Fichier PDF

Partage, hébergement, conversion et archivage facile de documents au format PDF

Partager un fichier Mes fichiers Convertir un fichier Boite à outils Recherche Aide Contact



10.1038@s41560 018 0133 0 .pdf



Nom original: 10.1038@s41560-018-0133-0.pdf
Titre: Impacts of fleet types and charging modes for electric vehicles on emissions under different penetrations of wind power
Auteur: Xinyu Chen

Ce document au format PDF 1.4 a été généré par Springer / , et a été envoyé sur fichier-pdf.fr le 17/05/2018 à 12:01, depuis l'adresse IP 24.203.x.x. La présente page de téléchargement du fichier a été vue 170 fois.
Taille du document: 2.1 Mo (9 pages).
Confidentialité: fichier public




Télécharger le fichier (PDF)









Aperçu du document


Articles
https://doi.org/10.1038/s41560-018-0133-0

Impacts of fleet types and charging modes for
electric vehicles on emissions under different
penetrations of wind power
Xinyu Chen1,6*, Hongcai Zhang   2,6, Zhiwei Xu2,6, Chris P. Nielsen3,6, Michael B. McElroy4,6* and Jiajun Lv5,6
Current Chinese policy promotes the development of both electricity-propelled vehicles and carbon-free sources of power.
Concern has been expressed that electric vehicles on average may emit more CO2 and conventional pollutants in China. Here,
we explore the environmental implications of investments in different types of electric vehicle (public buses, taxis and private
light-duty vehicles) and different modes (fast or slow) for charging under a range of different wind penetration levels. To do
this, we take Beijing in 2020 as a case study and employ hourly simulation of vehicle charging behaviour and power system
operation. Assuming the slow-charging option, we find that investments in electric private light-duty vehicles can result in an
effective reduction in the emission of CO2 at several levels of wind penetration. The fast-charging option, however, is counterproductive. Electrifying buses and taxis offers the most effective option to reduce emissions of NOx, a major precursor for
air pollution.

A

ccording to the Chinese Ministry of Environmental
Protection, 66 of the 74 largest cities in China failed to meet
the nation’s ambient air quality standards in 20141. The
power generation and transportation sectors represent the largest
sources of air pollutants and their precursors2–4. While the annual
power demand in China tripled from 2003 to 2016, the growth in
vehicle purchases increased even faster: private vehicle sales grew
from 2 million in 2003 to 24 million in 2016.
Electric vehicles (EVs) have been promoted aggressively by both
policymakers and industry stakeholders in China. With an emphasis on electrifying buses and taxis, a city-based government pilot
project, Ten Cities, Thousand Vehicles5, was initiated in 2009 and
expanded in 20146. Under this programme, the number of electric
buses in Beijing reached 2,672 in 2016 and is expected to double by
2017 and quadruple by 2020. For private EVs, strong incentives have
been introduced: a subsidy of up to US$9,500 has been provided
since 2010 for the purchase of such vehicles7 and battery EVs have
been exempt from sales tax (10% of the retail price) since 20147. As a
result, about 800,000 EVs were produced in China in 20178, a 134%
increase from the year 20159. Growth of EVs is projected to surpass
gasoline light-duty vehicles (LDVs) in Beijing by 201810. Along with
a rapid increase in EV sales, the charging infrastructure is also on
the rise. The annual growth rate for the number of public charging
piles averaged close to 90% over the past five years9. Approximately
40% of residential communities in Beijing are now equipped with
charging facilities11. A national fast-charging network is planned by
the State Grid Corporation of China (SGCC) for 2020.
In contrast to the flourishing EV market, the full fuel-cycle
environmental impacts of electrified LDVs have been deemed
negative compared to gasoline vehicles in areas dominated by coalfired electricity generation (for example, North China), in terms of
both emissions of conventional air pollutants (NOx, SO2, PM10 and

primary PM2.5) and also CO2 (refs 12,13). The existing analyses on the
environmental impact of EVs in China12,13 are based on the current
annual average energy mix for power generation and regional average emission inventories. The differences among more specific EV
development options have not been addressed.
The overall environmental impacts of EVs are influenced
strongly by both the types of vehicle and the timing of charging14.
Public buses, taxis and private LDVs differ significantly in terms
of emission factors (for different pollutants), fleet sizes and annual
average mileage. The emission factor for NOx from a diesel bus
(per kilometre travelled) is projected for example to be about 80
times greater in 2020 than the value from a private LDV15. The
differences in fleet type have not been considered in previous EV
studies for China or other countries12–14,16–19, but could be important for air pollution control. In addition, it has been realized in
the United States14,17,18 that the environmental impact of EVs on
the operation of power systems is highly dependent on the timing of charging, which depends both on fleet types and charging
strategies (fast versus slow). In the absence of actual statistical data,
the impacts of these considerations have not been discussed in the
Chinese context.
Increased investment in wind power could have an important
influence on the future of EVs in China. The installed capacity of
wind power in China had reached 188 GW by the end of 201720.
It is expected to climb to 200 GW by 2020, and to 400 GW by
203021. These wind investments have been concentrated mainly in
the northern part of China, and incorporating this variable power
resource in a power grid dominated by inflexible coal-fired units
has been a major challenge. In these regions, combined heat and
power (CHP) units account for approximately half of all thermal
generators16. The operational flexibility of power production is limited further by the demand for heat during the heating season16.

School of Engineering and Applied Sciences, Harvard University, Cambridge, MA, USA. 2Department of Electrical Engineering, Tsinghua University,
Beijing, China. 3Harvard China Project and School of Engineering and Applied Sciences, Harvard University, Cambridge, MA, USA. 4School of Engineering
and Applied Sciences and Department of Earth and Planetary Sciences, Harvard University, Cambridge, MA, USA. 5School of Electrical Engineering,
Xi’an Jiaotong University, Xi’an, Shaanxi, China. 6These authors contributed equally: Xinyu Chen, Hongcai Zhang, Zhiwei Xu, Chris P. Nielsen,
Michael B. McElroy and Jiajun Lv *e-mail: xchen@seas.harvard.edu; mbm@seas.harvard.edu
1

Nature Energy | www.nature.com/natureenergy

© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.

Articles

NATurE EnErgy

As a result, 17% of the total wind power potentially available from
China’s wind systems was curtailed—wasted—in 2016, resulting in
an economic loss of about US$3.7 billion (ref. 22).
With appropriately managed charging, EVs could facilitate wind
integration by providing a measure of flexibility in demand and/or
energy storage capacity. As we shall see, if the time for charging were
to overlap with the interval of peak power demand, additional coalfired capacity would be required, and benefits from this flexibility
would be reduced significantly, potentially increasing the requirements for curtailment of wind power. The impact of slow-charged
LDVs on wind power integration in the Inner Mongolia energy
system was found to be positive, on the basis of stylized charging
profiles19. However, the electrification of public fleets, options for
fast charging and implications with respect to reductions in air pollutants have not been considered to this point.
Here, we explore the environmental impacts of different electrified transportation options taking account of different strategies
for charging with respect to the accommodation of wind power.
Specifically, we investigate how various types of EV (buses, taxis and
private vehicles) and different modes of charging (fast and slow)
could affect the overall emission of greenhouse gases and selected
air pollutants (NOx, SO2, primary PM2.5) under different levels of
wind power penetration. Hourly simulations for interlinked electrified transportation, power and heating systems are conducted for
Beijing and the surrounding area, incorporating detailed statistics
for driving patterns and comprehensive modelling of energy systems. The results show that the vehicle charging mode for LDVs
will have a significant influence on power system operations and
the integration of wind power, especially at elevated wind levels.
Emissions of CO2 can be reduced by slow-charging LDVs at several
levels of wind penetration. Electric LDVs with fast charging, in contrast, will result in higher emissions of CO2, reflecting their negative
influence on wind integration. Electrifying public transportation
(buses and taxis) is most effective in reducing emissions of NOx, a
major precursor for air pollution.

Model and scenarios

The environmental impacts of different vehicle electrification strategies are quantified using a newly developed integrated energy
system optimization model, whose structure is presented in Fig. 1.
We begin using field investigations and regulatory reports to summarize information on driving patterns for buses, taxis and LDVs
in China. These driving patterns are then used to estimate probabilistic charging behaviour (arrival times, departure times and the
state of charge) for vehicles of different fleet types under different
charging conditions. The probabilities are employed in Monte Carlo
simulations to project the collective charging demand for vehicles
throughout the year, with 15-min time resolution. To quantify the
environmental impacts from EV charging, an optimization model is
developed to simulate the energy systems (covering power, heating

and electrified transportation) on an hourly basis for the entire
year. Hourly variations of power demand, wind power and charging
requirements for EV fleets are considered. The model accounts for
limitations on the possible range for power output, power generation changes over consecutive hours, and minimum times required
for restart or shut down of thermal units. It also accounts for variations in heating demand and limitations of power and heat production for individual CHP units. The detailed modelling formulation
is presented in the Methods.
Beijing, as the most populous metropolitan area in China with
the largest vehicle stock, is selected as a case study to evaluate the
impacts of EV options in 202023. Projections of vehicle populations for buses, taxis and private LDVs for the target year amount
to approximately 30 thousand, 66 thousand and 5.6 million, respectively. The projected annual kilometres travelled per vehicle (VKT)
for LDVs, taxis and buses correspond to 18,000 km, 58,000 km and
126,000 km, respectively (see Supplementary Fig. 1, Supplementary
Table 1 and Supplementary Note 1). Note that VKT values for public
buses and taxis are three and seven times higher than for LDVs15,24.
The study accounts for hourly power consumption, wind production and detailed energy system configurations and operations for
Beijing and its neighbouring areas, as detailed in the Methods.
Three vehicle electrification strategies are compared with a business-as-usual (BAU) reference in which conventional vehicle fleets
are powered solely by fossil fuels. These three strategies are: Pub-f,
electrification of public transportation vehicles, buses (non-natural
gas-propelled) and taxis, using fast charging; Prv-s, plug-in electric private LDVs serviced using slow-charging stations; and Prv-f,
which is the same as Prv-s but assuming service by fast-charging
stations. The value assumed for the number of electric private LDVs
is based on the number of new LDVs anticipated to be sold between
2015 and 2020, accounting for the projected growth in the vehicle
population and projected retirements and replacements for existing vehicles (detailed projections for growth and replacements of
vehicles are presented in Supplementary Notes 1 and 2). The government plan is to convert 37% of public buses to compressed natural gas; the scenario adopted here applies to the additional 63%. In
each scenario, penetration of wind power is allowed to vary from
0% to 40%. Wind curtailment, in addition to emissions of NOx, SO2,
primary PM2.5 and CO2 from combined on-road transportation and
electricity generation, is calculated for each electrification strategy,
as well as for combinations of options (Cmb-f, combining Pub-f and
Prv-f; Cmb-s, combining Pub-f and Prv-s). Details of the above scenarios are summarized in Table 1.

Characteristics of EV driving behaviour in Beijing

The temporal distribution of electricity demand from an individual
EV is determined by the time of its arrival and departure at a specific charging station, by the mode employed for its charging (fast/
slow), the state of charge (SOC, the ratio of the energy remaining

Integrated energy optimization model

Driving behaviours

Thermal unit

Monte Carlo
simulations

EV charging
Hourly output

Wind power

Wind curtailment
Fleet aggregation
for charging
characteristics

EV

Power
network

Power
balance

CHP

Heat balance

Operational cost
Emissions

Electric vehicle
simulation

Output

Fig. 1 | Modelling framework for the integrated energy system optimization. The model incorporates statistical driving behaviours to simulate the
aggregated charging characteristics of different EV fleets. The charging characteristics are then fed into the integrated energy system optimization model,
generating a variety of results for simulation, including wind curtailments and overall emissions.
Nature Energy | www.nature.com/natureenergy

© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.

Articles

NATurE EnErgy
Table 1 | Descriptions of the BAU and alternative scenarios for
2020 considered in this study
Scenario
abbreviations

Fleet type

Fuel type

EV charging
methodsa

BAU

Bus

Diesel

NA

Taxi

Gasoline

NA

Private vehicles

Gasoline

NA

Bus

Partly electric

Fast charging

Taxi

Electric

Fast charging

Private vehicles

Gasoline

NA

Bus

Diesel

NA

Taxi

Gasoline

NA

Private vehicles

Partly electric

Fast charging

Bus

Diesel

NA

Taxi

Gasoline

NA

Private vehicles

Partly electric

Slow charging

Bus

Partly electric

Fast charging

Taxi

Electric

Fast charging

Private vehicles

Partly electric

Fast charging

Bus

Partly electric

Fast charging

Taxi

Electric

Fast charging

Private vehicles

Partly electric

Slow charging

Pub-f

Prv-f

Prv-s

Cmb-f

Cmb-s

The detailed charging speeds for different vehicle types and charging locations are derived on the
basis of the Standards of EV Conductive Interface released in December 201540, summarized in
Supplementary Table 2.
a

in the battery to the rated capacity of the battery) and the energy
required for its next trip.
Here, we investigate first the above characteristics using statistics on driving behaviour obtained through field investigations of
public bus fleets and observations at commercial parking lots25, and
studies of transportation26,27 complemented by additional information from interviews with drivers. The statistical distributions of
arrival/departure times for private cars, buses and taxis for charging
on both weekdays and weekends are summarized in Supplementary
Table 2 and Supplementary Fig. 2. Based on the above information,
the likelihoods for an EV to physically connect to a certain charging spot at specific time intervals are illustrated in Fig. 2. The figure
indicates the probability of physical connectivity as a function of
time and location.
Buses are in service normally from about 5:30 to about 23:00,
with an average daily travel distance of approximately 160 km.
Since the expected driving distance for an electric bus per charge is
estimated at 85 km in existing pilot projects25, an electric bus must
be charged at least once during the day and once at night, using
fast-charging facilities at the bus terminal to minimize times when
the vehicle would be otherwise out of service. We assume that fastcharging of electric buses takes place between 10:00 and 16:30 and
between 23:00 and 5:30, as in ref. 25.
Fifty-five per cent of taxis in Beijing are driven by a single driver,
denoted hereafter as ‘single shift’; the remaining 45% are shared by
two drivers, denoted as ‘double shift’26. Reflecting the relatively long
daily travel distance of taxis in Beijing (averaging nearly 238 km
per day for single-shift taxis and 476 km per day for double-shift
taxis), single-shift and double-shift taxis are charged once and twice
per day, respectively, using fast charging. Here, we assume that the
one charge for single-shift taxis occurs between 20:00 and 22:00
(after rush hour), and that the two charges for double-shift taxis are
implemented between 2:00 and 4:00 (before the early morning shift)
and between 13:30 and 15:30 (before the afternoon shift).

Private LDVs in Beijing are employed mainly for commuting between homes and work places on weekdays, leaving home
between 6:00 and 9:30, arriving at work between 6:30 and 10:00.
The distribution of departure and arrival times is taken from ref. 27,
detailed in Supplementary Fig. 2. The majority of these vehicles
return home from work between 17:30 and 21:3027. They may be
used also for non-work purposes after work between 19:00 and
22:00 on weekdays27, with an average parking time estimated at
80 min based on field observations for nearly 10,000 private cars at
commercial centres in Haidian District, Beijing27. On weekends, we
assume that private vehicles are used solely for non-work purposes.
Besides parking at home at night, these vehicles may be parked also
at commercial establishments from 9:00 to 22:00.
Under the fast-charging scenarios (Prv-f and Cmb-f), EV owners must travel to fast-charging stations to recharge their batteries.
We assume that fast charging is concentrated mainly during lunch
hours (11:00–13:00) and after work (17:00–21:00). On weekends,
EV owners can either charge their vehicles at their residences
(9:00–21:00) or during trips to shopping or entertainment centres
(16:00–22:00). The detailed parameter settings for EV charging and
further assumptions are summarized in Supplementary Table 2.

Optimized charging

The impact of EV charging on power grid operation is evaluated on
an hourly basis throughout the year. Quarter-hourly driving behaviour for each EV is simulated first via Monte Carlo analysis based on
the probabilistic distributions for driving behaviours noted above.
Driving behaviours for EVs, together with hourly variations of wind
output and power demand, were incorporated in the energy system simulation model (see Methods) to determine the optimized
power consumption profile for the EV fleet and the corresponding
wind curtailment for the diverse EV scenarios and for the different
wind penetration levels over the course of the year. (The optimization model is detailed in the Methods.) Results for 10–16 March, for
a wind penetration level of 20%, are displayed in Fig. 3. The hourly
power demands for different EV strategies relative to Beijing electricity demand and wind generation are displayed in the panels on
the left, with corresponding wind power curtailment profiles, relative to the BAU case, illustrated on the right.
Power demand in Beijing peaks twice on workdays, at around
12:00 and 20:00, respectively, reaching a minimum between 3:00
and 7:00. Given that the fleet population is relatively small, the
charging power for electric buses and taxis has a limited impact on
the profile for total electricity demand (scenario Pub-f, Fig. 3a,f).
When fast charging is adopted for electric LDVs, the power demand
for charging LDVs overlaps significantly with the daily peak power
demand in Beijing, increasing daily peak power demand by 52.5%
(Fig. 3b). The increase in peak demand with the fast-charging scenario (Prv-f) requires the operation of additional coal-fired power
generators. These additional coal-fired generators are obliged to
keep producing power during off-peak hours, contributing further
to wind curtailment. If the same number of electric LDVs were
serviced using slow charging, the charging load could be allocated
mainly to off-peak periods (Fig. 3c), and could take advantage of
otherwise-curtailed wind power, thus lowering the emission intensity of associated power generation.

Emission results

Annual emissions of NOx, CO2, SO2 and primary PM2.5 for different
combinations of EV strategies and wind levels were analysed on the
basis of the hourly simulation results described above. The relative
contributions to the total emissions of NOx, CO2, SO2 and primary
PM2.5 from both the transportation and power generation sectors
under the BAU scenario (gasoline vehicles combined with 0% wind
power) for Beijing in 2020 are illustrated in Fig. 4. The potentials
for reduction of CO2 and NOx emissions corresponding to different

Nature Energy | www.nature.com/natureenergy

© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.

Articles

NATurE EnErgy
a 0.30

b 1.0

Bus

0.15
0

0

c 0.6

d 0.30

Taxi, single shift

0.3

0.15

0

0

Probability

e 0.50

0.1

0

0

g 0.2
0.1

0.15

0

0

Private LDVs, work, weekday

j 0.10

Private LDVs, home, weekday

0.5

0.05

0

0

k 0.8

Private LDVs, shopping, weekend

h 0.30

Private LDVs, home, weekend

i 1.0

Private LDVs, work, weekday

f 0.2

Private LDVs, home, weekday

0.25

Private LDVs, shopping, weekday

l 0.10

Private LDVs, home, weekend

Private LDVs, shopping, weekend

0.05

0.4
0

Taxi, double shift

0.5

3

6

9

12

15

18

21

24

0

3

Hours

6

9

12

15

18

21

24

Hours
Fast charging

Slow charging

Fig. 2 | Probabilities for physical connection to charging facilities for different vehicles with different charging options. a–k, Probabilities for connectivity
for different hours in a day for different fleet types, charging options, locations and weekday/weekends: bus using fast charging (a); taxi with double shift
(b); taxi with single shift (c); private LDVs using fast charging near the work place during weekdays (d) or at a shopping centre during weekends (f);
private LDVs using fast charging near home during weekdays (e) or weekends (g); private LDVs using slow charging at work during weekdays (h); private
LDVs using slow charging at home during weekdays (i) or weekends (k), private LDVs using slow charging at a shopping centre during weekdays (j) or
weekends (l). The colour indicates the charging mode and the horizontal axis indicates the hours in a day starting from midnight.

combinations of wind penetration and EV strategies relative to the
BAU scenario are summarized in Table 2. As will be indicated later,
electric LDVs, using slow charging, could take advantage of available wind power to accomplish an effective reduction in CO2 emissions. However, emissions would be significantly increased if these
vehicles were to rely on the fast-charging option, mainly attributed
to an increase of wind curtailment. Electrifying public transportation, though contributing to a minor change in total CO2 emissions
because of a limited fleet size, provides the most effective solution to reduce NOx emissions. Results for SO2 and primary PM2.5
are presented in Supplementary Note 3 and Supplementary Figs. 3
and 4. Sensitivity analyses of emission results for different populations of electrified LDVs are summarized in Supplementary Note 4
and Supplementary Tables 3–6.
The power sector represents the major source of CO2 emissions
(85% in Fig. 4). Shares of CO2 emissions by vehicles are proportional to the shares of fuel consumed: private vehicles contribute
10% to total emissions, buses and taxis 1%, and other vehicles,
mainly heavy trucks, 3%, as indicated in Fig. 4. Renewable energy
offers a major opportunity to reduce overall CO2 emissions in the
energy system: a 29% (19%) CO2 reduction at 40% (20%) wind penetration could be achieved, following the scenarios summarized in
Table 2. We note here that wind penetration is defined on the basis
of total power consumption. Since part of the power consumption
is imported from other regions, wind penetration in the local power
generation mix is slightly higher. The marginal CO2 reduction at
elevated wind penetration levels is low, limited by requirements for
increased wind curtailment. Noting that buses and taxis are responsible for only 2% of the total CO2 emissions, electrifying public
buses and taxis has a minor impact on total CO2 emissions.

Electric LDVs, using the slow-charging strategy, can contribute
to a further decease in CO2 emissions assuming that power is provided partly from wind. Converting gasoline LDVs to electric LDVs,
using the slow-charging strategy, would increase CO2 in emissions
in the absence of deployment of wind power, but would contribute
to a decrease in emissions if more than 30% of the power were supplied by wind. In addition to the reduced intensity of emissions in
the power sector at higher wind penetration, the reduction in wind
curtailment facilitated by using slow-charging EVs, as illustrated,
would contribute also to a decrease in CO2 emissions.
Substituting gasoline LDVs with fast-charging EVs, however,
will result in an increase in CO2 emissions as the penetration of
wind power rises. At 40% wind penetration, this substitution (gasoline to fast-charging EVs, Prv-f) will increase overall CO2 emissions by 10%, cancelling the contribution from as much as 12 GW
of wind generation. Charging the same number of EVs in different ways (fast versus slow) would result in a 11% difference in total
CO2 emissions at the 40% wind level, equivalent to the amount of
CO2 emitted annually by 5.8 million gasoline LDVs. Vehicle charging strategies are projected to have a significant impact on emissions of SO2 and primary PM2.5, as summarized in Supplementary
Figs. 3 and 4.
A different strategy is required to reduce NOx emissions, one
of the most important precursors for air pollution. The power
sector will be responsible for at most one-third of total NOx emissions, when the more stringent emission standards for thermal
generators are fully implemented28. Increasing wind penetration
would be less effective in reducing NOx emissions: NOx emissions
would be reduced by only 14% at a wind penetration level of 40%.
Consequently, the influence on NOx emissions of the choice of
Nature Energy | www.nature.com/natureenergy

© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.

Articles

NATurE EnErgy
Off-peak period

Baseline electricity demand

Curtailed wind without PEV charging

Electricity demand and vehicle charging

f 1.0

Capacity factor

Power (GW)

a 30

Pub-f
20
10
0

20

40

60

80

100

120

140

160

10

Power (GW)

60

80

100

120

140

160

30

Capacity factor

20

40

60

80

100

120

140

160

i
Cmb-f

20
10

30

20

40

60

80

100

120

140

160

0

20

40

60

80

100

120

140

160

40

60

80

100

120

140

160

40

60

80

100

120

140

160

40

60

80

100

120

140

160

Cmb-s

10
20

40

60

80
100
Hours

120

140

160

Prv-f
0.5

0

20

Prv-s

0.5

0

1.0

20

Cmb-f

0.5

0

20

j 1.0

20

0

Pub-f

0.5

h 1.0

Prv-s

10

0

Power (GW)

40

20

0

e

20

Capacity factor

Power (GW)

c 30

d

Capacity factor

Prv-f
20

0

Wind power curtailment

g 1.0

Capacity factor

Power (GW)

b 30

Electricity demand for PEVs

Wind

Curtailed wind with PEV charging

40

60

80

100

120

140

160

Cmb-s
0.5

0

20

Hours

Fig. 3 | Charging consumption profiles for EV fleets and corresponding implications for wind curtailment. a–j, Hourly EV charging demand (a–e)
and wind curtailment rates (f–j) for 10–16 March under the scenarios Pub-f (a,f), Prv-f (b,g), Prv-s (c,h), Cmb-f (d,i) and Cmb-s (e,j). PEV, plug-in
electric vehicle.

charging strategies for private EVs is relatively minor: adopting a
fast- or slow-charging strategy would impact total NOx emissions
by no more than 5%, despite significantly different rates for wind
curtailment for the two scenarios.
1% 1%
CO2

85%

10% 3%
1%

NOx

40%

15%

11%

33%
0.3% 0.9%

SO2

97.1%

1.6%
1.4% 0.4%

PM2.5

84.9%
Electricity generation

Bus

6.1%7.2%
Taxi

Private vehicle

Other vehicle

Fig. 4 | Relative contributions to emissions of power generation
and transportation sectors. Contributions of power generation and
transportation to CO2, NOx, SO2 and primary PM2.5 emissions by source for
Beijing in 2020 assuming zero contribution from wind. The contribution to
SO2 from taxis is less than 0.1% and is not shown in the figure.

Electrification of the public fleet provides the most effective
strategy to reduce NOx emissions. Public transportation, despite
the limited fleet size (30,000 buses and 66,000 taxis), is responsible
for 18% of total NOx emissions, equivalent to the contribution from
8.2 million private LDVs. The importance of the emissions of NOx
from the public transportation sector is related in part to the high
associated annual average VKT values and emission factors. The
NOx emission factor for public buses is on average 80 times higher
than that of gasoline LDVs (6.25 g km−1 for the public bus fleet versus 0.08 g km−1 for private LDVs)15. Electrification of the public fleet
alone would reduce total NOx emissions by 10%, equivalent to the
NOx emitted annually by all of the LDVs operational in Beijing. The
reduction in NOx emissions that would result from electrification
of the public fleets could rise to as much as 24% if combined with
exploitation of wind at a level of 40%.

Discussion

Whether an EV reduces or increases total CO2 emissions depends
largely on how and when the vehicle is charged, as discussed above.
Especially at elevated levels of wind penetration, EVs could cancel
out a large fraction of the environmental benefits from renewables if
the charging is not managed properly. These complications have not
been recognized in current policies addressing either EV promotion

Nature Energy | www.nature.com/natureenergy

© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.

Articles

NATurE EnErgy

Table 2 | Reduction percentage for CO2 and NOx under different combinations of wind penetration and EV strategies
CO2 reduction percentage

NOx reduction percentage

Wind level (%)

Wind level (%)

0

10

20

30

40

0

10

20

30

40

EV strategies
BAU (gasoline vehicles alone)

0

10

19

25

29

0

5

9

12

14

Pub-f (electrification of public vehicles)

−​0.2

10

19

25

30

10

15

19

22

24

Prv-f (electric LDVs with fast charging)

−​3.2

5

12

16

19

0.5

5

7

10

11

Prv-s (electric LDVs with slow charging)

−​2.7

8

18

25

30

0.7

6

10

14

16

Cmb-f (Pub-f +​ Prv-f)

−​3.4

6

12

17

20

11

15

18

20

22

Cmb-s (Pub-f +​ Prv-s)

−​2.7

8

18

26

31

11

16

21

24

27

The reduction baseline is chosen as total emissions from the scenario of gasoline vechiles combined with thermal generation. A negative value indicates an increase in emissions.

or wind power integration. To realize the potential environmental
benefits of EVs, it is important for the power system to transition
away from coal-fired generation. More importantly, incentives
should be introduced to promote charging of the EVs at off-peak
times, to take advantage of idled grid capacity and otherwise curtailed wind power. Time-of-use tariffs for EV charging—a lower
tariff at off-peak hours with a higher tariff at peaking hours—could
provide such an incentive. Tariffs for fast-charging and slow-charging modes should be differentiated to internalize environmental
impacts. A coherent plan for the charging infrastructure, with an
emphasis on slow-charging piles, will be critical if potential benefits
are to be realized from the early stage of the growing EV market.
Electrification of the public fleet, as indicated, serves as the most
effective strategy for the control of NOx emissions. The vehicle purchasing cost required to substitute 63% of the operating public buses
(about 20,000) would amount to about 20 billion RMB, assuming a
cost of 1 million RMB (US$0.16 million) per bus according to bidding in China in 201729. This expenditure would be equivalent to
the cost associated with as few as two or three gigawatt-level coalfired plants. The total investment for air pollution control in the
Beijing region amounted to 800 billion RMB over the period of 2013
to 201730. In this case, the purchasing costs to electrify the public
fleet could be accommodated at a cost equal to 2.5% of the total
budget. Considering the relative ease for deployment of charging
infrastructure at bus terminals, electrified public buses could offer a
potentially cost-effective means to address air pollution in Beijing.
The current target for Beijing, 5,000 electric buses by 201730, underestimates such effectiveness. A comprehensive economic feasibility
assessment for the deployment of a highly electrified public transportation system should be prioritized.

Methods

Overview of integrated energy system model and simulations. The wind
curtailment and emission results presented above are based on hourly simulation
of energy systems for the Jing-Jin-Tang (JJT) region covering Beijing and the
surrounding area (geographical coverage illustrated in Supplementary Fig. 5 and
Supplementary Note 5). As indicated in the introduction, the inflexible CHP units
are considered as one of the most important factors currently contributing to wind
curtailment. In addition to requirements for the charging of EVs, the optimization
model employed here accounts also for the operational properties of CHP units in
simulating the operation of power systems.
The integrated energy system optimization model (IESOM) simulates and
optimizes the charging of electrified transportation, the operation of power
systems and the interaction with heating systems through CHP units. For the
transportation sector, a Monte Carlo simulation is conducted for individual EVs
based on the statistics on driving behaviours with a time resolution of 15 min over
the course of a year. The resulting requirements for charging of individual EVs
are aggregated to describe the requirement for composite EV fleets. The charging
requirements for EV fleets, for different scenarios, are then taken into account in
the IESOM, to optimize jointly the charging for EVs and the production of energy
from both the power and heating sectors.

Extending the conventional unit commitment model, IESOM considers
additionally requirements for EV charging, as well as the operational limits for
CHP units. The scheduling and simulation account for hourly variations in energy
demand and power generation from wind, charging for EVs, operational flexibility
of individual power generators and CHP units, constraints for inter-regional
transmission and requirements for reserves. Reserves are provided by thermal
and CHP units, slow-charging EVs and wind power. Internal restrictions for
power and heat production from CHP units are modelled accounting for hourly
variations of heating demand. Annual fuel costs, emissions and wind curtailment
rates are aggregated from the hourly simulation results. The modelling structure is
summarized in Fig. 1.
Methodology for simulating and optimizing charging of EVs. Monte Carlo
simulations are applied first to model the charging behaviour for vehicles according
to the statistical distributions of different fleet types, as described in the main
text. The charging behaviours for individual vehicles are then formulated and
aggregated to define the collective charging requirements for EV fleets.
The arrival/departure times and initial SOCs for each EV in the fleet were
generated using Monte Carlo simulations specific to vehicle type, subject to the
appropriate probability distribution. In scenarios that include public vehicles, we
sampled the arrival times stochastically and the associated initial SOCs every day
for the 30,499 buses and 66,646 taxis, calculating their time-varying electricity
demand assuming that each vehicle begins charging instantaneously at its rated
charging power when it plugs in, continuing until its battery is fully charged. For
private vehicles undergoing slow charging, we first determined stochastically the
charging location (that is, home, work place or shopping centre) for each private
vehicle according to the type of day (workday or weekend), and then generated the
appropriate arrival/departure times and SOCs, subject to the specified probability
distributions. For fast-charging private vehicles, arrival times, initial SOCs and
charging load profiles are determined in the same manner as for public vehicles
after randomly simulating charging locations and times.
Given information on simulated arrival/departure times and charging
requirements for each vehicle, a method for aggregating the charging requirements
for individual EVs was adopted to schedule the charging of EVs when they are
connected to charging piles under the slow-charging mode. Instead of scheduling
the charging power for millions of EVs at each time period, the strategy
proposed aggregates the constraints for each individual EV, and optimizes the
overall charging power for the entire fleet to balance fluctuations in wind power
generation and minimize the cost for energy production. Pioneered and proved in
earlier publications31–33, this strategy could significantly improve the computational
efficiency of EV charging optimization.
Specifically, charging profiles for an individual EV must satisfy two constraints:
the maximum charging power allowed at the charging station (determined by
the rated charging power of the charging piles); and to reach the expected battery
energy level after a given time (determined by the charging behaviour described
above). We use two types of ‘boundary’ to define the aforementioned constraints.
First, the boundaries for the charging power (power boundaries): the upper
bound represents the maximum charging power of an EV when it is connected to
the station; the lower boundary represents its minimum charging power (zero).
Second, the boundaries for the cumulative charging energy (energy
boundaries): the upper boundary represents a charging strategy in which an
EV begins charging instantaneously at the maximum rate continuing until fully
charged; the lower boundary represents a charging strategy in which the charging
of an EV is delayed as much as possible, charging then at the maximum feasible
rate. Possible cumulative charging power and energy trajectories for an EV must
fall within these boundaries.
We utilize then the summation of power and energy boundaries for all
individual EVs to represent the EV fleet’s aggregate power and energy boundaries.
Nature Energy | www.nature.com/natureenergy

© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.

Articles

NATurE EnErgy
A detailed description of the boundary aggregation is presented in refs 32,33. For
the entire fleet, the overall charging trajectory must lie between the aggregated
upper and lower bounds for individual vehicles. Mathematically, the aggregated EV
charging power pte is bounded by the following:
pt− ≤ pte ≤ pt+ ,
et− ≤

1≤t≤T

t

∑ pτe Δt ≤ et+,

(1)

1≤t≤T

τ=1

where T is the number of time intervals; Δ​t is the duration between two
consecutive time intervals; pt− and pt+ are respectively the lower and upper bounds
for the aggregated charging power of the entire fleet at time t—the sum of the lower

+
and upper power bounds from the individual vehicles, respectively. et and et are
respectively the lower and upper bounds for the cumulative energy consumed by
the entire fleet by time t—the sum of the lower and upper energy bounds from the
individual vehicles, respectively. The aggregated boundaries for equation (1) are
incorporated in the simulation of the energy system as discussed in the following
section (see further detail in Supplementary Note 6).
Optimization for the integrated energy systems. Incorporating the charging
requirements for EV fleets, the IESOM minimizes the overall cost for power and
heat production, as well as the curtailed wind power. Besides the constraints
considered in the regular unit commitment model (range of thermal unit power
outputs, ramping limits, minimum on/off times, power balance and reserves),
constraints for charging requirements are also incorporated. The model considers
the detailed constraints for CHP units, including restrictions for power and heat
production, hourly ramping limits and minimum on/off times.
The decision variables for the integrated energy optimization model include
the hourly energy production and consumption across three sectors: aggregated
charging power for EV fleets, hourly power generation for different generators,
and power and heat production for CHP units. In scheduling thermal units, the
model accounts for continuous variables associated with dispatch and binary
e
e
variables associated with commitment decisions (online or offline). pi, t and ei, t
represent respectively the aggregated (slow) charging power and equivalent energy
h
w
level for the ith EV fleet at time t; pi, t and pi, t define the power production from the
c
c
ith conventional power plant and the ith wind farm at time t; pi, t and qi, t represent
power and heat production from the ith CHP unit at time t. The binary variables
c
Iih, t and Ii, t denote the online/offline status of the ith thermal and CHP units at
time t. The parameters and variables employed in the IESOM are summarized as
nomenclature in Supplementary Table 7.
The model minimizes wind power curtailment, as well as fuel and start-up
costs. The objective function includes the total fuel cost, start-up costs and penalties
for curtailment of wind power. The fuel cost considered in the objective includes
the fuel cost for conventional power plants and for CHP units, according to:
min f =

T

Nh

T

Nc

T −1

Nh

T −1

Nc

t =1

i =1

t =1

i =1

t =1

i =1

t =1

i =1

∑ ∑ Cih,t + ∑ ∑ Cic,t + ∑ ∑ Sih,t + ∑ ∑ Sic,t + C W

(2)

where Nh and Nc represent the numbers of conventional thermal power plants and
CHPs; Cic, t and Sic, t represent fuel costs and start-up costs for CHPs determined
according to equations (15) and (16); the fuel costs for power plants Cih, t depend
linearly on fuel consumption; Sih, t is the start-up costs for conventional power
plants34; CW defines the penalty term for wind curtailment:
T

CW = θ

Nw

∑ ∑ (P iw,̄ t −piw,t ) Δt
t =1

(3)

i =1

̄

w

where θ denotes the penalty factor for wind curtailment, P i, t is the wind power that
w
would be available from wind farm i at time t, pi, t is the actual power output and
Nw indicates the number of wind farms. Δ​t is set at 1 h hereafter. The Renewable
Energy Law in China mandated that the dispatch of wind power should be
prioritized when there is no technical constraint in energy systems, and the penalty
term CW ensured such priority during the optimization process35. The penalty factor
θ is set as 1,000 kg kWh−1 in the following simulation, sufficiently large to ensure
that the actual power output is as close to the available wind potential as possible.
The model includes constraints relating to charging of electric vehicles,
operations of CHP units and the system power balances and reserves. First,
constraints for the aggregated charging power of the EV fleet are incorporated. The
aggregated power consumption for the fast-charging vehicles at any given time t is
not controlled by the control centre of the grid companies. This portion of power
ef
consumption is taken as given, indicated hereafter as pi, t . When employing slowe
charging strategy, the aggregated charging power of EV fleet i, pi, t, is constrained by
the relevant power and energy boundaries as follows:
pi−, t

≤ pie, t

≤ pi+, t ,

∀ i, ∀ t

(4)

ei−, t ≤

t

∑ pie,τ Δt ≤ ei+,t,

∀ i, ∀ t

(5)

τ=1

where the power and energy boundaries ( pi−, t, pi+, t , ei−, t and ei+, t) are introduced in the
vehicle simulation section above.
EV fleets employing slow charging can provide upward spinning reserve
by reducing their aggregated charging power, pie, t to p e , while p e needs to be
i, t
i, t
constrained by the power and energy boundaries at time t:
pi−, t ≤ p e ≤ pie, t ,
i, t

∀ i, ∀ t

(6)

t −1

∑ pie,τ Δt ≥ ei−,t,

p e Δt +
i, t

∀ i, ∀ t

(7)

τ=1

They can also provide downward reserve by increasing their aggregated
charging power, pie, t to pie, t, while pie, t needs also to be constrained by the power and
energy boundaries at time t:

̄

̄

̄

pie, t ≤ pie, t ≤ pi+, t ,

̄

pie, t Δt +

∀ i, ∀ t

(8)

t −1

∑ pie,τ Δt ≤ ei+,t,

∀ i, ∀ t

(9)

τ=1

Second, constraints for the operation of CHP units and heating supplies are
also incorporated. Power output from a CHP unit is restricted by the requirement
for its heat production. The restrictions reflect internal technical limitations, such
as the steam pressure limits for individual turbines and fuel limits for boilers. All
of the possible combinations of heat and power output for CHP units satisfying
operational constraints constitute jointly a feasible operational area. For simplicity,
we assume that feasible operational areas for all CHPs are convex. The limitation
on power and heat production for a CHP unit can be represented by a convex
combination of the coordinates of the corner points35:

 c
 pi, t =


 q c =
 i, t


M

∑ αik,t xik
k=1
M


k=1

αik, t

(10)

yik

(xik, yik , cik) indicate the power output, heat production and fuel cost for the kth
M
corner point, and αik, t is the non-negative value constrained to satisfy ∑k=1 αik, t = Iic, t.
The binary variable, Iic, t, defines the on/off status of the CHP unit: Iic, t = 0 indicates
that the unit is shut down at time t. M represents the number of corner points.
Ramping constraints for a CHP unit, similar to conventional units presented in
ref. 34, are represented by:

̄

 pic, t −pic, t −1 ≤ Riu × Iic, t −1 + Siu (Iic, t−Iic, t −1) + pic (1−Iic, t)

 p c −p c ≥ − (Rid × Iic, t + Sid (Iic, t −1−Iic, t))
 i, t i, t −1

(11)

where Riu and Rid are the ramp-up and ramp-down limits for the ith CHP unit.
Siu and Sid are the ramping limits for the unit to start up or shut down, pic is the
maximum power output from the ith CHP unit, a constant number defined by
pic = max (xik), k = 1, …, M .
Power outputs are constrained additionally by:

̄

̄

̄

pic, t ≤ pic × Iic, t +1 + Sid (Iic, t−Iic, t +1)

(12)

Minimum on and off time constraints, similar to those for conventional
thermal units34, are formulated as follows:
Gi


(1−Iic, t) = 0


t =1
 t +UTi−1

Iic, v ≥ UTi (Iic, t−Iic, t −1) t = Gi + 1, …, T −UTi + 1


v
t
=

T


(Iic, v− (Iic, t−Iic, t −1)) ≥ 0 t = T −UTi + 2, …, T

v
t
=








Nature Energy | www.nature.com/natureenergy

© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.

(13)

Articles

NATurE EnErgy





(14)



where UTi and DTi represent the minimum on and off times of the ith CHP unit. Gi
is associated with the initial status of the ith CHP unit. If unit i is operating in the
initial time interval, Gi represents the minimum time intervals required before it
is capable of shutting down; otherwise, Gi indicates the minimum time intervals
required for restart.
Both fuel costs Cic, t and start-up costs Sic, t can be represented by the convex
combination formulation35:
Cic, t =

M

∑ αik,tcik

(15)

k=1

Sic, t ≥ λi (Iic, t−Iic, t −1)

(16)

c

where Si, t is a non-negative variable and λi is the start-up cost for the ith CHP unit.
According to current regulations in China, CHP units are required to be online
during the entire heating period. As such, flexibility constraint (11), assuming Rid
equals Riu, could be simplified as:
∣pic, t −pic, t −1 ∣ ≤ Rid

(17)

The heat demand is scheduled according to the requirements for the heating
system, balanced separately within each heating district35:
Nc

∑ ai,jqic,t = Qj,t

(18)

i =1

where Qj, t is the heating demand in heating district j at time t. ai, j indicates the
connectivity of the ith heating source to the jth heating district. When ai, j = 1, the ith
heating element is connected to the jth heating district. Otherwise, ai, j = 0.
Third, the power balance and reserve constraints must be satisfied also for
reliable system operations. The power demand must be equal to the sum of the
power outputs from all power generating units, CHP plants, wind farms and net
power imports, subtracting the power consumed by EVs (both slow-charging and
fast-charging vehicles) at every time interval:
Nh


i =1

pih, t +

Nc


i =1

pic, t +

Nw


i =1

piw, t +

NT


i =1

piT, t = Pt +

Ne


i =1

pie, t +

Nef

∑ pie,tf ,

∀t

(19)

i =1

where NT indicates the number of transmission corridors, Pt is the system load at
ef
e
time t, pi, t and pi, t are, respectively, the slow- and fast-charging load for the EV fleet
i at time t, and Ne and Nef are the number of EV fleets incorporating slow and fast
charging, respectively. piT, t is the power imported from the ith transmission corridor
at time t. The hourly rates for the transmitted power are fixed and determined
jointly by the annual contracting among different regional dispatch centres and
by the transmission capacity limitations, as detailed in Supplementary Table 8 and
Supplementary Note 5.
The uncertainties relating to potential errors in forecasting energy demand and
wind power generation are accounted for in the system reserve requirements:
Nh

Nc

NT

Ne

Nef

i =1

i =1

i =1

i =1

i =1

∑ Iih,t × pih̄ + ∑ pic,̄t + ∑ piT,t ≥ Pt + RStu−RWt + ∑ pie,t + ∑ pie,tf ,
Nh

Nc

NT

Ne

Nef

i =1

i =1

i =1

i =1

i =1

∑ Iih,t × pih + ∑ pic,t + ∑ piT,t ≤ Pt−RStd + ∑ pie,̄t + ∑ pie,tf ,

∀t

∀t

(20)

(21)

h

where Ii, t is the binary variable indicating the on/off status at time t of thermal
plant i, pih and p h are the nameplate capacity and minimal power output level for
i
the ith thermal unit, RS tu and RS td are the upward and downward reserve margin at
c
time t, respectively, RWt is the reserve contribution from wind power at time t, pi, t
and p c indicate the maximum and minimum power outputs for the ith CHP unit

̄

i, t

e
at time t. pi, t is the minimum charging load for EV fleet i at time t, as indicated
in equations (6) and (7); pie, t is the maximum possible charging load for EV fleet
i at time t, as indicated in equations (8) and (9). Note that EVs employing the
slow-charging strategy could provide both upward and downward reserves in this
formulation, in contrast to the case for fast-charging vehicles. In this formulation,
the slow-charging electric vehicles, CHP units, conventional thermal generators
and wind power jointly provide reserves to the system to allow for uncertainties
involved with power demand and renewable production. The curtailed wind power
can provide upward reserve, whereas the wind power is able to provide downward
reserve by curtailing its power generation.
The intra-regional constraints could be represented by inequalities employing
d.c. generation shift factors36. As the intra-regional transmission constraints are
minimal in the JJT region37,38 (detailed in Supplementary Note 5) such limitations
are not included in the following analysis.
Other constraints related to conventional thermal units include the maximum
and minimum generation limits, the ramping constraints and minimum on/off
time constraints. The constraints adopted the formulation introduced from ref. 34.
In sum, the optimization model presented above minimizes the overall costs
as well as the wind power curtailments indicated in equation (2), subject to
the constraints for EVs (equations (4)–(9)), for combined heat and power units
(equations (10)–(18)), for system energy balance and reserves (equations (19)–(21))
and for flexibility constraints of conventional thermal generators. The annual
simulation for the JJT area requires a total of approximately 10 million variables.

̄

Gi




Iic, t = 0



t
=
1


 t +DTi−1


(1−Iic, v ) ≥ DTi (Iic, t −1−Iic, t) t = Gi + 1, …, T −DTi + 1



v
t
=


T



(1−Iic, v− (Iic, t −1−Iic, t)) ≥ 0 t = T −DTi + 2, …, T



v
t
=



̄

Beijing and its surrounding JJT energy systems. Using the proposed optimization
model, we simulate, on an hourly basis, the JJT energy system covering both
Beijing and its neighbouring region. The total energy demand in the JJT area is
expected to increase from 339.6 TWh in 2014 to 455 TWh in 2020 (approximately
5% increase per year), of which 117.8 TWh is the anticipated power demand for
Beijing in 2020. A detailed description of the geographical coverage and energy
system configuration of JJT energy systems is presented in Supplementary Note 5
and Supplementary Fig. 5.
Hourly power demand values for both the JJT regional grid and Beijing city
are provided by the SGCC for the years 2012 and 2009, scaled proportionally
to the projected level for 2020. The unit information for the JJT regional power
systems is derived from SGCC (for units before 2014) and from Global Coal Plant
Tracker (for units constructed or planned after 2014). Hourly variations of heating
demand for eight major heating districts within the JJT area are considered.
Operational characteristics of combined heat and power units are provided by the
grid company. CHPs are must-run units during the heating season—starting in
November and ending in March. Inter-regional transmission corridors connecting
JJT and other regions, as well as the transmission capacities, are summarized in
Supplementary Table 8. The simulation set-up is detailed also in Supplementary
Note 5. The hourly power balances for two 14-day periods for the JJT system are
presented in Supplementary Fig. 6.
Hourly wind power data were derived from wind fields compiled using the
National Aeronautics and Space Administration-assimilated meteorological
database (GEOS-5)39. Wind-generated electricity employed in this analysis
is assumed to originate from wind farms deployed in suitable areas of Hebei
province, within the JJT area. The hourly variation of wind capacity factor is
assumed to be consistent with the data indicated by GEOS-5 for 200939.
Emissions from power generation and transportation sectors. The annual total
emissions from on-road conventional vehicles were determined by the product
of population size of vehicle fleet, the average annual VKT of individual vehicles
in the fleet and the average fleet emission factors for individual pollutants. The
detailed data on population size, VKT and average fleet emission factors are
summarized by fleet type in Supplementary Tables 1 and 9. Data for emission
factors by emission type and plant type are presented in Supplementary Table 10.
The total emissions from the power sector in the regional grid are aggregated
from hourly emissions of all thermal generators. We adopt emission factors
derived from the most recently available emission standards for thermal generation
plants28, standards that have become more stringent since 2012, with which all
power generation stations are required to comply by November 2014. The average
emission intensity for generating every kilowatt hour of electricity is calculated
on the basis of the total emissions divided by the annual total power demand
(excluding imports). For scenarios considering the deployment of EVs, the changes
in emissions from the power sector in Beijing account for both direct contributions
and indirect contributions. The direct contribution is the change in emissions
related to the power consumption in Beijing, assessed on the basis of the annual
average emission intensity and total Beijing power consumption. The indirect
contribution accounts for the changes of emissions in the regional grid resulting
from the charging of EVs in the Beijing area (largely due to changes in the wind
curtailment rate), which vary in response to differences in charging strategies and
vehicle types. The detailed description of emission factors, and calculations of
emissions for different scenarios are indicated in Supplementary Note 7.
Data availability. The hourly wind power capacity factors and variation of power
demand for typical days of each month, as well as the operational characteristics
Nature Energy | www.nature.com/natureenergy

© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.

Articles

NATurE EnErgy
for typical thermal and combined heat and power units, are available in
Supplementary Data 1. The other data that support the plots within this paper and
other findings of this study are available from the corresponding authors upon
reasonable request.

Received: 10 September 2016; Accepted: 15 March 2018;
Published: xx xx xxxx

References

1. Air Quality Review of Key Areas and 74 Cities in 2014 (in Chinese;
Ministry of Environmental Protection of People’s Republic of China, 2015);
http://www.mep.gov.cn/gkml/hbb/qt/201502/t20150202_295333.htm
2. Nielsen, C. P. & Ho, M. S. Clearer Skies Over China: Reconciling Air Quality,
Climate, and Economics Goals (MIT Press, Cambridge, MA, 2013).
3. Zhang, Q., He, K. & Huo, H. Cleaning China’s air. Nature 484, 161–162 (2012).
4. Wu, Y. et al. On-road vehicle emission control in Beijing: past, present and
future. Environ. Sci. Technol. 45, 147–153 (2011).
5. China Automotive Technology and Research Center (CATARC), Beijing
Guoneng Yingchuang Energy Information Technology Co., Ltd. & Editorial
Office of Year-book of Energy-saving and New-energy Vehicles 2011 Yearbook
of Energy-saving and New Energy Vehicles (China Economic Publishing
House, Beijing, 2011).
6. About the Support for Promotion of New-energy Vehicle in Shenyang,
Changchun etc (in Chinese; Ministry of Finance, 2014); http://jjs.mof.gov.cn/
zhengwuxinxi/tongzhigonggao/201402/t20140208_1041231.html
7. The Circular on Developing the Demonstration of Subsidies for Private Purchase
of New Energy Vehicles (in Chinese; Ministry of Finance, Ministry of Science
and Technology, Ministry of Industry and Information Technology &
National Development and Reform Commission, 2010).
8. Li, F. China has the most public EV charging stations worldwide
(China Daily, 2018); http://usa.chinadaily.com.cn/a/201801/11/
WS5a5759d9a3102c394518e9e1.html
9. Annual Report on the Development of Electric Vehicle Charging Infrastructure
in China (2016–2017) (in Chinese; National Energy Administration & China
EV Charging Infrastructure Promotion Alliance, 2017); http://www.nea.gov.
cn/136376732_14978397401671n.pdf
10. Wang, K. Beijing limits new car plates, boosts new energy vehicles
(China Daily, 2017); http://www.chinadaily.com.cn/a/201712/15/
WS5a33819aa3108bc8c6734ecb.html
11. Over 40% of Residential Areas are Equipped with Charging Facilities
(in Chinese; Commission of Housing and Urban-Rural Development of
Beijing, 2016) http://www.bjjs.gov.cn/bjjs/xxgk/xwfb/317878/index.shtml
12. Huo, H., Zhang, Q., Liu, F. & He, K. Climate and environmental effects of
electric vehicles versus compressed natural gas vehicles in China: a life-cycle
analysis at provincial level. Environ. Sci. Technol. 47, 1711–1718 (2013).
13. Wu, Y. et al. Energy consumption and CO2 emission impacts of vehicle
electrification in three developed regions of China. Energy Policy 48,
537–550 (2012).
14. Kim, J. & Rahimi, M. Future energy loads for a large scale adoption of
electric vehicles in the city of Los Angeles: impacts on greenhouse gas (GHG)
emissions. Energy Policy 73, 620–630 (2014).
15. Zhang, S. et al. Historic and future trends of vehicle emissions in Beijing,
1998–2020: a policy assessment for the most stringent vehicle emission
control program in China. Atmos. Environ. 89, 216–229 (2014).
16. Kang, C. et al. Balance of power: towards a more environmentally friendly,
efficient, and effective integration of energy systems in China. IEEE Power
Energy Mag. 11, 56–64 (2013).
17. Wang, J. et al. Impact of plug-in hybrid electric vehicles on power systems
with demand response and wind power. Energy Policy 39, 4016–4021 (2011).
18. Sioshansi, R. & Denholm, P. Emissions impacts and benefits of plug-in hybrid
electric vehicles and vehicle-to-grid services. Environ. Sci. Technol. 43,
1199–1204 (2009).
19. Liu, W., Hu, W., Lund, H. & Chen, Z. Electric vehicles and large-scale
integration of wind power—the case of Inner Mongolia in China.
Appl. Energy 104, 445–456 (2013).
20. Global Wind Statistics 2017 (Global Wind Energy Council, 2018); http://gwec.
net/wp-content/uploads/vip/GWEC_PRstats2017_EN-003_FINAL.pdf
21. China Wind Energy Development Roadmap 2050 (International Energy
Agency, 2011); https://www.iea.org/publications/freepublications/publication/
china_wind.pdf
22. Grid Integration of China’s Wind Power Generation in 2016 (in Chinese;
National Energy Administration of China, 2017); http://www.nea.gov.
cn/2017-01/26/c_136014615.htm

23. Beijing Municipal Bureau of Statistics Beijing Statistical Yearbook
(China Statistics Press, Beijing, China, 2014).
24. National Bureau of Statistics of China China City Statistical Yearbook
1998–2013 (China Statistics Press, Beijing, China, 1998–2013).
25. Luo, Z., Hu, Z., Song, Y., Xu, Z. & Lu, H. Optimal coordination of plug-in
electric vehicles in power grids with cost-benefit analysis—part II: a case
study in China. IEEE Trans. Power Syst. 28, 3556–3565 (2013).
26. Sun, R., Yu, H. & Du, Y. Time space distribution characteristics of
taxi shift in Beijing. J. Transp. Syst. Eng. Inf. Technol. 14, 221–228 (2014)
(in Chinese).
27. Beijing Transport Annual Report (in Chinese; Beijing Transportation Research
Center, 2013); http://www.bjtrc.org.cn/
28. Emission Standard of Air Pollutants for Thermal Power Plants GB 13223-2011
(Ministry of Environmental Protection of China, 2011).
29. Five Companies Got the Bidding for 914 Electric Buses in Qingdao with a Total
Amount of about 900 Million RMB (in Chinese; Electric Vehicle Resources
Network, 2017); http://www.sohu.com/a/196982793_233844
30. The Clean Air Action Plan 2013–2017 (in Chinese; Beijing Municipal
Government, 2013); http://fgcx.bjcourt.gov.cn:4601/law?fn=​lar830s997.txt
31. Subramanian, A., Garcia, M. J., Callaway, D. S., Poolla, K. & Varaiya, P.
Real-time scheduling of distributed resources. IEEE Trans. Smart Grid 4,
2122–2130 (2013).
32. Xu, Z., Callaway, D. S., Hu, Z. & Song, Y. Hierarchical coordination
of heterogeneous flexible loads. IEEE Trans. Power Syst. 31,
4206–4216 (2016).
33. Zhang, H., Hu, Z., Xu, Z. & Song, Y. Evaluation of achievable vehicle-to-grid
capacity using aggregate PEV model. IEEE Trans. Power Syst. 32,
784–794 (2017).
34. Carrión, M. & Arroyo, J. M. A computationally efficient mixed-integer linear
formulation for the thermal unit commitment problem. IEEE Trans. Power
Syst. 21, 1371–1378 (2006).
35. Chen, X. et al. Increasing the flexibility of combined heat and power for wind
power integration in China: modeling and implications. IEEE Trans. Power
Syst. 30, 1848–1857 (2015).
36. Wang, S. J., Shahidehpour, S. M., Kirschen, D. S., Mokhtari, S. & Irisarri, G.
D. Short-term generation scheduling with transmission and environmental
constraints using an augmented Lagrangian relaxation. IEEE Trans.
Power Syst. 10, 1294–1301 (1995).
37. Chen, X., Lu, X., McElroy, M. B., Nielsen, C. P. & Kang, C. Synergies of wind
power and electrified space heating: case study for Beijing. Environ. Sci.
Technol. 48, 2016–2024 (2014).
38. Geng, Z., Chen, Q., Xia, Q., Kirschen, D. S. & Kang, C. Environmental
generation scheduling considering air pollution control technologies and
weather effects. IEEE Trans. Power Syst. 32, 127–136 (2017).
39. McElroy, M. B., Lu, X., Nielsen, C. P. & Wang, Y. Potential for wind generated
electricity in China. Science 325, 1378–1380 (2009).
40. Connection Set for Conductive Charging of Electric Vehicles GB/T 20234.22015 (General Administration of Quality Supervision, Inspection and
Quarantine, 2015).

Acknowledgements

The research was funded in part by the Harvard Global Institute. Additional support
was provided by the Hui Fund of the Ash Center of the Harvard Kennedy School of
Government. This research is also supported by the State Key Laboratory on Smart
Grid Protection and Operation Control of NARI Group, through the open topic
project (20171613).

Author contributions

All authors contributed to all aspects of this work.

Competing interests

The authors declare no competing interests.

Additional information

Supplementary information is available for this paper at https://doi.org/10.1038/
s41560-018-0133-0.
Reprints and permissions information is available at www.nature.com/reprints.
Correspondence and requests for materials should be addressed to X.C. or M.B.M.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.

Nature Energy | www.nature.com/natureenergy

© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.


Documents similaires


Fichier PDF 101038s41560 018 0133 0
Fichier PDF double power split device for hybrid vehicle powertrain
Fichier PDF proximite de trafic et hta
Fichier PDF 20130820 supercapacitors as an alternative to batteries
Fichier PDF alliance sales results 2016 final
Fichier PDF difference between force and constant power control ever2012


Sur le même sujet..