Energy 172 (2019) 823e839

Contents lists available at ScienceDirect

Energy
journal homepage: www.elsevier.com/locate/energy

Evaluation of electric vehicle component performance over

eco-driving cycles*
Zhiming Gao a, *, Tim LaClair a, Shiqi Ou a, Shean Huff a, Guoyuan Wu b, Peng Hao b,
Kanok Boriboonsomsin b, Matthew Barth b
a
Oak Ridge National Laboratory, P.O. Box 2008, Oak Ridge, TN 37831, USA
b
University of California, Riverside, Center for Environmental Research & Technology, 1084 Columbia Avenue, Riverside, CA 92507, USA

a r t i c l e i n f o a b s t r a c t

Article history: Battery electric vehicles (BEVs) are a critical pathway towards achieving energy independence and
Received 6 October 2018 meeting greenhouse and criteria pollutant gas reduction goals in the current and future transportation
Received in revised form sector. Emerging connected and automated vehicle (CAV) technologies further open the door for
24 January 2019
developing innovative applications and systems to leverage vehicle efficiency and substantially trans-
Accepted 3 February 2019
Available online 4 February 2019
form transportation systems. Therefore, we present a simulation study of various BEV types and compare
the performance when driving on real-road drive cycles to highly optimized eco-driving cycles using
advanced CAV technologies. The results demonstrate that eco-driving has a high potential to reduce
Keywords:
Electric vehicle
energy consumption for all types of BEVs considered. The investigated BEVs include a compact vehicle, a
Eco-driving transit city bus, and a Class 7 delivery truck. The impact of eco-driving on conventional vehicles was also
Component efficiency compared to comparable BEVs. Compared to the BEVs, eco-driving provides a larger reduction in the
Powertrain modeling conventional vehicle's braking energy loss, and also provides conventional vehicles with greater re-
ductions in the engine mechanical energy output but the fuel savings did not show a consistent trend
among all the conventional vehicle types. As part of the study, a comprehensive EV powertrain model
was developed to account for key EV components and powertrain configurations.
Published by Elsevier Ltd.

1. Introduction independence and greenhouse and criteria pollutant gas reduction

goals for both current and future transportation [2,3]. The U.S. has
US Transportation represents over 3 trillion vehicle-miles driven become one of the world's largest plug-in electric vehicle (PEV)
annually, 11 billion tons of freight transported, and 70% of the na- markets for both BEVs and plug-in hybrid electric vehicles (PHEVs).
tion's petroleum consumption [1]. The transportation sector also In 2016, over 159.6 thousand passenger PEVs were sold in the U.S.
contributes more than 26% of greenhouse gas and 57% of NOx in the market, covering 21% of all PEV sales worldwide [4]. For BEVs only,
nation, which significantly impacts air pollution and climate the Volt has sold 113,489 vehicles over the period of the last seven
change [1]. Consequently, battery electric vehicles (BEVs) are years, followed by the Nissan Leaf with 103,600 units sold and the
considered to be a critical pathway towards achieving energy Tesla Model S with 94,623 sold [4]. Compared to light-duty (LD)
BEVs, there are few medium-duty (MD) and heavy-duty (HD) BEVs
on the road to date. This is because not all LD BEV technologies can
* be transferred directly to MD and HD applications, which typically
Notice: this manuscript has been authored by UT-Battelle, LLC under Contract
No. DE-AC05-00OR22725 with the U.S. Department of Energy. The United States require higher power demands [5] as well as considerably larger
Government retains and the publisher, by accepting the article for publication, battery capacity needs [6]. Moreover, the MD/HD vehicle market
acknowledges that the United States Government retains a non-exclusive, paid-up, covers a wide diversity of commercial vocations, operational pat-
irrevocable, world-wide license to publish or reproduce the published form of this
terns, and duty cycles [7]. However, federal and state governments,
manuscript, or allow others to do so, for United States Government purposes. The
Department of Energy will provide public access to these results of federally
OEMs and nationwide fleets have shown significant interest and
sponsored research in accordance with the DOE Public Access Plan. made substantial commitments for electrification of commercial
* Corresponding author. Oak Ridge National Laboratory, National Transportation vehicles, including transit buses, school buses, and MD and HD
Research Center, 2360 Cherahala Boulevard, Knoxville, TN 37932. trucks [6e13]. At present, three manufacturers, including Proterra,
E-mail address: [email protected] (Z. Gao).

https://doi.org/10.1016/j.energy.2019.02.017
0360-5442/Published by Elsevier Ltd.
824 Z. Gao et al. / Energy 172 (2019) 823e839

BYD, and New Flyer, already offer commercial battery electric point out that the optimal vehicle performance is not only depen-
transit buses for sale in the US [7]. Also GreenPower and Nova are dent on the low-level component control used, but the vehicle's
demonstrating their electric city bus prototypes [5]. Lion, Motiv area of operation should also be taken into consideration for the
Power systems, TransPower and Adomani have demonstrated design. Thus, Fiori et al. [36] developed a simple PHEV energy
battery all-electric school buses employing 80e400 kwh batteries consumption model intended to be used in a real-time in-vehicle or
[8e11]. Class 7e8 HD vehicle electrification is currently in the field smartphone-based eco-driving application. The approach includes
demonstration phase by manufacturers including Balqon, BYD, microscopic traffic simulation models which monitor the energy
Electric Vehicles International, Motiv Power System, TransPower, consumption in real-time and optimize energy efficiency [36]. In
ZeroTruck, Cummins and Tesla [12,13]. Significant growth of BEVs is this manner, automation and connectivity are integrated into the
expected in the future MD and HD electric vehicle markets. control strategy of EVs, which is very compelling.
To enhance the energy efficiency and market penetration of In fact, connectivity, automation, and electrification are
BEVs, extensive work has been carried out, particularly for BEV considered as three major pillars supporting a more efficient future
components, powertrain control and energy management [14e36]. transportation system and cleaner environment. Therefore, to
The motor and battery are the key components of a BEV. Numerous boost BEV energy efficiency, driving range and market adoption, it
motor technologies have been developed for modern vehicles, is desirable to adopt connected and automated vehicle (CAV)
including DC, induction, synchronous permanent magnet and technologies in optimizing BEV performance via eco-driving func-
synchronous brushed motors, and these technologies are already tions. Many studies have been conducted for eco-driving applica-
commercially available [14,15]. A fifth topology, the reluctance tions, particularly for vehicle safety and performance
motor, has also been proposed to deliver very high efficiency and improvements [39e48]. One of the most effective eco-driving
power density at potential low cost [16]. However, most of the EVs technologies is the eco-Approach and Departure (EAD) applica-
on the current market still employ a single induction or tion, which uses traffic signal phase and timing (SPaT) information
permanent-magnet motor and a traditional mechanical driveline from upcoming traffic signals along with information about the
with a differential [17]. For the battery, Li-ion technologies have equipped vehicle and preceding traffic to determine the most en-
been widely recognized for achieving a more practical energy ergy efficient speed trajectory to pass through an intersection
density and longer cycle life than Pb-acid and NiMH batteries [18]. [39e41]. In 2010, the NHTSA performed a preliminary analysis on
In current BEVs, the typical battery capacity ranges from 24 kWh for the benefits of SPaT and EAD. The study showed a 90% reduction in
passenger electric cars, 40e120 kWh for MD BEVs, and red light violations and up to 35% in energy savings for LD vehicles
80e320 kWh or more for HD E-trucks [7]. Nevertheless, battery [48]. Jiang reported a simulated eco-driving system for an isolated
technology is still one of the toughest barriers to BEV development signalized intersection enabled engine-powered LD vehicles to
and market penetration [19,20], in particular the high cost, low achieve from 2.02% to 58.01% in fuel savings and from 1.97% to
specific energy and short life cycle of current batteries compared to 33.26% in CO2 emissions reduction [46]. The University of California
liquid fuel and combustion engines. Battery production costs Riverside utilized a 2008 Nissan Altima to test an EAD application in
currently range around 350e500 $/kWh [21,22], representing a Real-World Traffic, indicating 6% energy for the trip segments un-
considerable cost penalty compared to conventional vehicles; the der light traffic conditions, as well as substantial CO/HC/NOx
energy density of Li-ion batteries is normally 0.36e0.875 MJ/kg, emissions reduction [39]. Conventional HD truck eco-approach and
compared to 46.4 MJ/kg for gasoline, which also leads to a signifi- departure algorithms for a signalized intersection have also been
cant mass penalty and/or payload reduction in EV applications explored, and the results show up to 16% fuel saving, which de-
[6,7]; current lithium batteries typically provide only 1000 to 2000 pends on road grade and vehicle speeds [40]. All these results show
deep charging cycles or around six years of daily BEV charge- that EAD is expected to significantly reduce or eliminate braking, as
discharge cycles [20], which indicates a rather expensive mainte- well as yielding benefits in the traffic throughput. For electric ve-
nance cost compared to typical maintenance of Internal Combus- hicles, there have also been many studies conducted to evaluate
tion Engine (ICE) powered vehicles. battery energy savings under EAD driving patterns. For example, Qi
The reported BEV powertrain control and energy management et al. evaluated the energy synergy of combining vehicle connec-
usually focus on appropriate battery energy management, efficient tivity, vehicle automation and vehicle electrification, by simulating
motor traction control, and regenerative braking control [23e38]. an EAD system for EVs with real-world driving data [42]. Zhang and
Battery energy management is an indispensable part of electric Yao [43] estimated that EVs following an eco-driving strategy at
vehicles to ensure an optimal and reliable operation of the battery signalized intersections achieved at least 8.01% in energy savings
packs [23]. The battery energy management studies support not relative to normal driving conditions, which was assessed using a
only battery fault estimation (including battery health monitoring microscopic, driving parameters-based energy consumption model
and life estimation) [24e26], but also vehicle toque or power de- based on chassis dynamometer test data. Flehmig et al. [44] re-
mand estimation [28,29], battery SOC and range estimation [29,30] ported an adaptive cruise control (ACC) simulation for estimating a
etc. Recent studies have even evaluated the effect of connecting the BEV's energy-optimal trajectory when following another vehicle in
vehicle energy management system with the electric grid to traffic. The results show a 2e4% saving compared to regular traffic
maximize energy efficiency [31]. Novel motor traction controls conditions. Such benefits have been confirmed by other research
improve BEV performance and energy savings by utilizing the [45] for BEVs in a connected vehicle environment using eco-
electric motor's quick and precise torque response. Consequently, approach at signalized intersections. However, none of these
various control strategies have been developed, including car studies addressed how eco-driving impacts BEV component per-
following, optimal slip ratio, and active fault-tolerant controls formance and efficiencies. There is inadequate information to
[32,33]. Similarly, control algorithms for the regenerative braking further improve BEV component and powertrain performance
system have been studied extensively in multiple scenarios for under eco-driving patterns. These are particularly important and
BEVs equipped with different types of components such as fly- attractive considerations for OEMs and automotive Tier 1 suppliers
wheels, ultra-capacitors and converters [34,35]. The results of these in designing and optimizing the EV components and powertrain
studies demonstrate that improvements in regenerative braking systems of future BEVs.
could increase BEV driving range by 8e25%, as is similarly To address the issues identified above, we present a compre-
confirmed by other results [37,38]. In addition, Tie and Tan [16] hensive simulation study for various BEV types and compare their
 Z. Gao et al. / Energy 172 (2019) 823e839 825

performance when driving on typical real-road drive cycles to 2.1.2. Electric motor and inverter model
optimized eco-driving cycles that comprise CAV-based eco-driving A map-based performance model was developed to account for
technologies such as EAD. The BEVs considered cover a broad range the energy consumption of the electric motor and inverter over the
of electric vehicles from light-duty to heavy-duty vehicles, entire motor operating conditions. The map-based model adopts
including a compact vehicle, a transit city bus, and a Class 7 delivery the efficiency maps of motors and inverters, which were generated
truck. The type of driving evaluated includes both city and highway from experimental data measured under steady-state conditions
conditions for the compact vehicle, city operations for the transit over a pre-defined matrix of speed and torque combinations. Fig. 2
bus, and highway driving conditions for the trucks. In addition, the shows the combined efficiency of the Nissan Leaf's motor and
results are compared to those of comparable conventional vehicles inverter. In the electric motor and inverter performance models, the
in order to fully understand the impact of eco-driving on vehicle motor's output mechanical power is defined by Eq. (2a), and the
component and powertrain system performance. As part of the inverter input electrical power is estimated using Eq. (2b). The
study, a comprehensive EV powertrain model was developed to inverter input electrical power is assumed to be equivalent to the
account for key EV components and powertrain configurations. The battery power output, defined in Eq. (2c). The motor and inverter
EV powertrain model employs data and models available from the efficiencies are based on interpolations of the efficiency maps that
public domain, as well as some existing ORNL component models, have been tabulated over a range of motor speed and torque. A
measurement data, and models adapted from Autonomie vehicle constraint given in the motor model is that the maximum torque
performance software [49]. delivery is not allowed to exceed the boundary of motor torque
output. The curve for the motor torque boundary shown in Fig. 2 is
tmot;pos bdry , and the motor regeneration torque boundary is
2. Methodology and assumptions
assumed to be tmot;neg bdry ¼  tmot;pos bdry . In addition, the electric
This section describes an EV powertrain model that considers motor and inverter model include a simple factor for scaling the
the tractive energy, braking forces, and vehicle component effi- power and torque outputs of simulated motors and inverters based
ciencies. The section also describes the assumptions for modeling on maintaining the equivalent performance efficiency and speed
LD/MD/HD battery electric vehicles, as well as the eco-driving range. This strategy allows scaling up or down the available motor
methodology. maps to approximate motor performance maps for larger or smaller
motors, whose maps may be not available, with reasonable
accuracy.
2.1. EV powertrain model
Wmot;me ¼ umot $tmot (2a)
A comprehensive EV model has been developed to account for
key EV components that impact driving efficiency: battery, motor,

Winv;ee ¼ Wmot;me ðhmot ðumot ; tmot Þ$hinv ðumot ; tmot ÞÞ (2b)
torque coupler, final drive, wheel, chassis, and accessory loads. The
simulation tool adopts various data and models available from the
public domain and from Autonomie vehicle performance software. Winv;ee ¼ Ibatt $Vbatt (2c)
A schematic of the model structure is presented in Fig. 1.
where the terms Wmot;me and Winv;ee are motor mechanical power
output and inverter electrical power input, respectively; umot and
2.1.1. Tractive force methodology
To evaluate the impact of aerodynamic and rolling resistance
tmot are motor speed and torque, respectively; hmot and hinv are
motor and inverter efficiencies, which are considered as functions
losses on vehicle energy consumption, the tractive power demand
of motor speed and torque, and they are estimated based on the
was used to account for vehicle forward acceleration, aerodynamic
interpolation of the motor and inverter maps; Ibatt and Vbatt are
loss, rolling resistance loss, and road grade. The tractive force
required at any time is given as follows [50,51]:
300 100
dV 1
Ftract ¼ m$ þ r$Cd $Af $V 2 þ m$g$Crr $cos q þ m$g$sin q efficiency
dt 2
95
(1) 250
42

90
where Ftract is the required vehicle tractive force; V is vehicle ve-
28

locity; r is air density; Cd is the aerodynamic drag coefficient; Crr is 200 , _

71

85
the rolling resistance coefficient; Af is the projected frontal area; q
57
Torque (Nm)

14

is the road grade; and m is the vehicle mass. The mass is deter-
150 80
mined from the curb weight of the EV and the passenger and cargo
42

weight. The curb weight includes the powertrain and drivetrain 71

75
systems (e.g. wheel, chassis, final drive, torque coupler, electrical
28

100 57
accessories, motor and high-voltage battery). The evaluation of
71 70
aerodynamic drag, rolling resistance, and road grade on the EV 42
14

57
tractive force demand is addressed in the chassis and wheel 50 28 42
component modules shown in Fig. 1. 65
14 28
14
0 60
1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
Speed (rpm)

Fig. 2. The combined efficiency map of Nissan leaf motor and inverter, rated at 280Nm
torque and 80 kW power, which was tested at ORNL under 375 V DC-link voltage,
Fig. 1. EV powertrain model configuration accounting for key EV components. 5 kHz switching frequency, and 65  C steady-state cooling temperature [52].
826 Z. Gao et al. / Energy 172 (2019) 823e839

battery current and voltage output.

Vbatt ¼ Nseries $Vcell (3f)
2.1.3. Li-ion battery model
The battery package is modeled based on an equivalent circuit of Ibatt ¼ Mparallel $Icell (3g)
multiple serial and parallel battery cells, each of which is assumed
to have the same performance. For the single battery cell modeling, where VSOC is the open circuit voltage which is a nonlinear function
an RC-based equivalent-circuit battery cell dynamic model was of SOC and normally measured as the steady-state open circuit
developed to address key transient response physics using elec- terminal voltage at various SOC points; Vtran;S and Vtran;L are volt-
trical circuit analog components based on the electrical network ages of short- and long-time step responses of RC networks,
consisting of a series resistor and two RC parallel networks, as respectively; Rtran;S , Rtran;L , Ctran;S , and Ctran;L represent short- and
shown in Fig. 3. The RC parallel networks represent short- and long- long-time constants of the step response of RC networks; Ro is a
time step responses. On the basis of experimental results, the use of series resistor; Vcell and Icell are the battery cell output voltage and
two RC networks, as opposed to one or three, provides the best current; Vbatt and Ibatt are the modeling voltage and current of the
tradeoff between accuracy and complexity [53e55]. The considered overall battery package with Nseries battery cells in series and
components include voltage sources, variable resistors, and ca- Mparallel battery cells in parallel; Dε is the temperature-dependent
pacitors. More specifically, the equivalent-circuit model accounts potential-correction term for the battery; aðIcell Þ is a charging/dis-
for open-circuit voltage, ohmic resistances in the connector, elec- charging rate factor; bðTcell Þ is a temperature factor; Tcell and Tamb
trodes and electrolyte, and two sets of parallel resistor-capacitor are battery cell and ambient temperature; mcell Cpcell and hA are
combinations to reproduce the effects of mass transport and the battery cell mass capacity and heat transfer rate, respectively.
electric double layer, respectively. In addition, the impact of tem- VSOC , Rtran;S , Rtran;L , Ctran;S , and Ctran;L , as well as DεðTcell Þ , aðIcell Þ,
perature and charging/discharging rate on battery SOC are also and bðTcell Þ, depend on individual battery type, design, and fabri-
considered. The details are described in Eqs. (3a)e3(e). Based on cation. VSOC , Rtran;S , Rtran;L , Ctran;S , and Ctran;L are normal functions of
the required overall battery package capacity and voltage, an SOC. DεðTcell Þ , aðIcell Þ, and bðTcell Þ are impacted with battery cell
appropriate combination of series and parallel connections is current and temperature. All of them can be derived from battery
determined to simulate the complete vehicle battery module, as test data. In the current study which considers a Li-ion battery, the
shows in Fig. 3 and Eqs. (3f)e(g). data of VSOC , Rtran;S , Rtran;L , Ctran;S , and Ctran;L are adopted from
dVtran;L Vtran;L Ref. [52], and DεðTcell Þ , aðIcell Þ, and bðTcell Þ are adopted from
I
¼ cell  (3a) Ref. [54].
dt Ctran;L Rtran;L $Ctran;L
The current battery model is capable of effectively simulating
both the steady-state and transient battery responses that have
dVtran;S I Vtran;S been observed in Li-ion and other batteries with appropriate
¼ cell  (3b)
dt Ctran;S Rtran;S $Ctran;S parameter inputs. The model has been validated with experimental
measurements from the open literature [53], and the simulated
Vcell ¼ VSOC  Icell $Ro  Vtran;L  Vtrans;S þ $εðTcell Þ (3c) voltage profiles using the model for a Li-ion battery subjected to
periodic pulse discharges and charges matched the experimental
Li-ion observations to within 1% except when the accumulated
dTcell V 2tran;L V 2tran;S charge level is below 5%. The primary reason for the discrepancy is
mcell Cpcell ¼ Icell $Ro þ þ  hAðTcell  Tamb Þ
dt Rtran;L Rtran;S due to large nonlinear changes in ohmic resistance and capacitance
(3d) at very low charge levels. This implies that future model im-
provements will be needed to simulate very low charge levels, but
ð there should be little impact on battery simulations when the
1
SOC ¼ aðIcell ÞbðTcell ÞIcell dt (3e) battery charge is maintained above 5%, which is outside the range
Ccell
of normal battery operation for EVs. The battery model has been
implemented into the current vehicle system model, and we have

+ + + +
1 - - - -
~ Vbatt

+ + + +
2 - - - -
Vcell

(a) (b)
+ + + +
N - - - -

1 2 M
Fig. 3. An equivalent-circuit battery cell dynamic model for simulating a battery package comprised of multiple serial and parallel battery cells. (a) Single battery cell configuration;
(b) battery package configuration.
 Z. Gao et al. / Energy 172 (2019) 823e839 827

confirmed that it can be used to simulate appropriately sized Li-ion  

battery packages through various parallel and/or series battery cell abrk;drv $twh;brk bdry $dspd $ddec
connections. tmot;dmd ¼ max ;
Rtc $Rfd
!
Wbatt;chg bdry þ Waccelec  
; tmot;regen bdry c twh;drv < 0
umot
2.1.4. Driver, drivetrain components and control model
(5b)
A driver model was employed to control the vehicle operation. A
proportional-integral (PI) control methodology was used to
and
manage the gap between the real and targeted vehicle speeds. The
driver's wheel torque demand is calculated based on Eq. (4a); and   
the pedal signal of acceleration and braking is estimated based on
tbrk;dmd ¼ max abrk;drv $twh;brk bdry 

  
Eqs (4b) and (4c), respectively; then the pedal signal is delivered to
 tmot;dmd $Rtc $Rfd ; 0 c twh;drv < 0 (5c)
the powertrain integrated control unit in order to meet the demand
for the targeted velocity.
where twh;accel bdry ¼ tmot;pos bdry $Rtc $Rfd ; twh;regen bdry ¼
ð tmot;neg $R $R ; W and W
bdry tc fd batt;chg bdry batt;dischg bdry are the bat-
twh;drv ¼ Ftract Rwh þ KP $DV þ KI $ DVdt (4a) tery charging and discharging power boundary, respectively; Rfd is
final drive ratio; Rtc is torque coupler ratio; dspd and ddec are factors
8 t     that consider the constraints during brake regeneration. We as-
>
> wh;drv
c twh;drv > 0 ∩ twh;drv < twh;accel bdry
<t sume that energy regeneration from braking occurs when the
wh;accel bdry
aaccel;drv ¼ vehicle deceleration has not exceeded a threshold (i.e. decelera-
>
>    
: 1 c twh;drv > 0 ∩ twh;drv  twh;accel bdry tion < -2 m/s2) and the vehicle speed remains above a given value
(i.e. vehicle speed >0.6 m/s). The constraints are used to distinguish
(4b) vehicle emergency braking from regenerative kinetic energy and
avoid very low kinetic energy regeneration. Waccelec is the electric
8  

>  twh;drv      accessory load, which is taken to be constant.
>
<   c twh;drv  0 ∩ twh;drv > twh;brk bdry In addition, the torque and speed of the wheel, final drive, and
twh;brk bdry 
abrk; drv ¼ torque coupler are given as follows:
>
>    
:
1 c twh;drv  0 ∩ twh;drv  twh;brk bdry
Vdrv
uwh ¼ (6a)
(4c) Rwh

DV ¼ Vtarget  Vdrv (4d) uwh $Rfd ¼ ufd (6b)

In Eqs. (4a)e(4d), aaccel; drv is driver acceleration demand;

abrk; drv is driver braking demand; twh;drv is driver wheel torque ufd $Rtc ¼ utc ¼ umot (6c)
demand; twh;brk bdry and twh;accel bdry are braking torque demand
boundary and acceleration torque demand boundary, respectively; twh;dmd ¼ tfd;dmd $hfd $Rfd (6d)
twh;brk bdry is negative and defined based on the individual vehicle
braking system, and twh;accel bdry is positive and defined as a tfd;dmd ¼ ttc;dmd $htc $Rtc (6e)
nonlinear function of vehicle speed and the selected motor and
drivetrain; Vtarget and Vdrv are the targeted and simulated driving
ttc;dmd ¼ tmot;dmd (6f)
vehicle speeds, respectively. The values of Kp and Kl are constants
for the PI speed control and depend on the driver behavior and the The forward-looking driving speed of the simulated vehicle is
simulated vehicle system; our suggested values are 1000 and 0.5, calculated based on the equation below.
respectively. Rwh is the effective wheel rolling radius. 
During vehicle acceleration, the motor torque demand depends dVdrv twh;dmd 1 1
¼ .   .  rC A V 2
not only on driver acceleration demand aacc;drv , but also is limited dt 2
m þ Itot Rwh $Rwh m þ Itot Rwh2 2 d f drv
by the operating boundary of the motors and batteries. The motor  
a  
torque demand is defined by the constraints below: brk;drv $twh;brk bdry 
þ mgCrr cos q þ mg sin q þ  . 
m þ Itot R2wh $Rwh
aaccel;drv $twh;accel bdry Wbatt;dischg bdry  Waccelec
tmot;dmd ¼ min ; ;
Rtc $Rfd umot (7)
!
  In Eq. (7), Itot addresses the total inertia of powertrain, which is
tmot;accel bdry c twh;drv > 0 given as below:
 
(5a) Itot ¼ ðImot $Rtc þ Itc Þ$Rfd þ Ifd þ nwh $Iwh (8)
During vehicle braking, the braking torque and motor regener-
ative torque demands are dependent on the constraints of the In Eqs. (6)e(8), uwh , ufd , utc , and umot are the speeds of the
regenerative conditions, as well as motor and battery operation wheel, final drive, torque coupler and motor; twh;dmd , tfd;dmd , ttc;dmd ,
boundaries, as given below. During vehicle braking, both the motor and tmot;dmd are the torque demands of wheel, final drive, torque
regenerative and braking torque demands are negative. coupler and motor; htc and hfd are torque coupler and final drive
828 Z. Gao et al. / Energy 172 (2019) 823e839

efficiencies; Imot , Itc , Ifd , and Iwh are the inertia of the motor, torque speed, torque and battery SOC (see Fig. 4(a)e(c), respectively),
coupler, final drive and wheel, respectively; nwh is number of implying the basic simulation assumptions for the EV powertrain
wheels on the vehicle. were indeed reasonable.
However, it is difficult to calibrate the E-bus and E-Truck model
directly in the absence of available experimental data. To confirm
2.2. EV configurations and model validation that the above assumptions give reasonable electric bus and truck
predictions, the Autonomie software was used to construct virtual
To identify the energy saving potential from LD to HD electric E-bus and E-truck simulations based on the specifications listed in
vehicles over eco-driving cycles, three battery electric powertrain Table 1. The battery energy consumption predicted by Autonomie
configurations were specified to account for a LD compact car, a MD was then compared with our models over the given drive-cycle
transient bus, and HD delivery truck, respectively. The selected scenario. Results indicated a less than 5% difference, implying
electric motor, battery and other key component parameters for the that these E-bus and E-Truck assumptions are also reasonably
three EVs are listed in Table 1. The LD electric car was created based accurate.
on a Nissan Leaf configuration, in which the used motor map was
generated from the measurement from a 2012 Nissan Leaf motor
2.3. Eco-driving methodology
rated at 80 kW (see Fig. 2). The specification of the motor and
battery for the E-bus and E-truck were estimated based on ORNL's
Coasting, instead of braking, is an efficient method for providing
commercial vehicle electrification evaluation tool (CVEET) which is
required decelerations. For example, appropriate coasting enables
a framework tool that uses a quantitative method of electric vehicle
vehicles to avoid stops at traffic signals without any delay or excess
energy consumption and appropriate charging considerations, as
energy consumption. Avoiding braking saves fuel since it eliminates
well as using the Fleet DNA statistical drive behavior data for
the need for subsequent accelerations to return to the same speed,
evaluating feasible or market-ready MD/HD BEV technologies [7].
or if extended coasting is used for vehicle deceleration it avoids
The battery sizes of the bus and truck are determined based on their
unnecessary energy dissipation of braking. A simplified eco-driving
75% service coverage [7]. Since the motor maps of HD vehicles are
model that provides similar speed profiles to an EAD application by
still rarely available, the motor maps used in the simulated E-bus
employing appropriate coasting was developed to minimize
and E-truck models were produced based on scaling of the Nissan
braking. The approach determines an appropriate coasting speed
Leaf's motor. This might be inaccurate due to the configuration
profile at locations where braking is required at stops or traffic
difference between LD car motors and HD vehicle motors, but it
signals. In the approach, the times of braking in the drive cycle can
would still be reasonable for estimating the general trend of HD
be identified using a tractive power evaluation (see Eq. (1)). The
electric vehicle performance over eco-driving cycles.
coasting deceleration rate is determined based on Eq. (9), derived
To confirm that the assumptions of the simulated vehicles
from Eq. (1), to develop an optimized eco-drive segment that
reasonably reflect the performance of electric vehicles, the LD EV
minimizes braking and stops.
simulations were carried out using the current model and
compared with the chassis dyno data of a 2012 Nissan Leaf dV 1 
following the regular urban dynamometer driving schedule acoast ¼ ¼ g sinqðxÞ  a þ b V þ c V2 : (9)
dt m
(UDDS), as collected by Argonne National Laboratory (ANL) [56]. For
the city drive cycle of 12 km (i.e. 7.45 miles) and 1372s, with a where a, b, c are the coastdown coefficients, which can be derived
temperature of 16.7  C (i.e. 72  F), the predicted and measured from Eq. (1) or obtained from coastdown experiments such as are
battery energy consumptions were 125.5 Wh/km (i.e. 202 Wh/ used prior to vehicle dynamometer testing. In the equation, the
mile) and 129.8 Wh/km (i.e. 209 Wh/mile), respectively. The grade, sin q(x) ¼ dh/dx, is shown explicitly as a function of position
simulation error is less than 4%. Moreover, the predicted transient since the elevation and grade are inherently distance-dependent.
motor performance also matches very well with the measurement A distance-based formulation of acceleration and speed can
of both motor speed and torque. Fig. 4 shows fairly good agreement efficiently determine the coasting speed profile that starts or ends
between the simulations and measurements of transient motor at a specified location. Briefly, an integral equation for the distance-

Table 1
Specifications of electrical vehicle components and key parameters.

Component Parameters LD Car Electric Bus Electric Truck

Aerodynamic drag coeff. Cd 0.32 0.7 0.62

Rolling resistance coeff. Crr 0.007 0.009 0.009
Frontal Area Af (m2) 2.01 9.0 9.0
Wheel tire Wheel radius (m) 0.3125 0.5 0.53
Inertia (kg-m2) 0.25 1.0 1.0
Final drive Final ratio 7.94 7.94 7.94
Inertia per wheel tire (kg-m2) <0.01 <0.01 <0.01
Torque coupler TC ratio 1.01 1.6 1.6
Inertia (kg-m2) <0.01 <0.01 <0.01
Motor Max power (kW) 84 212 265
Continuous power (kW) 42 106 132
Max torque (Nm) 280 700 874
Inertia (kg-m2) 0.03 0.06 0.08
Battery Capacity (kWh) 24 265 265
Peak chg power (kW) 148 1500 1500
Peak dis power (kW) 140 1580 1580
Normal Voltage (V) 374 581 581
Electric accessory Constant power (kW) 0.2 5.0 4.0
Vehicle mass Mass (kg) 1515 14407 15434
 Z. Gao et al. / Energy 172 (2019) 823e839 829

50 8000
40 Measurement Measurement
7000
Simulation Simulation
30
6000

Motor Speed (rpm)

Motor Power (kW)
20
10 5000

0 4000
-10 3000
-20
2000
-30
-40 1000

-50 0
0 200 400 600 800 1000 1200 1400 0 200 400 600 800 1000 1200 1400
Time (s) Time (s)
(a) Motor power (b) Motor speed
1.00
0.99 Measurement
Simulation
0.98
0.97
Battery SOC

0.96
0.95
0.94
0.93
0.92
0.91
0.90
0 200 400 600 800 1000 1200 1400
Time (s)
(c) Battery SOC
Fig. 4. Comparison of (a) motor power, (b) speed and (c) battery SOC for the simulated electric car over the cold-start UDDS cycle.

based speed profile can be obtained by multiplying Eq. (9) by dx ¼ V end point of the coasting intersects the original drive cycle at both
dt and integrating, and the speed profile will be determined as a the same time and location. It is apparent that starting at a later
function of position by numerically obtaining the root of the time and location results in overtaking the original drive cycle,
resulting equation. This coasting solution can be evaluated from while starting at an earlier time/position results in the distance of
any starting location and speed, but different end points or times the coast always trailing the original drive cycle. The case with
will of course be obtained for different starting locations. Fig. 5 equal time and distance is the desired ideal case for the coasting
shows several coasting results starting at different starting times/ optimization since it will provide a minimum change in headway
locations for the same drive cycle. In Fig. 5(a) the speed vs. time is throughout the drive cycle. The speed vs. time data in Fig. 5(a)
shown, while in Fig. 5(b), the distance vs. time is shown for the provides an insight into how this is achieved. Since the distance
same set of coasting solutions. traveled is the integral of the speed vs. time plot, we must have the
The curve shown in green corresponds to the case where the areas between the original drive cycle and the coasting segment

(a) Speed vs. time (b) Distance vs. time

Fig. 5. Speed (a) and distance (b) vs. time for several coasting solutions starting at different locations/times for the drive cycle.
830 Z. Gao et al. / Energy 172 (2019) 823e839

equal for the regions above and below the coast segment. For are matched. Figs. 6e8 show the comparison of the original driving
Fig. 5(a), the area between the black and green curves where they cycle and the eco-driving cycle. The key driving features are sum-
intersect from about 425s to 465s must equal the area between the maried in Table 2.
curves from 465s to 485s (all times are only approximate). The Figs. 6(a)e(b), 7(a)e(b) and 8(a)e(b) show elevation and speed
methodology allows arrival at a downstream location at the same profiles, respectively, as a function of distance for each of the three
time as occurred in the original cycle. This optimized coasting so- vehicles. Figs. 6(c), 7(c) and 8(c) show the vehicle speed profiles as a
lution eliminates braking for different positions where braking was function of time. It is clear that the three eco-driving cases repre-
present in the original drive cycle, yielding a highly optimized drive senting the EAD application eliminate stops significantly. In the
cycle. eco-driving cycles, the braking tractive energy (the integral of all
We note that when downhill grades are present, the coasting periods requiring braking power) is reduced by 65.4%, 46.9% and
solution can result in accelerations, so there may be periods of 70.8% for the passenger car, bus and delivery truck, respectively;
increasing speeds during the coasting evaluation. Furthermore, in also, the propulsive tractive energy (integral of all positive tractive
some situations with negative grades it may not be possible to powers) is reduced by 32.7%, 32.3% and 12.9%, respectively, for the
obtain an optimized solution with pure coasting that results in a passenger car, bus and delivery truck. Compared to the passenger
coasting end point at the same time and location as the original car and delivery truck, the braking tractive energy reduction of city
cycle. bus is slightly less. This is mainly a result of scheduled bus stops still
being included in the optimized cycle for the simulated E-bus. The
3. Results and discussion required bus stops significantly limit the braking reductions for the
case.
3.1. Eco-drive cycles With these optimized drive cycles, powertrain evaluations
comparing the energy consumption for the original and modified
To carry out eco-driving cycle simulations for a compact car, bus drive cycles were carried out to estimate the effect on the driving
and delivery truck, three typical driving cycles were selected using efficiency that the EAD application could provide and to quantify
on-road measurement data (see Table 2). The cycles selected how it impacts the efficiency of individual powertrain components.
include road grades determined from elevation lookup of GPS
location data. The passenger car case is a city driving cycle with a 3.2. BEV performance over eco-driving
total length of 17.5 km (i.e. 10.9 miles), which was measured in the
Knoxville, Tennessee. The bus case is a city driving cycle, 34.5 km Three battery electric vehicles were evaluatedea LD passenger
(i.e. 21.4 miles) in length, which was measured in Riverside, Cali- car, a MD transient bus, and a HD delivery truck. These vehicles
fornia. The regional delivery cycle is a combined highway and city represent completely different driving patterns. In the simulated
driving cycle that is 92.7 km (i.e. 57.6 miles) long, and was also passenger car, the battery energy consumption with and without
measured in the Knoxville area. The regional delivery truck typi- eco-driving was 107.0 Wh/km (i.e. 172 Wh/mile) and 146.7 Wh/km
cally picks up cargo from a warehouse, travels to downtown (i.e. 236 Wh/mile), respectively. The eco-driving allows the simu-
Knoxville (about 30 miles), and makes multiple deliveries. lated LD EV to achieve 27% battery energy savings. The detailed
The proposed eco-driving methodology is used to create eco- component energy losses are shown in Fig. 9, revealing that the key
driving optimized drive cycles which are used to estimate the po- component energy losses are aerodynamic drag, motor loss, rolling
tential benefits of eco-cruise and traffic signal EAD applications. For resistance, and drivetrain loss. Eco-driving reduces the aero-
our methodology, we identify the segments where braking occurs dynamic drag loss by 20%, motor loss by 51% and drivetrain loss by
(i.e. tractive power<0), and then use coasting for the deceleration 50%, but eco-driving does not affect the rolling resistance loss. In
prior to the start of braking. The distance traveled in the original addition, the eco-driving reduces frictional braking loss by 65% and
drive cycle is considered a constraint, as vehicle speeds over any battery loss by 39%, although the magnitude of these actual energy
given drive cycle represent constraints imposed by traffic condi- losses are relatively small compared to the top four component
tions. For example, the drive cycle speed may be limited by the energy losses. For example, the frictional braking losses are 5.4 Wh/
speed of a lead vehicle and cannot be overtaken during driving, but km (i.e. 8.7 Wh/mile) without eco-driving and 1.9 Wh/km (i.e.
it is acceptable to decelerate thereby increase the gap to the lead 3.1 Wh/mile) with eco-driving, respectively, representing 3.6% and
vehicle. The starting time and location for each coasting segment 1.8% of battery energy consumptions in the cases. However, there is
must be determined by iteration. A solution is accepted when no a different mechanism for the case with and without eco-driving in
overtake has occurred, the speed at the end of the segment matches frictional braking loss reduction. The eco-driving reduces frictional
that of the original drive cycle at the same time, and the positions braking loss via appropriate speed control and avoiding any braking

Table 2
Key driving characteristics of the passenger car, city bus and delivery truck drive cycles.

Vehicle Passenger Car City Bus Delivery Truck

Drive mode wo eco-drive w/eco-driving wo eco-drive w/eco-driving wo eco-drive w/eco-driving

Max Acceleration (m/s2) 2.60 2.66 1.29 1.25 1.28 1.27

Average Acceleration (m/s2) 0.45 0.32 0.28 0.27 0.17 0.15
Max deceleration (m/s2) 2.73 1.48 2.00 1.92 1.72 0.71
Average deceleration (m/s2) 0.42 0.19 0.51 0.14 0.18 0.12
Max Speed (mile/hr) 68.57 67.92 41.91 40.67 66.13 66.45
Average Speed (mile/hr) 29.09 29.09 13.51 13.51 40.52 40.53
Distance (mile) 10.89 10.89 21.77 21.75 57.67 57.68
Time (s) 1352 1352 5802 5802 5127 5127
Positive tractive energy (kWhr/km)a 0.135 0.091 1.015 0.687 1.090 0.950
Negative tractive energy (kWhr/km)a 0.048 0.017 0.561 0.298 0.155 0.045
a
Estimated based on EVs listed in Table 1.
 Z. Gao et al. / Energy 172 (2019) 823e839 831

360

Elevation (m)
(a)
340
320
300
280
140
Vehicle Speed (km/hr)

LD Vehicle wo eco-Driving
120 (b)
LD Vehicle w/ eco-Driving
100
80
60
40
20
0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Distance (km)
140
Vehicle Speed (km/hr)

LD Vehicle wo eco-Driving
120 (c)
LD Vehicle w/ eco-Driving
100
80
60
40
20
0
0 200 400 600 800 1000 1200 1400
Time (s)
Fig. 6. Passenger car speed profiles with and without eco-driving; (a) elevation profile vs. distance; (b) speed profile vs. distance; (c) speed profile vs. time.

losses, but the EV without eco-driving reduces frictional braking accessories. In addition, the eco-driving reduces braking loss by
loss via direct braking regeneration. The latter transfers a portion of 42%, battery loss by 40%, and final drive loss by 38%, but the net
regenerated kinetic energy to motor energy losses. Fig. 9 details the level of the energy losses is fairly small compared to the losses
impact of eco-driving on transient motor and battery energy con- related to the motor, rolling resistance, accessories, and aero-
sumption in the simulated LD EV. It is evident that eco-driving dynamic drag. In the E-bus simulation, eco-driving does not elim-
operation eliminates most braking regen, as the braking regener- inate most braking regen. This is because the E-bus still experiences
ation power becomes nearly zero. Instead, the EV operation frequent stopping for scheduled bus stops, which leads to un-
without eco-driving regenerates braking energy, but the regener- avoidable braking loss. In the E-bus cycle, the potential energy
ation process effectively doubles the motor energy loss (i.e. energy contribution is 0.24 kWh/mile, since the elevation decreases from
loss during regeneration and propulsion) for the portion of power 310 m at the start of the drive cycle down to 175 m at the end (see
that is regenerated. This explains why the eco-driving enables a Fig. 7).
significant decrease in motor energy loss (see Fig. 9(a) and (b)). Fig. 11 compares the E-truck energy loss with and without eco-
Fig. 10 shows the impact of eco-driving on the E-bus energy loss. driving. The eco-driving optimized drive cycle results in the
In the simulated transient bus, the battery energy consumption simulated delivery E-truck achieving an 8% battery energy savings.
without eco-driving is 1.17 kWh/km (i.e. 1.89 kWh/mile) compared The battery energy consumption with and without eco-driving are
to 0.92 kWh/km (i.e. 1.48 kWh/mile) for the E-bus with eco-driving. 1.25 kWh/km (i.e. 2.02 kWh/mile) and 1.37 kWh/km (i.e. 2.21 kWh/
The eco-driving cycle provides the simulated E-bus with 22% bat- mile), respectively. The smaller benefit in this case is mainly
tery energy savings. However, unlike the LD EV above, the top four because the delivery truck drives a significant distance on the
component energy losses of the E-Bus are motor loss, rolling highway, which does not provide substantial braking reduction
resistance, accessories, and aerodynamic drag due to unique transit opportunities (see the first 2000s shown in Figs. 8 and 11). In the
bus driving characteristics (see Fig. 10). The eco-driving reduces simulation, truck platooning is not considered. The detailed
motor loss by 38% and aerodynamic drag loss by 16%, but the eco- component energy losses from the E-truck are shown in Fig. 11. For
driving operation does not affect rolling resistance loss or the E-truck case, the top four component energy losses are
832 Z. Gao et al. / Energy 172 (2019) 823e839

350 reduces the energy loss percentage from the motor and final
drive. For example, in the passenger EV, the motor energy loss with
Elevation (m)

300
eco-driving is 19% of the total losses compared to 27% for the non
250 (a) eco-driving case; the final drive energy loss with eco-driving is 5%
200 compared to 8% for the non eco-driving case. However, the com-
150 bined energy loss percentage from rolling resistance and aero-
100 dynamic drag actually increases with eco-driving. The
100 contributions are 68%, 53%, and 72% for the passenger car, E-bus
Vehicle Speed (km/hr)

90 (b) Bus wo eco-Driving

80 Bus w/ eco-Driving
and E-truck, respectively, for the eco-driving case, compared to
70 55%, 43% and 68% for these vehicle without eco-driving. This means
60 that developing innovative technologies to reduce rolling resis-
50
40 tance and aerodynamic drag with eco-driving can provide even
30 greater improvements or potential opportunities in the EV perfor-
20 mance with eco-driving. In particular, rolling resistance will
10
0 become greater for EVs due to the battery mass penalty, so low
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 rolling resistance tire technology is even more attractive for
Distance (km) improving EV performance. In addition, the drivetrain efficiency
100 improvement is also important for further EV energy savings. The
Vehicle Speed (km/hr)

90 (c) Bus wo eco-Driving

80 Bus w/ eco-Driving
energy loss related to the drivetrain is around 5e6% with eco-
70 driving and 6e8% without eco-driving.
60
50
40 3.3. Discussion
30
20 One intriguing question that, to the authors' knowledge, had not
10
0 been addressed previously is whether eco-driving results in a larger
0 1000 2000 3000 4000 5000 6000 energy savings benefit for EVs or conventional vehicles. To address
Time (s) this, three conventional powertrain vehicle models, comparable in
size and performance to the EVs already considered, were created
Fig. 7. City bus speed profiles with and without eco-driving; (a) elevation profile vs.
using Autonomie [49] based on the basic vehicle performance pa-
distance; (b) speed profile vs. distance; (c) speed profile vs. time.
rameters shown in Table 1. The simulated conventional passenger
car is powered by a 5-speed automatic transmission and a 1.8-L
aerodynamic drag, rolling resistance, motor loss, and final drive stoichiometric gasoline engine with a peak power of 97 kW, and
loss. The eco-driving decreases aerodynamic drag loss by 5%, motor the simulated LD vehicle weight is 1414 kg which is 100 kg less than
loss by 25%, and final drive loss by 25%, but eco-driving does not the simulated electric car. The weight difference is based on that
affect rolling resistance loss. Also, the eco-driving reduces brake between the Nissan Leaf and Altima. The simulated conventional
loss by 50% and battery loss by 15%, but the net values of these transit bus is powered by a 6-speed automatic transmission and a
energy losses are smaller than for the other losses. In the E-truck 215-kW diesel engine; the simulated bus weight is 13,000 kg which
simulations, the braking loss associated with the limited delivery is 1407 kg less than the simulated E-bus. The weight difference is
stops does not result in a substantial impact on the total energy mainly due to the battery mass penalty estimated using ORNL's
consumption, since it represents less than 1% of the overall energy commerical vehicle electrification evaluation tool (CVEET) [7]. The
consumption for the complete drive cycle. Fig. 11(d) shows that the simulated conventional truck is powered by a 10-speed automated
8% battery energy savings is obtained during the city operation. The manual transmission and a 265-kW diesel engine; the simulated
eco-driving achieves very small benefits during highway operation. truck weight is 14,186 kg which is 1248 kg less than the simulated
It is evident from Figs. 9e11 that the component energy losses of E-truck. The weight difference is mainly due to the battery mass
different vehicle types vary significantly due to the different driving penalty also estimated using the ORNL commerical vehicle elec-
behaviors and vehicle characteristics. Fig. 12 summarizes each trification evluation tool (CVEET) [7]. These conventional vehicle
component's percent contribution to the total energy consumption powertrain models developed using Autonomie have been
for the passenger car, transit bus, and delivery truck with and described in our previous studies [57e63] and the models have
without eco-driving. For the EVs without eco-driving, the brake demonstrated reasonable accuracy. Therefore, we do not provide
energy loss is already quite small, around 1e4% except for the E-bus details of the models here to avoid lengthy descriptions.
which is 8% due to the requirement for scheduled bus stops. The The impact of eco-driving on the conventional vehicles is
low level of friction braking is due to the use of regenerative summarized in Fig. 13. The results indicate that eco-driving enables
braking in all EVs, which causes much of the kinetic energy to be a rather significant reduction in the conventional vehicles' braking
recovered. Nonetheless, the higher level of regeneration causes a energy loss. Frictional braking is reduced by 89%, 50% and 88% in
higher loss from the motor to be experienced, which is reduced the conventional passenger car, transit bus and HD truck, compared
with eco-driving. The eco-driving drive cycle optimization enables to 65%, 42% and 50% for the comparable E-car, E-bus and E-truck,
a brake loss to be reduced to around 1e2% for the E-passenger car respectively. This is reasonable because, unlike EV's braking regen
and E-truck cases, and 6% for the E-bus. The key component losses capability, conventional vehicles typically lost all braking energy
of EVs are aerodynamic drag loss, rolling resistance loss, motor loss, over non-eco-driving conditions; on the other hand, without bat-
final drive, as well as accessory loads for buses, with or without the tery mass penalty, conventional vehicles' light weight leads to less
eco-driving. The combined energy losses due to aerodynamic drag, frictional braking demand and improves frictional braking reduc-
rolling resistance, the motor and final drive are responsible for tion in eco-driving patterns. However, the results also show that
nearly 80e90% of the battery energy consumptionin all three of the eco-driving leads to frequent engine operation at low loads, which
vehicles. Compared to the non-eco-driving case, the eco-driving typically correspond to a low efficiency. The engine average effi-
ciencies are 22.4% without eco-driving vs. 17.0% with eco-driving
 Z. Gao et al. / Energy 172 (2019) 823e839 833

400
350 (a)

Elevation (m)
300
250
200
150
100
140
Vehicle Speed (km/hr)

120 (b)
100
80
60
40 HD Truck wo eco-Driving
20 HD Truck w/ eco-Driving

0
0 20 40 60 80 100
Distance (km)
140
Vehicle Speed (km/hr)

HD Truck wo eco-Driving
120 (c)
HD Truck w/ eco-Driving
100
80
60
40
20
0
0 1000 2000 3000 4000 5000
Time (s)
Fig. 8. Delivery truck speed profiles with and without eco-driving; (a) elevation profile vs. distance; (b) speed profile vs. distance; (c) speed profile vs. time.

for the conventional passenger car, 33.2% without eco-driving vs. respectively.
31.0% with eco-driving for the conventional transit bus, and 37.4% Clearly, eco-driving technologies enable improvements in en-
without eco-driving vs. 36.9% with eco-driving for the conventional ergy saving for both EVs and conventional vehicles, but the relative
truck. Fig. 14 compares the overall energy saving benefits of EVs and benefits are complex and depend on many factors. For conventional
conventional vehicles for the three classes of vehicles. Relative to vehicles, eco-driving seems to yield higher percentage savings for
EVs, eco-driving actually provides conventional vehicles with HD vehicles.
greater reductions in the net mechanical energy output. This is
reasonable since the drivetrain efficiency, including clutch and
4. Conclusion
transmission, of conventional vehicles, achieves higher efficiency
than EV motor. However, when including engine efficiency, the
A general EV simulation model that comprehensively addresses
conventional vehicle's fuel energy saving becomes complex. For the
component and system performance was developed to evaluate
conventional passenger car, the fuel energy savings due to eco-
energy consumption and component performance of LD/MD/HD
driving are less than for the EV in spite of the greater savings in
electric vehicles. The model enables consistent and meaningful
mechanical energy output. The main reason is the low efficiency of
comparisons between electric vehicle technology options by
the engine operating at lower loads experienced in the eco-driving
identifying energy efficiency trends and opportunities for future
case. Unlike the passenger car, the conventional HD truck is able to
research and development of vehicle components and powertrain
achieve a higher fuel energy saving and a greater reduction in en-
systems. The simulations were carried out to identify the potential
gine mechanical energy output simultaneously. In the HD truck, the
benefits of eco-driving for electrified passenger cars, E-buses and E-
average engine efficiency does not vary significantly with vs.
trucks.
without eco-driving, with predicted values of 36.9% and 37.4%,
The results demonstrate that eco-driving boosts EV energy
 Unit:Wh/km 4.4 4.3 Unit:Wh/km 2.7
Acc_loss 4.3 Acc_loss
Cd_loss 20.1 Cd_loss
40.2 43.7 Crr_loss Crr_loss
35.2
Brake_loss Brake_loss
5.8
Ptnl_enrg Ptnl_enrg
-0.1
FD_loss FD_loss
11.8 1.9
Mot_loss Mot_loss
37.1 Ess_loss 37.1 Ess_loss
-0.1
5.4 Overall battery energy consumption: 146.7 Wh/km Overall battery energy consumption: 107.0 Wh/km
(a): LD EV without eco-driving (b): LD EV with eco-driving
80 3.00
LD Vehicle wo eco-Driving
60 LD Vehicle w/ eco-Driving
2.50

Battery Energy (kWh)

40
Motor Power (kW)

2.00
20

0 1.50
0 200 400 600 800 1000 1200 1400
-20
1.00
-40
LD Vehicle wo eco-Driving 0.50
-60
LD Vehicle w/ eco-Driving

-80 0.00
Time (s) 0 200 400 600 800 1000 1200 1400
Time (s)
(c): Transient motor power (d): Battery energy consumption
Fig. 9. Comparison of the EV energy losses for the LD passenger car (a) without eco-driving and (b) with eco-driving; (c) transient motor power and (d) cumulative battery energy
consumption for non-eco-driving and eco-driving modes.

Unit:kWh/km 0.01 Unit:kWh/km 0.01

Acc_loss Acc_loss
0.23 0.24 0.23
Cd_loss Cd_loss
0.39
Crr_loss Crr_loss
0.16 0.06
Brake_loss Brake_loss
0.13
Ptnl_enrg Ptnl_enrg
-0.15
0.10 FD_loss FD_loss
-0.15 0.35 Mot_loss Mot_loss
Ess_loss 0.35 Ess_loss
0.09 0.05

Overall battery energy consumption: 1.17 kWh/km Overall battery energy consumption: 0.92 kWh/km
(a): E-bus without eco-driving (b): E-bus with eco-driving
250 45
Bus wo eco-Driving
200 40 Bus w/ eco-Driving
150 35
Battery Energy (kWh)
Motor Power (kW)

100
30
50
25
0
20
-50
15
-100
-150 10
Bus wo eco-Driving
-200 Bus w/ eco-Driving
5
-250 0
0 1000 2000 3000 4000 5000 6000 0 1000 2000 3000 4000 5000 6000
Time (s) Time (s)
(c): Transient motor power (d): Battery energy consumption
Fig. 10. Comparison of the EV energy losses for the transit city bus (a) without eco-driving and (b) with eco-driving; (c) transient motor power and (d) cumulative battery energy
consumption for non-eco-driving and eco-driving modes.
 Z. Gao et al. / Energy 172 (2019) 823e839 835

Unit:kWh/km 0.03 Unit:kWh/km 0.03

0.06 0.06
Acc_loss Acc_loss
0.25 0.19
Cd_loss Cd_loss
Crr_loss Crr_loss
0.01 0.06
0.08 Brake_loss Brake_loss
0.01 0.55 0.00
Ptnl_enrg 0.53 Ptnl_enrg
0.01
FD_loss FD_loss
Mot_loss 0.38 Mot_loss
0.38
Ess_loss Ess_loss

Overall battery energy consumption: 1.37 kWh/km Overall battery energy consumption: 1.26 kWh/km
(a): E-truck without eco-driving (b): E-truck with eco-driving
400 140
HD truck wo eco-Driving
300 120 HD truck w/ eco-Driving

200

Battery Energy (kWh)

100
Motor Power (kW)

100
80
0
60
-100
40
-200
HD Truck wo eco-Driving 20
-300
HD Truck w/ eco-Driving
-400 0
0 1000 2000 3000 4000 5000 6000 0 1000 2000 3000 4000 5000 6000
Time (s) Time (s)
(c): Transient motor power (d): Battery energy consumption
Fig. 11. Comparison of the EV energy losses for the delivery truck (a) without eco-driving and (b) with eco-driving; (c) transient motor power and (d) cumulative battery energy
consumption for non-eco-driving and eco-driving modes.

savings significantly. The eco-driving approach employed, which process of regeneration and subsequent acceleration to speed re-
provides reductions in braking similar to EAD technology, was sults in a double motor energy loss (i.e. there is inefficiency during
found to provide varying levels of energy savings for different types both regeneration and propulsion). This is a dominant effect of eco-
of vehicles due to differences in drive cycles and vehicle charac- driving that impacts the energy loss from EVs, and the results
teristics. For a passenger car EV, an E-bus and an E-truck, the model showed reductions in motor losses from 40.2 to 20.1 Wh/km (i.e.
predicts an overall energy savings from eco-driving of 27%, 22% and 64.7 to 32.3 Wh/mile) in the passenger EV, from 0.38 to 0.24 kWh/
8%, respectively. The key component losses of the EVs include km (i.e. 0.62 to 0.38 kWh/mile) in the e-bus and from 0.25 to
aerodynamic drag, rolling resistance, motor loss, and drivetrain 0.19 kWh/km (i.e. 0.40 to 0.30 kWh/mile) in the e-truck, repre-
loss, as well as accessory loads for buses. In all of these EV cases, the senting 50%, 39% and 25% reductions, respectively.
braking energy loss is very nearly eliminated. However, the eco- The impact of eco-driving on conventional vehicles was also
driving optimized drive cycles do not impact rolling resistance studied and compared to comparable BEVs. The observations show
loss. Rolling resistance becomes more important for EVs due to the that eco-driving provides a larger reduction in the conventional
mass penalty associated with excess battery weight, and since vehicle's braking energy loss compared to the comparable BEVs.
other losses are reduced while rolling resistance remains the same, This is due to the EVs employing regenerative braking. However,
its relative contribution to the total energy loss becomes more the eco-driving leads to frequent engine operation at low loads,
significant. Therefore, low rolling resistance tire technology can corresponding to lower efficiency. Compared to BEVs, eco-driving
have a significant impact on the efficiency of EVs with eco-driving. provides conventional vehicles with greater reductions in the
EV energy savings are also sensitive to motor and drivetrain effi- propulsion system (motor or engine) mechanical energy output.
ciency improvements. Eco-driving technology reduces the motor However, because of the effect of engine efficiency, the fuel savings
energy loss significantly by reducing the regenerative braking. achieved with eco-driving for the conventional vehicles are com-
Although regeneration captures kinetic energy which would plex. With the passenger car, the percent fuel energy savings with
otherwise be lost, there are still important energy losses from the eco-driving achieved for the conventional vehicle is less than the
motor. The portion of the initial kinetic energy that is returned after percent battery energy savings for the EV with eco-driving, in spite
a braking and reacceleration event is significantly reduced since the of the greater reduction in mechanical energy output. On the other
836 Z. Gao et al. / Energy 172 (2019) 823e839

Battery 3% Battery 1% Battery 2%

Motor 18%
Motor 27% Motor 33%
Final Drive 6%
Ptnl Enrg 1%
Brake <1%
Final Drive 8% Final Drive 8%
Battery Ptnl Enrg ~0% Ptnl Enrg -13%
energy Brake 4%
Brake 8%
Rolling Resist
28%
100% Rolling Resist
(wo eco- 25% Rolling Resist
30%
driving)
Aero drag 40%
Aero drag 13%
Aero drag 30%

Auxillar 20%
Auxillar 3% Auxillar 4%

Passenger Car Transit City Bus Delivery Truck

(a) No eco-driving
Battery 3% Battery 1% Battery 2%

Motor 15%
Motor 19% Motor 26%
Final Drive 5%
Final Drive 5% Ptnl Enrg 1%
Ptnl Enrg ~0% Final Drive 6% Brake <1%
Brake 2% Ptnl Enrg -17%
Battery Brake 6%

energy Rolling Resist

30%
Rolling Resist
100% 35% Rolling Resist
38%
(w/ eco-
driving)
Aero drag 14%
Aero drag 42%
Aero drag 33%

Auxillar 25%

Auxillar 4% Auxillar 5%
Passenger Car Transit City Bus Delivery Truck
(b) eco-driving
Fig. 12. The energy consumption percentage distribution among energy loss factors for the passenger car, transit bus, and delivery truck with and without eco-driving.
 Z. Gao et al. / Energy 172 (2019) 823e839 837

Fig. 13. Energy consumption distribution of the (a)e(b) conventional passenger car, (c)e(d) transit bus, and (e)e(f) delivery truck with and without eco-driving.

50% insights in this work. Thanks also go to the editors and reviewers
EV-Battery Energy Saving for their support and volunteered time.
45% Conventional Vehicle-Fuel Energy Saving
Conventional Vehicle-Engine Mech Energy Saving
40% References
Impact of Eco-Driving

35% [1] Vehicle Technology Office. https://www.energy.gov/eere/vehicles/vehicle-

technologies-office; 2018 [Accessed 20 September 2018].
30%
[2] Du J, Ouyang M, Chen J. Prospects for Chinese electric vehicle technologies in
25% 2016e2020: ambition and rationality. Energy 2017;120:584e96. https://doi.
org/10.1016/j.energy.2016.11.114.
20% [3] Gao Z, Lin Z, LaClair TJ, Liu C, Li JM, Birky AK, et al. Battery capacity and
recharging needs for electric buses in city transit service. Energy 2017;122:
15% 588e600. https://doi.org/10.1016/j.energy.2017.01.101.
[4] Block D, Brooker P. Prediction of electric vehicle penetration. 2018. http://fsec.
10% ucf.edu/en/publications/pdf/FSEC-CR-2069-17.pdf. [Accessed 20 September
2018].
5% [5] California Environmental Protection Agency Air Resources Board. Technology
assessment: medium- and heavy- duty battery electric trucks and buses.
0% 2018. https://www.arb.ca.gov/msprog/tech/techreport/bev_tech_report.pdf.
[Accessed 20 September 2018].
Passenger Car Transit Bus HD Truck [6] Gao Z, Lin Z, Franzese O. Energy consumption and cost savings of truck
electrification for heavy duty vehicle applications. Transport Res Rec: J
Fig. 14. Comparison of the percent energy savings of EVs and conventional vehicles Transport Res Board 2017;2628:99e109. https://doi.org/10.3141/2628-11.
resulting from eco-driving. [7] Gao Z, Lin Z, Davis SC, Birky AK. Quantitative evaluation of MD/HD vehicle
electrification using statistical data. Transport Res Rec: J Transport Res Board
2018;2672(24):109e21. https://doi.org/10.1177/0361198118792329.
hand, eco-driving provides the conventional HD truck with higher [8] TransPower. Economical electric school bus (EESB) final project report-June 2,
2014. 2018. http://www.transpowerusa.com/downloads/TransPower-
relative fuel energy savings and greater engine mechanical energy Economical-Electric-School-Bus-Final-Report.pdf. [Accessed 20 September
output savings simultaneously. 2018].
[9] Adomani. Help yellow go green. 2018. http://www.adomanielectric.com/
phoenix.zhtml?c¼254475&p¼irol-products. [Accessed 20 September 2018].
Acknowledgment [10] Lyden S. The state of all-electric trucks in the U.S. medium duty market. 2018.
http://zerotruck.com/wp-content/downloads/GRN_medium.pdf. [Accessed 20
This work was sponsored by the U.S. Department of Energy's September 2018].
[11] California HVIP. Hybrid and zero-emission vehicles. 2018. https://www.
Vehicle Technologies office and ARPA-E NEXT-CAR program. The californiahvip.org/docs/HVIP_EligibleVehicles.pdf. [Accessed 5 July 2018].
authors thank our colleagues for their helpful suggestions and [12] Calstart. Electric truck & bus grid integration opportunities, challenges &
838 Z. Gao et al. / Energy 172 (2019) 823e839

recommendations. 2018. http://www.westcoastelectricfleets.com/portfolio- and efficiency during deceleration for X-EV vehicles. SAE Int J Alt Power
items/electric-truck-bus-grid-integration-opportunities-challenges-and- 2013;2(2):350e61. https://doi.org/10.4271/2013-01-1473.
recommendations/. [Accessed 20 September 2018]. [39] Hao P, Wu G, Boriboonsomsin K, Barth MJ. Eco-approach and departure (EAD)
[13] Muller J. Cummins beats tesla to the punch, unveiling heavy duty electric application for actuated signals in real-world traffic. IEEE Trans Intell Trans-
truck. 2018. https://www.forbes.com/sites/joannmuller/2017/08/29/take- port Syst 2018;2018. https://doi.org/10.1109/TITS.2018.2794509.
that-tesla-diesel-engine-giant-cummins-unveils-heavy-duty-truck-powered- [40] Hao P, Boriboonsomsin K, Wang C, Wu G, Barth M. Connected eco-approach
by-electricity/#7195f52678f1. [Accessed 20 September 2018]. and departure (EAD) system for diesel trucks. TRB 2018. 18-06464.
[14] Thiemer H. Influence of automotive 42 V powernet on small PM DC motors. [41] Fredette D, Ozguner U. Dynamic eco-driving's fuel saving potential in traffic:
In: Electric machines and drives conference. IEMDC 2001; 2001. p. 591e3. multi-vehicle simulation study comparing three representative methods. IEEE
https://doi.org/10.1109/IEMDC.2001.939373. Trans Intell Transport Syst 2018;19(9):2871e9. https://doi.org/10.1109/TITS.
[15] Chau KT, Chan CC, Liu C. Overview of permanent-magnet brushless drives for 2017.2766767.
electric and hybrid electric vehicles. IEEE Trans Ind Electron 2008;55(6): [42] Qi X, Barth MJ, Wu G, Boriboonsomsin K, Wang P. Energy impact of connected
2246e57. https://doi.org/10.1109/TIE.2008.918403. eco-driving on electric vehicles. In: Meyer G, Beiker S, editors. Road vehicle
[16] Rahman KM, Schulz SE. Design of high-efficiency and high-torque-density automation 4. Cham: Springer; 2018. p. 97e111. https://doi.org/10.1007/978-
switched reluctance motor for vehicle propulsion. IEEE Trans Ind Appl 3-319-60934-8_10.
2002;38(6):1500e7. https://doi.org/10.1109/TIA.2002.805571. [43] Zhang R, Yao E. Eco-driving at signalised intersections for electric vehicles. IET
[17] de Santiago J, Bernhoff H, Ekergård B, Eriksson S, Ferhatovic S, et al. Electrical Intell Transp Syst 2015;9(5):488e97. https://doi.org/10.1049/iet-its.2014.
motor drivelines in commercial all electric vehicles: a review. IEEE Trans Veh 0145.
Technol 2012;61(2). 475-48, https://doi.org/10.1109/TVT.2011.2177873. [44] Flehmig F, Sardari A, Fischer U, Wagner A. Energy optimal adaptive cruise
[18] Etacheri V, Marom R, Elazari R, Salitra G, Aurbach D. Challenges in the control during following of other vehicles. In: 2015 IEEE intelligent vehicles
development of advanced Li-ion batteries: a review. Energy Environ Sci symposium (IV), Seoul; 2015. p. 724e9. https://doi.org/10.1109/IVS.2015.
2011;4(9):3243e62. https://doi.org/10.1039/C1EE01598B. 7225770.
[19] Offer GJ, Howey D, Contestabile M, Clague R, Brandon NP. Comparative [45] Qi X, Wang P, Wu G, Boriboonsomsin K, Barth MJ. Energy and mobility ben-
analysis of battery electric, hydrogen fuel cell and hybrid vehicles in a future efits from connected ecodriving for electric vehicles. IEEE Intell Transport Syst
sustainable road transport system. Energy Policy 2010;38(1):24e9. https:// (ITSC) 2017:1e6. https://doi.org/10.1109/ITSC.2017.8317744.
doi.org/10.1016/j.enpol.2009.08.040. [46] Jiang H, Hu J, An S, Wang M, Park BB. Eco approaching at an isolated signalized
[20] Onat NC, Noori M, Kucukvar M, Zhao Y, Tatari O, Chester M. Exploring the intersection under partially connected and automated vehicles environment.
suitability of electric vehicles in the United States. Energy 2017;121:631e42. Transport Res Part C 2017;79:290e307. https://doi.org/10.1016/j.trc.2017.04.
https://doi.org/10.1016/j.energy.2017.01.035. 001.
[21] Tie SF, Tan CW. A review of energy sources and energy management system in [47] Stephens TS, Gonder J, Chen Y, Lin Z, Liu C, Gohlke D. Estimated bounds and
electric vehicles. Renew Sustain Energy Rev 2013;20:82e102. https://doi.org/ important, factors for fuel use and consumer costs of connected and auto-
10.1016/j.rser.2012.11.077. mated vehicles. 2016. NREL/TP-5400-67216.
[22] Nykvist B, Nilsson M. Rapidly falling costs of battery packs for electric vehicles. [48] Huang X, Peng H. Speed trajectory planning at signalized intersections using
Nat Clim Change 2015;5(4):329e32. https://doi.org/10.1038/nclimate2564. sequential convex optimization. In: IEEE American control conference (ACC);
[23] Cuma MU, Koroglu T. A comprehensive review on estimation strategies used 2017. p. 2992e7. https://doi.org/10.23919/ACC.2017.7963406.
in hybrid and battery electric vehicles. Renew Sustain Energy Rev 2015;42: [49] Autonomie. 2018. https://www.autonomie.net/. [Accessed 20 September
517e31. https://doi.org/10.1016/j.rser.2014.10.047. 2018].
[24] Wang T, Martinez MJJ, Sename O. Hinf observer-based battery fault estimation [50] Gao Z, Smith DE, Daw CS, Edwards KD, Kaul BC, Domingo N, et al. The eval-
for HEV application. In: IFAC workshop on engine and powertrain control uation of developing vehicle technologies on the fuel economy of long-haul
simulation and modeling (E-COSM'2012); 2012. p. 7. trucks. Energy Convers Manag 2015;106:766e81. https://doi.org/10.1016/j.
[25] Singh A, Izadian A, Anwar S. Fault diagnosis of Li-Ion batteries using multiple- enconman.2015.10.006.
model adaptive estimation. In: IECON 2013-39th Annual conference of the [51] Gao Z, LaClair TJ, Smith DE, Daw CS. Exploring fuel-saving potential of long-
IEEE. Industrial Electronics Society; 2013. p. 3524e9. haul truck hybridization. Transport Res Rec: J Transport Res Board
[26] Onori S, Spagnol P, Marano V, Guezennec Y, Rizzoni G. A new life estimation 2015;2502:99e107. https://doi.org/10.3141/2502-12.
method for lithium-ion batteries in plug-in hybrid electric vehicles applica- [52] Burress T. Benchmarking state-of-the-art technologies. 2018. https://www.
tions. Int J Power Electron 2012;4(3):302e19. https://doi.org/10.1504/IJPELEC. energy.gov/sites/prod/files/2014/03/f13/ape006_burress_2013_o.pdf.
2012.046609. [Accessed 20 September 2018].
[27] Yu X, Shen T, Li G, Hikiri K. Regenerative braking torque estimation and [53] Chen M, Rincon-Mora G. Accurate electrical battery model capable of pre-
control approaches for a hybrid electric truck. In: IEEE American control dicting runtime and I-V performance. IEEE Trans Energy Convers 2006;21(2):
conference; 2010. p. 5832e7. https://doi.org/10.1109/ACC.2010.5530501. 504e11. https://doi.org/10.1109/TEC.2006.874229.
[28] Yu X, Shen T, Li G, Hikiri K. Model-based drive shaft torque estimation and [54] Gao L, Liu S, Dougal A. Dynamic lithium battery model for system simulation.
control of a hybrid electric vehicle in energy regeneration mode. ICCAS-SICE; IEEE Trans Compon Packag Technol 2002;25(3):495e505. https://doi.org/10.
2009. p. 3543e8. 1109/TCAPT.2002.803653.
[29] Ravey A, Wang R, Lukic S, Miraoui A. Distance estimation algorithm for plugin [55] Hammad N. Vehicle valve regulated lead acid battery modeling and fault
hybrid electric vehicle control strategy. In: IEEE transportation electrification diagnosis. 2010. SAE Technical Paper: 2010-01-0028, https://doi.org/10.4271/
conference and expo (ITEC); 2013. p. 1e4. https://doi.org/10.1109/ITEC.2013. 2010-01-0028.
6573504. [56] Downloadable Dynamometer Database. https://www.anl.gov/energy-
[30] Kim IS. Nonlinear state of charge estimator for hybrid electric vehicle battery. systems/group/downloadable-dynamometer-database; 2018 [Accessed 20
IEEE Trans Power Electron 2008;23(4):2027e34. https://doi.org/10.1109/ September 2018].
TPEL.2008.924629. [57] Daw CS, Gao Z, Smith DE, LaClair TJ, Pihl JA, Edwards KD. Simulated fuel
[31] Abushnaf J, Rassau A. Impact of energy management system on the sizing of a economy and emissions performance during city and interstate driving for a
grid-connected PV/Battery system. Electr J 2018;31(2):58e66. https://doi.org/ heavy-duty hybrid truck. SAE Int J Commercial Veh 2013;6(1):161e82.
10.1016/j.tej.2018.02.009. https://doi.org/10.4271/2013-01-1033.
[32] Hori Y, Toyoda Y, Tsuruoka Y. Traction control of electric vehicle: basic [58] Gao Z, Kim MY, Choi JS, Daw CS, Parks JE, Smith DE. Cold-start emissions
experimental results using the test EV“ UOT electric march”. IEEE Trans Ind control in hybrid vehicles equipped with a passive hydrocarbon and NOx
Appl 1998;34(5):1131e8. https://doi.org/10.1109/28.720454. adsorber. Proc IMechE Part D: J Automob Eng 2012;226(10):1396e407.
[33] Benbouzid MEH, Diallo D, Zeraoulia M. Advanced fault-tolerant control of https://doi.org/10.1177/0954407012443764.
induction-motor drives for EV/HEV traction applications: from conventional [59] Gao Z, Daw CS, Wagner RM, Edwards KD, Smith DE. Simulating the impact of
to modern and intelligent control techniques. IEEE Trans Veh Technol premixed charge compression ignition on light-duty diesel fuel economy and
2007;56(2):519e28. https://doi.org/10.1109/TVT.2006.889579. emissions of particulates and NOx. Proc IMechE Part D: J Automob Eng
[34] Wu J, Wang X, Li L, Du Y. Hierarchical control strategy with battery aging 2013;227(1):31e51. https://doi.org/10.1177/0954407012459137.
consideration for hybrid electric vehicle regenerative braking control. Energy [60] Gao Z, LaClair TJ, Daw CS, Smith DE, Franzese Q. Simulations of the fuel
2018;145:301e12. https://doi.org/10.1016/j.energy.2017.12.138. economy and emissions of hybrid transit buses over planned local routes. SAE
[35] Yoong MK, Gan YH, Gan GD, Leong CK, Phuan ZY, Cheah BK, et al. Studies of Int J Commercial Veh 2014;7(1):216e37. https://doi.org/10.4271/2014-01-
regenerative braking in electric vehicle. IEEE Sustainable Utilization and 1562.
Development in Engineering and Technology (STUDENT); 2010. p. 40e5. [61] Gao Z, Finney C, Daw C, LaClair T, Smith D. Comparative study of hybrid
https://doi.org/10.1109/STUDENT.2010.5686984. powertrains on fuel saving, emissions, and component energy loss in HD
[36] Fiori C, Ahn K, Rakha HA. Microscopic series plug-in hybrid electric vehicle trucks. SAE Int J Commercial Veh 2014;7(2):414e31. https://doi.org/10.4271/
energy consumption model: model development and validation. Transport 2014-01-2326.
Res Transport Environ 2018;63:175e85. https://doi.org/10.1016/j.trd.2018. [62] Gao Z, Curran SJ, Parks JE, Smith DE, Wagner RM, Daw CD, et al. Drive cycle
04.022. simulation of high efficiency clean combustions on fuel economy and exhaust
[37] Thomas J, Huff S, West B, Chambon P. Fuel consumption sensitivity of con- properties in light-duty vehicles. Appl Energy 2015;157:762e76. https://doi.
ventional and hybrid electric light-duty gasoline vehicles to driving style. SAE org/10.1016/j.apenergy.2015.03.070.
Int J Fuels Lubr 2017;10(3):672e87. https://doi.org/10.4271/2017-01-9379. [63] Gao Z, Daw CS, Smith DE, LaClair TJ, Parks JE, Jones PT. Comparison of parallel
[38] Rask E, Santini D, Lohse-Busch H. Analysis of input power, energy availability, and series hybrid power trains for transit bus applications. Transport Res Rec:
 Z. Gao et al. / Energy 172 (2019) 823e839 839

J Transport Res Board 2016;2570:97e106. https://doi.org/10.3141/2570-11. aaccel;drv : Driver acceleration demand

abrk;drv : Driver braking demand
Nomenclature b: A temperature factor
r: Air density
Dε: Battery temperature-dependent potential-correction
Af : Frontal area dspd and ddec : The constraint factors during brake regeneration
Cd : Aerodynamic drag coefficient hfd : Final drive efficiency
Crr : Rolling resistance coefficient hmot : Motor efficiency
Ctran;S : Capacity-related short-time constant of the step response of RC networks htc : Torque coupler efficiency
Ctran;L : Capacity-related long-time constants of the step response of RC networks hinv : Motor efficiency
Ftract : Required vehicle tractive force q: Road grade
hA: Battery cell heat transfer rate tfd;dmd : Final drive torque demand in powertrain
Ibatt : Battery package current output tmot : Motor torque
Icell : Single battery cell outlet current tmot;dmd : Motor torque demand in powertrain
Ifd : Final drive inertia tmot;pos bdry : Motor positive torque boundary
Imot : Motor inertia tmot;neg bdry : Motor negative torque boundary
Itc : Torque coupler inertia
ttc;dmd : Torque coupler torque demand in powertrain
Itot : Total inertia
Iwh : Wheel tire inertia
twh;dmd : Wheel torque demand in powertrain
Kp and Kl: PI control parameters twh;drv : Driver wheel torque demand
m: Vehicle mass twh;brk bdry : Wheel braking torque demand boundary
mcell Cpcell : Mass capacity twh;accel bdry : Wheel acceleration torque demand boundary
Mparallel : Battery cells in parallel ufd : Final drive speed
Nseries : Battery cells in series umot : Motor speed
Ro: Series resistor utc : Torque coupling speed
Rfd : Final drive ratio uwh : Wheel speed
Rtc : Torque coupler ratio
Rtran;S : Resistance-related short-time constant of the step response of RC networks Subscript
Rtran;L : Resistance-related long-time constants of the step response of RC networks
Rwh : Wheel radius accel: Acceleration
Tamb : Ambient temperature accelec: Electric accessary
Tcell : Battery cell temperature batt: Battery
Vcell : Single battery cell outlet voltage bdry: Boundary
V and Vdrv : Vehicle velocity brk: Brake
Vbatt : Battery package voltage output chg: Charging
Vdrv : Driving vehicle speed dischg: Discharging
VSOC : Open circuit voltage for single battery cell dmd: Demand
Vtarget : Targeted vehicle speed drv: Drive
Vtran;S : Voltage of short-time step response of RC networks ee: Electrical energy
Vtran;L : Voltage of long-time step response of RC networks fd: Final drive
Waccelec : Electric accessary load inv: Inverter
Wbatt;chg bdry : Battery charging power boundary me: Mechanical Energy
Wbatt;dischg bdry : Battery discharging power boundary mot: Motor
Winv;ee : Motor mechanical power output neg: Negative
Wmot;me : Inverter electrical power input pos: Positive
SOC: State of charge
Greek tc: Torque coupler
wh: Wheel

a: A charging/discharging rate factor