Applied Energy xxx (2015) xxx–xxx

Contents lists available at ScienceDirect

Applied Energy
journal homepage: www.elsevier.com/locate/apenergy

Energy consumption of electric vehicles based on real-world driving

patterns: A case study of Beijing q
Hewu Wang, Xiaobin Zhang, Minggao Ouyang ⇑
State Key Laboratory of Automotive Safety and Energy, China Automotive Energy Research Center, Tsinghua University, Beijing, China

h i g h l i g h t s

 The real-world driving cycles and driving range in Beijing are clearly studied.
 We assess energy consumptions of electric vehicles in real-world driving conditions.
 Shorter driving ranges and severe driving conditions bring EVs more fuel reduction.
 PHEVs with smaller batteries for 30–50 km CD range are preferred in Beijing.
 The impact of driving patterns on energy use of electric vehicles is quantified.

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

Article history: This study assesses the energy reduction associated with Hybrid Electric Vehicles (HEVs), Plug-in Hybrid
Received 12 September 2014 Electric Vehicles (PHEVs) and Battery Electric Vehicles (BEVs) compared to conventional vehicles (CVs)
Received in revised form 31 March 2015 for real-world driving conditions in a specific geographic region (Beijing, China). To understand the driv-
Accepted 18 May 2015
ing patterns in Beijing, a passenger car travel survey has been conducted since 2012, including over 1000
Available online xxxx
vehicles. The initial results from driving range distribution have been calculated. In this study, first, a
Utility Factor and the typical driving cycles based on 2000 days’ worth of Global Position System (GPS)
Keywords:
data are analyzed. Next, the real-world energy consumption of CVs, HEVs, PHEVs and BEVs are simulated.
Electric vehicle
Plug-in hybrid electric vehicle
Finally, the fuel consumption of vehicles under different driving patterns is compared to provide data on
Driving patterns the optimal electric vehicles and reliable test cycles for Beijing. We find that electric vehicles in Beijing,
Driving cycle including HEVs, PHEVs and BEVs, yield more fuel reduction benefits than in the U.S. because of the severe
Driving range driving conditions and short driving ranges. For PHEVs in Beijing, smaller batteries, corresponding to a
Fuel consumption 30–50 km Charging Depleting (CD) range, are preferred to meet the demands of most drivers and add less
extra cost to the vehicle. We also confirm that the Chinese current suggested label values based on NEDC
cycle underestimate the fuel consumption of vehicles and fuel reduction benefits of electric vehicles in
Beijing. This study addresses the importance of developing and using the real-world driving cycles in
designing and evaluating electric vehicles.
Ó 2015 Elsevier Ltd. All rights reserved.

1. Introduction 20 years to build a sustainable transportation system [2,3]. The

Automotive powertrain electrification can achieve low emission
and high energy efficiency to mitigate the energy shortage and air
pollution brought by transportation sectors [1]. The Chinese gov-
ernment has implemented research, development, demonstration,
and deployment programs for clean vehicle technologies for over
q
This paper is included in the Special Issue of Clean Transport edited by Prof.
Anthony Roskilly, Dr. Roberto Palacin and Prof. Yan.
⇑ Corresponding author. Tel./fax: +86 62783025.
E-mail addresses: wanghw@tsinghua.edu.cn (H. Wang), zhangxiaobin08@gmail.
com (X. Zhang), ouymg@tsinghua.edu.cn (M. Ouyang).
national strategy highlights new energy vehicles (NEVs), including
electric vehicles (EVs), fuel cell vehicles (FCVs), and plug in hybrid
electric vehicles (PHEVs), with more than 2 million produced and 5
million NEVs expected to be deployed in 2020. Facing severe traffic
congestion and air pollution, Beijing is actively promoting the
development of new energy vehicles to reduce Particulate
Matter2.5 (PM2.5) emissions from the transportation sector, which
currently accounts for 25% of the emissions [4].
However, due to the limitations of Li-ion batteries, EVs and
PHEVs have limited driving ranges and require longer charging
time compared to conventional vehicles [5]. For such
range-limited vehicles, driving patterns, including driving

http://dx.doi.org/10.1016/j.apenergy.2015.05.057
0306-2619/Ó 2015 Elsevier Ltd. All rights reserved.

Please cite this article in press as: Wang H et al. Energy consumption of electric vehicles based on real-world driving patterns: A case study of Beijing. Appl
Energy (2015), http://dx.doi.org/10.1016/j.apenergy.2015.05.057
2 H. Wang et al. / Applied Energy xxx (2015) xxx–xxx
conditions and driving range, become important issues that dictate
vehicle energy consumption [6], economic costs [7] and environ-
mental impacts [8]. Driving patterns are therefore being consid-
ered more and more in vehicle optimal designs [9] and controls
[10] as well as infrastructure construction [11].
Driving conditions directly determine the energy consumption
of on-road vehicles, and include driving speed, acceleration, idling
time, etc. To address the impact of driving conditions on fuel econ-
omy and emissions, standard driving cycles have been developed
based on regional real-world driving data. In U.S., the FTP and
HWFET cycles were originally developed to reflect city and high-
way driving conditions in the 1970s. Then, the Environmental
Protection Agency (EPA) adopted the EPA-5 cycle to test conven-
tional vehicles, as well as alternative fueled vehicles in 2006. In
Europe, the NEDC cycle is supposed to represent the typical usage
of a car in Europe to assess the emission levels and fuel economy.
Japan adopted the JC08 cycle for vehicle testing since 2007. China
has also adopted the NEDC cycle for vehicle fuel economy testing.
Driving conditions are changing rapidly and vary in different
regions, mainly because of the development of vehicle powertrain
technologies, increases in the vehicle population, strict regulations
and traffic infrastructure. Significant deviations between the test
results from standard driving cycles and real-world performance
has been observed, and raise questions on the impact of the driving
cycles on vehicle energy consumption [12]. Real-world driving
cycles were generated based on recorded driving profiles for vehi-
cle testing or simulation [13,14].
Driving range affects the average energy consumption of PHEVs
and BEVs. For PHEVs capable of being fueled by liquid fuels and
grid electricity, the daily driving range is essential to estimating
the contribution from both energy sources [15]. For BEVs with a
limited All Electric Range (AER), not all trips can be substituted
by BEVs. Therefore, BEV utilization depends heavily on the daily
driving range. The distribution of the driving range is mainly devel-
oped by questionnaires survey and onboard instrumentation
records (usually GPS data loggers), and the impact of driving range
on PHEV and BEV energy consumption is evaluated based on this
distribution [16]. The Utility Factor (UF) is introduced to represent
the percentage of electric driving range contained in the total driv-
ing range [17,18]. The National Household Travel Survey (NHTS) of
the US conducted national questionnaire survey on driving pat-
terns including average daily driving range and provides compre-
hensive data for transportation researchers [19]. The GPS devices
were also adopted by the National Renewable Energy Laboratory
(NREL) in regional travel surveys, such as the study on 407 vehicles
in the greater Chicago area in 2007, and 1,325 vehicles in Atlanta in
2011 [20]. In Europe, the Danish National Travel Survey and
on-board GPS data have been adopted to extract driving distances
and driving time periods [21]. In China, the annual driving mileage
was intensively studied [22,23], and the distribution of daily driv-
ing ranges in specific regions has been studied since 2012 [24].
The impact of driving conditions and driving range on vehicle
energy consumption are closely coupled. Driving conditions
directly determine the energy consumption per kilometer of driv-
ing, which will also affect the range of a BEV or the charge deplet-
ing (CD) range of a PHEV. The available electric range and the
distribution of the driving range will then determine the utilization
of electricity and fuel [17]. These two factors may act together or in
conflict on the ability to reduce fuel usage in electric vehicles.
However, as discussed above, current studies mainly focus on the
evaluation of driving conditions or driving range, respectively
[7,13,16,18,25]. To assess the average electricity and fuel consump-
tion from vehicles of a specific region, the driving conditions and
driving range should be obtained and evaluated together.
To understand the driving patterns in Beijing, a passenger car
travel survey has been conducted since 2012. Preliminary results
Please cite this article in press as: Wang H et al. Energy consumption of electric vehicles based on real-world driving patterns: A case study of Beijing. Appl
Energy (2015), http://dx.doi.org/10.1016/j.apenergy.2015.05.057
taken from 112 vehicles’ GPS driving data have been compiled
focusing on the driving range distribution [24,26,27]. This study
focuses on the energy reduction potential of HEVs, PHEVs and
BEVs compared with conventional vehicles based on driving pat-
terns in Beijing. The paper is organized as follows. First, the UF
and typical driving cycles based on 2000 days’ worth of GPS data
are analyzed and compiled. Second, the real-world energy con-
sumption of CVs, HEVs, BEVs and PHEVs are simulated, and the
coupled effect of driving conditions and driving range is examined
to provide insight into PHEV battery sizing. Finally, the fuel con-
sumption under different driving patterns are compared to provide
conclusions on optimal electric vehicles and reliable test cycles for
Beijing.
2. Driving patterns based on GPS survey
The driving patterns in Beijing, including the driving range, and
driving conditions are studied based on systemic and real-world
collected driving data. The Utility Factor is adopted to describe
the usage of electricity of PHEV in the total driving range, and four
representative driving cycles are built. This is the basis of evalua-
tion on vehicle energy consumption.
In the Beijing passenger car travel survey, GPS loggers were
adopted to collect travel data and 112 volunteer vehicles were
observed from June 2012 to March 2013. The recorded data of
2003 travel days comprises nearly 10,000 km for 4892 trips.
2.1. UF for PHEVs and BEVs based on the distribution of the driving
range
The collected data are processed to extract the distribution of
the daily driving range and the single trip range, which is denoted
as set S and set T separately.
S ¼ fd1 ; d2 . . . dNs g
T ¼ ft 1;1 ; t 1;2 ; t2;1 . . . ti;j g i ¼ 1:2 . . . Ns
In the set S, di is a daily driving range of a vehicle and Ns is the num-
ber of total travel days. In the single trip set T, ti, j is the jth trip of a
vehicle in the ith travel day. The average daily driving range in
Beijing is 39.6 km, and the average single trip range is 16.4 km.
2.1.1. Utility factor for PHEVs
The Utility Factor is proposed to describe the fraction of charge
depleting (CD) range and charge sustaining (CS) range of a PHEV by
the SAE. The SAE standard J2841 assumes that a PHEV begins each
travel day fully charged, and starts out in the CD mode. When the
trip exceeds the CD range, the vehicle switches to charge sustain-
ing (CS) mode. Thus UF indicates the percentage of CD range, and
(1-UF) indicates the percentage of CS range in the total driving
range. Based on this assumption, the standard UF for a PHEV in
Beijing (UFPHEV) can be calculated, as is shown in Fig. 1.
P
minðdk ; RCD Þ
dk 2s
UF PHEV ðRCD Þ ¼ P :
dk 2s dk
RCD is the CD range of the PHEV, and dk is from the daily driving
range set S.
The distribution percentage of the daily driving range shows
that the vast majority of daily driving ranges are less than
100 km and tend to be short. The UF is 0.4218 for a 20 km CD range
and 0.7361 for a 50 km CD range, meaning that a PHEV with a
20 km CD range can drive solely on electricity for 42.18% of its
the daily trips and if the CD range is increased to 50 km, the per-
centage rises to 77.06%. Compared with the UF in the U.S., the
Beijing UF is larger at the same CD range, with a maximum
observed difference of 23% at 50 km. This may cause a difference
 H. Wang et al. / Applied Energy xxx (2015) xxx–xxx
Fig. 1. Standard UF of Beijing and U.S. are compared.
in the energy consumption for PHEVs, which will be examined in
the following sections.
2.1.2. Utility factor for BEVs
To evaluate the BEV usage, we need to know the percentage of
the electric range in the total daily range, which can be defined as
the utility factor for a BEV UFBEV(AER). The UFBEV derivation
approach is similar to that of a PHEV, but each trip must be exam-
ined to see whether can be driven by the BEV:
P P  P
t i;j 2di t i;j þ tr
UF BEV ðAERÞ ¼ P
ti;j 2T t i;j
X Three standard driving cycles, the NEDC, UDDS and HWFET
t i;j < AER; t r < AER
t i;j 2di
In which ti, j is from the single trip set T, and tr is the round trip
which starts from home and sequentially back home. Assuming that
a BEV begins each travel day fully charged, trips can only be driven
P
by a BEV when the total range of all the trips in a day ti;j 2di t i;j or the
round trip does not exceed the AER.
The UFBEV is shown in Fig. 2 and is 0.7508 for the 100 km AER,
meaning that a BEV with a 90 km AER can drive 75.08% of the daily
trips. Increasing the AER can enable the BEV to meet a greater tra-
vel demand. When the AER is greater than 200 km, the UFBEV is
over 90% and increases slowly. The variable UFBEV is used to calcu-
late the electricity consumption of BEVs.
Fig. 2. UF of Battery Electric Vehicles in Beijing is calculated.
Please cite this article in press as: Wang H et al. Energy consumption of electric vehicles based on real-world driving patterns: A case study of Beijing. Appl
Energy (2015), http://dx.doi.org/10.1016/j.apenergy.2015.05.057
3
2.2. Real-world driving cycles
A general analysis of the overall data reveals that real-world
driving conditions change significantly in different regions and
over time. Based on the variations in the driving conditions,
Beijing is divided into a downtown area (within the 5th ring road)
and a suburban area by geography. A typical day is divided into the
peak hours (7:00–9:00 a.m. and 5:00–7:00 p.m.) and non-peak
hours by time. To determine the typical cycles, the recorded data
were classified into four clusters by geography and time. The four
typical driving cycles were then generated from each cluster of the
recorded data with the TRAFIX method. The method compares
speed codes of micro-trips with the overall survey data and select
mandatory cycles according to the characteristics [28]. They are
Beijing Downtown Peak Hour cycle (BDPH), Beijing Downtown
Non-peak Hour cycle (BDNH), Beijing Suburb Peak Hour cycle
(BSPH), and Beijing Suburb Non-peak Hour cycle (BSNH). The four
driving cycles are illustrated in Fig. 3.
To identify the differences in these cycles, several cycle
parameters are calculated and compared. Table 1 shows the
parameters reflecting the driving conditions over various regions
and time. The BDPH cycle has the lowest average speed of
15.72 km/h along with the largest acceleration, deceleration and
the longest idle time of 40.1%. The BDPH cycle represents the
worst driving conditions in the central area of Beijing during peak
hours, whereas during non-peak hours, the traffic reduces so that
the idle time decreases to 30%. In the suburban areas, the average
speed increases by more than 40% (22.28 km/h) due to the
decrease in idle time compared with the BDPH cycle. However,
the differences between the peak and non-peak hours in the sub-
urban areas are negligible with the exception of the maximum
speed.
cycles, are selected as references. Because the NEDC cycle is the
standard test cycle for Chinese fuel regulations, this study is
intended to examine the difference in Beijing real-world driving
conditions and any corresponding bias towards the energy con-
sumption in vehicles. The UDDS and HWFET cycle are adopted to
calculate the energy consumption of an identical vehicle in the
U.S. using the EPA proposed method.
Compared with Beijing cycles, the average and maximum speed
of NEDC cycle are much higher, which imply better driving condi-
tions. However, the average accelerations and decelerations are
65% and 109% (1.04 m/s2 and 1.389 m/s2) larger than the other
Beijing cycles, which is not consistent with reality. For urban areas
of the US, the UDDS cycle reflects better the driving conditions for
higher speeds and smaller average accelerations.
3. Vehicle platform for typical Chinese powertrain technologies
To assess the fuel consumption for different powertrain tech-
nologies, representative vehicle models are essential besides the
UF and driving cycles. Thanks to government policies, there are
many electric vehicles emerging in Chinese market. The AER of
BEVs varies from 150 km of BAIC E150 EV to 300 km of BYD e6
EV. PHEVs are represented by BYD Qin plug-in with a 50 km CD
range, and the Roewe 550 plug-in with 58 km CD range. These real
vehicles provide a baseline for the component sizes for typical
vehicle models.
We used the vehicle simulation tool Autonomie [29] developed
by Argonne National Laboratory to model typical midsize electric
vehicles including BEVs, HEVs and PHEVs. To ensure these vehicles
are physically and functionally equivalent for consumers, the mod-
els should have the same common components and meet the same
performance demands, which are referred from market surveys
4 H. Wang et al. / Applied Energy xxx (2015) xxx–xxx

Fig. 3. Four typical driving cycles associated with Beijing traffic are developed.

Table 1
Parameters for seven different driving cycles are compared.

Units BDPH BDNH BSPH BSNH NEDC UDDS HWFET

Distance km 4.49 6.68 8.45 8.75 17.70 11.97 16.52
Max. speed km/h 81.00 81.00 87.00 98.00 120.00 91.23 96.38
Avg speed km/h 15.72 19.97 22.28 23.39 33.60 31.50 77.68
Avg acceleration m/s2 0.65 0.66 0.61 0.6 1.04 0.50 0.19
Avg deceleration m/s2 0.73 0.68 0.60 0.65 1.389 0.58 0.22
Idle time % 40.1 29.71 25.55 26.22 23.1 18.92 0.65

Table 2 conditions through iterative steps. For a BEV, a 50 kW h battery

Performance demands of typical midsize vehicles.
Performances 0–100 km/h Gradeability with Max speed
acceleration time 400 kg load at 90 km/h
Values 8.6s ± 0.1 s >6.5% >150 km/h
and related studies [8,30] (see Table 2). The common components
include a glider mass of 900 kg, accessories, tires, etc.
The same performance demands were maintained for a typical
midsize car possessing different powertrain styles. All the vehicles
should have enough power to ensure the acceleration and grade-
ability satisfies driver’s needs.
The HEVs and PHEVs adopted an input power split configura-
tion with the same planetary gear set. For PHEVs and BEVs, the bat-
tery size is a key factor that influences the performance. As the
electric efficiency changes with the driving conditions, the AER of
a BEV and RCD of a PHEV will change with driving conditions. For
a PHEV, we carefully choose the battery size for 10, 20, 30 and
40 mile (16, 32, 48 and 65 km) CD ranges in Beijing driving
is employed to ensure a 200 km AER. A conventional vehicle using
a Honda Accord configuration is also included for reference.
To maximize the utilization of electricity, the EV/CS control
strategy is implemented for the PHEVs. PHEVs are initially fully
charged, first operate in charge depleting (CD) mode and the
engine is forced to turn off. The power of motors and batteries is
designed to meet the test driving cycles in the CD mode. When
the SOC reaches 30%, the vehicle switches to charge sustaining
(CS) mode. The HEV and PHEV in the CS mode adopt the similar
rule-based strategy. In the strategy, a power threshold depending
on the battery SOC determines the engine on to propel the vehicle
and recharge the batteries to sustain the SOC. For the PHEV, the
blended control strategy is proven to reduce the average fuel con-
sumption by 4% over unknown driving ranges [31]. However it has
not been applied in real products, so simulations of these vehicles
is left for future work. Regenerative braking is used for electric
vehicles. All the vehicles are designed using an iterative process
to meet all these performance requirements and the control strate-
gies perform well. The component sizes in each vehicle model are
as follows (see Table 3).

Table 3
Component sizes of typical vehicles are listed.

Component size CV HEV PHEV10 PHEV20 PHEV30 PHEV40 BEV

Engine max power (kW) 140 73 73 73 73 73
Motor 1 (kW) 66 67 70 73 77 148
Motor 2 (kW) 30 30 30 30 30
Battery capacity (kW h) 2.184 4.530 8.855 13.282 17.709 50.4
Battery pack mass (kg) 22 64.26 60.78 118.79 178.19 237.59 613
Vehicle mass (kg) 1569 1606 1599 1658 1721 1783 1950

Please cite this article in press as: Wang H et al. Energy consumption of electric vehicles based on real-world driving patterns: A case study of Beijing. Appl
Energy (2015), http://dx.doi.org/10.1016/j.apenergy.2015.05.057
 H. Wang et al. / Applied Energy xxx (2015) xxx–xxx 5

4. Real-world energy consumption Fuel consumption of a HEV in 7 driving cycles

HWFET 4.8
The cycle-specific energy consumption of all the vehicle models
over the seven driving cycles are compared to show the impact of UDDS 4.51
the different driving cycles throughout the vehicle simulations. NEDC 4.89
Then, the real-world average energy consumption is calculated:
BSNH 5.07
For the CV, HEV and BEV, which have a single energy source, the
average results for Beijing and the adjusted results for the U.S. BSPH 5.19
are calculated,. Additionally, for the range-sensitive PHEVs that
BDNH 5.42
have dual energy sources, the UF is considered in the calculation.
BDPH 5.84
4.0 4.5 5.0 5.5 6.0
Fuel consumption (L/100km)
4.1. Cycle-specific energy consumption
Fig. 5. Fuel consumption of a HEV over seven driving cycles.
The fuel consumption of a midsize conventional vehicle over
seven driving cycles is illustrated in Fig. 4. Generally speaking,
the CV is intensively sensitive to the driving conditions, and exhi- Electricity consumption of a EV in 7 driving cycles
bits greater fuel consumption in the Beijing test cycles. The stan-
HWFET 17.933
dard deviation of fuel results is 1.09 L/100 km. The fuel
consumption of the BDPH cycle is 14.85 L/100 km, which is 66%, UDDS 16.43
60% and even 133% higher than that of NEDC, UDDS and HWFET NEDC 17.86
cycles, respectively. These comparisons imply that the low speeds,
BSNH 17.91
aggressive accelerations and frequent stops associated with Beijing
city driving conditions shift the engine into a low-efficiency zone BSPH 17.71
and increase the fuel consumption. This analysis is consistent with BDNH 18.32
reality. Fuel consumption in midsize cars such as the Volkswagen
CC, Toyota Camry and Honda Accord are reported to be approxi- BDPH 20.12
mately 9 L/100 km in the NEDC cycle [32], whereas drivers report 12.0 14.0 16.0 18.0 20.0 22.0
the consumption to be over 11 L/100 km in real use [33]. Electricity consumption (kWh/100km)
The results for HEVs are shown in Fig. 5. The mean value of the
fuel results in the 7 cycles is 5.1 L/100 km and the standard devia- Fig. 6. Electricity consumption of an EV over seven driving cycles.
tion is 0.44 L/100 km. The HEV generally has a higher fuel economy
and is less sensitive to driving conditions than the CV. The
rule-based control strategy adjusts the ratio of electric and engine Fuel consumption of PHEVs in CS mode
6.5
power to keep the engine operating in a high efficiency zone during
fuel consumption (L/100km)

bad driving conditions. The fuel consumption of the BDPH cycle is

6.0
5.84 L/100 km, which is 19% and 29.5% higher than that of NEDC
(4.89 L/100 km) and UDDS cycles (4.51 L/100 km), respectively. 5.5
The peak hours increase the fuel consumption by 8% in downtown
areas, and 2% in suburban areas. 5.0
Next, the electricity consumption in BEVs are illustrated in
Fig. 6. The mean value of the test results is 18.04 kW h/100 km 4.5
and the standard deviation is 1.09 kW h/100 km. BEVs are less
4.0
sensitive to driving conditions. The electric motor permits a high PHEV10 PHEV20 PHEV30 PHEV40
efficiency over a wide range of speeds and torques, and regenera-
tive braking makes BEVs less sensitive to decelerations. During BDPH BDNH BSPH BSNH NEDC UDDS HWFET
the BDPH cycle, the energy consumption is 20.12 kW h/100 km, Fig. 7. Fuel consumption of PHEVs in the CS mode.
which is 12.7% and 22.4% higher than that of NEDC
(17.86 L/100 km) and UDDS cycles (16.43 L/100 km), respectively.
In the suburban areas, the electricity consumption deceases to
approximately 17.9 kW h/100 km.
Fuel consumption of a CV in 7 driving cycles For PHEVs, the fuel consumption in the CS mode and electricity
HWFET 6.37 consumption in the CD mode is compared in Figs. 7 and 8, respec-
tively. The fuel consumption in the CS mode is similar to an HEV,
UDDS 9.27
with high fuel economy and moderate standard deviation of
NEDC 8.96 0.50 L/100 km through driving cycles, whereas the electricity con-
BSNH 11.23 sumption in the CD mode is similar to a BEV. As the CD range
increases, the battery weight, engine and motor power, and vehicle
BSPH 11.46
mass increase, which corresponds to subsequent increases in the
BDNH 12.89 fuel and electricity consumption. For the PHEV10, the fuel con-
BDPH 14.85 sumption during the BDPH cycle is 5.81 L/100 km, whereas the fuel
4.0 6.0 8.0 10.0 12.0 14.0 16.0
consumption for the PHEV40 is 6.08 L/100 km. The electricity con-
sumption during the BDPH cycle is 17.61 kW h/100 km for the
Fuel consuption (L/100km)
PHEV10 and 18.38 kW h/100 km for the PHEV40. For the same
Fig. 4. Fuel consumption of a Conventional Vehicle over seven cycles. vehicle in a downtown area, the fuel consumption at peak hours

Please cite this article in press as: Wang H et al. Energy consumption of electric vehicles based on real-world driving patterns: A case study of Beijing. Appl
Energy (2015), http://dx.doi.org/10.1016/j.apenergy.2015.05.057
6 H. Wang et al. / Applied Energy xxx (2015) xxx–xxx

Electricity consumption of PHEVs in CD mode 90% of their daily travel. For simplicity, UFBEV(200) is set to 100% in
19.0 this study.
electricity consumption (kWh/100km)

18.0

17.0
4.2.2. Average energy consumption in the U.S.
The average energy consumption of the same vehicle in U.S. is
16.0
calculated based on the EPA 2006 regulations [34]. To correct test
15.0 laboratory values to ‘‘real world’’ estimates, the derived 5-cycle
fuel economy method is applied to adjust for higher speed/aggres-
14.0
sive driving, air conditioning use and colder temperatures. This
13.0 method is reasonable for vehicles built after MY2011 if the error
12.0 is acceptable and suitable for alternative fueled vehicles, therefore,
PHEV10 PHEV20 PHEV30 PHEV40 in this study; the values are used as a reference.
BDPH BDNH BSPH BSNH NEDC UDDS HWFET The equation for EPA combined fuel economy is as follows:
1
Fig. 8. Electricity consumption of PHEVs in the CD mode. City MPG ¼ 0:003259þ 1:1805
FTP FE
1
Highway MPG ¼ 0:001376þ 1:3466 ;
HWFET FE
1
increases by approximately 11% over non-peak hours, whereas in MPGcomb ¼ 0:55 þ 0:45
MPGc MPGh
suburb areas, the change is negligible over time.
where FTPFE and HWFETFE represent the fuel economy in FTP and
HWFET cycles, and are in miles per gallon. The MPGcomb reflects
4.2. Real-world energy consumption of CVs, HEVs and BEVs
the average fuel consumption under U.S. conditions, which is com-
bined from city and highway fuel economy results.
4.2.1. Average energy consumption in Beijing
In order to calculate the average energy consumption, the per-
centage of each driving cycle that a vehicle may encounter is 4.2.3. Comparisons of CVs, HEVs and BEVs
needed. The weighting factor for a specific driving cycle is defined The real-world fuel consumption of CVs, HEVs and BEVs in
as wi. Beijing and the U.S. is illustrated in Fig. 9, which show the electric-
P ity consumption converted to equivalent gasoline. Fuel consump-
t 2DC j t i tion of CVs in Beijing is 12.43 L/100 km, which is 22.3% higher
wj ¼ P i DC j # T; j ¼ 1; 2; 3; 4:
t k 2T t k
than in the U.S. EVs and HEVs in Beijing exhibit 2.068 L/100 km
gasoline equivalent and 5.34 L/100 km fuel consumption, which
DCj represents a set of trips that meet a specific driving cycle is a 30.7% and 14.4% fuel reduction over that in the U.S. BEVs and
according to the parameters in Table 1. Based on the driving pattern HEVs exhibit a 83.4% and 57.0% energy reduction compared to con-
survey, peak hour periods account for 48% of the total driving range, ventional vehicles in Beijing. The comparisons imply that the driv-
and vehicles travel 44% in downtown areas and 56% in suburban ing patterns in Beijing are worse for conventional vehicles under
areas within the geographical scope of Beijing. The variable wi is severe driving conditions, but provide comparative advantages
shown in Table 4. for HEVs and BEVs.
For CVs and HEVs, which are not sensitive to driving range, the
average energy consumption can be calculated as:
4.3. Real-world energy consumption of PHEVs
P4 P 
j¼1 FC j t i 2DCj t i X
4
EC av e ¼ P ¼ wj FC j : The energy consumption of PHEVs differs from that of CVs or
t k 2S t k j¼1
HEVs because PHEVs either work in the CD or the CS mode, and
the energy consumption is mixed. First, consider the impact of
where FC j indicates the fuel consumption of a vehicle in the jth of driving conditions and follow the method used on the HEVs and
the four representative driving cycles, so EC_ave is the average fuel BEV. The adjusted cycle-specific fuel and electricity consumption
consumption of CV or HEV. is illustrated in Figs. 10 and 11. Note that the driving conditions
For a BEV with an AER of over 200 km, the utility factor do not change with the PHEV’s operating mode; therefore, wj is
UFBEV(AER) is considered to calculate the electricity use: the same for both the CD and the CS mode.
!
X
4
EC av e ¼ wj EC j  UF BEV ðAERÞ
j¼1

where EC j is the electricity consumption of a BEV in the jth driving

cycle. UFBEV is derived above. In fact, for Beijing city drivers whose
driving range is short, a BEV with 200 km AER can meet more than

Table 4
The variable wi is derived from the Beijing travel survey.

Driving Beijing Beijing Beijing Beijing suburb

cycles Downtown Downtown suburb peak nonpeak hours
peak hours nonpeak hours hours
Weighting 0.21 0.23 0.27 0.29
factor
wi
Fig. 9. Real-world fuel consumption of CVs, HEVs and BEVs in Beijing and the U.S.

Please cite this article in press as: Wang H et al. Energy consumption of electric vehicles based on real-world driving patterns: A case study of Beijing. Appl
Energy (2015), http://dx.doi.org/10.1016/j.apenergy.2015.05.057
 H. Wang et al. / Applied Energy xxx (2015) xxx–xxx 7

Adjusted Cycle-specific fuel consumption of PHEVs in where FC CDj ; FC CSj are the average fuel consumption in the jth driving
CS mode cycle in CD and CS mode separately, and EC CDj ; EC CSj are the average
8.0 -10%
electricity consumption in the jth driving cycle in CD and CS mode
Fuel consumption L/100km

7.5
-15%
separately. UFPHEV is already derived.
7.0
To calculate the energy consumption in the U.S., we simply

Increasing rate
-20%
6.5
replace the average cycle results with the EPA combined average
6.0 -25% results. The UF from Beijing is derived from the driving pattern sur-
5.5
-30%
vey from 2012 to 2013, and the UF from the U.S. is based on the
5.0 NHTS 2001 tests. These values are adopted to calculate the fleet
4.5
-35% average fuel and electricity consumption in PHEVs.
4.0 -40%
The real-world fuel and electricity consumption of PHEVs in
PHEV10 PHEV20 PHEV30 PHEV40 Beijing and the U.S. is illustrated in Figs. 12 and 13. The average
B average EPA Combined B_a vs EPA city fuel consumption in Beijing is 3.32, 2.15, 1.47 and 1.06 L/100 km
for 10, 20, 30 and 40 mile CD ranges, respectively. As the CD range
Fig. 10. Adjusted cycle-specific fuel consumption of PHEVs. increases, the cycle-specific fuel consumption increases, as shown
in Fig. 10, and the average fuel consumption decreases in both
As the CD range increases, the fuel consumption in the CS mode Beijing and the U.S. due to the increase in the electric range over
increases both in Beijing and the U.S., yet PHEVs in Beijing exhibit the total driving range. However, the electricity consumption
approximately 13% lower fuel consumption than that in the U.S., increases from 5.91 kW h/100 km to 13.38 kW h/100 km as the
which is closer to HEV conditions. Electricity consumption in the CD range increases from 16 km to 65 km.
CD mode increases with the CD range, and PHEVs in Beijing con- The average fuel consumption in Beijing is 33.9%, 49.6%, 59.8%
sume 30% less electricity than in the U.S. and 66.4% lower than that of the U.S., respectively, which is much
For a PHEV fully charged at the beginning of each day, if the larger than the cycle-specific fuel reduction of 13% (see Fig. 10).
driving range is taken into consideration, the daily driving range The comparisons reveal the coupled effect of driving conditions
can be divided into the CD and the CS driving range. and the driving range, which is illustrated in Fig. 14.
 Firstly, drivers in Beijing tend to drive shorter ranges than in the
di if di 6 RCD US, thus a PHEV can operate in CD mode for more proportion in the
dCDi ¼
RCD if di > RCD total driving range. Secondly, PHEV with the same batteries has
 :
0 if di 6 RCD different CD ranges in Beijing and US. That‘s because the
dCSi ¼
di  RCD if di > RCD cycle-specific electricity consumption in Beijing is 30% lower,
meaning that PHEVs operating in Beijing have a 43% longer CD
Where dCDi is the driving range in the CD mode of the ith day, and dCSi range for the same battery. The longer CD range enables PHEVs
is the driving range in the CS mode, RCD is the CD range of the PHEV. to drive more ranges on electricity, which magnifies the gap
Considering the driving range and driving conditions together, between the UF of Beijing and the U.S. For example, the PHEV30
the average fuel consumption of a PHEV can be calculated from: has 50 km CD range in Beijing, but a 34 km range in the U.S.
P P P P Thus, the UF is 0.77 in Beijing and 0.41 in the US and the gap
4 4
j¼1
wj FC CDj  dCDi þ
j¼1
wj FC CSj  dCSi becomes 0.36. The driving conditions and driving range in Beijing
FC ave ¼ P
d
dk 2S k together contribute to the fuel reduction. As to the average elec-
!
P4 X
4 tricity consumption shown in Fig. 13, although the cycle-specific
¼ UF PHEV ðRCD Þ  ð j¼1 wj  FC CDj Þ þ ð1  UF PHEV ðRCD ÞÞ  wj  FC CSj : electricity consumption in Beijing is 30% lower than in the U.S.,
j¼1
the percentage of the electric driving range is much larger.
Likewise the average electricity consumption of fleet PHEVs is: Therefore, the PHEV10 in Beijing consumes 48.3% more electricity
than in the US, and the increasing rate for the PHEV40 is 13.6%.
!
X
4 These comparisons also shed light on the optimal design of the
EC ave ¼ UF PHEV ðRCD Þ  wj  EC CDj batteries and CD range of PHEV for Beijing. Because of the short
j¼1 driving range and the coupled effect of driving conditions and driv-
!
X
4
ing range, a PHEV in Beijing has a longer CD range and can drive on
þ ð1  UF PHEV ðRCD ÞÞ  wj  EC CSj : electricity for more proportion of the daily driving range. It implies
j¼1
that a PHEV, with a smaller battery, can achieve a reasonable CD

Adjusted Cycle-specific electricity consumption of PHEVs

30.0 -25%
electricity consumption kWh/100km

28.0 -27%

26.0 -29%
-31%
24.0
Increasing rate

-33%
22.0
-35%
20.0
-37%
18.0
-39%
16.0 -41%
14.0 -43%
12.0 -45%
PHEV10 PHEV20 PHEV30 PHEV40
B average EPA Combined B_a vs EPA city

Fig. 11. Adjusted cycle-specific electricity consumption of PHEVs. Fig. 12. Real-world average fuel consumption of PHEVs in Beijing and the U.S.

Please cite this article in press as: Wang H et al. Energy consumption of electric vehicles based on real-world driving patterns: A case study of Beijing. Appl
Energy (2015), http://dx.doi.org/10.1016/j.apenergy.2015.05.057
8 H. Wang et al. / Applied Energy xxx (2015) xxx–xxx

Real-world average electricity consumption of Fuel consumption and reduction rate of different vehicles
PHEVs 18 100%

Fuel consumption L/100km (Ge)

Electricity consumption kWh/100km

18.0 50% 16 90%

16.0 40% 14 80%
30% 70%

Reduction rage
14.0 12

Increasing rate
20% 60%
12.0 10
10% 50%
10.0 8
0% 40%
8.0 6
-10% 30%
6.0 4
-20% 20%
4.0 -30% 2 10%
2.0 -40% 0 0%
0.0 -50% CV HEV PHEV10 PHEV20 PHEV30 PHEV40 EV
PHEV10 PHEV20 PHEV30 PHEV40
beijing EPA combined
B average EPA Combined B_a vs EPA city Reduction rate in Beijing Reduction rate in US

Fig. 13. Real-world average electricity consumption of PHEVs in Beijing and the U.S. Fig. 15. Fuel consumption and reduction rate of vehicles in Beijing and the U.S.

range in the Beijing driving conditions, satisfying most of the travel

demands of Beijing drivers. A small battery can also help reduce
the purchasing cost and life-cycle GHG emissions associated with
battery manufacturing and extra energy consumption due to bat-
tery weight [35]. Based on this, a PHEV20 with 8.85 kW h batteries
is recommended in Beijing because the 32 km CD range can meet
the average daily driving range of 35 km and the UF is 62.9%.
The fuel reduction from electric vehicles depends on the driving
conditions and driving ranges in a specific region. The case study
for Beijing reveals that electric vehicles are superior in large cities
which have severe driving conditions due to large vehicle popula-
tions, concentrated traveling and people who mainly drive short
ranges to commute to the workplace or public places.

5. Comparisons and discussions of the impact of driving 5.2. Fuel consumption bias associated with the NEDC cycle
patterns
The NEDC cycle is the current standard driving cycle and has
5.1. Comparisons of fuel consumption in Beijing been widely adopted by Chinese manufacturers in evaluating the
fuel economy of conventional and electric vehicles [32].
The fuel consumption of various vehicles in Beijing and the U.S. Consumer’s choices are also guided by the fuel economy results
is illustrated in Fig. 15. The fuel reduction rate for electric vehicles released by the Ministry of Industry and Industrial Technology of
are calculated and compared. The reduction rates reveal that elec- China (MIIT).
tric vehicles in Beijing use less fuel than in the U.S., and the max- To examine the reliability of NEDC results, they are compared
imum fuel reduction margin is 24.7% for the PHEV20. Due to the with real-world values to reveal any bias in the fuel estimation
severe driving conditions of Beijing, conventional vehicles in (see Fig. 16). The real-world fuel consumption of the CVs, HEVs
Beijing have the highest fuel consumption of 12.43 L/100 km. and BEVs in Beijing is 38.8%, 9.3% and 3.1% higher than the NEDC
However, electric vehicles are less sensitive to driving conditions results, respectively. For PHEVs, the rates are 17.8%, 20.9%, 25.3%
and the short driving range is more suitable for range-limited elec- and 26.2%, respectively. This reveals that the label values released
tric vehicles. Therefore, electric vehicles in Beijing show more by MIIT of China underestimate the fuel consumption of vehicles in
potential for fuel reduction. This analysis justifies the promotion Beijing, especially for conventional vehicles and PHEVs.
and demonstration policies for electric vehicles sponsored by the For fuel reduction, the NEDC underestimates the fuel reduction
Beijing government, and recommends the PHEV20, instead of a benefits associated with electric vehicles in Beijing. The underesti-
BEV to achieve the maximum fuel reduction goal. mation of HEVs, PHEV10s and BEVs is 11.6%, 4.7% and 5.7%,

Fig. 14. CD range and UF in Beijing and the U.S.

Please cite this article in press as: Wang H et al. Energy consumption of electric vehicles based on real-world driving patterns: A case study of Beijing. Appl
Energy (2015), http://dx.doi.org/10.1016/j.apenergy.2015.05.057

Fuel consumption bias of NEDC in Beijing
18
Fuel consumption L/100 km (Ge)
16
14
12
10
8
6 [4] The Clean Air Initiative Plan, 2013 [Internet]. Beijing Municipal Commission of
4 beijing.gov.cn/gzdt/gggs/t1322955.htm>.
2 20.0%
0 0.0%
CV HEV PHEV10 PHEV20 PHEV30 PHEV40
beijing
reduction rate in NEDC
Fig. 16. Fuel consumption bias of NEDC cycle in Beijing is shown.
respectively. This fuel consumption and reduction bias of the NEDC
cycle may confuse consumers about the real fuel economy of elec-
tric vehicles, and mislead policies designed to promote electric
vehicles in Beijing. This result addresses the importance of devel-
oping and using the real-world driving cycles in China to design
and evaluate electric vehicles.
6. Conclusions
The impact of real-world driving patterns on the fuel consump-
tion of vehicles is examined in this study to design and apply opti-
mal powertrain configurations in a specific region. Based on a
two-year driving pattern survey and travel data, we calculated
the real-world fuel consumption for electric vehicles in Beijing
and compared them to the U.S. values. A series of conclusions were
reached based on this analysis.
1. Electric vehicles in Beijing, including HEVs, PHEVs and BEVs
yield more fuel reduction benefits than in the U.S., and the max-
imum fuel reduction margin is 24.7% for the PHEV20. The driv-
ing patterns in Beijing are characterized by low speeds, long idle
times, severe speed changes and short driving ranges. Statistics
show that CVs are heavily affected by these driving conditions,
whereas electric vehicles are less sensitive. Moreover, the short
driving range contributes to the utilization of range-limited
electric vehicles, such as BEVs and PHEVs. The comparative
advantages of electric vehicles in Beijing justify the promotion
of their deployment and infrastructure construction by the gov-
ernment. We believe that electric vehicles are superior in large
cities with similar driving conditions.
2. For PHEVs, the coupled effect of the driving range and the driving
cycle are examined in this study. For PHEVs in Beijing, smaller
batteries corresponding to a 30–50 km CD range are preferred
to meet the demands of most drivers and add less extra cost to
the vehicle. More battery than is needed will cause further reduc-
tion in the fuel economy and increase of purchasing cost.
3. The NEDC cycle is regarded as the standard test cycle for
Chinese regulations, and manufacturers use this test when
describing fuel economy results in advertisements. However,
we find that this common practice may underestimate vehicle
fuel consumption. On the other hand, the results underestimate
the fuel reduction benefits associated with electric vehicles,
which may influence the penetration rates of electric vehicles.
This result addresses the importance of developing and using
real-world driving cycles in specific regions to design and eval-
uate electric vehicles.
Please cite this article in press as: Wang H et al. Energy consumption of electric vehicles based on real-world driving patterns: A case study of Beijing. Appl
Energy (2015), http://dx.doi.org/10.1016/j.apenergy.2015.05.057
H. Wang et al. / Applied Energy xxx (2015) xxx–xxx 9
References
140.0%
[1] Wilbanks TJ, Greene DL. The importance of advancing technology to america’s
120.0% energy goals. Energy Policy 2010;38(8).
[2] Wang H, Ouyang M. Transition strategy of the transportation energy and
100.0% powertrain in China. Energy Policy 2007;35(4):2313–9.
[3] Gong H, Wang MQ, Wang H. New energy vehicles in China: policies,
80.0%
demonstration, and progress. Mitig Adapt Strat Glob Change
60.0% 2013;18(2):207–28.
40.0% Transport. [updated 2013 Sep 2, cited 2014 Jun 25], 2013. <http://zhengwu.
[5] Van Noorden R. A better battery Chemists are reinventing rechargeable cells to
drive down costs and boost capacity. Nature 2014;507:26–8.
[6] Society of Automotive Engineers, Recommended Practice for Measuring the
EV
Exhaust Emissions and Fuel Economy of Hybrid-Electric Vehicles, Hybrid
NEDC cycle reduction rate in Beijing Committee, 1999, J1711.
beijing VS NEDC
[7] Neubauer J, Brooker A, Wood E. Sensitivity of plug-in hybrid electric vehicle
economics to drive patterns, electric range, energy management, and charge
strategies. J Power Sources 2013;236:357–64.
[8] Elgowainy A, Han J, Poch L, et al. Well-to-wheels analysis of energy use and
greenhouse gas emissions of plug-in hybrid electric vehicles. Argonne National
Laboratory (ANL); 2010. Report No.: ANL/ESD/10-1.
[9] Xu L, Ouyang M, Li J, et al. Optimal sizing of plug-in fuel cell electric vehicles
using models of vehicle performance and system cost. Appl Energy
2013;103:477–87.
[10] Pearre NS, Kempton W, Guensler RL, et al. Electric vehicles: how much range is
required for a day’s driving? Transport Res Part C: Emerg Technol
2011;19(6):1171–84.
[11] Kelly JC, Macdonald JS, Keoleian GA. Time-dependent plug-in hybrid electric
vehicle charging based on national driving patterns and demographics. Appl
Energy 2012;94:395–405.
[12] Ericsson E. Independent driving pattern factors and their influence on fuel-use
and exhaust emission factors. Transport Res Part D: Transport Environ
2001;6(5):325–45.
[13] Gonder J, Markel T, Thornton M, et al. Using global positioning system travel
data to assess real-world energy use of plug-in hybrid electric vehicles.
Transport Res Record: J Transport Res Board 2007;2017(1):26–32.
[14] Moawad A, Singh G, Hagspiel S, et al. Impact of real world drive cycles on PHEV
fuel efficiency and cost for different powertrain and battery characteristics.
EVS24 International Battery, Hybrid, and Fuel Cell Electric Vehicle Symposium,
Stavanger, Norway, 2009.
[15] Karabasoglu O, Michalek J. Influence of driving patterns on life cycle cost and
emissions of hybrid and plug-in electric vehicle powertrains. Energy Policy
2013;60:445–61.
[16] Lin, Zhenhong, David L. Greene. Significance of daily VMT variation over time
and among drivers on assessment of PHEV energy impact. In: Proceedings of
90th Annual Meeting of the Transportation Research Board (Compendium of
Papers, CD-ROM). No. 11–2242. 2011 Jan 23–27; Washington, D.C.
[17] SAE. Utility Factor Definitions for Plug-In Hybrid Electric Vehicles Using 2001
U.S. DOT National Household Travel Survey Data, Hybrid Committee, J2841,
2009.
[18] Duoba M. Developing a utility factor for battery electric vehicles. SAE Int J
Altern Powertrains 2013;2(2):362–8.
[19] National Household Transportation Survey [Internet], [cited 2014 Jun 25].
<http://nhts.ornl.gov/>.
[20] Transportation Secure Data Center data [Internet]. The National Renewable
Energy Laboratory. [cited 2014 Jun 25]. <http://www.nrel.gov/
vehiclesandfuels/transportationdata/>.
[21] The Danish National Travel Survey (2012) Latest dataset. 2012.
[22] Hao H, Wang H, Ouyang M. Fuel conservation and GHG (Greenhouse gas)
emissions mitigation scenarios for China’s passenger vehicle fleet. Energy
2011;36(11):6520–8.
[23] Huo H, Zhang Q, He K, et al. Vehicle-use intensity in China: current status and
future trend. Energy Policy 2012;43:6–16.
[24] Hou C, Wang H, Ouyang M. Survey of daily vehicle travel distance and impact
factors in Beijing. Advances in Automotive Control, vol. 7, Part 1 p35-40. 7th
IFAC Symposium on Advances in Automotive Control, 2013. Tokyo, Japan,
2013. <http://dx.doi.org/10.3182/20130904-4-JP-2042.00002>.
[25] Raykin L, Roorda MJ, Maclean HL. Impacts of driving patterns on tank-to-wheel
energy use of plug-in hybrid electric vehicles. Transport Res Part D: Transport
Environ 2012;17(3):243–50.
[26] Wu L. A GPS-based Research on the Driving Patterns of Private Passenger
Vehicle in Beijing [dissertation]. Tsinghua University, 2013.
[27] Wang H, Wu L, Hou C. A GPS-based Research on Driving Range and Patterns of
Private Passenger Vehicle in Beijing.EVS27, Barcelona, Spain, November 17–20.
2013.
[28] Hung WT, Tong HY, Lee CP, et al. Development of a practical driving cycle
construction methodology: a case study in Hong Kong. Transport Res Part D:
Transport Environ 2007;12(2):115–28.
[29] Vijayagopal R, Rousseau A. System Analysis of Multiple Expert Tools. SAE
2011-01-0754, SAE World Congress, Detroit, April 2011.
[30] Plotkin SE, Singh MK. Multi-path transportation futures study: vehicle
characterization and scenario analyses. Argonne National Laboratory (ANL),
2009. Report No.: ANL/ESD/09-5.
10 H. Wang et al. / Applied Energy xxx (2015) xxx–xxx

[31] Hou C, Xu L, Wang H. Energy Management of Plug-in Hybrid Electric Vehicles
with Unknown Trip Length. J Franklin Instit. Forthcoming 2014.
[32] The website of automobile fuel consumption of China [Internet]. MIIT China.
[cited 2014 Jun 25]. <http://chinaafc.miit.gov.cn/n2050/index.html>.
[33] qcwp.com [Internet]. Real fuel consumption report; c2008-2014[cited 2014
Jun 25]. <http://beijing.qcwp.com/>.
[34] EPA. Fuel Economy Labeling of Motor Vehicles: Revisions To Improve
Calculation of Fuel Economy Estimates; Final Rule. Environmental Protection
Agency; 2006 Dec. Report No.: EPA420-R-06-017. 2006.
[35] Michalek JJ, Chester M, Jaramillo P, et al. Valuation of plug-in vehicle life-cycle
air emissions and oil displacement benefits. Proc Natl Acad Sci 2011;108(40):
16554–8.

Please cite this article in press as: Wang H et al. Energy consumption of electric vehicles based on real-world driving patterns: A case study of Beijing. Appl
Energy (2015), http://dx.doi.org/10.1016/j.apenergy.2015.05.057