You are on page 1of 6

Proceedings of The Canadian Society for Mechanical Engineering International Congress 2016

CSME International Congress 2016


June 26-29, 2016, Kelowna, BC, Canada

A COMPARISON OF VEHICLE SIMULATION SOFTWARE AND DYNAMOMETER


RESULTS FOR BATTERY ELECTRIC VEHICLES

Kieran Humphries Alexei Morozov


Department of Mechanical Engineering Centre for Intelligent Machines
McGill University McGill University
Montreal, Canada Montreal, Canada
kieran.humphries@mail.mcgill.ca alexvit@cim.mcgill.ca

Abstract— The accuracy of simulation software when modeling simulation based on the basic forces on a moving vehicle. A
electric vehicle performance is of great importance in the design particular vehicle driving cycle, a standard trace of velocity over
and testing of such vehicles. In order to reduce energy consump- time, is selected and vehicle specifications are input into the
tion and improve range and performance, simulation software is simulator, including component masses and efficiency tables.
often used prior to prototyping to save time and cost. However, The simulator calculates the force required to move the vehicle
the accuracy of simulation models may not be as good as that at each time step of the selected driving cycle via the road
of prototypes and may depend on the software and accuracy load equation. Component efficiencies are used to calculate the
of specifications used for analysis. In the presented research, a amount of energy used at each of these steps. The total energy
comparison of simulation results with published dynamometer used is the sum of the energy used at each step. This way,
data for currently available electric vehicles has been completed. vehicle models can be evaluated and component changes can
ADVISOR 2003, a simulation software package, was used to be modeled quickly. The purpose of this study is to evaluate
simulate several passenger vehicles and the results were com- the accuracy of two simulation programs when compared with
pared to those published by the United States Argonne National actual dynamometer testing performed on production vehicles
Laboratory. After considering the validity of the simulation data, by the Argonne National Laboratory (ANL) in the United
an additional vehicle model of a Class 4 electric delivery truck States. This data is available to the public as the Downloadable
was tested and the results are discussed. Dynamometer Database1 .
All vehicles tested by the ANL are passenger vehicles,
Keywords- electric; vehicle; simulation; dynamometer; soft-
therefore in the paper the analysis is presented with reference
ware
to these passenger vehicles in the database. Furthermore, a
case study is presented with a Class 4 medium-duty electric
I. I NTRODUCTION
delivery truck. The Class 4 vehicle weight class (14001-16000
Electric and hybrid vehicles are an important tool to reduce lb or 6350-7260 kg) is an advantageous segment for fleet
carbon dioxide emissions and mitigate climate change. Electric electrification because it is popular for delivery vehicles in and
vehicles have reduced well-to-wheel emissions when compared around urban centers which have a relatively short driving range.
with their conventional counterparts and they have no local This limited range is an advantage because a smaller, lighter,
tailpipe emissions. The emissions associated with such vehicles and less expensive battery pack than otherwise needed can
are due to electricity generation at power plants. This electricity be used in these vehicles, thus improving their payback time
can be generated using conventional combustion power plants, when compared to the cost of conventional trucks. Low-speed
hydroelectric facilities, or renewable wind and solar collectors. cycles with many starts and stops can also lead to an advantage
Using renewable energy, it is possible to almost entirely elimi- for electric vehicles due to their improved efficiency in these
nate the emissions associated with electric vehicle operation. situations and the potential for regenerative braking [1].
Simulation is crucial in the design of modern vehicles
II. BACKGROUND
and allows improvements in efficiency and performance at
significantly lower cost and time than traditional prototyping. The vehicle architecture for this study is standard in terms
In the current paper, a comparison of the functionality and of battery electric vehicles, with an electric motor powering the
accuracy ADVISOR to dynamometer data is presented. The 1 http://www.transportation.anl.gov/D3/
type of simulation used in this study is a vehicle efficiency

1 Copyright © 2016 by CSME


drive wheels through a single gear reduction and a differential. these test procedures and analyzed in order to obtain fuel use and
No shifting transmission is required in this case because of the emissions data. However, in the case of electric vehicles there
large range of operation of the electric motor. Most electric are no tailpipe emissions and the energy consumption is instead
vehicles on the market use this configuration, including the measured using voltage and current sensors. Using the voltage
popular Nissan Leaf, the Mitsubishi i-MiEV, and the Ford Focus and current output of the battery, the battery power output can
Electric. All vehicles considered in this study use lithium-ion be calculated [2]
battery packs to provide energy for the electric motor. The
electric motors are permanent magnet synchronous motors that Pbattery = Vbattery × Ibattery (1)
run on AC electricity produced from the DC battery using a
power electronic inverter. Each of these components has an where Pbattery is the battery output power in Watts, Vbattery is
efficiency that must be included in simulation, although the the battery voltage in Volts, and Ibattery is the battery current in
inverter and motor efficiencies are bundled together. Amps. The total energy use can be calculated from the battery
power using [2]
Electric vehicles are more efficient than their gasoline coun- ∫
terparts mostly because of the efficiency of their electric motors E = Pbattery · dt (2)
when compared with internal combustion engines. This means
that they use less energy to travel the same distance as a
conventional vehicle. However, fossil fuel is very energy dense where E is the energy used in Joules. Since test data is sampled
and contains enough energy for a conventional vehicle to travel a at discrete time steps in both the dynamometer testing and in
great distance between refueling. Batteries are less energy dense vehicle simulations, the integral in (2) becomes a summation
and this means that the efficiency of electric vehicles is of great in practice. Finally, the energy consumption in Wh/km can be
importance in increasing vehicle range. Vehicle testing is used to calculated using
measure the vehicle energy consumption over standard driving E
cycles and predict their real-world efficiency. EC = (3)
D · 3600
A. Vehicle Testing where EC is the energy consumption in Wh/km, and D is the
distance traveled in km.
Most vehicle testing for efficiency and fuel consumption
uses standard driving cycles, which are traces of speed versus B. Simulation Software
time to be followed by the test vehicle, whether on a track
or in a laboratory setting. These vehicle test cycles are also Efficiency calculations in simulation software are based on
used in simulation. The automotive industry has developed the road load equation below [3]
various test cycles which simulate city and highway conditions
and are used to produce fuel economy ratings for vehicles FR = FAD + FRR + Fgx (4)
sold to consumers. The latest iteration of the United States
where FR is the road load force, FAD is the aerodynamic drag
Environmental Protection Agency’s fuel economy testing for
force, FRR is the tire rolling resistance, and Fgx is the force of
passenger vehicle rating purposes uses four different driving
gravity in the vehicle direction of travel. When each component
cycles over five different tests which simulate city, highway,
is expanded to include its components, the equation becomes [3]
high speed/acceleration, cold temperature, and air conditioning
conditions. The city test and cold temperature test both use the ρ
FR = Af CD v 2 + mg cos β [Crr0 + Crr1 v] + mg sin β (5)
UDDS driving cycle, the highway test uses the HWFET cycle, 2
the high speed/acceleration test uses the US06 cycle, and the
air conditioning test uses the SC03 cycle. Table I lists several where ρ is the air density, Af is the vehicle frontal area, CD
parameters of each of these driving cycles. is the aerodynamic drag coefficient, v is the vehicle speed, m
is the vehicle mass, g is the acceleration due to gravity, β is
TABLE I. Cycle Parameters the angle of the road measured from horizontal, Crr0 is the
first coefficient of rolling resistance, and Crr1 is the second
Vehicle Driving Cycles coefficient of rolling resistance. This equation is used to estimate
Name Time Dist Max Spd Avg Spd Max Accel Avg Accel the forces on the vehicle which, along with the tractive force
- s km km/h km/h m/s2 m/s2 provided by the vehicle itself, determine the vehicle acceleration
UDDS 1369 11.99 91.3 31.5 1.48 0.50 and speed throughout a simulated driving cycle. The driving
HWFET 765 16.51 96.4 77.6 1.43 0.19 cycle is broken up into discrete time steps of usually less than 1
US06 600 12.89 129.2 77.2 3.76 0.67 second and the forces are calculated at each step in order for the
SC03 600 5.76 88.2 34.5 2.28 0.50 vehicle to maintain the proper speed trace. The energy used by
the vehicle is calculated using the required tractive force and the
component efficiencies at each step and taking the summation
Tailpipe emissions can be collected when applicable during of all these steps.

2 Copyright © 2016 by CSME


ADVISOR (ADvanced VehIcle SimulatOR) is a Matlab- for the Nissan Leaf and Ford Focus Electric but the validity of
Simulink modeling and simulation plugin developed by the such data is unknown. Alternatively, motor maps were estimated
United States National Renewable Energy Laboratory (NREL), using the Argonne data for wheel force and battery power. The
a US government department [4, 5]. The package is used for wheel power output can be estimated using
the analysis of performance, fuel economy, and emissions of
conventional, electric, hybrid electric, and fuel cell vehicles. The Pwheel = Fdyno · vdyno (6)
latest ADVISOR version was released in 2004 though many
where Pwheel is the power output at the wheel, Fdyno is the
researchers still use ADVISOR in their projects. The vehicle
force applied by the dyno, and vdyno is the speed of the dyno.
component specifications needed in the road-load calculation
Combining equations (1) and (6) we get an equation for the
are input into ADVISOR by the user and the vehicle simulation
motor/inverter efficiency:
is run over the prescribed driving cycle to determine the energy
consumption. Pout Pwheel Fdyno · vdyno
ηdrivetrain = = = (7)
Pin Pbattery Vbattery · Ibattery
C. Regenerative Braking
The gearbox and differential efficiencies were estimated as
Regenerative braking is the use of methods to recapture approximately 98% each, and removed from the efficiency via
energy while decelerating the vehicle. In a conventional vehicle, the following equation in order to obtain the motor/inverter
this energy would be lost as heat through the friction brakes. efficiency only
With electric or hybrid electric vehicles, this energy can be
recaptured in the electric motor and sent to the battery for ηdrivetrain
ηmotor = (8)
storage. During braking, the electric motor provides a negative ηgearbox · ηdif f
torque on the wheels, slowing the vehicle while generating The resulting partial efficiency maps, an example of which is
electricity. Many strategies exist for the control of regenerative shown in Fig. 2, were calculated for various driving cycles and
braking and certain regulations and best practices must be were used to estimate the overall efficiency map of each vehicle
followed in order to maintain vehicle stability. For example, the motor through interpolation. However, these maps showed a
brake balance between the front and rear wheels of a vehicle is significantly lower efficiency than expected. Permanent magnet
of importance for wheel lockup and stability. If the rear wheels motors of this type in previous simulations show an overall
are locked before the front wheels, this can cause an unstable efficiency of over 90% during some driving cycles. However,
condition and lead to a spin [6]. the motor maps created through this procedure did not attain
overall efficiencies in this range during driving cycle testing.
III. M ETHOD
A. Modeling and Simulation Calculated Motor Efficiency
150

ADVISOR has a number of basic block diagrams available 90

in the software including a battery electric model that was used 100
80
for the vehicles in this study (Fig. 1) which did not need to
be changed for the vehicle simulations in this test. Vehicle 70
50
Motor Torque (Nm)

parameters including motor maps and regenerative braking 60

Efficiency (%)
strategy were modified. The vehicle parameters were mostly
50
obtained from Argonne and DOE data, as mentioned in the 0

previous section. Tests in ADVISOR were set up for each of the 40


driving cycles in the standard testing procedure with appropriate -50
30
accessory loads (higher loads for the SC03 air conditioning).
20
-100
B. Vehicle Specifications 10

Vehicle specifications for all test vehicles were obtained -150 0


through on-line manufacturer information or from the ANL -1000 0 1000 2000 3000 4000 5000 6000 7000

data and the U.S. Department of Energy, Vehicle Technologies Motor Speed (RPM)

Office, Advanced Vehicle Testing Activity (AVTA) Data and Figure 2. The estimated motor efficiency for the Nissan Leaf motor during
Results. Table II shows a summary of the vehicle specifications several UDDS cycles
for the simulations which were input into the two software
packages as the vehicle setups.
Since the motor maps created in this way were of such
Unfortunately, motor maps and regenerative braking strategy low quality (perhaps due to the fact that the dynamometer
are not usually published for vehicles as they are company data included all four wheels in the tractive force calculation
secrets. Unverified motor maps were obtained on the Internet rather than just drive wheels and this may necessitate further

3 Copyright © 2016 by CSME


Figure 1. ADVISOR battery electric vehicle block diagram

TABLE II. Vehicle Specifications for all test vehicles

Vehicle Specifications
Make - Mitsubishi Nissan Ford BMW Workhorse
Model - i-MiEV Leaf Focus EV i3 P42 BEV
Configuration - RWD FWD FWD RWD RWD
GVWR kg 1510 1902 2085 1620 6373
Test Mass kg 1304 1498 1791 1443 5637
Frontal Area m2 2.23 2.27 2.26 2.38 7.02
Drag Coefficient - 0.35 0.29 0.295 0.29 0.70
Wheelbase m 2.55 2.70 2.65 2.57 4.01
Tire Rolling Radius m 0.28 0.31 0.32 0.34 0.393
Crr0 - 7.41e-3 9.66e-3 9.22e-3 7.41e-3 8.00e-3
Crr1 s/m 4.64e-4 7.56e-5 2.94e-4 4.66e-4 1.20e-4
Motor Type - PMSM PMSM PMSM PMSM PMSM
Motor Peak Power kW 49 80 107 125 150
Motor Peak Torque Nm 196 253 245 250 575
Motor Base Speed RPM 2400 3000 4200 4775 2400
Motor Max Speed RPM 9900 10000 8750 11400 12000
Final Drive Ratio - 7.065:1 7.9:1 7.82:1 9.7:1 14.76:1
Battery Capacity kWh 16 24 23 21.6 100

calculations), it was decided to use the available on-line maps used by Argonne to program the dynamometer, the simulation
for the Nissan Leaf and Ford Focus Electric and to scale these software is accurately estimating the required force for the
maps for the other two test vehicles. The less than ideal results vehicle to follow the cycle trace.
of this solution are discussed later in this paper.
Fig. 4 shows a comparison of the battery power output, the
A generic lithium-ion type battery model was used in AD- opposite end of the drivetrain from the force at the wheel. The
VISOR but was modified to reflect the correct mass, battery power output of the battery on the dynamometer was measured
architecture, and voltage from Table II. This is another area by Argonne using voltage and current sensors at the battery.
where detail is scarce in terms of actual battery performance The power output of the battery in the simulation is calculated
charts, and the simulation accuracy could be improved if such based on the efficiency specifications of each component block
data were available. as well as the regenerative braking strategy and accessory load.
The accessory load was estimated using the load on the vehicle
C. Argonne Dynamometer Test Data when the wheels were stopped but the key was on. For example,
the power use at standstill for the Leaf was approximately
Fig. 3 shows the measured force that was applied by the 180 Watts, which was used as the accessory load during this
dynamometer during testing of the Nissan Leaf on a UDDS driv- simulation. The close agreement in this example shows that the
ing cycle compared with the simulated tractive effort that was model has similar efficiency to the real vehicle in this case.
calculated by ADVISOR. This graph shows that the simulation
software is calculating a similar load to what the dynamometer When tested on other driving cycles, this vehicle simulation
provided, as expected. Based on the coast down test constants model similarly matched the dynamometer data. However, dur-

4 Copyright © 2016 by CSME


Tractive Force Estimates #10 4 High Voltage Battery Power
4000 5
Dyno Tractive Effort Dyno HV Battery Power
3000 Simulation Tractive Effort 4 Simulation Battery Power

2000 3

1000 2

Power(W)
Force(N)

0 1

-1000 0

-2000 -1

-3000 -2

-4000 -3
0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200
Time(s) Time(s)

Figure 3. An example of the dynamometer force applied during Argonne Figure 4. An example of the battery power calculated from Argonne test data
testing compared to the calculated force required in ADVISOR simulation for compared to the calculated battery power in ADVISOR simulation for a
a Nissan Leaf during the UDDS cycle Nissan Leaf during the UDDS cycle

ing cold weather SC03 testing, the accessory load on the vehicle Energy Consumption of Vehicles
220
was changed to match the base load visible in the dynamometer
Energy Consumption (Wh/km)
Mitsubishi i-MiEV
tests for this cycles using the same method of checking the 200 Nissan Leaf
Ford Focus Electric
average load when the vehicle was stopped.
180 BMW i3

The amount of regenerative braking versus friction braking


160
in simulation software can be tuned to match that of the actual
road vehicles. For example, if the regenerative braking in the 140
models is too low, the power returned to the battery during
simulation will not match the power data from the dynamometer. 120

In this way, the braking strategy can be reverse engineered and 100
input into the simulated vehicles. Up to 90% of the braking force UDDS HWFET US06 SC03
was found to come from regenerative braking in order to match Cycles

this data. Figure 5. Results of energy efficiency tests in graphical format

IV. R ESULTS
subishi MiEV during the HWFET and US06 tests as well as
A. Energy Consumption Results
the BMW i3 for all tests there are large percent differences
Using equation (2), the overall energy use was calculated between testing and simulation. For these two vehicles, no
from the simulations and from the dynamometer results. This motor efficiency data was available on-line. Therefore the motor
total energy divided by the distance traveled gives the energy efficiency maps were created by stretching and scaling the
consumption, shown in Table III and represented graphically efficiency table for the Nissan Leaf motor, and this may have
in Fig. 5. This energy consumption calculated from the dy- led to significant error since the specific motor design can have
namometer data (also available in the ANL data sheets) as a large influence on efficiency.
well as the energy consumption from the simulation program
are shown in the table for comparison purposes. The percent B. Performance Results
difference between the dynamometer energy consumption and
The performance results during certain dynamic tests were
that from each simulation is shown beside each test row.
tested as well, namely the top speed and acceleration time from
The results in Table III show that the simulation model 0-60 mph (0-96.6 km/h). These are standard tests that are used
results can be very close to the dynamometer results, meaning to qualify the performance of road vehicles for driveability and
the simulation models are accurately representing the vehicles, marketing purposes. The simulation results in Table IV show
for example for most of the Nissan Leaf and Ford Focus Electric good agreement with the published figures, as expected since it
data. However, there are cases with significant discrepancy is significantly simpler to estimate these performance metrics.
between the testing and simulation. For example, for the Mit- This is because the performance metrics do not depend on

5 Copyright © 2016 by CSME


TABLE III. Dynamometer and Simulation Results for all test vehicles

Make Mitsubishi i-MiEV Nissan Leaf Ford Focus Electric BMW i3 Class 4 BEV

Test Type Dyno Sim Diff Dyno Sim Diff Dyno Sim Diff Dyno Sim Diff Dyno Sim
UDDS (Wh/km) 105 106 1% 108 109 1% 127 130 2% 98 114 16% - 656
HWFET (Wh/km) 122 155 27% 128 135 5% 139 141 1% 116 161 39% - 744
US06 (Wh/km) 172 215 25% 171 190 11% 189 199 5% 153 217 42% - 1194
SC03 (Wh/km) 130 132 2% 147 160 9% 148 159 7% 121 149 23% - 738

efficiency tables but on the maximum torque and power output showed significantly lower efficiency than expected, leading to
of the motor as well as the overall gear ratio of the drivetrain. the belief that there was a missing factor in the calculations used
These are widely published figures found in vehicle literature to create these maps. Also, the dynamometer itself may have a
and noted in Table II. resistance torque that needs to be included in the calculation.
Further study of this phenomenon should be completed in
TABLE IV. Published and Simulated Performance Data
order to attempt to create better quality motor maps for these
Vehicle Top Speed (km/h) Acceleration Time (s)
vehicles, to be used in future simulation validation. Accessory
Published Simulation Published Simulation
loads should also be studied in more detail in order to improve
Mitshubishi i-MiEV 134 148 14.9 13.9
accuracy.
Nissan Leaf 146 144 10.6 10.4 The simulation of the Class 4 delivery vehicle, however, used
Ford Focus EV 137 136 10.9 10.0 the motor map of an experimental electric motor design and the
BMW i3 150 151 6.5 6.5 efficiency map was known exactly. This made the simulation
task much easier and potentially more accurate, as shown by
the results from the previous cases where the Nissan Leaf and
V. C ASE S TUDY: E FFICIENCY A NALYSIS OF C LASS 4 Ford Focus Electric showed good agreement between testing
E LECTRIC D ELIVERY T RUCK IN S IMULATION and simulation due to the availability of their motor maps.
S OFTWARE
ACKNOWLEDGMENT
This section details the results of the new Class 4 medium-
duty electric delivery truck simulations in ADVISOR. For prac- The research work reported here was supported by a grant
tical purposes, a 2004 GM Workhorse P42 chassis with a under the Automotive Partnerships Canada Project APCPJ418901-
body made by ENOVA Systems was selected. This truck is 11. We also gratefully acknowledge the support of our industrial
similar to those used by Purolator, DHL, UPS and other courier partners: Linamar, TM4 and Infolytica.
companies, and has a Gross Vehicle Weight Rating (GVWR) of
6350 kg (about 14000 lb). The motor used is an experimental R EFERENCES
motor designed at McGill which has a power rating of 200 kW, [1] M. O’Keefe, A. Simpson, K. Kelly, and D. Pedersen, “Duty cycle
sufficient to move such a heavy vehicle at highway speeds even characterization and evaluation towards heavy hybrid vehicle applications,”
with only a single speed gearbox and an overall ratio of 6.4:1. SAE Technical Paper, 2007, pp. 01–0302.
The results of this vehicle’s simulation runs are available [2] H. Lohse-Busch, K. Stutenbeerg, M. Duoba, E. Rask, F. Jehlik, and
in Table III and show similar trends to those from the other G. Keller, “Chassis dynamometer testing reference document,” Tech. rep.,
vehicles, however with consumption values several times larger Argonne National Laboratory, United States Department of Energy, 2013.
than those of the smaller vehicles. Energy consumption for [3] I. Husain, Electric and Hybrid Vehicles: Design Fundamentals, Second
other tested cycles is as follows: NYCC 821 Wh/km, OCC 660 Edition, Taylor & Francis, 2011.
Wh/km, HTUF4S 621 Wh/km. These cycles better represent [4] K. B. Wipke, M. R. Cuddy, and S. D. Burch, “ADVISOR 2.1:
actual daily driving for a delivery vehicle of this type. A user-friendly advanced powertrain simulation using a combined
backward/forward approach,” Ieee Transactions on Vehicular Technology,
VI. C ONCLUSION vol. 48 (6), 1999, pp. 1751–1761.
[5] T. Markel, A. Brooker, I. Hendricks, V. Johnson, K. Kelly, B. Kramer,
Currently, some vehicle simulations showed good agreement M. O’Keefe, S. Sprik, and K. Wipke, “ADVISOR: a systems analysis tool
with the dynamometer data, while others showed a large dis- for advanced vehicle modeling,” Journal of Power Sources, vol. 110 (2),
crepancy between the dynamometer and the simulations. As 2002, pp. 255–266.
discussed in the Results section, this may be due to the use of [6] M. Ehsani, Y. Gao, and A. Emadi, Modern electric, hybrid electric, and fuel
scaled motor maps for several vehicles instead of motor effi- cell vehicles : fundamentals, theory, and design, Boca Raton: CRC Press,
ciency maps created for the vehicles in question. As previously 2010.
mentioned, the motor maps created from dynamometer data

6 Copyright © 2016 by CSME

You might also like