Received: 2 March 2020 Revised: 28 May 2020 Accepted: 9 June 2020

DOI: 10.1002/er.5700

SPECIAL ISSUE RESEARCH ARTICLE

Electric vehicle energy consumption modelling

and estimation—A case study

Ilyès Miri1 | Abbas Fotouhi1 | Nathan Ewin2

1
School of Aerospace, Transport and
Manufacturing, Cranfield University,
Summary
Cranfield, UK Electric vehicles (EVs) have a limited driving range compared to conventional
2
Delta Motorsport Ltd, Silverstone, UK vehicles. Accurate estimation of EV's range is therefore a significant need to
eliminate “range anxiety” that refers to drivers' fear of running out of energy
Correspondence
Abbas Fotouhi, School of Aerospace, while driving. However, the range estimators used in the currently available
Transport and Manufacturing, Cranfield EVs are not sufficiently accurate. To overcome this issue, more accurate range
University, College Road, Cranfield,
Bedford, MK43 0AL, UK.
estimation techniques are investigated. Nonetheless, an accurate power-based
Email: [email protected] EV energy consumption model is crucial to obtain a precise range estimation.
This paper describes a study on EV energy consumption modelling. For this
purpose, EV modelling is carried out using MATLAB/Simulink software based
on a real EV in the market, the BMW i3. The EV model includes vehicle
powertrain system and longitudinal vehicle dynamics. The powertrain is mod-
elled using efficiency maps of the electric motor and the power electronics'
data available for BMW i3. It also includes a transmission and a battery model
(ie, Thevenin equivalent circuit model). A driver model is developed as well to
control the vehicle's speed and to represent human driver's behaviour. In addi-
tion, a regenerative braking strategy, based on a series brake system, is devel-
oped to model the behaviour of a real braking controller. Auxiliary devices are
also included in the EV model to improve energy consumption estimation
accuracy as they can have a significant impact on that. The vehicle model is
validated against published energy consumption values that demonstrates a
satisfactory level of accuracy with 2% to 6% error between simulation and
experimental results for Environmental Protection Agency and NEDC tests.

KEYWORDS
battery, electric vehicle, energy consumption estimation, modelling, simulation

Abbreviations: AC, Alternative Current; ANN, Artificial Neural Network; DC, Direct Current; ECN, Equivalent Circuit Network; EPA,
Environmental Protection Agency; EU, European Union; EV, Electric Vehicle; EVs, Electric Vehicles; FTP-75, Federal Test Procedure-75; GPS,
Global Positioning System; HWFET, Highway Fuel Economy Test; Li-ion, Lithium-ion; MLR, Multiple Linear Regression; NEDC, New European
Driving Cycle; PCA, Principal Component Analysis; PI, Proportional-Integral; PID, Proportional-Integral-Derivative; REx, Range Extender; RWD,
Rear-Wheel-Drive; SoC, State-of-Charge; SoH, State-of-Health; US, United-States; WLTP, Worldwide Harmonised Light Vehicle Test Procedure.

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided
the original work is properly cited.
© 2020 The Authors. International Journal of Energy Research published by John Wiley & Sons Ltd

Int J Energy Res. 2021;45:501–520. wileyonlinelibrary.com/journal/er 501

502 MIRI ET AL.

1 | INTRODUCTION proposed concept. The authors believe that same tech-

nique can be applied to other types of EV and the general
Global warming and environmental pollution have led outcomes of this study do not depend on the vehicle type
to more severe regulations on CO2 and other pollutant used here. As an important part of this study, the pro-
emissions. In this context, electric vehicles (EVs) have posed vehicle model is validated against experimental
become an alternative to conventional vehicles as they data obtained from the literature.
offer a zero-emission alternative. Besides, they are Two main approaches are used for EV modelling3:
cheaper to be recharged as electricity is cheaper than pet- (a) Forward approach also called “dynamic approach”
rol/diesel and also energy recovery is possible from or cause-effect method, and (b) Backward approach also
regenerative braking in EVs. called “quasi-static approach” or effect-cause method.
However, the EVs' market penetration rate is not very The forward approach is based on equations of the
quick because of their limited range, charging time, powertrain components behaviour and the dynamic inter-
battery replacement cost, and other limitations related action between the components. This approach requires
to infrastructure. This study is particularly related to one a driver controller to set the start of the calculations.
of these restrictions, the limited EV range. This limitation Therefore, the driver behaviour can be studied using this
causes an issue called “range anxiety” that refers to approach. The controller is implemented to model the
drivers' fear of running out of energy while driving.1 This driver that has to press/release either the accelerator or
phenomenon can be limited by increasing the battery the brake pedal in order to reduce the error between the
capacity and/or the number of charging stations. How- actual speed and the speed from a drive cycle.4 The driver
ever, both solutions are expensive, and will not improve model provides the torque demand to match the drive
the confidence of drivers in the remaining driving range cycle speed profile. Thereafter, from the driver set-point,
estimation. Nowadays, range estimators are not suffi- the energy required to overcome the opposing forces act-
ciently accurate because they are mainly based on vehicle ing on the vehicle is computed. The backward approach
historical data. Thus, they can lead to major estimation considers a reference speed profile, as input, to determine
errors and cannot be fully trusted by drivers. “Range anx- the forces acting at the wheels and then processes back-
iety” can be reduced by improving the range estimation ward through the powertrain. Subsequently, the model
to increase drivers' confidence. For EV range estimation, computes the motor torque and the energy required from
an accurate estimation of the EV's energy consumption is the battery to power the electric motor.4 The advantages
vital and is therefore the purpose of this study. of the forward method are that the driving speed profile
In this study, the energy flow is only considered does not need to be known5 and that it can be easily and
inside the vehicle so, the energy flow between the grid rapidly used for prototyping and hardware in the loop
and vehicle is out of the framework. Generally, the EV testing.6 Besides, it is suitable to identify the interactions
energy consumption refers to the sum of: between components that can affect the energy consump-
tion and the performance of the vehicle. Even though it
• Energy that is required at the wheels to propel the requires more computational effort to solve the model's
vehicle, differential equations, the forward approach is more accu-
• Energy losses along the powertrain, and rate than the backward approach.4
• Energy that is required for the operation of the auxil- Vehicle energy consumption is affected by several
iary devices. factors that can be divided into two main categories7,8:
(a) Internal Factors associated with the vehicle itself
New techniques are required for more accurate EV (vehicle design parameters, characteristics, efficiency and
energy consumption/range estimation aiming to reduce inertia of the vehicle components, auxiliary devices
“range anxiety” and increase the driving range. In fact, usage, etc.), and (b) External Factors associated with driv-
higher range can be achieved by giving more confidence to ing conditions (environmental and traffic conditions, road
the drivers, enabling them to extend the use of their vehicle type and conditions, driving behaviour, etc.). To develop
on a single charge. This idea comes from knowing that an accurate EV energy consumption estimation model,
nowadays, most of the drivers only use about 70% of the the impact of all factors must be carefully examined
estimated remaining battery energy due to a lack of because road slope, for instance, has a major impact on
confidence.2 that.9 The external factors demonstrates different level of
In this study, EV energy consumption estimation variability with regard to real world driving conditions.
is the main focus and it is performed based on vehicle Depending on their variability and their predictability,
modelling using MATLAB/Simulink software. The BMW they are classified in three categories10,11: Stable, Dynamic
i3 is selected as the case-study here to demonstrate the but easy to predict and Dynamic and difficult to predict.
MIRI ET AL. 503

For range estimation, most of the car manufacturers that includes acceleration data while a third model is pro-
use an approach based on analysis of a short history of posed for instantaneous energy consumption estimation
energy consumption to predict it in the near future. In while driving.15 In Reference 16 an improved MLR
that method, it is assumed that the rate of energy con- energy consumption model based on the extraction of
sumption remains unchanged in a short prediction hori- real-world data and speed profile prediction using Neural
zon. However, this approach is not accurate since it does Networks is presented. That model also considers the
not consider the changes in driving conditions that may energy consumption of auxiliary devices. The proposed
occur.10 EV energy consumption estimation models can model performs well even in the existence of changes in
be classified in three main categories: Analytical, Statisti- driving behaviour and environmental conditions.16 Fur-
cal and Computational models.7 thermore, Principal Component Analysis (PCA) is used
Analytical models work based on longitudinal vehicle in Reference 7 to study the impact of each factor sepa-
dynamics and electric motor losses estimation from avail- rately. PCA consists of transforming a set of correlated
able efficiency maps.2,12,13 Longitudinal vehicle dynamics variables into a set of new uncorrelated variables using
is modelled from the vehicle dynamics theory to calculate an orthogonal transformation. The EV energy consump-
the required power at the wheels to overcome the opposing tion is then estimated as a polynomial combination of
forces. In some studies, the regenerative braking is mod- each variable weighted depending on its relative impor-
elled as a linear function of vehicle speed to estimate the tance.7 Statistical models demonstrate a good applicabil-
energy recovered while braking or driving downhill2 or it is ity since they require less computational effort than
modelled as a function of vehicle deceleration.12 The model analytical models however, they are less accurate.
developed in Reference 12 considers the instantaneous EV Computational models based on artificial neural net-
speed and acceleration to provide an accurate second-by- works (ANN) are developed to determine relationships
second energy consumption estimation. Contrary to,2 the between a number of affecting factors and EV energy con-
model not only considers the motor efficiency but also the sumption.7,17,18 This type of models are used to estimate
efficiency of other powertrain components. The impact of EV energy consumption as a function of the input factors,
auxiliary devices is considered as well for an improved esti- where a weight is determined for each factor depending on
mation.12 The model introduced in Reference 13 expresses its relative importance using training algorithms.17 This
the relationship between EV power, speed, acceleration approach performs well in fitting nonlinear relationships
and road grade to determine the required power at the between input and output variables.7 ANN can also be
wheels. The model can be either used for instantaneous used for prediction of driving behaviour by classifying driv-
energy consumption estimation or energy consumption ing patterns using Global Positioning System data. This
prediction over a trip for eco-route planning.13 The existing method is found to be a powerful approach as it is data-
regenerative braking models are improved in this study by driven and self-adaptive.18 As part of this method, cluster-
considering the limitations coming from the battery and ing techniques are used to recognize similar patterns in a
the electric motor. set of data in order to gather data with similar proper-
Statistical models are based on the analysis of real- ties.19-23 It is applicable to driving pattern recognition and
world driving data to derive empirical relationships driving behaviour analysis which is used to improve the
between different factors and EV's energy consumption. accuracy of energy consumption estimation models.24 The
For this purpose, regression models that consider both computational models found in the literature do not take
EV dynamic behaviour and powertrain efficiency, have into account all the factors affecting the energy consump-
been developed.14-16 For example, the model developed tion such as auxiliary devices that can affect significantly
in Reference 14, works based on historical and real-time the vehicle energy consumption.
data analysis in order to derive polynomial combinations The selection of an estimation approach depends on
of EV speed, acceleration and battery State-of-Charge the targeted application. In general, statistical and compu-
(SoC) under different operation modes. In Reference 15, tational models require more computational effort than
three regression models using multiple linear regression analytical models. However, they are more accurate as they
(MLR) method, have been developed with different levels work based on data analysis and probabilistic prediction.
of detail. Road characteristics, traffic conditions, driving In addition, analytical models can only reflect changes in
style and environmental conditions are well considered vehicle behaviour as they are based on vehicle dynamics
to update the estimation models for an improved accu- and physical modelling. Using analytical models, it is diffi-
racy. The first model is used for energy consumption cult to take into account factors associated with driving
estimation over a trip for route planning and does not conditions such as environmental and traffic conditions.10
consider weight variation and acceleration, assuming However, some hybrid methods that are both physics-
constant EV speed. It is improved with a second model based and data-driven have been developed in References
504 MIRI ET AL.

25 and 26. In Reference 26, a simulation tool called “Auto- • Estimation of the power consumption of auxiliary
nomie” is described that is developed by Argonne National devices.
Laboratory. That simulation tool works based on vehicle • Estimation of the efficiency of electric motor and
parameters and data analysis and it is used for vehicle inverter from efficiency maps.
energy consumption calculation. Autonomie has demon- • Precise modelling of a regenerative braking strategy
strated good accuracy against test data and is widely used currently used on the market.
by the industry. Such hybrid methods combine the advan-
tages of both analytical and data-driven models. Unlike conventional vehicles, the range of EVs is lim-
Energy consumption estimation models can be used ited even if the battery capacity has been increased in the
for various applications: newly available vehicles. Therefore, the use of highly pre-
cise range estimators is still a major issue in EVs. How-
• Estimation before a trip for route planning as part of ever, the current range estimators work on the basis of
an eco-routing system. For a targeted destination, the vehicle's historical data analysis and are therefore not
system determines the best route by minimising the very accurate. For range estimation, an accurate model of
energy consumption based on the current traffic and the EV's energy consumption is essential. Such a model
environmental conditions.7 can be implemented in EV range estimators to assess the
• Estimation second-by-second to provide dynamic infor- energy consumption of any EV model.
mation about the vehicle energy consumption. The main objective of this paper is to introduce an accu-
• Eco Approach and Departure application to provide rate modelling approach for EV energy consumption esti-
recommendations to reduce the energy consumption mation. In order to demonstrate the proposed concept and
when approaching signalised intersections. Examples validate the results, a case-study on BMW i3 has been cho-
are calculating the optimum speed to reach the next sen as a typical EV in the market. So, the goal is to model
traffic signal on a green light or to come to a stop in the target EV including its powertrain system and longitudi-
the most efficient way and display it to the driver.27 nal dynamics and then validate it using the available data.

In this study, a model-based EV energy consumption

estimator is proposed and validated for a case-study on 2 | VEHICLE MODELLING
BMW i3. For this purpose, a forward EV powertrain model
is developed using MATLAB/Simulink software. The pro- Architecture of the EV energy consumption estimation
posed model considers the power consumption of auxiliary model, developed in this study, is presented in Figure 1. The
devices in contrast to the work presented in Reference 2. consumed energy, Econs, is calculated as per unit of distance
The power consumption of auxiliary devices is estimated (Wh/m) derived from the battery power output Pbat29:
from average values found in the literature, which is dis-
cussed in this paper. This estimation is included in the EV E bat
E cons = , ð1Þ
model since auxiliary devices can have a significant impact d
ð ð !
on the vehicle energy consumption. In addition, the effi- 1
ciency values of power electronics and electric motor are E bat = Pb − out ðτÞ dτ − Pb − in ðτÞ dτ # ,
traction braking 3600
estimated from the efficiency maps whereas they are
assumed to be constant in previous studies such as the ð2Þ
work performed in Reference 28. The efficiency values are
interpolated over the entire range of the electric motor RTotal # V Vehicle
Pb − out = and Pb − in = α # Pregen , ð3Þ
using efficiency maps available in the literature. Further- ηPowertrain
more, the regenerative braking strategy that is modelled in
this study works based on the series brake system configu- where Ebat is the battery energy output in (Wh), d is the
ration used on the BMW i3. This strategy considers several distance travelled in (m), RTotal is the total resistance forces
factors affecting the regenerative braking capability of the opposed to the vehicle motion in (N), VVehicle is the vehicle
EV such as vehicle speed, acceleration and battery SoC. As speed in (m/s), ηPowertrain is powertrain efficiency (includ-
a consequence, the braking strategy tends to be more accu- ing power electronics, electric motor and transmission), α
rate than previous models developed for example in Refer- is the percentage of the braking energy that can be recov-
ences 2 and 12. ered (0 < α < 1), that is also called regenerative braking
Summarising the aforementioned literature, the nov- factor and Pregen is the regenerative power that is calculated
elties of this study are as follows: based on X BRegen and motor's limitations as follows:
MIRI ET AL. 505

FIGURE 1 Forward electric vehicle model architecture

X BRegen :r d for SoC level above 6% which is sufficient for the proof
T Br_demanded = , ð4Þ
G # ηG of concept in this study. BMW i3 is a rear wheel drive
(RWD) EV with one electric motor at the rear axle. The
PBr_demanded = T Br_demanded :ωmotorðsÞ , ð5Þ power transmission between the motor and the wheels is
achieved by a single-speed automatic transmission sys-
tem. The vehicle specifications are presented in Table 1.
Pb − out and Pb − in are respectively the power provided In addition, the efficiency maps of the inverter and the
by the battery for vehicle motion and the power electric motor that are used in the BMW i3, are shown in
regenerated to charge the battery considering electric Figure 2. The torque and power curves of the electric
motor braking capabilities in generator mode. motor are also shown in the same figure.
As aforementioned, the battery power output Pbat is
divided into two main parts:
2.2 | Vehicle model
• Power that is used to propel the vehicle (Pb − out): the
battery must supply this power to overcome the oppos- Since this study aims at EV energy consumption estima-
ing forces and any power losses along the powertrain tion, only the powertrain system and the longitudinal
system (Power out). vehicle dynamics are modelled. The lateral dynamics is
• Power that is regenerated during braking (Pb − in): part neglected as it does not have a major impact on vehicle's
of the braking energy can be recovered from regenera- energy consumption. Three main power flows are consid-
tive braking by operating the motor in generator mode ered in the proposed model:
and charging the battery (Power in).
• Energy flow from the battery pack to the wheels to
In the following sections, individual components of the propel the vehicle.
proposed model are explained in more details. In order to • Energy flow from the wheels to the battery pack dur-
simulate the model, numerical values of BMW i3 are used ing energy recovery by regenerative braking.
as a popular EV in the market. The proposed model is then • Energy flow from the battery pack to the auxiliary sys-
validated against the available data for that particular EV. tems via the 12 V battery.

The vehicle model is developed in MATLAB/

2.1 | Vehicle specifications Simulink and consists of a combination of different sub-
systems listed in below:
The 2014 BMW i3 60Ah Range Extender (REx) is consid-
ered to be modelled as a case-study. For simplicity, the • Driving cycle subsystem including the reference speed
REx is not modelled since it is only used for battery SoC that vehicle must follow. This subsystem is the input of
below 6%.30 The vehicle model is therefore applicable the model.
506 MIRI ET AL.

TABLE 1 2014 BMW i3 60Ah Range Extender specifications31-33

VEHICLE BODY
Curb weight (EU) 1390 kg
Curb weight (US) 1420 kg
Aerodynamic drag coefficient 0.3
Frontal area 2.38 m2
Wheelbase 2570 mm
Static weight distribution 44.9/55.1 Front %/Rear %
(empty car)
Drivetrain Rear wheel drive (RWD)
POWERTRAIN
Number of motor(s) 1
Motor type Permanent magnet AC synchronous electric motor (BMW hybrid synchronous motor)
Motor operating range 0-11 400 rpm
Maximum power/at rpm 125/4775 kW/rpm
Maximum torque/at rpm 250/0-4475 Nm/rpm
Maximum regenerative brake 55 kW
power
TRANSMISSION
Type Single-speed automatic transmission
Simple fixed gear ratio 9.7:1
Tyres model Bridgestone Ecopia EP600
Front/rear tyres size 175/70 R19
Front/rear tyres radius 0.3638 m
BATTERY
Chemistry Lithium-ion
Battery configuration 8 Modules (96 Cells Connected in Series)
Nominal cell voltage 3.7 V
Nominal cell capacity 60 Ah
Nominal battery pack voltage 355.2 V
Nominal battery pack capacity 60 Ah
Nominal battery pack energy 22 kWh
PERFORMANCE
Top speed 150 km/h
Acceleration (0-100 km/h) 7.9 s
Driving modes Comfort
Eco Pro
Eco Pro +
Electric range (NEDC) 170 km
Electric range (EPA combined) 115 km
Energy consumption (NEDC) 13.5 kWh/100 km
Energy consumption (EPA 117 mpge
combined) 29 kWh/100mi

Abbreviations: EPA, Environmental Protection Agency; NEDC, New European Driving Cycle.
MIRI ET AL. 507

FIGURE 2 A, BMW i3 inverter efficiency map,34 B. electric motor efficiency map, and C. electric motor torque and power curves35
[Colour figure can be viewed at wileyonlinelibrary.com]

• Driver model that is responsible for controlling the • Transmission model for calculating the tractive force
vehicle motion by providing accelerator and brake by considering the energy losses while transmitting the
commands in the model. torque from the motor to the driving wheels.
• Brake system and the controller that is designed for • Battery subsystem that is designed to calculate energy
distributing the braking force/torque demand between demand from the battery pack by considering the limi-
friction and regenerative brakes. tations of battery in terms of voltage and current
• Electric motor and inverter model for computing the boundaries.
energy losses by considering the efficiency of the motor • Auxiliary subsystem that is designed to calculate the
and the inverter. power demand from auxiliary devices.
508 MIRI ET AL.

• Longitudinal vehicle dynamics subsystem to calculate As shown in Figure 4, the proposed driver model con-
the opposing forces and to update vehicle's velocity at sists of two subsystems: (a) the driver controller, and
each simulation time step. (b) accelerator and brake commands. PID controllers are
the most widely used controllers at industry as they are
Figure 3 illustrates the whole vehicle model including easy to be implemented. Besides, PI controllers have been
all the above-mentioned subsystems and interactions found to be widely used for driver modelling in several
between them. In the following sections, individual sub- previous studies. Therefore, a PI controller is chosen to
systems are explained with more details. model the driver in this study, as follows:
" #
1
PI ðsÞ = P + I # , ð6Þ
2.2.1 | Driver model s

The driver model aims to represent a human driver's where P and I are proportional and integral gains respec-
behaviour in the most realistic possible way. However, tively. The driver's pedal command tahen splits into brake
driving behaviour is a difficult phenomenon to be mod- and accelerator commands according to its sign as shown
elled because it depends on subjective factors such as in Figure 4. The driver's command is scaled between −100
driver's physical conditions and mood. In this study, and 100% corresponding to fully pressed brake pedal and
a simplified driver model is considered which is just fully pressed accelerator pedal respectively. Both com-
responsible to minimise the error (∆V) between the drive mands are then normalized between 0 and 1, corresponding
cycle (reference desired speed, Vdesired) and the actual to fully released and fully pressed pedal respectively.
vehicle's speed (Vactual). Depending on the sign of ∆V, In order to tune the PI controller's gains, New
driver's acceleration or brake command is generated to European Driving Cycle (NEDC) simulation case-studies
make the vehicle to follow the reference speed profile. are performed. During real drive cycle tests, the test driver
When ∆V is positive, an acceleration command (DA) is must follow a reference speed profile such as NEDC with
generated, meaning that the driver must press the acceler- maximum 2 km/h error.36 Therefore, to get a similar
ator pedal to increase vehicle's speed. On the other hand, behaviour to a human driver from the model, a maximum
when ∆V is negative, a brake command (DB) is generated. allowed deviation of 2 km/h (±2 km/h) is considered in
In that situation, the driver has two options, either press the simulations as well. The PI controller is therefore
the brake pedal to brake the vehicle using the frictions tuned according to the aforementioned criterion. A sensi-
brakes or release only the accelerator pedal to slow down tivity analysis is also conducted as shown in Table 2.
the vehicle by dynamic braking. The choice between one From this analysis, it can be concluded that when the
of these two options depends on the braking strategy. controller's gains are changed, the energy consumption

FIGURE 3 BMW i3 model in MATLAB/Simulink [Colour figure can be viewed at wileyonlinelibrary.com]

MIRI ET AL. 509

FIGURE 4 A, Driver model and B, driver accelerator and brake commands subsystem in Simulink

TABLE 2 Driver controller sensitivity analysis

Distance Energy consumption Maximum speed error (+) Maximum speed error (−)

Tuning setup (km) (Wh/km) (km/h) (km/h)

P = 10 and I = 1 11.02 129.44 7.15 −8.35
P = 60 and I = 1 11.02 126.97 1.36 −1.57
P = 60 and I = 2 11.02 127.64 1.42 −1.52
P = 100 and I = 1 11.02 127.12 0.86 −0.95

varies slightly while the variation of the speed error is vehicle and the adhesion between the tyres and the
substantial. As a result, because the actual vehicle speed road29,37:
must remain in the allowable range (±2 km/h) the con-
$ %
troller parameters are chosen as follows: P = 60 and I = 2. X BMAX = φ # Z f + Z r = φ # M Vehicle # g, ð7Þ
Reference velocity, actual vehicle velocity and the error
between them are demonstrated in Figure 5 for NEDC
simulation case-study. where φ is the adhesion coefficient between the tyres and
the road, Zf and Zr are the normal loads on front and rear
axles in (N) respectively, MVehicle is the vehicle mass in
2.2.2 | Braking strategy model (kg) and g is the acceleration due to gravity in (m/s2). Typ-
ical values of φ are around 0.8 on dry or wet asphalt and
In EVs, dynamic braking by the electric motor enables concrete surfaces.29 In the proposed EV model, the brak-
recharging the battery while driving. In order to estimate ing force is distributed between the friction and regenera-
the braking force required to slow down the vehicle, the tive brakes as follows:
maximum available braking force XBMAX must be deter-
X BFriction = X BMAX # DBFriction , ð8Þ
mined, which depends on the normal load acting on the
510 MIRI ET AL.

F I G U R E 5 Reference and
actual speed profiles on the New
European Driving Cycle [Colour
figure can be viewed at
wileyonlinelibrary.com]

X BRegen = X BMAX # DBRegen , ð9Þ • If PBr_demanded > PMax_Regen:

8
< PMax_Regen ,if ωmotorðsÞ 6¼ 0
>
where X BFriction and X BRegen are the friction and the regener- ωmotorðsÞ
TBr_Limited =
ative braking forces in (N) respectively, and DBFriction and >
: 0, if ωmotorðsÞ = 0
DBRegen are the friction and regenerative brake commands.
Regenerative braking is only effective at driven axles • If PBr_demanded ≤ PMax_Regen:
and is more efficient at the front axle due to the load 8
< PBr_demanded , if ωmotorðsÞ 6¼ 0
>
transfer from rear to front during braking giving ωmotorðsÞ
TBr_Limited =
more grip and increasing the normal load at the >
: 0, if ωmotorðsÞ = 0
front. In general, 65% of the braking energy goes to
the front axle. In addition, regenerative braking at
the rear axle is more limited by the legislation where ωmotor is the motor speed in (rpm). Thereafter, the
because the rear is more critical regarding wheel available electric motor brake command EMAvailable_BC is
locking.29 Furthermore, regenerative braking is found derived from the above limitation and the maximum avail-
to be limited by several factors such as battery charg- able braking force as follows:
ing power limitation and SoC, vehicle speed and
vehicle deceleration. A maximum regenerative brak- T Br_Limited # G # ηG
EM Available_BC = , ð10Þ
ing power is set to protect the battery since the bat- r d # X BMAX
tery charging power is limited for battery protection.
For the BMW i3, the regenerative braking power is where G is the single-speed gear ratio, ηG is the transmis-
limited to 55 kW at the wheels,32 which lead to a sion efficiency and rd is the dynamic tyre radius in (m).
limit of about 53 kW at the electric motor considering At low speeds, regenerative braking is inefficient so,
the transmission efficiency of 97%. it is disabled for safety reasons as it may cause the vehicle
The demanded motor braking power PBr_demanded is to brake when the vehicle is started. Thus, regenerative
compared to the maximum regenerative motor braking braking is set to zero at low speeds and it is progressively
power PMax_Regen to derive the limited motor braking tor- increased for smooth operation.38 Figure 6 shows the
que TBr_Limited as follows: speed-dependent regeneration factor depending on
MIRI ET AL. 511

threshold speeds u1 and u2 that must be determined. The where DB is the driver's brake command, DBFriction and DBRegen
regeneration factor is set to 0 below u1 and is increased are the friction and the regenerative brake commands
linearly between u1 and u2 up to 1. Usually, u1 and u2 respectively and EMAvailable_BC is the electric motor's
are set to 10 and 20 km/h respectively.38 available braking command. As a result, a regenerative
Above a certain level of deceleration, the electric motor braking strategy based on a series brake system is used in
is unable to brake the vehicle because the braking torque the proposed EV model by considering the limiting fac-
demand is too high, and the friction brakes must therefore tors mentioned above.
be used. The deceleration limit is set at 0.7g where g is the
acceleration due to gravity. Above this limit, regenerative
braking is disabled as shown in Figure 6. 2.2.3 | Power electronics and electric
Regenerative braking also depends on the battery machine model
SoC. It is disabled for SoC above 95% to avoid recharging
the battery when it is fully charged as shown in Figure 6. In this section, another sub-system of the proposed EV
The BMW i3 uses a series brake system because it is model is explained that is electric motor and power electron-
possible to brake most of the time by only releasing the ics. Since the goal of EV modelling is energy consumption
accelerator pedal to recover as much kinetic energy as estimation in this study, more focus here is on the efficiency
possible during the braking phases. The brake pedal is of the electric motor and power electronics as it affects the
thus only necessary for a complete stop or emergency overall energy consumption significantly. Energy losses due
braking.39,40 A series brake system is thus considered to the power electronics increase the energy that the battery
based on the following algorithm: has to provide to the electric motor and also reduce the
( energy effectively recovered from regenerative braking. The
DBRegen = DB
• If DB < EMAvailable_BC: on-board charger is not considered in the model since the
DBFriction = 0 energy loss between the grid and the EV battery is neglected
( in this study. Thus, only the inverter and the converter
DBRegen = EM Available_BC
• If DB ≥ EMAvailable_BC: are modelled here. The inverter efficiency is computed in
DBFriction = DB −EM Available_BC Simulink using a 2D lookup table that is prepared according

F I G U R E 6 Regenerative braking factors as a function of vehicle speed, vehicle deceleration and battery State-of-Charge [Colour figure
can be viewed at wileyonlinelibrary.com]
512 MIRI ET AL.

to the BMW i3 inverter efficiency map shown in Figure 2. by a 12 V battery that is charged by the high voltage
Since no specific information was available about the con- battery via a DC/DC converter. The power consump-
verter technology used in the BMW i3 in the public domain, tion of the auxiliary devices can significantly affect the
the converter efficiency is assumed to be 90% as the average overall EV's energy consumption. That is why, they
DC/DC converter efficiency is around 90%.41 must be included in the vehicle model for more accu-
Motor torque in Nm, motor speed in rpm and motor racy. The power demand of auxiliary devices is calcu-
efficiency must be taken into account in the vehicle model lated as follows:
too as they affect the vehicle energy consumption. The tor-
que demand is the input of the electric machine model PAc
PDem = , ð12Þ
whereas the output torque from the motor, by considering ηDC=DC # η12V_bat
motor and inverter efficiencies, is the output. The torque
demand TDem is derived from the driver model as follows: where PDem is the power demand in (W), PAc is the
power consumption of the auxiliary devices in (W) and
T Dem = T Max # DA , ð11Þ ηDC/DC and η12V_bat are the DC/DC converter and the
12 V battery efficiencies respectively. The energy con-
where TMax is the maximum available torque in (N) and sumption of the auxiliary devices depends on several fac-
DA is the driver acceleration command. TMax is equal to tors such as the ambient temperature. However, for the
the output torque from the motor divided by the efficiency sake of simplicity, average values are extracted from
of the motor and the inverter. The motor torque the literature as stated in Table 3. The table includes the
is computed in Simulink using a 1D lookup table that is main auxiliary devices in an EV and their average power
prepared according to the electric motor torque curve consumption.
shown in Figure 2. The motor efficiency is computed for a Thereafter, the effective power consumption of the
given motor speed and a given torque demand using a 2D auxiliary devices are estimated based on the devices acti-
lookup table that is prepared according to the available vated during the tests. For instance, during NEDC homol-
electric motor efficiency map. The electric machine model ogation tests, lights and auxiliary devices must be switched
developed in MATLAB/Simulink is shown in Figure 7. off, except those required for testing and day-time opera-
tion of the vehicle.36 From the list of auxiliary devices
shown in Table 3, it is assumed that only the driving con-
2.2.4 | Model of auxiliary devices trol and energy management systems are activated during
NEDC tests. Figure 8 shows the battery energy consump-
Nowadays, there are more and more auxiliary devices tion with and without auxiliary load on the NEDC. The
in vehicles for safety and comfort. They are powered battery energy consumption increases by 9% with a load

FIGURE 7 Electric machine model in Simulink

MIRI ET AL. 513

TABLE 3 List of the main electric vehicle auxiliaries42,43

Operation Power
Subsystem mode (W)
Air-conditioner Continuous 500
Audio Continuous 35
Driving control Continuous 150
Energy management system Continuous 150
Head and tail lamps Continuous 120
Parking, turn and interior lamps Intermittent 50
Horn Intermittent 10 FIGURE 9 Battery electrical circuit model (Thevenin model)
Power steering Continuous 400
Power windows Intermittent 80 According to the literature, the charging/discharging effi-
Window defroster Continuous 250 ciency of the Li-ion and the Lead Acid battery packs are
Wipers Continuous 40 assumed to be 95% and 80% respectively.29
According to the literature, the two most widely
used battery modelling techniques are the electrochemi-
cal and the equivalent circuit network (ECN) modelling
techniques.44 Electrochemical cell modelling approach
is the most accurate approach however, it requires signif-
icant computational effort because of its complexity. On
the other hand, ECN modelling approach is roughly
accurate, and it can be used in real-time applications too.
The high voltage battery is therefore modelled using The-
venin model that is the most famous ECN model shown
in Figure 9. The model consists of an internal voltage
source (VOC), an ohmic resistance (RO) and polarisation
resistance (R1) and capacitance (C1).
From the above electrical circuit, the battery terminal
voltage Vt is derived as a function of the current load IL
from Kirchhoff's Laws:

F I G U R E 8 Battery energy output on the New European

V t = V OC −RO # I L −V 1 , ð13Þ
Driving Cycle for different auxiliaries loads [Colour figure can be
viewed at wileyonlinelibrary.com]
dV 1 1 1
=− V 1 + I L: ð14Þ
dt R1 # C 1 C1
around 300 W. Therefore, auxiliary devices have a major
impact on energy consumption and must be considered as The model parameters VOC, RO, R1 and C1 are defined
accurately as possible. as a function of the battery SoC. The input of battery
model is the total power demand for propulsion and aux-
iliary devices that takes into account the energy losses
2.2.5 | Battery model along the powertrain. On the other hand, the outputs of
the model are battery terminal voltage and current
There are two main energy storage systems in the BMW obtained by the following equations:
i3: the high voltage Lithium-ion battery pack used to pro-
pel the vehicle and the low voltage (12 V) Lead Acid bat-
V Bat_Pack = V Cell # N Cells_series , ð15Þ
tery that powers the auxiliary devices. In this Section,
dynamic charging/discharging characteristics of the high
I Bat_Pack = I Cell # N Cells_parallel , ð16Þ
voltage battery pack is modelled to determine its operat-
ing voltage and SoC with a satisfactory level of accuracy.
The charging/discharging efficiency of both batteries is where NCells_series and NCells_parallel are the number of
also considered as it affects the EV energy consumption. cells in series and in parallel respectively, VCell and
514 MIRI ET AL.

VBat_Pack are the single cell and the battery pack termi- current demand is adjusted according to the charg-
nal voltage in (V) and ICell and IBat_Pack are the single ing/discharging battery efficiency. Subsequently, the
cell and the battery pack's current (A). The battery Thevenin model has been modelled in MATLAB/
model developed in MATLAB/Simulink is shown in Simulink based on Equations (11) and (12) as shown
Figure 10. in Figure 10.
The single cell power demand PCell,dem is derived by The battery SoC is updated at each time step using
dividing the total power demand Pdem by the number of the current integration method, also known as “Coulomb
cells NCells. Thereafter, the single cell current demand counting,” presented in Equation (16). Although this
ICell,dem is derived from PCell,dem as follows: method is not useable in a real application (because of
measurement noise, etc.), it is quite useful and accurate
' pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi(
V OC − ðV OC 2 −4 # RO # PCell,dem Þ in simulation environment.
I Cell,dem = : ð17Þ
2 # RO "ð t #
I Cell,dem ðτÞ
SoC = SoC0 − dτ , ð18Þ
t0 Ccell
The single cell current demand is calculated con-
sidering a power limitation to protect the battery con-
sidering the cut-off voltage of the battery cell. The where SoC is the battery state-of-charge in (%), SoC0 is
current demand is positive in traction mode while it the initial battery state-of-charge in (%), Ccell is the single
is negative in regenerative mode. After being derived cell capacity in (Ah) and ICell,dem is the single cell current
from the power demand, as explained above, the demand in (A).

FIGURE 10 A, Battery model and B, single cell model (Thevenin model) built in MATLAB/Simulink
MIRI ET AL. 515

2.2.6 | Transmission model PWheels = ðRθ + RR + RA + RI + RT Þ # V Vehicle , ð22Þ

Transmission system aims to transfer the torque between where:

the motor and the driving wheels. Efficiency of the trans-
mission system affects the EV energy consumption and Rθ = M Vehicle # g # sinα, ð23Þ
is defined in both traction and regenerative modes, as
follows: RR = CRR # M Vehicle # g # cosα, ð24Þ

T wheels # ωwheels 1
ηg_traction_mode = , ð19Þ RA = # ρ # AF # Cd # ðV Vehicle −V wind Þ2 , ð25Þ
T motorðsÞ # ωmotorðsÞ 2

T motorðsÞ # ωmotorðsÞ RI = RIa + RIε = δ # M Vehicle # a, ð26Þ

ηg_regenerative_mode = , ð20Þ
T wheels # ωwheels
ð1 − ηG Þ
RT = ðRR + RA + Rθ + RI Þ # : ð27Þ
ηG
where ωwheels and ωmotor(s) are the wheels and the motor
speed respectively in (rad/s), Twheels is the torque at the
driving wheels in (Nm) and Tmotor(s) is the motor torque In the above equations, VVehicle is the vehicle speed
in (Nm). in (m/s), and Vwind is the wind speed that has a positive
The transmission model is derived based on the fol- sign when it is tailwind and a negative sign when it is
lowing equation: headwind. CRR is the coefficient of rolling resistance,
MVehicle is the vehicle mass in (kg), g is the acceleration
T motorðsÞ # G # ηG due to gravity (m/s2), α is the angle of inclination of the
FT = , ð21Þ
rd road in (% ), ρ is the air density in (kg/m3), AF is the vehi-
cle frontal area in (m2), Cd is the drag coefficient, a is
where FT is the tractive force in (N), G is the single the vehicle acceleration in (m/s2) and δ is the coefficient
speed gear ratio, rd is the dynamic tyre radius in (m) and of rotary inertia that is assumed to be around 1.15 (typi-
ηG is the transmission efficiency that is assumed to be cal value). CRR can be approximated as a linear function
97%.38,45 of vehicle speed for a passenger car on concrete roads
for most of the tyre inflation pressure. This approxima-
tion provides a satisfactory estimation for speeds up to
2.2.7 | Longitudinal vehicle dynamics 128 km/h.29
" #
The following opposing forces are considered in the pro- V Vehicle
CRR = 0:01 # 1 + : ð28Þ
posed model as a common technique in the literature37 100
as shown in Figure 11:

• Gradient Resistance Force Rθ due to the road inclina- 3 | VEHICLE MODEL

tion with regard to the horizontal plane. VA L I DATI O N
• Rolling Resistance Force RR mainly due to the friction
between the tyres and the road. In this section, the proposed model is validated against
• Aerodynamic Drag Force RA due to the friction published values in the literature. For this purpose,
between the vehicle body and the air. energy consumption values on NEDC and Environmen-
• Inertia Resistance Force RI related to the forces required tal Protection Agency (EPA) cycles are used as explained
for the linear acceleration of the vehicle RIa and the in the following.
increase of the rotational speed of the rotating compo-
nents RIε.
• Transmission Resistance Force RT related to the losses 3.1 | Tests characteristics
between the motor and the wheels due to the transmis-
sion efficiency ηG. NEDC and EPA test procedures consider a combination
of urban and extra-urban driving patterns to take into
Thereafter, the power required at the wheels PWheels account the vehicle behaviour under different driving
to overcome the opposing forces is derived as follows: conditions. Both tests are performed indoors on a chassis
516 MIRI ET AL.

FIGURE 11 Forces acting

on the moving vehicle [Colour
figure can be viewed at
wileyonlinelibrary.com]

dynamometer, so there is no wind or variation in the Fuel Economy Test (HWFET) (highway cycle), SC03 Sup-
slope of the road.36,46 plemental FTP (use of air-conditioning) and US06 Sup-
plemental FTP (high speeds and accelerations).
Tests on FTP-75, HWFET and US06 cycles are run
3.1.1 | NEDC procedure without activated auxiliary devices, except those required
in the US for usual day-time operation: head and tail
The NEDC is the former cycle used for EU homologation lamps, driving control device and energy management
tests. It was replaced by Worldwide Harmonised Light system. For the SC03 test, air conditioning is activated in
Vehicle Test Procedure drive cycle that is more represen- addition to the other devices to measure the impact of air
tative of real driving conditions. However, the homologa- conditioning on the vehicle energy consumption.46
tion of the 2014 BMW i3 60Ah REx was carried out on
the NEDC as it was still applicable before 2014. NEDC is
a combined five-cycle test with four elementary urban 3.2 | Energy consumption calculation
cycles and an extra-urban cycle as shown in Figure 12.
Before testing, the vehicle is fully charged and is then In the NEDC test procedure, energy consumption Econs is
run twice over the cycle. Among the auxiliary devices calculated using Equation (27).36
listed in Table 3, only the driving control device and the
energy management system are assumed to be activated E
E cons = , ð29Þ
during the NEDC test because only the auxiliary devices Dtest
necessary for normal day-time operation of the vehicle
shall be activated.36 where E is the energy consumed in (Wh) and Dtest is the
distance covered in (km) during the test.
In the EPA test procedure, combined energy con-
3.1.2 | EPA procedure sumption, CombinedFC, is calculated as a combination of
the city and the highway energy consumption, CityFC
The US homologation is based on the EPA test proce- and HighwayFC, respectively46:
dure. For this purpose, the vehicle is fully charged
" #
the day before and then it is driven over the cycle during 0:89 0:11 0:18
CityRunningFE = 0:82 # + +
the test until battery is fully discharged.46 The test con- FTP US06 FTP
" #
sists of a combination of 4 cycles shown in Figure 12: 1 1
Federal Test Procedure-75 (FTP-75) (city cycle), Highway + 0:133 # 1:083 # − , ð30Þ
SC03 FTP
MIRI ET AL. 517

FIGURE 12 Speed profiles used in the simulations47

1 1 (Wh/km), and CityRunning_FE and HighwayRunning_FE are

CityFC = # , ð31Þ
0:905 CityRunning_FE the city and the highway fuel economy values
in (km/Wh).
" #
0:79 0:21
HighwayRunningFE = 1:007 +
US06 HWFET
"
1 1
# 3.3 | Simulation results and model
+ 0:133 # 0:377 − , ð32Þ validation
SC03 FTP
1 1 According to the literature, the combined energy con-
HighwayFC = # , ð33Þ
0:905 HighwayRunning_FE sumption of 2014 BMW i3 60 Ah REx is 135 Wh/km on
the NEDC31 while it is 117 mpg (179 Wh/km) on EPA
CombinedFC = 0:55 # CityFC + 0:45 # HighwayFC , ð34Þ cycles.32 In this study, simulation case-studies are per-
formed considering five different drive cycles: NEDC,
where FTP, US06, SC03 and HWFET are the energy FTP-75, HWFET, SC03 and US06. For each cycle, two
consumption values for the corresponding cycles in cases are simulated:
518 MIRI ET AL.

• Case 1 No auxiliary device is activated. the test. The increase of the error may be due to the fact
• Case 2 Some auxiliary devices are activated as described that the auxiliary load is overestimated in this study and
in Section 3.1, based on the devices activated during was less than 300 W during the NEDC test. Besides, the
each driving test which is described by the homologation approximation is not precise as it is based on the power
procedure.36,46 consumption of usual auxiliary devices which might be dif-
ferent for the BMW i3. For EPA cycles, the error between
The energy consumption values obtained from the sim- the simulation and experimental results decreases signifi-
ulations are compared to the above values found in the lit- cantly from 10.6% to 1.1% when auxiliary devices are
erature as presented in Tables 4 and 5. For NEDC cycle, included in the model, which leads to a higher accuracy of
the error between simulation and experimental results is the model. As aforementioned, EPA cycles are run with
about 3% when auxiliary devices are not included in the some auxiliary devices turned on such as air-conditioning
model whereas the error increases up to 5.9% when the and lamps, which explains the significant error between
power consumption of the auxiliary devices is assumed to simulation and experimental results when those devices
be 300 W. When the auxiliary devices are included in the are assumed to be turned off.46 Overall, the model demon-
model, the vehicle energy consumption from the simula- strates a satisfactory level of accuracy in view of NEDC
tion model becomes higher than the actual value from and EPA simulation results.

TABLE 4 Simulation results on NEDC and EPA cycles

4 | CONCLUSIONS
Energy consumption Case 1 Case 2
NEDC Cycle The main purpose of this study was to develop an accu-
NEDC 130.67 142.66 Wh/km rate computer-based model to estimate EV energy con-
EPA Cycles
sumption along with a given driving cycle. The BMW i3
was selected to be modelled as a case-study to prove the
FTP-75 114.58 130.95 Wh/km
concept. A forward vehicle simulation model was devel-
SC03 111.69 147.42 Wh/km oped in MATLAB/Simulink, including the powertrain
US06 191.69 201.93 Wh/km system and the longitudinal vehicle dynamics. The
HWFET 139.99 146.28 Wh/km powertrain model was implemented using accurate effi-
EPA city ciency maps of both the electric motor and the inverter.
1/City running FE 118.44 137.52 Wh/km The powertrain system also includes transmission and
(fuel economy) battery where the Thevenin equivalent circuit battery
City FC (fuel consumption) 130.87 151.96 Wh/km
model was used. Moreover, the resistance forces opposed
to the vehicle motion were modelled in the longitudinal
EPA highway
vehicle dynamics. A driver model was developed using a
1/Highway running FE 176.30 187.18 Wh/km
PI controller to control the vehicle's speed. In addition, a
Highway FC 194.81 206.83 Wh/km regenerative braking strategy that models the behaviour
EPA combined of a real braking controller was developed to distribute the
Combined FC 159.6 176.6 Wh/km braking torque demand between the friction and regenera-
tive brakes. Finally, the model was validated using the pub-
Abbreviations: EPA, Environmental Protection Agency; FTP-75,
licly available data from BMW and other reliable sources
Federal Test Procedure-75; HWFET; Highway Fuel Economy Test;
NEDC, New European Driving Cycle. in the literature. As a novelty of this work, power

TABLE 5 Comparison between experimental and simulation results on EPA and NEDC cycles

Simulation

Tests Case 1 Case 2

NEDC Energy consumption (Wh/km) 135 Energy consumption (Wh/km) 131 143
Error (%) −3.0 +5.9
EPA Energy consumption (Wh/km) 179 Energy consumption (Wh/km) 158 176
Error (%) −15.6 −2.2

Abbreviations: EPA, Environmental Protection Agency; NEDC, New European Driving Cycle.
MIRI ET AL. 519

consumption of the auxiliary devices was also estimated Transp Res Part D: Transp Environ. 2018;64:36-52. https://doi.
from average values found in the literature to be included org/10.1016/j.trd.2017.08.008.
in the proposed model as they can have a significant 8. Pan C, Dai W, Chen L, Chen L, Wang L. Driving range estima-
tion for electric vehicles based on driving condition identifica-
impact on energy consumption. In addition, the powertrain
tion and forecast. AIP Adv. 2017;7(10):105206. https://doi.org/
efficiency is estimated with better accuracy compared to 10.1063/1.4993945.
other studies, which only consider the electric motor effi- 9. Moure, C., Roche, M. and Mammetti, M. (2014). Range estima-
ciency. In this study, the inverter efficiency is calculated tor for electric vehicles. Paper presented at: 2013 World Electric
from an available efficiency map of the BMW i3 inverter. Vehicle Symposium and Exhibition, EVS 2014, pp. 1–15.
This more accurate estimation of the powertrain efficiency https://doi.org/10.1109/EVS.2013.6914917
leads to an improved estimation of the EV energy con- 10. Jiquan Wang (2016) Battery electric vehicle energy consump-
tion modelling, testing and prediction: a practical case study.
sumption. The model has demonstrated a satisfactory level
Eindhoven: Technische Universiteit Eindhoven. https://doi.
of accuracy with less than 6% error between the simulation
org/10.1111/nhs.12017
results and test data for both EPA and NEDC tests. 11. Kim, E., Lee, J. and Shin, K. G. (2013). Real-time prediction of
As a perspective to future research of this study, the battery power requirements for electric vehicles. Paper pres-
proposed model can potentially be used as a base for EV ented at: 2013 ACM/IEEE International Conference on Cyber-
range estimation. For this purpose, additional information Physical Systems, ICCPS 2013, pp. 11–20. https://doi.org/10.
about the road such as traffic and weather conditions and 1109/ICCPS.2013.6603995
also driver's characteristics should be added. In addition, 12. Fiori C, Ahn K, Rakha HA. Power-based electric vehicle energy
consumption model: model development and validation. Appl
the battery model can be improved by considering the
Energy. 2016;168:257-268. https://doi.org/10.1016/j.apenergy.
effects of battery SoC and state-of-health (SoH) which 2016.01.097.
have a major impact on battery efficiency and energy con- 13. Wu X, Freese D, Cabrera A, Kitch WA. ‘Electric vehicles’
sumption. In addition, the inertia of the vehicle's rotating energy consumption measurement and estimation. Transp Res
components such as the wheels, brakes and rotor can be Part D: Transp Environ. 2015;34:52-67. https://doi.org/10.1016/
also calculated and included in the model to improve its j.trd.2014.10.007.
accuracy. 14. Zhang R, Yao E. ‘Electric vehicles’ energy consumption estima-
tion with real driving condition data. Transp Res Part D:
Transp Environ. 2015;41:177-187. https://doi.org/10.1016/j.trd.
ORCID
2015.10.010.
Abbas Fotouhi https://orcid.org/0000-0002-5402-8629 15. De Cauwer C, Van Mierlo J, Coosemans T. Energy consumption
prediction for electric vehicles based on real-world data. Ener-
R EF E RE N C E S gies. 2015;8(8):8573-8593. https://doi.org/10.3390/en8088573.
1. Nilsson, M. (2011), Electric Vehicles The Phenomenon of Range 16. De Cauwer C, Verbeke W, Coosemans T, Faid S, Van Mierlo J.
Anxiety. https://cordis.europa.eu/docs/projects/cnect/5/249105/ A data-driven method for energy consumption prediction and
080/deliverables/001-ThephenomenonofrangeanxietyELVIRE.pdf energy-efficient routing of electric vehicles in real-world condi-
2. Hong, J., Park, S. and Chang, N. (2016). Accurate remaining tions. Energies. 2017;10(5):10. https://doi.org/10.3390/en10050608.
range estimation for Electric vehicles. Paper presented at: Pro- 17. Kriesel D. A brief introduction on neural networks. Lancet.
ceedings of the Asia and South Pacific Design Automation 2007;346:1712. https://doi.org/10.1016/S0140-6736(95)92880-4.
Conference, ASP-DAC. IEEE, January 25-28, pp. 781-786. 18. Diaz Alvarez A, Serradilla Garcia F, Naranjo JE, Anaya JJ,
https://doi.org/10.1109/ASPDAC.2016.7428106 Jimenez F. Modeling the driving behavior of electric vehicles
3. Luigi Mapelli F, Tarsitano D. Modeling of full electric and using smartphones and neural networks. IEEE Intell Transp
hybrid electric vehicles. New Generation of Electric Vehicles. Syst Mag. 2014;6(3):44-53. https://doi.org/10.1109/MITS.2014.
London, UK: IntechOpen; 2012:207-236. https://doi.org/10. 2322651.
5772/53570. 19. Saurav, K. (2016). An Introduction to Clustering and different
4. van Mierlo, J., Maggetto, G. and van den Bossche, P. (2004). methods of clustering, Analytics Vidhya. https://www.
Simulation methodologies for innovative vehicle drive systems. analyticsvidhya.com/blog/2016/11/an-introduction-to-clustering-
http://etec.vub.ac.be/publications/A71712.pdf and-different-methods-of-clustering/. Accessed: April 26, 2019.
5. Horrein L, Bouscayrol A, Delarue P, Verhille JN, Mayet C. For- 20. Fotouhi A, Yusof R, Rahmani R, Mekhilef S, Shateri N. A
ward and backward simulations of a power propulsion system. review on the applications of driving data and traffic informa-
IFAC Proc Vol. 2012;45:441-446. https://doi.org/10.3182/ tion for vehicles' energy conservation. Renew Sust Energ Rev.
20120902-4-FR-2032.00078. 2014;37:822-833.
6. Hauer KH, Moore RM. Fuel cell vehicle simulation– part 1: 21. Montazeri M, Fotouhi A, Naderpour A. Driving segment simu-
benchmarking available fuel cell vehicle simulation tools. Fuel lation for determination of the most effective driving features
Cells. 2003;3(3):84-94. https://doi.org/10.1002/fuce.200332102. for HEV intelligent control. Veh Syst Dyn. 2012;50(2):229-246.
7. Qi X, Wu G, Boriboonsomsin K, Barth MJ. Data-driven decom- 22. Montazeri-Gh M, Fotouhi A, Naderpour A. Driving patterns
position analysis and estimation of link-level electric vehicle clustering based on driving feature analysis. Proc Inst Mech Eng
energy consumption under real-world traffic conditions. C J Mech Eng Sci. 2011;225(6):1301-1317.
520 MIRI ET AL.

23. Montazeri-Gh M, Fotouhi A. Traffic condition recognition of these. https://www.unece.org/fileadmin/DAM/trans/main/

using the k-means clustering method. Sci Iran. 2011;18(4): wp29/wp29regs/2015/R101r3e.pdf
930-937. 37. Larminie, J. and Lowry, J. (2012) Electric Vehicle Technology
24. Lee CH, Wu CH. A novel big data modeling method for Explained, Electric Vehicle Technology Explained. https://doi.
improving driving range estimation of EVs. IEEE Access. 2015; org/10.1002/9781118361146
3:1980-1993. https://doi.org/10.1109/ACCESS.2015.2492923. 38. Genikomsakis KN, Mitrentsis G. A computationally efficient simula-
25. Ye, F., Wu, G., Boriboonsomsin, K., Barth, M. J. (2016). A tion model for estimating energy consumption of electric vehicles in
hybrid approach to estimating electric vehicle energy consump- the context of route planning applications. Transp Res Part D: Transp
tion for ecodriving applications. IEEE. https://ieeexplore.ieee. Environ. 2017;50:98-118. https://doi.org/10.1016/j.trd.2016.10.014.
org/document/7795633 39. Green Car Congress (2013) More details on BMW's i3; electric
26. Islam, E., Moawad, A., Kim, N., Rousseau, A. (2018) An Exten- and connected. https://www.greencarcongress.com/2013/07/
sive Study on Sizing, Energy Consumption, and Cost of bmwi3-20130710.html. Accessed June 17, 2019.
Advanced Vehicle Technologies. https://www.autonomie.net/ 40. Goodwin, A. (2015) 2014 BMW i3 with range extender review:
pdfs/ANL_BaSce_FY17_Report_10042018.pdf BMW's lightweight EV looks strange, drives strangely, CNET.
27. Amstrong, C. (2015) Eco-Approach and Departure at Signal- https://www.cnet.com/roadshow/reviews/2014-bmw-i3-
ized Intersections. https://rosap.ntl.bts.gov/view/dot/34781/ review/. Accessed June 17, 2019.
dot_34781_DS1.pdf 41. Abatte, N. (2008). DC-DC converter for Hybrid Electric Vehicle
28. Fotouhi, A., Auger, D. J., Propp, K., Longo, S. (2015). Simula- and EV. Arrow Times, p. 7. https://pdfs.semanticscholar.org/
tion for prediction of vehicle efficiency, performance, range 697d/cf2b439aa90638413daa8e3c4aec591fd51f.pdf
and lifetime: A review of current techniques and their applica- 42. Schaltz E. Electrical vehicle design and modeling. Electric Vehi-
bility to current and future testing standards. (November 2014), cles - Modelling and Simulations. London, UK: IntechOpen;
p. 9.1–9.1. https://doi.org/10.1049/cp.2014.0959 2012. https://doi.org/10.5772/20271.
29. Ehsani M, Gao Y, Gay S, Emadi A. Modern Electric, Hybrid 43. Chan CC, Chau KT. Modern Electric Vehicle Technology.
Electric, and Fuel Cell Vehicles. Fundamentals, Theory, and Oxford, UK: Oxford University Press; 2001.
Design. Florida: CRC Press LLC; 2005. 44. Fotouhi A, Auger DJ, Propp K, Longo S, Wild M. A review on
30. i3 Guide (2014). I3's Range. http://www.i3guide.com/range. electric vehicle battery modelling: from Lithium-ion toward
html. Accessed May 31, 2019. lithium-Sulphur. Renew Sust Energ Rev. 2016;56:1008-1021.
31. BMW Group (2013). The BMW i3 Experience. https://www. https://doi.org/10.1016/j.rser.2015.12.009.
bmw.no/content/dam/bmw/marketNO/bmw_no/brosjyre/ 45. Hayes, J. G., De Oliveira, R. P. R., Vaughan, S. and Egan, M. G.
i3PSLEAL_web.pdf (2011). Simplified electric vehicle power train models and
32. BMW Group (2014). BMW U.S. Media Information Technical range estimation. Paper presented at: 2011 IEEE Vehicle Power
Data 2014 BMW i3. https://www.press.bmwgroup.com/usa/ and Propulsion Conference, VPPC 2011 IEEE, pp. 1–5. doi:
article/detail/T0149790EN_US/the-all-new-bmw-i3?language= https://doi.org/10.1109/VPPC.2011.6043163
en_US 46. Environmental Protection Agency (2006) Fuel Economy Label-
33. Idaho National Laboratory (2016). 2014 BMW i3 with Range ing of Motor Vehicles: Revisions To Improve Calculation of
Extender (REx) Advanced Vehicle Testing – Baseline Vehicle Fuel Economy Estimates. https://www.govinfo.gov/content/
Testing Results. pp. 3–7. https://avt.inl.gov/sites/default/files/ pkg/FR-2006-12-27/pdf/06-9749.pdf
pdf/phev/fact2014bmwi3rex.pdf 47. T. J. Barlow I. S. McCrae, and P. G. Boulter, S. L. (2009). A
34. Oak Ridge National Laboratory (2016) Oak Ridge National reference book of driving cycles for use in the measurement
Laboratory Annual Progress Report for the Electric Drive Tech- of road vehicle emissions. https://www.gov.uk/government/
nologies Program. https://info.ornl.gov/sites/publications/ uploads/system/uploads/attachment_data/file/4247/ppr-354.pdf
Files/Pub71072.pdf
35. BMW Group (2014) The Hybrid-Synchronous Machine of the
New BMW i3 & i8, pp. 1–9. http://hybridfordonscentrum.se/
How to cite this article: Miri I, Fotouhi A,
wp-content/uploads/2014/05/20140404_BMW.pdf
36. United Nations Economic Commission for Europe (2013) Agree-
Ewin N. Electric vehicle energy consumption
ment Concerning the Adoption of Uniform Technical Prescrip- modelling and estimation—A case study. Int
tions for Wheeled Vehicles, Equipment and Parts which can be J Energy Res. 2021;45:501–520. https://doi.org/10.
fitted and/or be used on Wheeled Vehicles and the Conditions 1002/er.5700
for Reciprocal Recognition of Approvals Granted on the Basis