Introduction

From reading manufacturer press releases, and seeing the new vehicles being offered, it is
becoming evident that advanced technology vehicles, such as hybrids, will become more
commonplace in the near future. A hybrid vehicle combines two forms of propulsion in order to
optimize efficiency and fuel economy. The incremental cost of hybrid vehicles is expected to
decrease as volumes increase, and as public acceptance increases. There are also other types of
advanced vehicles, such as clean diesel, and fuel cell vehicles, which may claim a portion of the
market in the future. It is important to understand the possible effects that these vehicles will
have on the fleet and the environment.

The goal of this project is to develop a model, which can simulate a variety of future or advanced
technology vehicles for the purpose of supporting policy analysis within EPA as well as
estimating future emissions inventories. The Physical Emissions Rate Estimator (or PERE) is
expected to generate fuel (or energy) consumption rates for the new EPA emissions inventory
model, MOVES (MOtor Vehicle Emission Simulator). PERE is in a spreadsheet format and
should be usable for most users, who have a nominal amount of background information on
hybrids and fuel cells.

The current report describes the modeling of fuel and energy consumption by conventional and
advanced technology light duty vehicles using physical principles. The model development and
validations are for the following light duty technology types: conventional gasoline, advanced
gasoline (lean burn), diesel, "mild" parallel hybrid, "full" parallel hybrid, and fuel cell hybrid
vehicles. With the exception of fuel cells, these are the technologies, which have the most short-
term promise, i.e.  they are already sold in significant numbers and are likely to experience
growth. The number of other advanced technology combinations that could exist are enormous.
Such examples are hydraulic hybrids, series hybrids, plug-in hybrids, hydrogen internal
combustion engine, pure electric, etc. This report does not model all of the possible
combinations, however the hope is to demonstrate that PERE is a robust model that can
accommodate most of these technologies if the need arises.

Though the model is based on mathematical and physical principles, it is intended to be
aggregate, and is probably not appropriate for engineering or product development. Thus it is
designed to model a typical vehicle of technology type, rather than a specific vehicle. Energy
flows within vehicle components are modeled using overall  systems efficiencies. The validations
in this report are conducted with certification fuel economy data. Where appropriate,
simplifications and approximations are made using physical constraints, or based on publications
in the literature.

For the purposes of modeling the future fleet, our goal is to allow as many of the significant
assumptions as possible to be under the control of the user. However, default values will be
presented in this paper. An attempt will be made to justify the assumptions in each case.

The report begins  by describing conventional vehicles, both gasoline and diesel. It then goes on
to briefly examine advanced engines. Hybrid vehicles are modeled and validated followed by

-------
fuel cell vehicles. The report caps off with a sensitivity analysis and describes how PERE rates
might feed into MOVES.

The final form that the model takes for the integration may be different from what is presented in
this report.
Conventional Vehicles

Vehicle Model
As Figure 1 shows, the model is basically a "backwards-looking" model in that it takes a driving
cycle (second-by-second speed vs. time) input and outputs the energy consumption required to
follow the drive trace. The power demand is based on overcoming inertia, road grade, tire
friction, and aerodynamic loss. This is essentially the VSP equation shown in numerous
publications.
    ACTIVITY
   INPUT V,A,
    Road grade
  Vehicle Param:
    M, Cr, CdA
VSP or
Power Demand
^

Internal
Combustion Engine
                                                                       Fuel
                                                                   Consumption
Figure 1.Conventional internal combustion engine vehicle model flow.

The power demand (in Watts) is the brake power or, VSP*m:
where:
           Pb = VSP*m = mv[a(l+e) + g*grade + gCR] + 0.5pCDArv3

v: is vehicle speed (assuming no headwind) in m/s (or mps)
a: is vehicle acceleration in m/s2
e: is mass factor accounting for the rotational masses (~0.1)
g: is acceleration due to gravity (9.8 m/s2)
grade: is road grade
CR: is rolling resistance (-0.009)
p: is air density (-1.2 kg/m3)
CD: is aerodynamic drag coefficient (-0.30)
Ar:  is the frontal area in meters2 (-2.4 m2)
m: is vehicle mass in metric tonnes.
                                                                                    (1)
The rotating mass term, £, is assumed to be 0.1 [Jimenez, 1999]. However it increases at lower
gears [Gillespie, 1992]. The rolling resistance for radial tires, CR, can range from 0.008 - 0.013
for a majority of the on- road passenger car tires. This value can be larger depending on state of
inflation, temperature, ground surface type, and speed (though this effect is minor) [Bosch, 2000;
Gillespie, 1992]. Heavy-duty trucks tend to have lower resistance. The aerodyanamic drag, CD,

-------
varies according to the body shape, and is sometimes provided by the manufacturers. The frontal
area, Ar, of the vehicle is also sometimes supplied. Where it is unknown, PERE uses 93% of the
box frontal area based on a number of vehicles examined by the author. This is calculated using
the dimensions available for the vehicles:

                                      Ar = (H-GC)*W*0.93                            (2)
where: H is the vehicle height
       GC is the ground clearance
       W is the width.

Alternately to Equation 1, one could also use:

                               Pb = Av + Bv2 + Cv3 + mv[a + g*grade]                 (3)

where the coastdown coefficients: A, B, C, are rolling, rotating and aerodynamic resistive
coefficients, respectively. These are determined from track coast-down tests and are available
from the EPA certification database. The "A" coefficient is equivalent to the tire rolling
resistance terms of Equation 1. "B" tends to be small, and describes higher order rolling
resistance factors in addition to rotating friction losses. The "C" term represents the air drag
coefficient terms. These two equations are not necessarily identical, and it is possible that by
preferring Equation 1 over 3, that PERE could overestimate fuel economy (underestimate energy
consumption) somewhat. When available, the coastdown (track) coefficients should be used in
Equation 2, since the road load is closer to the test conditions. This will be discussed in greater
detail in the  sensitivity section.
Gasoline Spark Ignited Internal Combustion Engine
The internal combustion engines (ICE) converts chemical into mechanical energy. This
combustion process, as well as the losses inherent to it, is a critical element to the modeling of
vehicles. Modern vehicles powered with gasoline have been the subject of several models in the
literature. This paper remains consistent with the approach developed by Ross and An [1993];
Earth et al. [1999]; and Weiss et al. [2000] as well as other authors. The formalism is described
well in Ross [1997], which will be briefly reviewed here.

The basis for the engine model lies in the linear relationship between brake power and fuel
consumption, both represented in mean effective pressure (mep in bar) units. The power
equivalent of fuel is:

                                     Pf = FR*LHV                                   (4)

where FR is the fuel consumption rate in grams per second
       LHV is the lower heating value of the fuel (approximately 43.7 kJ/g for gasoline)

In mep [kPa], this becomes:

                               fuel mep = 2000*Pf/(VN)                               (5)

-------
where  V is the engine displacement in liters
       N is the engine speed (rps)

One can model it as follows:
                               Fuel mep = k + bmep/77                                 (6)
or
                               fuel mep = (fmep + bmep)/77                             (7)

where k is the engine friction related term (in friction mep terms, k = fmep/rj) and 77 is a measure
of the indicated or thermal efficiency of the engine, i.e., it is the fraction of the energy in the fuel
which is converted to useful work. This efficiency describes engine properties such as
compression ratio, fuel mixing, fuel type, cylinder temperature, valve timing, spark timing,
combustion chamber geometry, etc. [Muranaka, 1987; Ross, 1997]. Ideally, this is limited by
engine heat cycle and thermodynamic 2nd law losses at the upper limit.

The indicated thermal efficiency is not to be confused with "mechanical" efficiency, or even
with "overall" engine efficiency (discussed below), k and fmep reflect the mechanical losses in
the  engine mainly from rubbing, pumping, and auxiliary load losses. Specifically, the rubbing
losses can originate from valves, gears, piston rings, lubricant, piston, crankshaft, etc [Milington
& Hartles, 1968]. The pumping losses are derived from the difference between the intake
cylinder and ambient pressures. The throttle plate, and intake, exhaust and valve geometry on
gasoline engines cause this loss in "volumetric efficiency" [Patton, et al., 1989]. As an engine
"breathes" better, the pumping losses decrease and its volumetric efficiency increases. Thus an
engine with 4 valves per cylinder is typically more fuel  efficient than one with 2 valves. Also
diesel engines lack throttle plates, so they have improved volumetric efficiency over gasoline
engines. Throttling results in about 25% of the friction losses [Ross, 1997]. The product of
mechanical efficiency (which we have not quantified yet) and thermal efficiency is the overall
engine efficiency, which will be discussed in more detail below.

Given the fuel mep, to obtain a mass fuel rate, one only needs engine displacement and speed.
The bmep, is equivalent to brake power and can be determined from road load using the
ubiquitous road load Equations 1 & 3. Equation 7 is equivalent to the one used by Weiss , et al.
[2000]:

                                      imep = fmep + bmep                             (8)

where imep is the indicated mep equivalent to fuel mep*77.

This relationship has been plotted on Figure 2 below for a series of 10 different engines at many
different operating conditions, where the wide-open throttle points have been omitted [Nam &
Sorab, 2004]. According to the figure the indicated or thermal efficiency (inverse slope) is 77 ~
0.40, while k ~ 4.24 bar and fmep ~ 1.72 bar for this diverse family of engines. This procedure
for  estimating the engine friction analytically is referred to as the "Willans Line" method
[Millington & Hartles, 1968]. It is similar to experimentally motoring the engine down. The
largest uncertainty associated with this methodology lies in the slight curvature of the lines

-------
especially at high load. This is why the high load points were omitted, though there is still some
curvature [Pachernegg, 1969]. Despite the limitations, the method is quite robust as the lack of
scatter in Figure 2 demonstrates. For the scale of modeling in this report, this estimate is quite
sufficient.

It has been found that the indicated efficiency has not changed significantly over the preceding 3
decades [Nam & Sorab, 2004, Weiss, et al. 2000]. It should be noted that this holds for engines
operating at stoichiometry, where there is exactly the right quantity of air to combust the fuel
completely. Older vehicles can run rich under various high power driving conditions, and those
need to be modeled separately [Goodwin,  1996]. It is assumed that present and future vehicles
(which have to meet SFTP standards) will not undergo significant periods of enrichment.
However future vehicles may have improved efficiency if lean burn engines (or other advanced
technologies) become more prevalent (this will be discussed more later).
  Q.
  CD
     -2
                            bmep (bar)
Figure 2. Fuel mean effective pressure as a function of brake mep for 10 engines from 4 different
manufacturers, omitting wide open throttle points [Nam & Sorab 2004].

It was mentioned earlier that "indicated thermal" efficiency is NOT the same as "overall" engine
efficiency. The latter includes the effect of mechanical or frictional losses. This can be
demonstrated most dramatically in Figure 2. The overall (or "brake") efficiency of the engine at
a given operating (load) point, is the x-axis value (output energy) divided by the y-axis value
(input energy). For example, at bmep = 2 bar, the fuel mep =10 bar, meaning the overall engine
efficiency is 20%. Also at bmep = 8 bar, fuel mep = 24 bar, meaning the engine efficiency is
33%. It is clear that the overall  efficiency increases with the load (omitting wide-open throttle
and fuel enrichment) while the  indicated thermal efficiency (inverse slope) remains constant at
roughly 40%. Thus the indicated efficiency is the efficiency limit of the engine. The product of
the indicated and mechanical efficiency is the overall engine efficiency. Thus the mechanical
efficiency is dependent on load and engine speed, and the most efficient operating points are
those with low engine speed and high load. This highlights the advantage that hybrids have over
conventional vehicles: the engine operates more at higher loads and thus higher efficiencies,
while the battery (or other energy storage medium) operates during low engine efficiency modes

-------
(allowing it to shut down). To calculate a final overall vehicle (or "pump-to-wheel" efficiency, it
is also necessary to factor in transmissions and final drive losses. Accessories can also play a
minor role.

Though the scatter in Figure 2 is small, there is some systematic relationship, which can explain
some of the variation. It is well known that the friction (or offset) is dependent on the engine
speed. This is demonstrated in Figure 3. The base friction term can thus be modeled
                                      k = k0 + kiN
(9)
where: k0 = 3.283 bar and ki = 0.000515 bar/rpm for the present gasoline engines (multiply by
100 to get kPa, then divide by 2000 to get the proper units for the fuel rate equation). The friction
at idle is significantly higher (about 50%). There is a slight curvature evident in the figure and
some references include higher order terms, but these are more necessary for higher revving
engines [Yagi, et al., 1990].
            1000
                   2000
                          3000
                           rpm
                                 4000
                                        5000
                                               6000
Figure 3. Engine friction as a function of engine speed for 10 engines. [Nam & Sorab 2004].

Though the indicated efficiency has changed little over the years, the friction has decreased over
the past 30 years at a rate of roughly 10% per decade [Nam and Sorab, 2004; Sandoval and
Heywood, 2003]. Various technologies have helped accomplish this: overhead cam, multiple
valves per cylinder, improved lubricants, variable valve timing, etc. Additionally, engine
efficiency can be improved with "advanced" technologies such as lean burn gasoline (e.g.,
Honda HX), Atkinson cycle (as in Toyota Prius), Direct Injection Spark Ignition (DISI - no
current model in US), Homogeneous Charge Compression Ignition (HCCI), or with variable
compression ratio. Fuel economy can also be improved with variable displacement, cylinder
deactivation,  or adding a turbocharger with engine downsizing in the future. There may come a
day when engines will be "variable everything". Some of these technologies will be discussed
further below.

Despite the many different models and manufacturers of gasoline engines (neglecting advanced
engines), there is remarkable homogeneity in their performance characteristics (within  a certain

-------
threshold). As a result, it is possible to develop "rules of thumb", which give rough estimates, of
engine efficiency, friction, peak torque and peak power, given only its displacement (and model
year). Figure 4 shows generic peak torque and power curves for a 2.0 liter gasoline engine
[Weiss et al., 2000]. The curves can be scaled to different engine sizes using relationships in
Sandoval and Heywood [2000]. While not all 2.0 liter engines will have the same shape and peak
values, the general trends appear in most engines and this relationship is sufficient for our
modeling needs. These peak curves are mainly used to determine transmission down-shift points
and for cut-point determination for hybrids (where engines may be significantly downsized),
therefore the PERE output is not highly sensitive to the shape of the curves.
ocn -r-
^ou
9nn
^uu
0) 1 en
3 I OU
E
o inn
J5 I UU
EID
OU
:

:
:
: i

» <


i

> » '
•
i

> » '
i
•
i



> ft '
i
i




• ,
•


•\ on


;

• torque (N-m)
• power (kW)
,
i i i i
;
I
, , , , Z
1 £.\J
100
80 ,_
§
60 |
40 °-
20
n
U
0 1000 2000 3000 4000 5000 6000 7000
rpm
Figure 4. Engine peak torque and power for a "generic" 2.0L gasoline engine.
Transmission
Another element required to calculate total vehicle efficiency is transmission and final drive
efficiency. All vehicle powerplants must connect to the wheels via a transmission. Sometimes
this is as simple as a single gear reduction (as in many fuel cell and electric vehicles). More often
though, there are multiple gears to take advantage of the engine's limited operating range. The
multi-gear transmission model used for PERE is based on Thomas and Ross [1997]. While
transmissions and their shift strategies vary tremendously between vehicles, the overall fuel
consumption rate is not highly sensitive to this sub-model as long as 'sensible' parameters are
employed.
The engine speed is:
                              N = (N/v)top*(60rps/rpm)*(g/gtop)*v
(10)
where  N is engine speed in rps
       (N/v)top is the rpm to speed ratio in top gear
       v is vehicle speed in mph
       (g/gtop) is the gear ratio at the various gears

Typical values (with shift points) are shown in tables in Appendix B. (N/v)top is assumed to be a
constant, but a future version of PERE may make it dependent on engine size. Smaller engines
                                           10

-------
tend to rev higher than larger engines. One could also calculate engine speed from final drive
ratio and tire radius.

In order to meet the power demand of some driving cycles, it is necessary to downshift. The
algorithm  for downshifting is as follows:

If TorqueDemand > MaxTorque
       Then Downshift

The algorithm is repeated until the Torque demand is met, or gear 1, whichever comes first.
There is no consideration for shift or ride quality in this simplified model.

All  the multi-gear vehicles in this study are assumed to be 5-speed manual transmission for
simplification. The overall efficiency used by PERE for a manual transmission is approximately:
fltrans = 0.88, which includes final drive losses [Thomas & Ross, 1997; Weiss et al., 2000].
Typical auto transmission efficiency would be about 75% using a single value. There is
additional  significant loss in automatic transmissions during gear shifts, due to the slippage in the
torque converter. The model could be slightly improved by making the efficiency dependent on
load and/or rpm [Park, et al.,  1996].

Manual transmissions range in efficiency from 87-99%. Automatic transmissions range 85 -
95% when locked, but can drop to 60 - 85% unlocked. An overall 1.5% improvement in
transmission efficiency could correspond to a 0.1 km/L increase in fuel economy [Greenbaum, et
al.,  1994; Kluger, et al., 1995; Bishop, et al., 1996]. Manual transmissions have similar
efficiencies to continuously variable transmissions (CVT), so can be seen as equivalent to
advanced transmissions.

Vehicles whose drivetrains run off of motors do not require complex transmissions employing
gear shifts. These motor  driven vehicles usually only require a single gear due to the large
operating range of motors. Single gear transmissions naturally tend to be very efficient and are
assumed to be 95% efficient in this report.

If transmissions efficiency is  added to Equations 4 through 6, we would obtain the fuel rate
equation used in PERE [Nam, 2003]:

                       FR = 9 [kNV + (Pb/Tit + Pacc)/Ti] / LHV                         (11)
where
       (p:     is the fuel air equivalence ratio (mostly =1)
       k:     is the engine friction (can depend on engine speed).
       N:     is the engine speed, depending mostly on vehicle speed
       V:     is the engine displacement volume
       r|t:     is a transmission and final drive efficiency (-0.88)
       r|:     is a measure of the engine indicated efficiency (-0.4).
       Pace:  is the power draw of accessories such as air conditioning.  This is a function of
             ambient temperature, T, and humidity (and engine speed for AC). Without AC, it
             is some nominal value - 0.75kW.
                                           11

-------
       LHV:  is the factor lower heating value of the fuel (~43.7kJ/g).
Diesel Internal Combustion Engine
Diesel engines are different from gasoline engines in several important ways. They run on diesel
fuel, which has higher density, energy density, and carbon content. The fuel is self-ignited by
compression rather than with a spark (hence it is also known as compression ignition engine).
The increase in compression ratio required for compression ignition improves the thermal
efficiency. The engines can also combust fuel in an environment that is lean of stoichiometry. A
stoichiometric fuel air mixture has just enough air to completely combust (or  oxidize) all of the
fuel, so a lean mixture has more air, and rich mixture has more fuel. Naturally, lean mixtures
result in higher fuel efficiency. Modern diesel engines inject the fuel directly in the chamber with
electronically controlled bursts, which increases the pumping efficiency (decreases friction) by
doing away with the losses inherent to a throttle. Unfortunately, diesel engines have higher NOx
and particulate matter (PM) emissions. The excess NOx emissions is due to the lack of
aftertreatment system, and the PM emissions is  due to the combustion mechanism with fuel
droplets. Advanced diesel engines complying with federal Tier 2 emission standards will require
aftertreatment technologies, which may require the fuel combustion to run at stoichiometry on
occasion so that NOx and PM can be stored, then treated. This will in turn,  have a small negative
effect on fuel economy.

In PERE, Diesel indicated efficiency is assumed to be 0.45. This is 12.5% more efficient than
current gasoline engines [Wu & Ross 1997]. Weiss et al. [2000] employ an indicated efficiency
of 0.51 for advanced diesel engines.

The friction in diesel  engines is typically lower than their gasoline counterparts [Heywood, 1988,
Weiss, et al., 2000, Wu&Ross 1997] - though this is mostly in the speed independent term (ko in
Eq. 8). However, Ball et al.  [1986] measured a diesel engine to have similar (total) friction
compared to an equivalent gasoline engine at higher operating conditions. In a diesel engine, the
pumping losses are significantly lower, but the piston/crank assembly as well as auxiliary load
(mainly from high pressure fuel  injector) losses are higher. This hints that the speed dependent
term (ki) should be higher for diesels, or that higher order terms in Equation 9 are necessary. For
PERE, we assume that ko  is 25% lower, which is consistent with many of the  references above,
but we also assume that ki is higher (-.00072 bar/rpm) to match overall friction at higher speeds.
Whatever the particular values chosen for ki, this is a small correction to the output fuel
consumption rate (see Figures 2 & 3). The output is not be very sensitive to this term.

The transmission model for diesel powertrains is identical to that described above, with the
exception that diesel engines tend to rev lower.  The (N/v)t0p is estimated to  be 75% that of
gasoline. This figure is obtained by comparing peak rpm power values  from an assortment of
vehicles, which have  a diesel option (US and European). E.g. VW Jetta, VW Beetle, Ford Focus,
Ford Mondeo, & Peugeot 607. See Table 1.

The peak torque and power curves are genetically determined by Weiss et al.  [2000]. Figure 5
shows the relationship for a 2.0L turbocharged diesel. The torque curves are flatter than that of
gasoline engines. This high low-end torque gives diesel engines the advantage in towing as well
                                           12

-------
as decent 0-30 mph acceleration compared to their naturally aspirated gasoline equivalents (of
same displacement). However the peak power tends to be lower, thus the overall 0-60 mph
acceleration may be lower depending on the vehicle.
ann
OUU
9^0
£.-j\j
9DD

•



* i




. '
! *
4


4 nn
i • :
>
» :
1
• torque (N-m)
• power (kW)


0 1000 2000 3000 4000
rpm

:

1 UU
80
60 fc
40 S.
20
t n
5000
Figure 5. The peak torque and power curves for a "generic" 2.0L diesel engine.

Diesel engines are ideally suited for turbo charging. A turbo charger employs the exhaust heat
energy to spin a turbine, which pumps more air charge into the engine cylinders. This efficient
use of energy generates more power with a smaller engine. As a result virtually all light and
heavy duty diesels employ turbo chargers. Unfortunately, the variety of turbocharger sizes makes
generating peak torque and power curves, given only engine displacement, a difficult exercise.
This would matter more if we were constructing the model using a performance basis (e.g. 0-60
mph time), rather than a fuel economy basis.

It is useful to conduct a comparison of gasoline and diesel light duty vehicles currently sold.
Comparing the current technologies can help us to extrapolate how they will compare in the
future. Table 1 shows the power and weight of various vehicles in the American and European
markets.

Based on the table, we note that diesel engines tend to be heavier than their gasoline counterparts
(to accommodate higher pressures), thus they often need to be downsized in order to  fit into the
same frame. According to the ratios of these same vehicles, the turbocharged diesel engines
should be approximately 5% smaller. Due to the many differences between gasoline and diesel
engines, it is impossible to do a perfect apples-to-apples comparison. However, based on these
vehicles, the equivalent vehicle weight  of a diesel is assumed to be 4% higher than the gasoline
vehicle.

Table 1. Diesel vehicles used for comparison
                                           13

-------
model
Ford Focus
Diesel
Ford Focus
Gasoline
Ford Mondeo
Diesel
Ford Mondeo
Gasoline
peugot 607
diesel
peugot 607
aasoline
VW Beetle
Diesel
VW Beetle
Gasoline
VWJetta
Diesel
VW Jetta
Gasoline
max torque
200N-m@
2000rom
160Nm@
4400rom
245N-m@
1900rom
190Nm@
4500rom
250N-m@
1750rom
217Nm@
3900rom
210N-m@
1900rom
172Nm@
3200rom
240Nm@
1800rpm
165Nm@
2600rpm
max power
66kW@
4000ram
85kW@
5500rom
85kW@
4000rom
107kW@
6000mm
79kW@
4000ram
116kW@
5650rom
66kW@
3750rom
85kW@
5400rpm
75kW@
4000rpm
86kW@
5200rpm
displace
ment (h
1.753
1.8
1.998
1.999
1.997
2.231
1.896
1.984
1.9
2.0
curb weight
(kci)
1258
1200
1498
1376
1500
1455
1270
1222
1347
1331
* All sources are either from manufacturer website, Car&Driver, or Road&Track
Advanced Internal Combustion Engine (ICE)

In this report "advanced" technology is defined as a vehicle (or component) that is improved
over those in most vehicles currently. Advanced ICE vehicles might employ one or more non-
standard engine or driveline technologies meant to improve thermal, and/or driveline
efficiencies, in order to improve fuel economy and performance. Examples include: lean-burn
gasoline engines, variable displacement, direct injection gasoline, and continuously variable
transmission (CVT).

In order to model future vehicles, it is educational to study current technologies comparing them
with their predecessors.  Two areas which have seen marked improvement in the past, are engine
friction and power. The  former likely leading to the latter. Typically, improvements in (overall)
engine efficiency can either improve fuel economy, or increase engine size in order to increase
power performance (thus maintaining the same fuel economy). The latter has been the main
trend over the past few decades, though there is evidence to show that specific power (power per
unit displacement) has also improved [Chon and Heywood, 2000].

Based on the research of Chon and Heywood [2000] the relationship between engine power and
model year has been established for the past 15 years. This trend has been projected forward by
Weiss et al. [2000] and is shown in Figure 6. The friction reduction projections have also been
added to the figure, based on the research of Sandoval and Heywood [2003] and Nam and Sorab
[2004]. Advanced diesel trends are assumed to be the same, though their starting values may be
different.
                                          14

-------
     250
                 -frnep (kPa)
           • peak torque (Nm)
                      peak power (kW)
        1990
2000
   2010
Model Year
2020
2030
Figure 6. Projected gasoline engine trends over time [Chon & Heywood, 2000; Nam & Sorab,
2004].

Beyond incremental improvements in engine friction, it is difficult to predict how engines will
change in the next 30 years. It is conceivable that production gasoline engines can have
improved indicated efficiencies. This could be accomplished with the trend of "variable
everything". The Atkinson cycle engine (as in the current Toyota Prius) is a good example of
variable valve timing being used to increase efficiency at the cost of power. The lean burn
engines of Honda (currently installed in the Civic HX, Hybrid, and Insight models) improve
efficiency by burning gasoline under lean conditions at low loads.

The Honda Lean burn engine improves the best bsfc (best brake specific fuel consumption, or
"sweet spot") over conventional engines by approximately 20%, and decreases friction by 8%
[Ogawa, et al., 2003]. PERE models this with an indicated efficiency of-0.48, with an 8%
friction improvement. During higher engine speeds, the engine reverts to stoichiometric
operation, where the engine is still more efficient than conventional engines (0.44). The engine
speed breakpoint is not provided in the reference, so in PERE it is assumed to be 50% of the
peak power rpm (6100 rpm*0.5).

There is less information on the Toyota Atkinson engine. PERE assumes that the efficiency is
15% greater than that of conventional engines, until the rpm cut point is reached [Heywood,
1988]. The peak torque values are decreased correspondingly. Beyond that point, it behaves as a
conventional engine. This is similar to the lean burn engine.

Gasoline Direct Injection (GDI) or Homogeneously Charged Compression Ignition (HCCI)
engines are also a promising alternative, though these are still mainly laboratory engines.  In the
near future, there may be engines with variable displacement, variable compression ratio,
cylinder deactivation, etc. coming on to the scene. Moreover the current  move toward 42  Volt
power systems on the vehicle will allow for Integrated Starter Generator systems to  start and stop
the car during idle.  This has been referred to as "minimal hybrid" in some references [An and
Santini, 2004].
                                          15

-------
In addition to technical advances in engine design, it is possible that tire rolling resistance and
vehicle aerodynamics can improve in the future. Currently passenger cars have tire rolling
resistance of approximately 0.009. It is unlikely that these will improve significantly without
sacrificing other performance measures. The aerodynamic resistance and vehicle frontal area of
vehicles has room for improvement, but here, form tends to follow function, or design. Current
cars have coefficients of drag approximately 0.30 - 0.35 (Toyota Camry Cd = 0.30). Vans and
SUVs have Cd ~ 0.35 - 0.40 and pickups range from 0.40 - 0.45 [Gillespie, 1992]. The
exceptional 2004 Toyota Prius has a Cd of 0.26, and the Honda Insight has the best in class
coefficient of 0.25.
Hybrid Vehicles

Currently the most successful technology implemented commercially to improve the overall
efficiency of vehicles is to hybridize their drivetrains, by adding an electric propulsion path to
the wheels.

There are a number of electric hybrid vehicles on the road today: The Honda Insight was the first
to be introduced in the United States in 1999. It was followed by the Toyota Prius and the Honda
Civic (the Prius was first sold in Japan in 1997). Other manufacturers have also announced that
they will release hybrid models of various kinds within the next few years.

There are several different kinds of hybrid vehicles under development. All share the  common
trait that there are two power sources that drive the vehicle forward during different operating
modes. Most methods hybridize an internal combustion engine with an energy storage device
such as a battery, ultracapacitor, or even hydraulic pump (fuel cell hybrids will be discussed in
the next section). The hybrid configuration is usually either series or parallel. Series hybrids run
off electric power only, and the batteries are recharged by the engine, which can run in an
efficient mode.  Series hybrids suffer from the disadvantage that both a large battery and motor
are required, while losses are incurred in both charging and discharging the battery. Parallel
hybrids  run the  drivetrain alternately with the engine, battery or both. This allows for  a smaller
battery/motor than the series hybrid and for the engine to run efficiently. The disadvantage
compared to the series configuration is that the system requires more sophisticated strategy,
packaging and transmission. Some models require a separate generator. The Honda hybrids are
parallel, while the Toyota Prius is a combination series/parallel.

The spectrum of hybrids soon to be on the road is broad. General Motors is proposing a
"minimal" hybrid version of its Sierra/Silverado truck, which is expected to improve fuel
economy by 12%. By 2007, they announced a displacement on demand (cylinder deactivation)
hybrid system for the Tahoe and Yukon line [GM, 2004]. Ford is set to release a full hybrid
version of the Escape compact SUV in 2004. Honda announced that a hybrid version  of the
Accord will be sold in 2004, which will have Variable Cylinder Management (VCM) in its V6
engine [Honda,  2004]. Toyota announced the Highlander hybrid in 2005, while the Lexus 400
hybrid is expected in 2004 [Toyota, 2004]. All of the hybrids mentioned above are in  (mostly)
parallel  configuration and store secondary energy in a battery. The Honda, Toyota, and GM
                                           16

-------
models also mainly use "advanced" gasoline engines. Many of the hybrids recharge the batteries
by employing regenerative energy from the brakes.

With the promise of hybrids hitting many of the markets segments in the near future, it is
imperative that any modeling of future vehicles include this class of vehicle. There have been a
number of models developed, which captures the behavior of hybrid vehicles. ADVISOR
(developed atNREL) is widely used in the research community [Kelly, 2001]. It models
conventional as well as hybrid and fuel cell vehicles. ADVISOR is a relatively sophisticated
model that can be run stand-alone or on the MATLAB® platform. Unfortunately, the model
could not easily be integrated into MOVES for several reasons. Namely, the intensive data and
effort required to calibrate the model is prohibitive for trying to model broad sections of the fleet
required. It was thus necessary to adjust the approach so that a model could be run on the PERE
spreadsheet. From there, it may either be run in spreadsheet mode, or programmed directly to
link with MOVES in the future. The basic framework for PERE is similar to ADVISOR though:
starting with a driving cycle input, calculate the road loads, then  distribute the power and losses
to the various energy conversion and storage media. Many other  models exist as well, but the
hybrid architecture in PERE is roughly based on the simple control logic described in Weiss et
al.,  [2000] at MIT.
Strategy
Figure 7 shows the PERE flow chart for the parallel hybrid design. The battery can be any
energy storage device (ultracapacitor or hydraulic). The logic control is relatively simple [Weiss
et al., 2000]. When power demand is less than the hybrid threshold (2kW in the MIT case), the
car runs on battery alone. Beyond the threshold, the car only runs on the gasoline engine. When
power demand is greater than the peak engine power, then the battery assists or provides the
needed boost. When braking, part of the energy is recaptured into the batteries. Each of the
power paths described has its unique losses. This strategy is modified slightly in PERE to require
that the hybrid threshold be set such that the state of charge on the battery is sustained over the
course of the Federal Test Procedure driving cycle (city and highway), i.e. the battery ends the
test with the same charge with which it started. Other differences in the approaches will be
described below. The parallel power and state of charge algorithms are shown in Appendix A.
                                           17

-------
    ACTIVITY
    INPUT V,A,
    Road grade
  Vehicle Param:
    M, Cr, CdA
                          VSPor
                       Power Demand
Transmission/
  Gearbox
                                                                                Fuel or Energy
                                                                                 Consumption
Figure 7. Parallel hybrid (gasoline/battery) flow chart.
Actual hybrid strategies can get much more sophisticated, where the battery can get recharged by
the engine directly, during operation (as is the case for the Toyota Prius THS - Toyota, 2003).
This system is a series/parallel architecture and is especially appropriate for stop and go driving.
However, the PERE methodology takes the basic approach  of distributing energy in the system
defined by the strategy above, and taking into account losses. Thus, whether the energy is stored
in the battery, or converted by the engine, the same overall energy is required, therefore the
overall predicted fuel economy should be similar to what is measured (within 10%). The main
omissions of PERE at this time is that it does not include a cold-start module. Cold start factors
will be inserted by MOVES, not PERE. Hybrid vehicle fuel consumption can be significantly
larger during cold start to heat the catalyst to  light-off temperatures and also to charge up the
batteries  or capacitors (if needed). Cold start  factors are discussed in greater detail in the
Sensitivity section.

Hybrid threshold was chosen in order to give close to net zero state of charge over the FTP and
HWY driving cycles. More explicitly, this usually means that the battery discharges during city
type driving, and recharges during highway type driving. This loosely simulates a "charge
sustaining" strategy ('pseudo-charge sustaining strategy'). For all the hybrids modeled in this
report, this threshold fell between 2 and 4 kW. Values are in Appendix B. This strategy implies
that the batteries are only being recharged through regenerative braking, and not the engine.
Motor/Generator/Inverter
The conversion between electrical and mechanical energy is done with a motor. In reverse this is
called a generator. For hybrids, we will use the term "motor" to indicate both devices. In PERE
the motor operates both forwards and backwards at different times depending on the driving,
                                            18

-------
with different losses. Weiss et al. [2000] estimates the motor system efficiency to be 76% with
an additional 15% loss during regenerative braking. The motor efficiency (alone) is likely higher,
approaching 90% for a permanent magnet brushless DC motor, though it depends on the size of
the motor [Laramie and Dicks, p. 279 2000, Ogawa, et al., 2003]. However, the MIT value
includes other system losses (such as from an inverter, which can have efficiencies of-94%).
Moreover, motor efficiency depends somewhat on the operating mode, but PERE also assumes a
fixed value for simplicity.

The motor power and torque curve is shown in Figure 8. It is scalable with peak power [Weiss et
al., 2000]. The motor in the model is connected to the transmission, not directly to the wheels
(except in the fuel cell).
60 kW Motor
94D -r an
9nn
/uu <
•^ -\Rr\
^
Q) 1 9D
E
O on
AC\
1

• • <
i


i

' •
i
i




i
•
i
»
»
<

• Torque (Nm) :
• Pnwpr tk\l\l}


1
•
1
>
•

_
\
1 .
:
*
cc
DO
Cf) 	 ,
^
oc'T"'
^ai
200.25. The 2004  Toyota Prius has a
rating of- 0.47. This definition does not necessarily imply that mild hybrids are worse, or less
efficient than full hybrids. In reality, however, the size of the motor is meaningless without a
battery (or ultracapacitor) that can manage the current flows in and out of it. The effect of
hybridization naturally depends on the hardware, and what functions are built in, moreso than the
ratio of any two numbers.
                                           19

-------
Energy Storage Devices - Batteries, Ultracapacitors and Hydraulics
Batteries on hybrid vehicles tend to be of the nickel metal hydride (NiMH) variety. They have
relatively high energy densities and can withstand many charge/discharge cycles, while being
affordable. Compared to a conventional vehicle, however, batteries add significant cost.

In PERE, each second the power draw (discharge) from the battery is adjusted by the discharge
efficiency (95%) as well as the motor efficiency. To be more accurate, this efficiency really
depends on the state of charge of the battery, but we will approximate it as constant. During
brake recharge, the rate includes a regenerative brake efficiency of 85% [Weiss et al., 2000].
There is added loss due to fact that front wheel drive hybrids only can recapture braking power
from the  front wheels (not the rear). Thus up to approximately 75% of the braking energy is
available for recapture [An and Santini, 2004]. However, the new Prius brake by wire system can
recapture significantly more [Toyota, 2003]. After the losses are included, the power is
integrated over each second of the driving cycle, and then the discharge (positive) and recharge
(negative) totals are added to give a final state of charge (SOC), as a  fraction of the kWhr rating.
The final state of charge over the FTP driving cycle is calibrated to be close to zero in our model.
Unfortunately, it may not be true for other driving cycles. A more sophisticated charge sustaining
strategy is a possibility for future version of PERE.

The battery is assumed to have a rating of 0.936 kWhr on the mild hybrid, the nominal rating for
the Honda Insight (the Civic is 0.864). For the full hybrid, the battery is assumed to be similar to
the '04 Prius, or 1.31kWhr. However, because we have a  'pseudo-charge sustaining strategy' and
the vehicle is not assumed to drive on any extremely aggressive driving cycles, this number does
not affect the model. In reality it is more important to define the limitations of the hybrid based
on the total electrical power generated from the battery/motor system. PERE simplifies this, by
placing emphasis on the motor system only. This assumes that manufacturers will  integrate a
"properly" sized battery for the motor. If PERE is required for high performance (0-60mph) runs,
this, and  other, portions of the model may require revision.

Replacing a battery with an ultracapacitor changes the model slightly. The discharge and
recharge  efficiencies would be different, as would its ability to recapture energy from higher
power braking events.

Hydraulic hybrids combine internal combustion engines with an hydraulic energy  storage device.
This storage device stores the energy in pressurized gases. The hydraulics would respond similar
to an ultracapacitor except that the link would be mechanical, rather than electrical. This would
increase recapture efficiency. The  EPA is developing this technology with industry partners
through cooperative research and development agreements, including implementation of the
technology on a number of prototype vehicles [Alson et al., 2004].

Options for including hydraulic and ultracapacitor hybrids in MOVES are to model these
different  energy storage devices as separate engine technologies (source bins), using PERE or
existing fuel consumption rates on these technologies [e.g. Alson et al., 2004]; or, to create a
"generic  hybrid" engine technology and modify the hybrid component of PERE to account for
hybrids with battery, hydraulic, or ultracapacitor storage devices.   These options are currently
under consideration.
                                           20

-------
Vehicle Weight and other Specifications
Hybrid vehicles are usually able to downsize the engine in order to save weight and fuel further,
however the added components of the motor, inverter, battery, and other connecting components
more than compensate for the engine weight loss. Previous studies have summed the weights of
the components based on power density in order to estimate a vehicle weight. This method has
limitations in that other structural and component weights may be omitted. This report uses a
different approach. The weights of existing hybrids are compared with their conventional vehicle
counterparts. Table 2 shows the weights of various hybrid vehicles. In particular the Honda Civic
Hybrid is compared with the Civic DX, and the Dodge Durango hybrid is compared with the
Durango 4WD. While the Durango is no longer planned for production, it is still useful to
compare specifications. The Prius and Insight obviously have no direct partner, so no direct
weight comparison could be done. The weight ratio of hybrid to non-hybrid is found to be 1.07,
or a 7% increase. This is based on the ratio of test weights, which is curb weight + 300 Ibs. The
diesel hybrid is assumed to have a 4% increase over its gasoline hybrid counterpart. As more
hybrid versions of vehicles are released, this mass increase ratio estimate can be improved.

Table 2: Hybrid Vehicle Specification in Comparison to Conventional (if exists). All sources are
from Manufacturer, Car&Driver or Road&Track websites.

;o
.a
><
Conventional
model
Honda Insight
Honda Civic
Hybrid
2003 Toyota
Prius
2004 Toyota
Prius
Dodge
Durango
Honda Civic
DX
Honda Civic
HX
Toyota Camry
SE
Dodge
Duranqo 4wd
max torque
89.5N-m@
2000rpm
118Nm@
43300rpm
111N-m@
4200rpm
111Nm@
4200rpm
305N-m
149Nm@
4500rpm
151Nm@
4500rpm
220Nm@
4000rpm
475Nm@
SOOOrpm
max
power
50kW@
5700rpm
63.4kW@
5700rpm
52kW@
4500rpm
57kW@
SOOOrpm
130.6kW
86kW@
6100rpm
87.3kW@
6100rpm
117kW@
5600rpm
186kW@
4200rpm
displace
ment (I)
0.995
1.339
1.497
1.497
3.9
1.668
1.668
2.4
5.9
curb weight
(kg)
853
1213
1254
1311
2389
1111
1111
1425
2251
Motor
Power
10.4kW@
SOOOrpm
10.kW@
SOOOrpm
33kW@
1540rpm
50kW@
1540rpm
66.4kW




Battery Capacity
(kWhr)
0.936
0.864
1.31
1.31
?




Many of the other vehicle attributes that aren't published had to be approximated. Often
coefficients of drag and weights are available, but rolling resistance is not. Frontal area is
approximated from the published dimensions (Equation 2) if a direct measurement is not
provided.

The table of hybrid coefficients are also in Appendix B. Many of the variables may have a model
year relationship into the future. E.g. component weights, Cr, Cd, Pace, motor efficiency,
regeneration braking efficiency, discharge/recharge efficiency, etc. However, most of these are
expected to drop somewhat and level off in the near future. The accessory term (Pace) is the only
                                           21

-------
term expected to rise in the next few years with the addition of more electrical loads (and
luxuries).
Accessories
Currently, PERE does not model air conditioning or heat in hybrid (or fuel cell vehicles).
Therefore the accessory loads are a relatively small portion of the power required (-0.75 kW).
However it is not insignificant, if the accessories were run completely on the batteries, there
would be less power remaining for accelerations. In addition it is expected that accessory power
will increase in the future.

In PERE, the accessories  are run off the battery when the battery is on a discharge cycle.
Otherwise it runs off the generator (engine). The algorithms are listed in the Appendix A.

Air conditioning and heating are potentially large loads for hybrid vehicles. From a fuel
economy reference, the accessory loading can be large. In MOVES, the effect of air conditioning
will be applied to the base AC-off energy consumption rates developed by PERE.
Model Calibration
Since the PERE hybrid strategy model is based on Weiss, et al. [2000], the model is 'calibrated'
to the model "MIT hybrid" vehicle. The MIT report gives most of the vehicle parameters
required for the model. Figure 9 compares the fuel economy results of PERE with the MIT
hybrid. Figure 10 compares the state of charge of the batteries following the FTP test. Since the
models are not identical, differences are expected. However, the fuel economies are an excellent
match. This indicates that the modeling methodologies are quite similar.
    90
    80
    70
    60
  O)
  Q.
  >, 50
  o

    40
    30
    20
    10
                                       DPERE
                                       • MIT
                Combined
FTP
HWY
                                           22

-------
Figure 9. Fuel economy of a model year 2020 hybrid car compared to MIT study [Weiss et al.,
2000].






0 -

0 002 -

-0 004 -
_n DDR -

• PERE
• MIT








discharging



FTP H\ \/Y

recharging


Figure 10. State of Battery compared to MIT study [2000].

The differences b/w MIT and PERE models are listed below:
1- PERE does not include bag 1 cold start. The MIT report does not mention if cold start effects
   are included in their model.
2- MIT transmission model is unknown, (minor effect)
3- Hybrid threshold is determined differently
4- Battery recharge model probably is not identical.
5- PERE adds additional loss from front wheel drive regenerative braking only.
6- Engine friction model is slightly different.
Validation Results
It is difficult to conduct validation results with real hybrid vehicles, when the model is only an
approximation of a hybrid vehicle. Actual hybrid vehicles have much more sophisticated control
strategies, as well as efficiencies that depend on various operating conditions. However, the
energy flows still follow a similar physical and logical progression. Given a power demand the
fuel consumption rate should be similar if the loss terms are characterized sufficiently well.

The validation is conducted on a number of vehicles (hybrid and non-hybrid), whose
specifications could be found on either the manfuacturer's, or a trade magazine website (such as
Car and Driver and Road and Track). All weights are test weights (curb weight plus SOOlbs). The
list of vehicles is on Table 3. The Honda Civic HX is the lean burn engine version of the Civic
series.

Table 3. Validation vehicles - all model years are 2004 unless otherwise specified.
                                           23

-------
    Mfr         Model
  Toyota      Camry
   VW       Jetta gas
   VW      Jetta Diesel
  Honda      Civic DX
  Honda      Civic HX
  Honda    Civic Hybrid
  Honda      Insight
  Toyota      Prius '01
  Toyota	Prius '04
The validations are conducted in comparison to unadjusted EPA certification fuel economy
numbers. The rated fuel economy numbers for production vehicles are adjusted based on real
world driving correction factors: adjusted FTP = unadjusted FTP*0.9 and adjusted HWY =
unadjusted HWY*0.78. PERE's output should be compared with the dynamometer measured
fuel economy, which are the unadjusted numbers, since PERE is modeling the physical loads on
the vehicle as imposed by the dynamometer directly. The combined fuel economy (miles per
gallon) is a weighted average 0.45 (FTP) and 0.55 (HWY), calculated in the following standard
fashion [Schaefer, 1994]:

                       Combined F.E. = 1/(0.55/FTP + 0.45/HWY).                    (12)

Figures 11, 12, and 13 show the validations for the vehicles (city, highway and combined
respectively). The model performs especially well for conventional vehicles. There were several
other conventional vehicles (light cars and trucks) validated, the Camry, Civic, and Jetta are the
only ones shown. All of the conventional vehicles matched rated fuel economy to within 5% (on
city and highway) with the exception of the Volkswagen Jetta (gasoline). There were also
several conventional vehicles validated to dynamometer tests in previous studies, using a similar,
but older version of PERE [Nam, 2003]. Although the Jetta (1.9L TDI) results were within 5%,
the reason for the discrepancy with the Jetta (gas) is unknown. For a small fraction of vehicles,
the fuel economy is not accurately captured by PERE. This is most likely due to many  different
factors unique to each make and model. It is also interesting to note that PERE slightly
overestimates fuel economy from the city for many of the vehicles. This is not too surprising
since cold start factors are not included.

The "advanced vehicle" validations begin with the Honda Civic HX, with the lean burn engine.
The predictions for this vehicle is very good (within 5%). This suggests that the powertrain
model is correctly modeled, based on the description of the lean burn engine by Ogawa et al.
[2003]. All of the advanced vehicles are accurately capture by PERE for highway driving (within
5%). With the exception of the Toyota Prius, the city estimates are also all within 10%. PERE
overpredicts the fuel economy on the Honda Hybrid on the city cycle by around 9%. This is
likely due, in part, to the lack of a cold start module in the present version of PERE. Hybrids are
especially expected to have significantly higher fuel consumption during the cold start first phase
(bag 1) of the FTP. This is due to the fact that the engine has to stay on in order to ensure that the
catalyst lights off in a timely fashion (after which it can start breaking down the criteria
pollutants). Many conventional engines also run slightly rich during the start up period. Presently
PERE shuts the engine down during decelerations and idle, which would tend to  significantly
                                          24

-------
under predict fuel consumption (over predict fuel economy) during start-up for hybrids, and to a
lesser extent conventional engines. Changing PERE so that the engine remains on for the first
two minutes would be relatively simple. Despite this, PERE still underpredicts city fuel economy
for the Prius by 12% and 18% for the 2001 and 2004 models respectively. This demonstrates
how different the strategy of the Prius is from the rudimentary strategy employed in PERE. The
Toyota Hybrid System (THS) is a series/parallel hybrid design. A powersplit device routes some
of the power from the engine to recharge the batteries, thus the motor is used much more
frequently than in parallel hybrid [Toyota, 2003] to supplement power. The hybrid  system can be
fine-tuned to give optimal fuel economy. The brake-by-wire system on the 2004 Prius also helps
to regain maximal energy from regenerative braking. In this way, the Prius can and does get
better fuel economy in the city than on the highway (a difficult condition for PERE to duplicate).
Moreover, the specifications (efficiency improvements) for the Prius Atkinson cycle engine are
unknown.
   0.000
          PERE  Toyota   VW
           MIT   Camry   Jetta
                         gas
 VW   Honda  Honda  Honda  Honda  Toyota  Toyota
Jetta  Civic DX Civic HX  Civic   Insight Prius '01 Prius '04
Diesel                 Hybrid
Figure 11. Fuel Economy validation for the FTP city (UDDS) compared to unadjusted EPA
figures.
                                          25

-------
   90.000


   80.000


   70.000

  5
  ieo.ooo
  ISD.OOO
  o
  o
 ^40.000
  ;so.ooo
   20.000
   10.000
    0.000
DPERE
• EPA unadjusted
           PERE   Toyota    VW    VW    Honda  Honda   Honda  Honda  Toyota  Toyota
            MIT   Camry   Jetta    Jetta   Civic DX Civic HX  Civic   Insight  Prius '01 Prius '04
                           gas   Diesel                  Hybrid

Figure 12. Fuel Economy validation for the FTP (FEWY) compared to unadjusted EPA figures.
   10.000
    0.000
           PERE   Toyota    VW    VW    Honda  Honda   Honda  Honda  Toyota  Toyota
            MIT   Camry   Jetta    Jetta   Civic DX Civic HX  Civic   Insight  Prius'01 Prius'04
                           gas   Diesel                  Hybrid

Figure 13. Fuel Economy validation for the FTP/HWY Combined compared to unadjusted EPA
figures.
                                             26

-------
It has been demonstrated that PERE can capture the fuel consumption behavior of gasoline,
diesel, and "generic" parallel hybrid vehicles. The vehicle parameters required from the user are
model year, weight, body type (size, which can be approximated), engine displacement, motor
power, and fuel type. With the exception of body type, these are the primary "source" inputs for
MOVES.

This simplification of hybrid control strategy modeling could result in under-predicting fuel
economy.  A more optimized control strategy would be to run the battery at low load and allow
the engine to recharge the battery at higher loads, where the engine can operate more efficiently
(as the Prius does). This was not done because it was difficult to put in spreadsheet format. If
PERE is coded, then a more sophisticated control algorithm could be added.
Fuel Cell Vehicles

While, there are several types of fuel cells, the hydrogen Proton Exchange Membrane (PEM)
fuel cell is the fuel cell of choice for vehicle applications. A fuel cell is a power source that runs
on hydrogen to make current across a Proton Exchange Membrane. Each cell produces a small
amount of current and (like a battery) these are stacked together to produce a relatively efficient
power pack. There are other types of fuel cells, but PEM cells are the most commonly researched
today. Most prototype vehicles are equipped with Hydrogen PEM stack fuel cells. Hydrogen is
the primary energy source, which produces energy when combined with oxygen. This reaction
produces water, which is the only tailpipe emissions. Some fuel cell systems use a reformer to
break hydrocarbons from liquid fuels into gaseous hydrogen, but this technology appears to be
falling out of favor for light duty vehicle use due to its size and relative inefficiency.

In PERE, fuel cell vehicles are modeled as "series hybrid" vehicles. This means that there is only
one mechanical path. However, there is still a parallel electrical power path, in that either the
battery or fuel cell can drive the motor. This model is similar to the one  proposed by Weiss et al.
[2000 & revised in 2003] (there is still a hybrid threshold required). The power flow chart is
shown on Figure 14. It is essentially the same as the hybrid model above, but the engine is
replaced with a fuel cell, and there is no transmission (only a single gear driven by the high
speed motor). The fuel cell system efficiency (from the same reference) is shown on Figure  15.
The fuel cell efficiency curve is a projection and reflects an optimistic estimate. The figure
shows the efficiency points (scaled to a  72 kW cell) including a 7th order polynomial fit for
PERE. The fit allows for ease of interpolation in a spreadsheet.

Again in this version of PERE, there is no cold start modeled, the effects of which can be
significant. We also use the battery, similar to a parallel hybrid, to 'launch' the vehicle below the
hybrid threshold (initial acceleration). This avoids the very low fuel cell system efficiencies at
low load (see Figure 15). The model is similar in structure to the hybrid presented above. The
"calibration" comparison to the MIT results is  shown in Figure 16. The fuel economy figures are
calculated in miles per gallon gasoline equivalent by converting the work equivalent hydrogen
into gasoline mass (with their respective lower heating values). This vehicle is found to have a
tank to wheel efficiency of 50%. The tank to wheel efficiency is simply  defined here as the
                                           27

-------
useful energy out/fuel energy in. The useful energy out is only the (positive) energy required for
the car to follow the driving cycle. It would be slightly lower if the accessories were included.
   ACTIVITY
   INPUT V A
   Road grade
  Vehicle Param:
   M, Cr, CdA
Figure 14. Power flow chart for the hydrogen hybrid fuel cell vehicle.
                             72kW Fuel Cell
               10      20     30      40     50
                            Net Power Output (kW)
60
70
80
                                                                                ,th
Figure 15. Fuel cell integrated system efficiency [Weiss et al, 2000, 2003] including a 7  order
polynomial fit.
                                            28

-------

1 An

1 on

"oT
S inn
£
>s
E sn
o °u
c
o
o
LU
— on
3
LL
>in
on
n























































DPER
• MIT









E

































































                Combined
FTP
HWY
Figure 16. Fuel economy comparison between PERE and Weiss et al [2003]. The units are in
miles per gallon gasoline equivalent.
Validation
It is difficult to conduct a validation of the fuel cell model since there isn't a tremendous amount
of data available for the prototype vehicles that exist. Some specifications have been collected
for a number of manufacturer prototypes. They are listed on Table 4 along with the conventional
vehicle body on which the vehicle is built (the Daimler Chrysler F-Cell vehicle weight is
unknown). The Honda FCX has a unique body, so it does not have a conventional vehicle
equivalent (though it does have a similar electric vehicle). The average increase in weight of the
first 3 vehicles compared to its equivalent conventional vehicle is 23%. This increase is expected
to be less for larger (medium and heavy  duty) vehicles.

Table 4. List of some fuel cell vehicles. (References are all from information sheets distributed
during ride&drives by manufacturer representatives, or from website)
                                           29

-------

 n pj
c
0)
'o n =;
£ °'5
LLI
- n 4
E u-4
»
> 03
w ""*
;5j n o
O V.£.
~S>
3 0 1
4
n

AA
A

A

t

i






L A






A ,






^ A







A






L A






A .




0 10 20 30 40 50 60 70 80
Net Power Output (kW)
Figure 17. Fuel Cell System efficiency for an 80kW stack [Nelson, 2003].
                                           30

-------
Validations can only be conducted on the Honda PCX, since this is the only vehicle whose fuel
economy equivalent numbers are known.

In order to model the Honda PCX, some of the energy storage parameters had to be adjusted
since the vehicle is equipped with an ultracapacitor, rather than a battery and does not have a
DC-DC converter. The motor system, and recharge efficiencies are higher. The Honda PCX has
an adjusted EPA fuel economy rating of 51/46 miles per kilogram hydrogen (city/highway). A
kg of hydrogen is roughly equal to a gallon of gasoline in energy. In comparison, PERE's
predictions are shown on Figure 18. The discrepancy is within 5%.
    70
    60
    50
  O)
  O)
  Q.
    40
  E

  I 30
    20
    10
     0
               Combined                   FTP                     HWY

Figure 18. Fuel economy (in miles per gallon gasoline equivalent) for the Honda FCX.
Sensitivity

This section is devoted to a sensitivity analysis of PERE. The analysis determines which
parameters are most important to characterize accurately. A static sensitivity is performed in
relation to a change in city vs highway fuel consumption. It is necessary to do a separate analysis
since the parameters affect the results from two different driving types in a disparate fashion. The
simple static sensitivity test assumes a 10% variation in each parameter (though some of the
parameters are correlated). The change in the fuel consumption result is ranked and compared. In
order to perform a more complete uncertainty analysis, it is necessary to quantify the variation in
the parameters. This is a potentially difficult undertaking, since some of the parameters are
estimates (and not measurements). However, a more complete analysis will be performed in a
future report.

-------
Table 5 shows the sensitivity results, in % difference, for a typical modern passenger car
modeled using PERE. To get the relative sensitivity, divide by 10. TRLHP is the Track Road
Load Horsepower. It is a single quantity on which the (track) road load coefficients (A,B,C) are
based in MOVES. PERE does not typically use TRLHP, but MOVES does. As expected, the
parameters most affecting fuel consumption tend to be different for city vs highway driving. It is
not surprising though, that the efficiency, weight, and engine displacement all dominate the top
ranking. Fortunately, these quantities are typically known inputs for the model, or in the case of
indicated efficiency, should not vary much. For those advanced engines where the efficiency is
unknown, it clearly poses a problem for the model. The high ranking of the transmission
efficiency might support the need for a speed or load based transmissions model. Aside from
efficiency and N/v, however, PERE seems to be relatively insensitive to the other transmission
parameters. However, it is important to make sure that the gear ratios are not systematically off
from the actual gear ratios. It is interesting also to note that the model is quite insensitive the
peak torque curves. It would be more crucial if high  acceleration (full throttle) performance
modeling was being performed. Since the activity in MOVES describes representative "real-
world" driving, these driving modes are not stressed in the model, nor is it in PERE. Though it is
not included the table, the model was not sensitive (<1%) to the rotational term in Equation 1 (e).
                                        parameters used to model a typical conventional
Table 5. Static sensitivity test for the PERE
passenger car. Parameters are adjusted 10%.
RANK
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
Parameter
P/T indicated eff (eta)
Engine Displ (L)
trans eff
Vehicle wgt (kg)
kO (N indep friction kJ/l
N/v (rpm/mph)
TRLHP
Shift point 3-4
g/gtop 4
Shift point 2-3
Shift point 4-5
CrO (rolling resistance)
g/gtop 5
Cd (drag coeff)
A (frontal area mA2)
Nidle (rpm)
k1 (N dependent fric)
Pace (accessory - kW)
g/gtop 1
g/gtop 2
g/gtop 3
Shift point 1-2 (mph)
torque curve up 10%
Error City
4.93
4.57
4.46
3.98
3.72
3.31
1.95
1.31
1.26
1.24
1.11
1.05
0.96
0.93
0.93
0.90
0.85
0.52
0.38
0.22
0.16
0.04
0.03

RANK
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
Parameter
P/T indicated eff (eta)
trans eff
TRLHP
Engine Displ (L)
Cd (drag coeff)
A (frontal area mA2)
Vehicle wgt (kg)
kO (N indep friction kJ/l
Nidle (rpm)
g/gtop 2
g/gtop 1
Shift point 2-3
Shift point 1-2 (mph)
N/v (rpm/mph)
g/gtop 3
CrO (rolling resistance)
Shift point 4-5
Shift point 3-4
k1 (N dependent fric)
g/gtop 4
Pace (accessory - kW)
g/gtop 5
torque curve up 10%
Error Hwy
5.92
5.55
5.14
3.49
3.11
3.11
3.01
2.75
2.27
2.25
2.25
2.22
2.21
1.89
1.88
1.88
1.46
0.91
0.74
0.67
0.40
0.23
0.00
The sensitivity results for a "typical" full hybrid are presented in Table 6. Here the motor is
power split such that it can bypass the transmission. This improves the efficiency of the vehicle.
This is a fictional hybrid vehicle designed using PERE only, but assuming the same body and
                                           32

-------
overall power as the previous vehicle. The transmission parameters have been omitted from this
analysis. The hybrid threshold was adjusted to maintain the pseudo-charge sustaining strategy
with each run. It is clear from this table that the engine parameters are still more important for
fuel economy modeling compared to the electric parameters. Since the small engine is operating
at much higher loads, the model is much more sensitive to changes than the conventional
vehicle. Again, since the vehicle is able to meet the drive cycle (sometimes with the help of the
motor), the battery limit parameters are not important for the model.

                                        parameters used to model a fictional full electric
Table 6
hybrid
RANK
1
2
3
4
5
6
7
8
9
10
11
12
13
14
. Static sensitivity test for the PER
)assenger car.
Parameter
P/T indicated eff (eta)
Vehicle wgt (kg)
Engine Displ (L)
overall motor efficiency
CrO (rolling resistance)
Cd (drag coeff)
A (frontal area mA2)
torq curve up 10%
Regen Brake Eff
FWD power frac
Motor peak power (kW)
fmepO (N indep friction k,
Pace (accessory - kW)
fmepl (N dependent fric)
Error City
8.11
7.18
5.42
3.38
3.32
3.12
3.12
2.96
1.99
1.99
1.08
1.01
0.73
0.39
RANK
1
2
3
4
5
6
7
8
9
10
11
12
13
14
Parameter
P/T indicated eff (eta)
Cd (drag coeff)
A (frontal area mA2)
Vehicle wgt (kg)
CrO (rolling resistance)
Engine Displ (L)
fmepO (N indep friction k,
overall motor efficiency
fmepl (N dependent fric)
Pace (accessory - kW)
Regen Brake Eff
FWD power frac
torq curve up 10%
Motor peak power (kW)
Error Hwy
8.52
3.47
3.47
2.35
2.21
2.08
1.81
0.61
0.49
0.44
0.30
0.30
0.24
0.05
A sensitivity analysis is not conducted on the fuel cell hybrid, however, it is obvious from the
above two studies that the fuel cell system efficiency curve is the most important parameter. The
fuel cell peak power also being of moderate importance.
Road Load and Track Coefficients
Finally a brief comparison was run between using the road load coefficients Cr, Cd, A (area), etc,
vs using the track (or target) coast-down coefficients (A, B, C). The fuel consumption was
compared using both techniques for the conventional and hybrid vehicles. This is essentially a
sensitivity analysis of the B term. It was found that the choice of methods made little difference
for city fuel consumption results, but were significant for highway. For the city, the fuel
consumption was underpredicted by about 2% (or less). For the highway cycle, the fuel
consumption was underpredicted by 5% for conventional vehicles and 8% for hybrids on
average. This implies that at high speed, the load curves from the estimated coefficients tend to
be lower than those measured on the track.  This could be due to the lack of a first order speed
dependent term (B in force units), which tends to be quite small (or sometimes negative) on track
tests. Additionally, there could be higher order rolling resistance terms  [Gillespie,  1992]. It may
be further aggravated in hybrids because of motor drag and their lack of a true neutral gear.

With production vehicles, the issue of choosing the method of calculating road loads is simple,
since most of the track A, B,  C coefficients are publicly available. Unfortunately, this cannot be
done with vehicles, which have not yet been produced. It is impractical to combine Equations 1
                                           33

-------
and 3 in order to back-derive physical quantites from track A, B, and C coefficients, since the
values are determined from empirical fits to coastdown data. Thus, it is necessary to have a
methodology for estimating the effect of the B coefficients for future vehicles. Two methods of
estimating its magnitude are compared here. The first method involves fitting to the known
parameters.

The following coast-down force equations are based on Equations 1 and 3:

                              FRL = mgCR + 0.5pCDArv2                              (13)

                              FT = A + Bv + Cv2                                     (14)

Where the subscripts "RL" and "T" stand for "Road load" and "Track" respectively.
The equations are compared for the validation vehicles in the study. Then a B "correction" term
is added into Equation 13 and empirically fitted to 14.

                       FRL = mgCR + Bv + 0.5pCDArv2                                (15)

The second method is to use the higher order rolling resistance term from Gillespie [1992]
modified slightly here:

                              CR = CRO*(l+v/44.7)                                   (16)

Where CRois the base rolling resistance (typically -.009). An estimate of B is

                              B'= MgCRO/44.7   [N/mps]                             (17)

The speed dependent parameter in the denominator is not precise. It is estimated to double the
rolling resistance factor at 100 mph. In other references, the speed dependent rolling resistance is
embedded in the C parameter [Petrushov, 1997]. Unfortunately, this approach does not take other
vehicle (non-tire) frictional factors into account.

Table 7 shows the fitted B coefficients along with their 1 standard deviation uncertainties. The
third column displays the percentage underprediction in fuel consumption from the use of the
road load Equation 1  compared to using the track Equation 3. The last column (B') shows the
estimated value of B using Gillespie's method. First it is clear that the B term has a significant
effect on fuel consumption. In all but the Camry, the difference is over 5% on the highway cycle.
Comparing the fitted  B with the estimated B', there is some agreement, but the trends break
down for certain vehicles (e.g. Camry). Unfortunately, the tires (and rolling resistance) used for
testing these vehicles were unknown. Moreover, the lack of correlation with some vehicles only
supports the notion that other factors are inherent to the B term, such as rotating friction in
bearings, transmission, and final drive, tire windage, motor drag etc. Because of the motor drag
(and other hybrid differences), the official fuel  economy numbers for certain hybrids may not
necessarily match what one would achieve in the real world [Rechtin, 2003].
                                             34

-------
Table 7. Fitted B coefficients and 1 standard deviation. CV = Conventional Vehicle, HEV =
Hybrid Electric Vehicle.

Civic DX
Camry
Jetta
Civic hybrid
Insight
Prius 01
avg CV
avgHEV
avg ttl
Bfit
(N/mps)
2.01
0.67
2.85
2.32
1.37
3.11
1.84
2.27
2.06
sigma
0.19
0.07
0.11
0.11
0.15
0.35
0.13
0.20
0.16
HWY Fuel
Cons Effect %
8.60
1.48
8.87
8.18
6.27
10.46
6.31
8.30
7.31
B1
2.44
3.09
2.89
2.37
1.73
2.13
2.81
2.08
2.44
This subject clearly requires further research. However, if a vehicle's A, B, C track coefficients
are unknown, PERE will assume Gillespie's methodology, until an improved methodology is
presented.

The subjects of dynamometer, track, and road load coefficients are complex. This brief study,
only begins to scratch at the surface. A more full discussion of this topic is beyond the scope of
this report.
Application to MOVES

PERE has been developed fundamentally to produce energy consumption and emission rates for
MOVES for vehicle categories (known as "source bins" in MOVES terminology) where in-use
data is insufficient to populate these rates directly. Thus PERE will produce these rates for most
of the advanced technology vehicles (the focus of this report). The potential for PERE to
compliment MOVES in this way is discussed in EPA [2002 -MOVES GHG emission analysis
report].

Presently PERE is a standalone spreadsheet model capable of modeling on a finer scale than
MOVES requires. PERE also has the potential to model a wider assortment of technologies than
MOVES currently has. For example, PERE modeled a fuel cell hybrid with an ultracapacitor
(above), but this particular technology is not modeled in MOVES. This flexibility allows for
more source bins to be filled in the future as the need arises. However, the development time for
each technology type using PERE is significant, so MOVES will model some advanced
technologies directly using energy and emissions rates measured from other studies (e.g.
hydraulic hybrids and some alternative fuels).

The process of how PERE results could be integrated into MOVES is the subject of this section.
The final procedure is still subject to change, but fundamentally consists of the following steps:
                                            35

-------
Step 1: Choose the Source Bin
The first step is to determine what kind of vehicle PERE will model. In MOVES, these are
defined by "source bins" defined by vehicle characteristics which differentiate energy
consumption and emissions.  For energy consumption, source bins have 5 dimensions: fuel type,
engine technology, model year group, engine size, and loaded weight [Beardsley et al., 2004].
An example of some of the combined fuel types and engine technologies to be included in
MOVES are in Table 8, along with the primary source of the energy consumption rates. PERE
will be used to fill a majority of these technologies, but not all. In other cases, energy rates will
be derived from already-published analyses. While this report focuses on light-duty, heavy-duty
applications will be modeled as well with a version of PERE currently under development based
on the principles presented in this report.

Table 8. Examples of advanced technologies, as defined by fuel types and engine technologies,to
be included in MOVES. "C" is conventional, "A" is advanced, and "1C" is internal combustion.
This list is not final.
   •  Gasoline conventional (1C)
   •  Gasoline Advanced 1C
   •  Gasoline Hybrid -CIC Mild
   •  Gasoline Hybrid -CIC Full
   •  Gasoline Hybrid -AIC Mild
   •  Gasoline Hybrid -AIC Full
   •  Diesel Fuel conventional (1C)
   •  Diesel Fuel Advanced 1C
   •  Diesel Fuel Hybrid -CIC Mild
   •  Diesel Fuel Hybrid -CIC Full
   •  Diesel Fuel Hybrid -AIC Mild
   •  Diesel Fuel Hybrid -AIC Full
   •  Compressed Natural Gas (CNG) conventional (1C)
   •  Liquid Propane Gas (LPG) conventional (1C)
   •  Ethanol (E85 or E95) conventional (1C)
   •  Methanol  (M85 or M95) conventional (1C)
   •  Gaseous Hydrogen Advanced 1C
   •  Hydrogen -Fuel Cell
   •  Hydrogen Hybrid -Fuel Cell
   •  Electricity electric only

Not all of the engine technologies are represented fully across the other 5 dimensions, for
example, it is highly unlikely that long haul trucks will be hybridized in the future, due to the
relatively minor effect it has on highway fuel economy compared to the already  efficient diesels.
However, hybridization is quite promising for buses. Due to the sheer number of tests required,
simplifications are necessary. Examples of these are described in greater detail below.

Step 2: Define the Vehicle Specifications
Based on the engine technology  source bin, the vehicle parameters such as weight, engine size,
vehicle shape etc,  can be defined. The weight and engine size values are simply the central value
                                            36

-------
of the bin. Thus PERE effectively models an "average" vehicle in the source bin. For the source
bins on the "edge" of the matrix (<2000 Ibs; >130,0001bs; <2.0L; and >5.0L), their results will
be extrapolated - see Step 6.

Since engine size, has limited meaning for hybrids and no meaning for fuel cells, it is necessary
to define a power surrogate for engine size. This has already been mentioned in this report [Chon
and Heywood, 2000]. Though it is not perfect, we will take the convention that motor power +
engine peak power = total peak power. The fuel cell definition of power is simpler since there is
only one motor and the battery and fuel cell should be sized optimally within the motor
limitations.

Because the source bins do not have a dimension for body type, the estimations for will be
estimated based on the weight. Lighter light duty weights tend to be (compact) passenger cars,
then midsize cars, luxury, compact pickups, SUVs, minivans on up to medium light duty trucks
(heavy-duty trucks are discussed in a separate report). The variation in body types will certainly
lead to variation (or uncertainty) in the emission rates.

Step 3: Run PERE on the FTP if Modeling a Hybrid
Once the vehicle specifications have been finalized for the source bin, the next step is to run the
model  over the FTP driving cycle (city and highway). This is to determine the hybrid threshold
for charge sustaining. It also gives the estimated fuel economy of the vehicle. If a hybrid is not
modeled, or if fuel economy is not required, this step can be skipped.

Step 4: Define the Driving Cycles
The output to PERE is second-by-second energy consumption, therefore the next step is to input
the driving cycles. The driving cycles input into PERE help determine the "binned" energy
consumption rates for MOVES (see Figure 1). Henceforth "binned" refers to operating mode bin
and not source bin, which has already been defined.

The binned energy consumption and emissions rates are dependent on the driving cycle input. It
is important to capture a representative sampling of real-world driving.  14 driving cycles for
light duty applications have been selected for this purpose from MOVES. These cycles represent
a broad spectrum of driving, from very low to very high speeds. The cycles and their average
speeds are shown on Table 9. When merged, the cycles run 6,981 seconds. It is also important to
artificially set the accelerations to zero in between the cycles since they do not all start from rest.

Table 9. The 14 MOVES light-duty driving cycles used as input to PERE.
                                             37

-------
CYCLE
FWYHI1
FWYAC
FWYD
FWYE
FWYF
FWYG
RAMP
ARTAB
ARTCD
ARTEF
NYCC
FWYHI2
FWYHI3
LOWSPEED1
avq spd
63.2
59.7
52.9
30.5
18.6
13.1
n/a
24.8
19.2
11.6
7.1
68.2
76.0
2.5
It should be noted that for hybrids, the change in driving cycles may not necessarily maintain the
battery state of charge.

Step 5: Run PERE and Bin Output
After PERE is run, the rates must be binned into the respective operating modes. These are the
17 VSP and speed operating mode bins as defined in Koupal [2003]. The binning procedure is
straightforward and can be programmed in any script.

Figure 19 shows a sample of predicted fuel consumption as a function of VSP for a fictional
hybrid passenger car. This is an indication of how an output from PERE would be translated into
an emissions rate in MOVES. The uncertainty bars are from PERE generated variations within a
bin, and do not adequately reflect the true uncertainty of the emission rates, which will be added
in future versions of PERE.
                                            38

-------
        2.0
        1.5
        1.0
         .5
      .
    O
    o
    LJJ
    O
    §  o.o
                                  1
B

                                                                   inn
                                                                 D
          N= 258  357 111  171  121  85  48   15 151  283 281  144 62  46  173 301  34
             .00    11.00   13.00   15.00  21.00  23.00   25.00   33.00   36.00
                1.00    12.00  14.00   16.00   22.00  24.00  26.00  35.00

           VSPBIN3

Figure 19. Hybrid fuel consumption as a function of VSP bin.

Step 6: Repeat for the Rest of the Source Bins
Rather than run PERE for every operating mode bin within every advanced technology source
bin, it is more efficient to extrapolate from one bin to another along a dimension. A dimension
could be weight or power (engine displacement), so the rates could be interpolated between the
endpoints (e.g. lightest and heaviest weight class. To demonstrate the validity of the linear
extrapolation technique, PERE was run on a conventional gasoline powered passenger car for a
variety of weights and engine displacements. The driving cycles employed were the city and
highway portions of the FTP. The base car is a 35001b, 2.5L passenger car. For the different
engine sizes, the VSP values are identical, however, this is not true along the weight axis. Here
the second-by-second fuel consumption as well as the VSP values are binned.

Figure 20 shows the fuel consumption rates as a function of VSP bin. Each color represents a
separate engine displacement, as labelled along the right. It is evident that given the minimum
and maximum points, that the other could be linearly interpolated.
                                             39

-------
                                21   23   25   33   36
                    12    14    16    22    24    26    35
          VSPBIN25

Figure 20. Fuel Rate (g/s) as a function of VSP
bin for engine sizes 1.5 - 5.0L.

       weight varied instead of engine
Figure 21 shows the same relationship but with vehicle weignt varied instead ot engine
displacement. Despite the curvature between the VSP bins, within a bin, an interpolation is quite
reasonable.
                                                             4500
                                                             5000
        N =
            0    11    13   15   21    23   25   33   36
              1     12    14    16    22    24    26    35

         VSPBIN

Figure 21. Fuel Rate (g/s) as a function of VSP bin for weights ranging from 2000 - 5000 Ibs.
                                              40

-------
Given the rates in any two cells near the edge of the matrix, the rest of the emission rates can be
interpolated or extrapolated. Figure 22 shows a matrix and defines an example bin filling
strategy. The red bins are possible engine size (power) and weight combinations. The green bins
are filled using PERE and the black arrows represent the direction of interpolation or
extrapolation. The rest of the bins can be filled interpolating from the edge bins. In this case,
PERE needs to be run a minimal of four times. Note that each of the bins below has 17 operating
mode bins to fill.
Figure 22. A sample source bin matrix hole filling strategy.
When choosing bins to fill, the absolute edge bins (<2000 Ibs; >130,0001bs; <2.0L; and >5.0L)
should not be run since the masses and or engine sizes are not well defined. This is the reason for
choosing the "outer" bins in Figure 22, but not the "edge" bins.

If it is easier to fill each bin with PERE, rather than interpolating, this can also be done.

Step 7: Insert Estimated Cold Start Energy/Emission Factors
This is discussed in more detail in the next section. Cold start factors will be approximated
within MOVES, not PERE.

Cold Start

Modern vehicles are capable of eliminating nearly all of the criteria pollutants during hot running
periods. However, during cold starts, the engine and the catalytic converter are cold, thus
preventing the breakdown of emissions. For the vehicles meeting the cleanest standards, the cold
start emissions account for most of the vehicle's operating emissions. However, the impact on
fuel consumption is not as dramatic.

Due to the lack of data on which to base a model, cold start fuel consumption factors are still
elusive for advanced technology vehicles. However, there are steps PERE can take in order to
estimate the effects of cold start.
                                             41

-------
In order for criteria pollutants to be treated in the catalytic converter, the catalyst must first
obtain a high enough temperature to break down the pollutants. Before this can happen, the
engine must warm up to a point where its exhaust is hot enough to heat the catalyst. When this
temperature is reached, this condition is referred to as "light-off. In order for the catalyst to
light-off, the heat from the engine exhaust must heat the catalyst, therefore the engine must be
running until this point. At this time, PERE cannot predict at what time light-off will occur, this
must be done empirically. Typical gasoline vehicles light-off somewhere around 2 minutes into
the first portion (bag 1) of the FTP. There are advanced technologies, which help minimize this
light-off time, but they usually  come at a cost. For example, the engine may run rich briefly, or
electrical power may be needed to pre-heat the catalyst. In either of these cases, energy is being
consumed. Additionally, the engine runs less efficiently while it is cold.

In PERE, the fuel strategy on the full hybrid vehicle (used in the sensitivity study) is adjusted so
that the engine remains on for the first 2 minutes of the test. The battery power is set to zero for
the duration, so that vehicle is acting much  like a conventional vehicle, with the exception that
the regenerative brakes are still active. The  fuel never goes to zero in this scenario, but is
minimal at engine idle. The additional energy (fuel) consumed in this period is assumed to
balance any alternative technologies such as electrically heated catalyst.

The fuel consumption from this modified hybrid vehicle increases by 3.2% on the city cycle,
while the fuel economy decreases by 3.1%. This corresponds to 11% fuel consumption increase
on bag 1 vs bag 3 (10% lower fuel economy). This is likely to be a conservative estimate to use
for the model. By subtracting the hot running bag 3 from the bag 1,  this corresponds to a cold
start factor of 22.3 grams of fuel per start. This start factor will depend on the type of hybrid, size
of engine, and other vehicle parameters. It is hoped that in time, data will be collected, which can
improve upon this estimate. Kim and Lee [2004] measured the cold vs hot fuel economy for the
(pre-2004) Toyota Prius. Their report indicates that the city cold start fuel economy was 12.4%
worse than a hot start test. However, their test vehicle may have had a different calibration than
one in the United States.

Including cold start, the efficiency could be improved over the one mentioned above. It would be
advantageous to run the engine at higher loads  during the cold start  period. This serves the 3-fold
advantage of heating up the catalyst quickly, running the engine at a more efficient mode, and
recharging the batteries so that more of the rest of the cycle can be run on electrical power alone.
Since the light-off time would be shortened, the cold start fuel consumption factor may not
increase significantly.
Acknowledgments

The author would like to thank the following people for the help and support: Bob Giannelli,
John Koupal, Joe McDonald, James Warila, and and Tad Wysor from the EPA. Also Feng An
and Dan Santini from Argonne National Laboratory for their valuable feedback.
                                             42

-------
References

Alson, J., D. Barba, J. Bryson, M. Doorlag, D. Haugen, J. Kargul, J. McDonald, K. Newman, L.
Platte, M. Wolcott, "Progress Report on Clean and Effective Automotive Technologies Under
Development at EPA," EPA document number: EPA420-R-04-002, 2004.

An, F., and M. Ross, "The Use of Fuel by Spark Ignition Engines," SAE 930329, 1993.

An, F., D.,J. Santini, "Mass Impacts on Fuel Economies of Conventional vs. Hybrid Electric
Vehicles," SAE 2004-01-0572, 2004.

Beardsley, M., D. Brzezinski, R. Gianelli, J. Koupal, S. Srivastava, "MOVES 2004: 1999 Highway
Vehicle Population and Activity Data," EPA Document to be published, 2004.

Barth, M., F. An, T. Younglove, G. Scora, C. Levine, M. Ross, T. Wenzel, "Comprehensive Modal
Emission Model (CMEM), version 2.0 User's Guide," 1999.

Bishop, J.S., M.A. Kluger, "Proposed Efficiency Rating for Automatic Transmissions,"  SAE 960425,
1996.

Bosch, "Automotive Handbook - 5th Edition," Stuttgart, Robert Bosch GmbH, 2000.

Car and Driver Website: www.caranddriver.com

Chon, D.M., J.B. Heywood, "Performance Scaling of Spark-Ignition Engines: Correlation and
Historical Analysis of Production Engine Data," SAE 2000-01-0565, 2000.

DaimlerChrysler publication: "F-Cell, Driving the Future," October, 2003. (http://www.mercedes-
benz.com/com/e/home/innovati on/laboratory/fuelcell/impraxistest/aclassfcell/index2.html)

Ford Motor Company pamphlet, "Ford Focus Hydrogen Fuel Cell Electric Vehicle," 2003

General Motors pamphlet, "Hydrogen 3, GM Fuel Cell Technology," 2003.

General Motors News Website: "GM's Hybrid Timeline,"
http://www.gm.com/company/gmabilitv/adv tech/300  hvbrids/hyb timeline.html, 2004

Gillespie, T., "Fundamentals of Vehicle Dynamics," Society of Automotive Engineers, 1992.

Goodwin, R.W., M. Ross, "Off-Cycle Exhaust Emissions from Modern Passenger Cars with
Properly-Functioning Emissions Controls," SAE 960064, 1996.

Greenbaum, J.J., M.A. Kluger, B.E. Westmoreland, "Manual Transmission Efficiency Trends and
Characteristics," SAE 942274, 1994.

Heywood, J.B., "Internal combustion engine fundamentals," McGraw-Hill, New York, 1998.
                                           43

-------
Honda Motor Company publication (pamphlet): "PCX Fuel Cell Power

Honda Motor Company website: "Remarks by Takeo Fukui, President and CEO, Honda Motor Co.,
Ltd. — North American International Auto Show,"
http://www.hondacars.com/info/news/article.asp? ArticleID=2004010562646&Categorv=currenthond
a, 2004

Jimenez-Palacios, J., "Understanding and Quantifying Motor Vehicle Emissions with Vehicle
Specific Power and TILDAS Remote Sensing," PhD Thesis, Massachusetts Institute of Technology,
Cambridge, MA, 1999.

Kelly, K.J., M. Zolot, G. Glinsky, A. Hieronymous, "Test Results and Modeling of the Honda Insight
using ADVISOR," SAE 2001-01-2537.

Kluger, M. A., J. J. Greenbaum, D.R. Mairet, "Proposed Efficiency Guidelines for Manual
Transmissions for the Year 2000," SAE 950892, 1995.

Koupal, J., "Emission Rate Development," Presentation at FACA technical review meeting Dec 2,
2003, Ann Arbor, MI.

Laramie, J., A. Dicks, "Fuel Cell Systems Explained," John Wiley & Sons Ltd., West Sussex,
England, 2000.

Milington, B.W., E.R. Hartles, "Frictional Losses in Diesel Engines," SAE 680590, 1968.

Muranaka, S., Y. Takagi, T. Ishida, "Factors Limiting the Improvement in Thermal Efficiency of S.I.
Engine at Higher Compression Ratio," SAE 870548, 1987

Nam, E.K., J. Sorab, "Friction Reduction Trends in Modern Engines," SAE 2004-01-1456, 2004.

Nam, E.K., "Proof of Concept Investigation for the Physical Emission Rate Estimator (PERE) for
MOVES," EPA document number: 420-R-03-005, February, 2003.

Nelson, D.J., "Automotive Fuel Cell Systems," Lecture  Slides, Society of Automotive Engineers,
2003.

Ogawa, H., M. Matsuki, T. Eguchi, "Development of a Power Train for the Hybrid Automobile - The
Civic Hybrid," SAE 2003-01-0083.

Pachernegg, S.J., "A Closer Look at the Willans-Line,"  SAE 690182, 1969.

Park, D.H., T.S. Seo, D.G. Lim, H.B. Cho, "Theoretical Investigation on Automatic Transmission
Efficiency," SAE 960426, 1996.
                                           44

-------
Patton, K.J., R.G. Nitschke, J.B. Heywood, "Development and Evaluation of a Friction Model for
Spark-Ignition Engines," SAE 890836, 1989.

Petrushov, V.A., "Coast Down Method in Time-Distance Variables," SAE 970408, 1997.

Rechtin, M., "EPA mpg test doesn't work for hybrids," Automotive News, November 24, 2003.

Road and Track website: www.roadandtrack.com

Ross, M., "Fuel Efficiency and the Physics of Automobiles," Contemporary Physics, vol. 38, number
6, pp 381-394, 1997.

Sandoval, D., J.B. Heywood, "An Improved Friction Model for Spark Ignition Engines," SAE 2003-
01-0725,2003.

Schaefer, R.M., "Testing of an Electric Vehicle on a Clayton Water-Brake Chassis Dynamometer,"
EPA document: EPA/AA/TDG/94-01, 1994.

Thomas, M., M. Ross, "Development of Second-by-second Fuel Use and Emissions Models Based on
an Early 1990s Composite Car," SAE 971010, 1997.

Toyota Motor Coporation, "Toyota FCHV Book," Tokyo, Japan. March, 2003.

Toyota Motor Coporation, Public Affairs Division, "Toyota Hybrid System, THS II," Tokyo, Japan,
May 2003.

Toyota Motor Coporation Press release: "Toyota Stages World Premiere of All-new Highlander
Hybrid SUV at 2004 North American International Auto Show," website:
http://www.toyota.co.jp/en/event/auto_shows/2004/naias/toyota_suv.html, 2004

U.S. Department of Energy, U.S. Environmental Protection Agency, "Fuel Economy Guide," 2004.
(www.fueleconomy.gov)

Weiss, M.A., J.B. Heywood, E.M. Drake, A. Schafer, F.F. AuYeung, "On the Road in 2020 - A Life
Cycle Analysis of New Automobile Technologies," Massachusetts Institute of Technology  Energy
Laboratory Report # MIT EL 00-003. October, 2000.

Weiss, M.A., J.B. Heywood, A. Schafer, V.K. Natarajan, "Comparitive Assessment  of Fuel Cell
Cars," Massachusetts Institute of Technology Energy Laboratory Report # MIT LFE 2003-001 RP.
February, 2003.

Wu, W., M. Ross, "Modeling of Direct Injection Diesel Engine Fuel Consumption," SAE 971142,
1997.

Yagi, S., Y. Ishibasi, H. Sono, "Estimation of Total Engine Loss and Engine Output in Four Stroke
S.I. Engines," SAE 900223, 1990.
                                           45

-------
Appendix A: PERE Algorithms

Key:
FR    Fuel rate (g/s)
KNV  Friction term
LHV  Lowere heating value of fuel
Pace    Accessory power
Pbatt    Battery power
Pd     Power demand
Peng    Engine power
PFC    Fuel Cell power
Pengmax Maximum power of engine
Pfcmax  Maximum power of fuel cell
Pmotmax Maximum power of motor
Pmin    Minimum regeneration power
Pmot    Motor power
Pth    Hybrid threshold power
77      Engine indicated efficiency
??disch  Battery discharge efficiency
       Fuel cell efficiency
       Final drive efficiency (for motor only)
T?FWD  Front wheel  drive power fraction
7?mot    Motor efficiency
77tran    Transmission efficiency
fyegen  Regenerative braking recharge efficiency
Parallel Hybrid Algorithm

Engine Power (nested if statements)
       IF OR(v < 2, Pd < Pth) THEN Peng = 0                    'idle & decel engine shut-off
              IF Pd > Pengmax THEN Peng = Pengmax                'Can't 6XC6ed max
                    ELSE Peng = Pd
Fuel Rate
       IF Pd = 0 THEN FR = 0                                 ' fuel shut-off
              IF Pd > Pengmax THEN FR = [KNV + Peng /rjTfoJ/LHV  'no Pacc - slightly overpower
                    ELSE FR = [KNV + (Peng lritran + PaCc)/7?]/LHV
Battery discharge (Pbatt > 0)
       IF AND(Pd <= Pth, Pd > 0) THEN Pbatt = (Pd /7?mot7?FD + Pacc)/7?d1Sch   'launch
              IF Pd > Pengmax + Pmotmax THEN Pbatt = Pmotmax              'Can't 6XC6ed max
                    IF Pd > Pengmax THEN Pbatt = [(Pd - Pengmax)/7?mot7?FD + Pacc]/7?disch    'assist
                           ELSE Pbatt = 0
                                             46

-------
Battery recharge (Phot < 0)
       IF Pd <= - Pmotmax/77FWD THEN Pbatt = -Pmotmax??mot??regen??FD         'Can't 6X066(1 max
              If Pd < - Pmin/r/Fwo THEN Pbatt = Pdr/FWDr/mot?7regen77FD      ' can't fall below min
                     ELSE Pbatt = 0
Fuel Cell Hybrid Algorithm

Fuel Cell Powerplant
       IF OR(v < 2, Pd < Pth) THEN PFC = 0               'idle & decel fc shut-off
              IF Pd > Pmotmax THEN PFc = Pfcmax           'can't exceed max
                      ELSE P PFC = [Pd/(7?mot7?FD) + Pace]/ ??FC
Battery (ultracapacitor) discharge
       IF AND(Pd <= Pth, Pd > 0) THEN Pbatt = (Pd/7?mot7?FD + Pace)/Hdisch  'launch
              IF AND (Pd > Pfcmax, Pd < Pmotmax) THEN Pbatt = [(Pd - Pfcmax)/7?mot7?FD + Pacc]/7?diSch
                      IF Pd <= 0 THEN Pbatt = Pacc/7?dlSch
                             ELSE Pbatt = 0
Battery recharge
       Same as for hybrid (above)
                                               47

-------
Appendix B: Tables of
Vehicle
Model Year
Vehicle wgt (kg)
CrO (rolling resistance)
Cd (drag coeff)
A (frontal area m"2)
Pace (accessory - kW)
A(N)
B (N/mps)
C (N/mps"2)
Engine
Engine Displ (L)
kO (N indep friction kJ/Lrev)
k1 (N dependent fric)
P/T indicated eff (eta)
Transmission
N/v (rpm/mph)
Nidle (rpm)
trans eff
Shift point 1-2 (mph)
Shift point 2-3
Shift point 3-4
Shift point 4-5
g/gtop 1
g/gtop 2
g/gtop 3
g/gtop 4
g/gtop 5
Fuel
LHV (kJ/g)
density gas (kg/L)
Motor
overall efficiency
Regen Brake Eff
FWD power frac
Motor peak power (kW)
min regen (kW)
Motor Energy (kWhr)
Battery
Initial SOC
Batt Energy (kWh)
min SOC
max SOC
discharge eff
Hybrid
hybrid threshold (kW)
MIT
2020
1154
0.006
0.22
1.8
1




1.11
0.153
0.00155
0.405

35.6
700
0.88
18
25
40
50
4.04
2.22
1.44
1.00
0.90

43.7
0.737

0.76
0.85
0.7
30
2.8
1.8

0.56
1.8
0.2
0.8
0.95

1.75
parameters for validation vehicles.
Camry
2004
1565
0.009
0.3
2.4
0.75
127.36
0.9578
0.4374

2.4
0.164
0.00155
0.405

35.6
700
0.88
18
25
40
50
4.04
2.22
1.44
1.00
0.90

43.7
0.737















Jetta
2004
1467
0.009
0.3
2.11
0.75
111.25
3.6834
0.3764

2
0.164
0.00155
0.405

35.6
700
0.88
18
25
40
50
4.04
2.22
1.44
1.00
0.90

43.7
0.737















Jetta TDI
2004
1483
0.009
0.3
2.11
0.75
111.25
3.6834
0.3786

1.9
0.123
0.00215
0.45

26.7
700
0.88
18
25
40
50
4.04
2.22
1.44
1.00
0.90

41.7
0.856















Civic DX
2004
1239
0.009
0.3
2.14
0.75
105.47
5.4276
0.2670

1.7
0.164
0.00155
0.405

35.6
700
0.88
18
25
40
50
4.04
2.22
1.44
1.00
0.90

43.7
0.737















Civic HX
2004
1224.2664
0.009
0.3
2.14
0.75
105.47
5.4276
0.2670

1.7
0.15088
0.00155
0.48

35.6
700
0.88
15
25
40
50
4.04
2.22
1.44
1
0.9

43.7
0.737















Divic Hybrk
2004
1354
0.008
0.28
2.14
0.75
125.58
-0.9000
0.4474

1.35
0.15088
0.00155
0.48

35.6
700
0.88
18
25
40
50
4.04
2.22
1.44
1.00
0.90

43.7
0.737

0.76
0.85
0.75
10
2.8
1.8

0.56
0.936
0.2
0.8
0.95

2.2
Insight
2004
989
0.008
0.26
1.92
0.75
53.76
2.2837
0.3013

1
0.15088
0.00155
0.48

35.6
700
0.88
18
25
40
50
3.461
1.75
1.096
0.86
0.71

43.7
0.737

0.76
0.85
0.75
10
2.8
1.8

0.56
0.936
0.2
0.8
0.95

1.5
Prius '01
2001
1390
0.009
0.26
2.11
0.75
86.33
2.2355
0.4144

1.5
0.164
0.00155
0.46575

35.6
700
0.88
18
25
40
50
4.04
2.22
1.44
1.00
0.90

43.7
0.737

0.76
0.85
0.7
33
2.8
1.8

0.56
1.8
0.4
0.8
0.95

2.18
Prius '04
2004
1447
0.007
0.26
2.33
0.75
88.63
1.3849
0.3645

1.5
0.164
0.00155
0.46575

35.6
700
0.88
18
25
40
50
4.04
2.22
1.44
1.00
0.90

43.7
0.737

0.76
0.85
0.95
50
2.8
1.8

0.56
1.3104
0.4
0.8
0.95

2.9
48

-------