Applied Energy 115 (2014) 582–590

Contents lists available at ScienceDirect

Applied Energy

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

Electrical consumption of two-, three- and four-wheel light-duty electricvehicles in India

0306-2619/$ - see front matter � 2013 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.apenergy.2013.10.043

⇑ Corresponding author.E-mail address: samveg@berkeley.edu (S. Saxena).

Samveg Saxena ⇑, Anand Gopal, Amol PhadkeEnvironmental Energy Technologies Division, Lawrence Berkeley National Laboratory, United States

h i g h l i g h t s

�Model electrical consumption of 2-, 3- and 4-wheelers in India.� Average city energy use is 33 Wh/km for scooters, 61 Wh/km for 3-wheelers.� Average city energy use is 84 Wh/km and 123 Wh/km for low and high power 4-wheelers.� The increased energy use from air conditioning is quantified.� Energy use from variations in vehicle mass and motor efficiency are quantified.

a r t i c l e i n f o

Article history:Received 9 August 2013Received in revised form 18 October 2013Accepted 27 October 2013Available online 20 November 2013

Keywords:Electric vehiclesPowertrainTransportationVehicle to gridIndia

a b s t r a c t

The Government of India has recently announced the National Electric Mobility Mission Plan, which setsambitious targets for electric vehicle deployment in India. One important barrier to substantial marketpenetration of EVs in India is the impact that large numbers of EVs will have on an already strained elec-tricity grid. Properly predicting the impact of EVs on the Indian grid will allow better planning of newgeneration and distribution infrastructure as the EV mission is rolled out. Properly predicting the gridimpacts from EVs requires information about the electrical energy consumption of different types ofEVs in Indian driving conditions. This study uses detailed vehicle powertrain models to estimate per kilo-meter electrical consumption for electric scooters, 3-wheelers and different types of 4-wheelers in India.

The powertrain modeling methodology is validated against experimental measurements of electricalconsumption for a Nissan Leaf. The model is then used to predict electrical consumption for several typesof vehicles in different driving conditions. The results show that in city driving conditions, the averageelectrical consumption is: 33 Wh/km for the scooter, 61 Wh/km for the 3-wheeler, 84 Wh/km for thelow power 4-wheeler, and 123 Wh/km for the high power 4-wheeler. For highway driving conditions,the average electrical consumption is: 133 Wh/km for the low power 4-wheeler, and 165 Wh/km forthe high power 4-wheeler. The impact of variations in several parameters are modeled, including theimpact of different driving conditions, different levels of loading by air conditions and other ancillarycomponents, different total vehicle masses, and different levels of motor operating efficiency.

� 2013 Elsevier Ltd. All rights reserved.

1. Introduction

India is one of the world’s most rapidly growing economies, andis the third largest vehicle market in the world. Annual demand ofvehicles is rapidly growing in India, with 2020 annual projectedsales of 10 million passenger vehicles, 2.7 million commercial vehi-cles, and 34 million two-wheelers. India currently imports about85% of its oil and is projected to reach 92% by 2020, creating a signif-icant challenge for the balance of payments and the energy securityof the country [1]. Based on the pressing challenges with growth invehicle sales and energy security facing the country, the Central

Government of India has released the National Electric MobilityMission Plan (NEMMP) [1] which establishes a pathway for thewidespread deployment of hybrids, plug-in hybrids, and electricvehicles in India. The NEMMP calls for the deployment of 5–7 mil-lion EVs (hybrids and full EVs) on the road by 2020, and the Govern-ment of India has committed Rs 22,500 cr (approximately $4.1Billion USD) to this initiative. Similar goals for widespread electricvehicle adoption have been set by other governments around theworld [2–4].

The rapid deployment of plug-in hybrid and fully electric vehi-cles (collectively called plug-in vehicles, PEVs, in this paper) calledfor in the NEMMP places significant demands on an alreadystrained electricity grid in India [5]. However, since range anxietyis a significant consumer perception barrier to EV deployment

S. Saxena et al. / Applied Energy 115 (2014) 582–590 583

[6], the absence of reliable charging points (which require a stableelectricity grid) in India will make it difficult to achieve the tar-geted levels of EV market penetration. Additionally, if the electric-ity grid is unable to accommodate PEV charging, it is possible thatdiesel generators will be used to provide the unmet electricity de-mand. Although this local distributed generation solution mayaccommodate PEV charging demand in the interim, it is not aneffective way to decouple the Indian transportation sector fromoil and can still lead to urban air quality problems.

The Government of India has recently joined the Electric VehicleInitiative (EVI) of the Clean Energy Ministerial, which seeks to facil-itate the deployment of 20 million EVs by 2020. Under this initia-tive, Lawrence Berkeley National Laboratory is supporting theNEMMP in assessing the real-world costs, benefits and environ-mental impacts of EV uptake in India; this publication is the firstin a series of studies in this effort.

To properly plan for the rapid deployment of PEVs in India,there is a need for finely resolved temporal and spatial predictionsof PEV charging load on the electricity grid. The ability to properlyforecast PEV charging load is essential for utility grid operators toensure that adequate generation capacity is available at the correcttimes, and ensure that distribution infrastructure can accommo-date substantial PEV charging. Several studies [7–13] have devel-oped methods to estimate PEV charging load for the USelectricity grid. The most rigorous of these studies [14–16] followa three-step methodology (listed below) to predict temporally re-solved PEV charging load profiles. A modeling tool, called V2G-Sim, has been developed at Lawrence Berkeley National Laboratoryto streamline the simulation of vehicle-grid interactions and thistool is available for use in potential research collaborations [17].

1. Estimating the time when vehicles are plugged in: Survey data isused to provide information on how drivers use their vehicles,including number of vehicle trips per day, time of departureof each trip, trip travel length, arrival time of each trip, typeof vehicle, etc. In the United States, a common data source forthis information is the National Household Travel Survey(NHTS) [18], however other data sources have also been used.

2. Estimating the amount of energy required to charge the vehiclebattery: Typically, a simplified vehicle model is used to esti-mate: (a) how much of the vehicle battery is depleted duringeach trip, and (b) how much energy is required during charging.A standard approach in prior studies [14–16] is to assume aconstant value for electrical consumption (kWh/km or kWh/mile) depending on the type of vehicle (i.e. car, van, SUV, truck,etc.) that is being modeled. More accurate estimates of batterydepletion while driving can be obtained with detailed vehiclephysics models (such as models used in other papers), howeverthis approach may be prohibitively computationally expensivewhen attempting to model hundreds, thousands, or millionsof PEVs on an electricity grid.

3. Estimating charging rates while a vehicle is plugged in: Using esti-mates of when different vehicles will plug in for charging fromstep 1, how much charging is required from step 2, and infor-mation about the charging rate (i.e. level 1, level 2, or DC fastcharger), number of PEVs and any smart charging strategies,aggregate charging load profiles are estimated for a large num-ber of vehicles within a given region (i.e. utility service territory,state, or country).

Successful implementation of the NEMMP within the prescribedtimeline requires immediate planning and infrastructure deploy-ment to ensure that the Indian electricity grid can cope with theadded charging load from large numbers of PEVs. Thus, the 3-stepanalysis methodology described above must be applied to the In-dian context, however much of the required data for India is not

available in published studies. For instance, for step 1 better datais required to characterize vehicle usage patterns in India. For step2, average electrical consumption (Wh/km) numbers are requiredfor vehicles specifically in the Indian context (i.e. for vehicle sizesrepresentative of typical Indian vehicles driving in Indian trafficconditions). The use of prior published electrical consumption val-ues does not adequately account for typical Indian vehicles or forthe influence of driving and usage factors (i.e. from dense traffic,or the use of power-consuming devices like an air conditioner).Electrical consumption data for scooters, 3-wheelers, and small4-wheelers has previously been unavailable in the literature, par-ticularly for the Indian context where driving conditions will bedifferent than in developed countries and air conditioning load willbe a significant factor. For the Indian context in particular, it maybe inappropriate to use prior published Wh/km values becausetwo-wheelers and ultra-compact four-wheelers that are typicalin India are significantly smaller and lighter than the US market,and typical driving conditions are different in India with more fre-quent stopping, lower average speeds and potentially more suddenacceleration and deceleration [19].

In support of the NEMMP and as a step towards predicting thecharging load of PEVs on the Indian electricity grid, the results pre-sented in this study will enable better estimates of PEV chargingload on the Indian electricity grid. Specifically, the results of thisstudy provide Wh/km values that are representative of typicalvehicles in India, driving in conditions representative of Indianroads. The results of this study can then be used in Step 2 of the3-step methodology above to estimate temporally resolved PEVcharging loads on the Indian electricity grid.

2. Specific objectives

In support of the India National Electric Mobility Mission Plan,this study provides critical data to enable detailed predictions ofPEV temporal charging load profiles for the Indian electricity grid.Detailed vehicle powertrain modeling is used for:

1. Providing estimates of average electrical consumption(Wh/km) for vehicles that are representative of typicalIndian two-, three- and four-wheel vehicles over drivecycles that are representative of Indian driving conditions.

2. Providing correlations for the Wh/km results that accountfor variations in vehicle use, such as variability in vehiclemass, the use of air conditioners, and variations in power-train component efficiency.

3. Vehicle models

3.1. Vehicle powertrain models

A detailed vehicle powertrain model is used to estimate electri-cal consumption for four types of vehicles, with specifications foreach vehicle listed in Table 1. The powertrain models are createdin the industry standard Autonomie powertrain modeling platform.

3.2. Drive cycles

Given that energy consumption of a vehicle depends signifi-cantly on driving patterns [19–25], several different drive cyclesare chosen. Five drive cycles are chosen based on Indian drivingconditions, including a New Delhi cycle [26], Pune cycle [27], themodified Indian drive cycle (MIDC) [28], and an Indian urban andIndian highway cycle. Additionally three US certification cyclesare also included for comparison purposes, the EPA UDDS, HWFETand US06 cycles [29]. Figs. 1–4 compare the characteristics of each

Table 1Vehicle specifications used in powertrain models.

Scooter 3-Wheeler Low power 4-wheeler High power 4-wheeler

Base vehicle mass (kg) 150 500 898 1493Motor max power output (kW) 1.5 5.46 19 80Final drive ratio 6.3805 6.3805 6.8737 7.9377Usable battery capacity (kWh) 2.16 4.25 6.54 16.7Tire size 1000 � 300 1000 � 4.500 P155/70R13 P205/55 R16Drag coefficient 0.60 0.35 0.335 0.28Frontal area (m2) 1.25 2.40 2.0 2.50Baseline electrical accessory & AC load (W) 50 100 200 200Estimated range in City (km) 64–71 60–80 70–95 123–138Estimated range on Highway (km) N/A N/A 34–76 73–136Top speed (km/h) 50 73 117 120

Fig. 1. Velocity characteristics of US and Indian drive cycles.

Fig. 2. Stopping/idling characteristics of US and Indian drive cycles.

Fig. 3. Acceleration characteristics of US and Indian drive cycles.

Fig. 4. Deceleration characteristics of US and Indian drive cycles.

584 S. Saxena et al. / Applied Energy 115 (2014) 582–590

drive cycle in terms of velocity, stopping/idling, acceleration anddeceleration characteristics. The values in these figures are nor-malized by the average values across all driving cycles to alloweasier comparisons.

Fig. 1 compares the velocity characteristics of the US and Indiandrive cycles. The plot shows that driving conditions on the Indiancycles involve lower maximum speed, lower mean speed, and low-er mean driving speed1. Even the speeds on the Indian highway cy-cle are considerably lower than the speeds on the US highway cycles.

1 Mean speed is defined as the average of all velocities over the drive cycle. Meandriving speed is defined as the average of all non-zero velocities.

Fig. 2 compares the stopping and idling characteristics of the USand Indian drive cycles. As expected, the results show that stop fre-quency and fraction of total time stopped are much higher on thecity cycles as compared with the highway cycles. Of particularimportance, Fig. 2 shows that stop frequency is much higher inthe Indian city cycles than the US city cycle. The total fraction oftime stopped is highest in the Pune cycle, followed by the US citycycle.

Fig. 3 compares the acceleration characteristics of the US andIndian drive cycles. The highest acceleration values are seen inthe high speed US highway cycle (US06). Comparing the US and In-dian city cycles, it is seen that greater maximum acceleration and

S. Saxena et al. / Applied Energy 115 (2014) 582–590 585

maximum acceleration from stop values are encountered in the In-dian city cycles, however the average acceleration is higher in theUS city cycle.

Fig. 4 compares the deceleration characteristics of the US andIndian drive cycles. The results show that maximum decelerationand maximum deceleration to stop are higher in Indian city condi-tions than US city conditions, however higher levels of averagedeceleration are seen in the US city cycle.

Summarizing the results in this section, Figs. 1–4 compared thedrive cycle characteristics for the US and Indian drive cycles. It wasgenerally observed that the Indian drive cycles involve lower driv-ing speeds, greater frequency of stopping, and higher levels ofmaximum acceleration and deceleration. These results suggest thatdriving in India may involve more severe stop-and-go conditions,and previous studies [19,25] have found that these types of drivingconditions create unique opportunities for achieving greater levelsof fuel savings with vehicle electrification.

Fig. 5. Model validation: comparison of modeled and measured electrical con-sumption for a Nissan Leaf.

3.3. Parametric variations

In addition to the vehicle speed profiles while driving, otherparameters will significantly influence vehicle energy consump-tion as well. The vehicle modeling that is discussed in this papercaptures the impact on vehicle energy usage from several parame-ters that will change with different vehicle designs, usage patterns,and driving conditions. For warm climates like India, ancillarycomponents such as vehicle air conditioning load will have asignificant impact on energy consumption [30–31]. Loading thevehicle with more passengers or cargo will also impact energyconsumption. Additionally, variations in powertrain componentefficiency will also impact energy consumption. Table 2 lists therange of parameter variations that were explored using the vehiclepowertrain models for their impact on vehicle energyconsumption.

Table 3Electrical consumption range of each vehicle.

Electrical consumption (Wh/km)

Avg city Avg hwy Range

Scooter 33 38 31–403-Wheeler 61 85 53–97Low power EV 84 133 70–192High power EV 123 164 101–224

3.4. Model validation

To ensure that the electrical consumption estimates presentedin the results section of this paper are reasonable, the same mod-eling methodology is followed to create a powertrain model for aNissan Leaf electric vehicle, for which there are well documentedvalues of electrical consumption under various driving conditions.

A vehicle powertrain model was constructed with specificationsresembling a Nissan Leaf, and electrical consumption model esti-mates were compared against published measurement data [32]for the EPA UDDS, Highway, and US06 drive cycles over a rangeof total vehicle mass. Fig. 5 shows a comparison of the modeledand measured electrical consumption values for a Nissan Leaf.

The modeled and experimentally measured electrical consump-tion values plotted in Fig. 5 show that the vehicle powertrain mod-el reasonably predicts both the trends and absolute values ofelectrical consumption for a range of different vehicle masses forall three drive cycles. The largest difference in absolute valuesbetween the model and the experimental measurements is11.50%, which occurs for the lowest vehicle mass on the highwaycycle. It is typically the case that increased vehicle mass leads toincreased energy consumption, however the experimental

Table 2Range of parameter variations explored for their impact on vehicle energy consumption.

Scooter 3-Wheele

Ancillary loading (i.e. A/C) (kW) 0.0–0.30 0.0–0.50Vehicle mass (kg) 150–300 500–800Motor efficiency (%) 55–90 55–90

measurements on the highway cycle do not display this expectedtrend. This may be due to experimental error because obviouslythe vehicle mass will have a significant influence on the electricityconsumption of a vehicle. This expected trend is indeed seen forthe UDDS and US06 experimental measurements, thus the datapoints at the lowest mass values for the highway cycle seem higherthan expected. As a result of the overall agreement of trends andabsolute values shown in Fig. 5, the modeling methodology is con-sidered accurate enough for the purposes of this study.

4. Results

4.1. Baseline electrical consumption estimates

Table 1 lists the vehicle specifications that were used in thepowertrain models for an electric scooter, electric 3-wheeler, lowpower EV 4-wheeler and high power EV 4-wheeler. These power-train models provide the electrical consumption per kilometer esti-mates over several different drive cycles in Fig. 6 and Table 3. Thereare several numerical values in Fig. 6 which are crossed out (partic-ularly for the electric scooter and 3-wheeler). These crossed outvalues denote that the vehicle was unable to perform on the drivecycle, either because the drive cycle requests speeds which arehigher than the maximum speed capability of the vehicle, orbecause acceleration profiles are demanded which exceed thecapabilities of the powertrain components. Thus, these crossed

r Low power 4-wheeler High power 4-wheeler

0.20–3.0 0.20–4.0898–1200 1493–180055–90 55–90

Fig. 6. Electrical energy consumption rate for different types of EVs on differentdrive cycles.

Fig. 7. Variation of vehicle electrical consumption with different ancillary compo-nent loading.

Fig. 8. Variation of vehicle electrical consumption with different total vehiclemasses.

586 S. Saxena et al. / Applied Energy 115 (2014) 582–590

out values should not be given much weight but instead simplyconsidered for reference.

The results in Fig. 6 show that electrical consumption per kilo-meter is highest for the 4-wheelers and lowest for the electricalscooter, which comes as no surprise given the differences in vehi-cle mass. For the vehicles which are capable of sustaining highwayspeeds (i.e. only the 4-wheelers), electrical consumption is signifi-cantly higher for high speed highway driving.

Table 4Coefficients for equation of fit for impact of ancillary component loading (kW) on vehicle

UDDS HWFET US06

2 Wheeler m 42.35b 32.83R2 1.00

3 Wheeler m 36.47b 65.70R2 1.00

4 Wheeler low power m 35.30 15.43 15.73b 87.86 117.48 189.49R2 1.00 1.00 1.00

4 Wheeler high power m 34.22 14.21 14.70b 128.27 142.42 220.64R2 1.00 1.00 1.00

4.2. Impact of parameter variations on vehicle electricity consumption

The electrical consumption was calculated for several vehicleson several US and Indian drive cycles in Section 4.1, with specifica-tions defined in Table 1. Vehicles on the road, however, will rarelyhave exactly the same specifications as those defined in Table 1,thus this section explores how different parameters will impactelectrical consumption of each vehicle.

4.2.1. Ancillary component and air conditioning loadsFor hot climates like India, energy use by air conditioners will

have a significant impact on the electricity consumption of a vehi-cle. Additionally, other ancillary components (like vehicle controlelectronics, radio, and lights) will consume energy. Fig. 7 presentsthe impact on vehicle electricity consumption from different levelsof loading by ancillary components. As two- and three-wheelerstypically do not have an enclosed cabin they will not have air con-ditioners, and thus their maximum loading from ancillary compo-nents will be lower. Thus, in Fig. 7 the modeled range of energyconsumption from ancillary components for the two- and three-wheel vehicles is much lower than for the four-wheelers.

Fig. 7 shows that for each vehicle on all the different drive cy-cles, vehicle electricity consumption (Wh/km) increases linearlywith increasing loading from ancillary components. It is particu-larly important to note that the slope of this linear increase is dif-ferent across the different drive cycles. The equation of fit for therelationship between ancillary component loading and vehicleelectricity consumption follows the form of Eq. (1), where x is

electricity consumption (Wh/km).

India urban India highway Delhi Pune MIDC

48.13 61.62 57.59 40.5928.85 30.40 28.84 33.971.00 1.00 1.00 1.00

45.85 22.75 59.28 55.40 34.4250.62 67.15 46.94 50.46 69.801.00 1.00 1.00 1.00 1.00

47.11 23.72 60.64 56.86 34.4967.81 80.78 57.03 69.33 89.801.00 1.00 1.00 1.00 1.00

46.14 22.88 59.57 55.73 33.38112.28 118.40 88.84 113.99 124.401.00 1.00 1.00 1.00 1.00

Table 5Coefficients for equation of fit for impact of vehicle mass (kg) on vehicle electricity consumption (Wh/km).

UDDS HWFET US06 India urban India highway Delhi Pune MIDC

2 Wheeler m 0.03 0.02 0.02 0.03 0.02b 29.94 28.12 30.88 27.64 32.73R2 1.00 0.95 0.95 0.98 0.96

3 Wheeler m 0.07 0.07 0.05 0.04 0.07 0.04b 34.43 22.27 45.54 32.00 23.64 51.03R2 1.00 1.00 1.00 1.00 1.00 1.00

4 Wheeler low power m 0.07 0.05 0.07 0.07 0.06 0.04 0.07 0.05b 30.87 77.49 127.74 18.62 31.96 32.06 22.60 48.43R2 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00

4 Wheeler high power m 0.06 0.04 0.07 0.06 0.05 0.04 0.06 0.05b 44.05 83.86 114.44 31.90 42.89 39.43 37.38 60.27R2 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00

Fig. 9. Variation of vehicle electrical consumption with different average motoroperating efficiency.

Fig. 10. Powertrain architecture for electric vehicle models.

S. Saxena et al. / Applied Energy 115 (2014) 582–590 587

the electrical loading in kW, m is the slope, and b is the y-intercept(in this case, the electrical consumption if there was no loadingfrom ancillary components). Values for m and b for each vehicleon each drive cycle are listed in Table 4.

y ¼ mxþ b ð1Þ

The fitting equation parameters in Table 4 shows that the slopeof each fitting equation for a given drive cycle is not sensitive tovehicle type. The slopes are generally higher for lower speed driv-ing conditions. These results suggest that ancillary component

Table 6Coefficients for equation of fit for impact of motor efficiency (%) on vehicle electricity con

UDDS HWFET US06

2 Wheeler m �54.11b 75.97R2 0.99

3 Wheeler m �125.9b 165.3R2 0.99

4 Wheeler low power m �198.8 �189.0 �319.0b 248.5 271.7 441.8R2 0.99 0.99 1.00

4 Wheeler high power m �304.2 �231.1 �438.0b 361.7 326.1 568.6R2 0.99 0.99 0.99

loading has a greater impact on vehicle electricity consumptionat lower speed driving conditions (i.e. city driving), and is not sen-sitive to vehicle type.

4.2.2. Variations in vehicle, passenger and cargo massIndividual vehicles are bound to be loaded with different mass

due to variations in the number of passengers or cargo being car-ried. Fig. 8 shows the impact of variations in total vehicle massfor each vehicle on each drive cycle. The 3- and 4-wheelers willhave greater carrying capacity and thus the range of vehicle massesmodeled is larger for these vehicles.

Fig. 8 shows that vehicle electricity consumption is also linearlydependent on vehicle mass, with increased electricity consumptionfor greater total vehicle mass. The equation of fit relating changesin vehicle electrical consumption with changes in vehicle mass fol-lows the form of Eq. (1) as well, with x being vehicle mass in kg.Table 5 lists the coefficients for the equations of fit for variationsin vehicle mass. The fitting coefficients in Table 5 suggest thatthe impact of vehicle mass on vehicle electricity consumption is

sumption (Wh/km).

India urban India highway Delhi Pune MIDC

�50.93 �49.53 �51.55 �54.3469.54 70.89 69.97 77.540.99 0.99 0.99 0.99

�106.2 �118.2 �74.5 �99.9 �118.0135.6 161.4 110.2 131.3 164.60.99 0.99 0.99 0.99 0.99

�286.7 �266.0 �167.1 �274.9 �251.1328.7 322.9 219.4 321.1 318.90.99 0.99 0.99 0.99 0.99

�268.7 �266.0 �167.1 �274.9 �251.1328.7 322.9 219.4 321.1 318.90.99 0.99 0.99 0.99 0.99

588 S. Saxena et al. / Applied Energy 115 (2014) 582–590

fairly consistent across the different drive cycles and across the dif-ferent vehicles (especially the 3-wheeler and both 4-wheelers).

4.2.3. Variations in average motor efficiencyThe values chosen for the motor efficiency maps used for the

baseline vehicle simulations (in Section 4.1) were established tofit the Nissan Leaf model validation results in Section 3.4. Electricvehicles released in the Indian market, however, may use differenttypes of motors with different efficiency operating profiles, thusthis section explores the impact of changes is motor operating effi-ciency. Fig. 9 shows the variation of vehicle electricity consump-tion with average motor operating efficiency for the differentvehicles driving on the different drive cycles.

As expected, the results in Fig. 9 show that vehicle electricityconsumption decreases as a more efficient motor is used. A partic-ularly interesting result, however, is that changes in motor effi-ciency have very little impact on electricity consumption for thesmaller vehicles, especially the two-wheeler. This result is of sig-nificant importance as it suggests that the use of less expensivemotors, which may be less efficient, can be used to lower the costof electric scooters while having minimal impact on vehicle elec-tricity consumption (and thus vehicle range). For the larger vehi-cles and for higher speed driving conditions (i.e. on highways),however, motor efficiency impacts electrical consumption signifi-cantly and thus better motors must be used. The results for the lar-ger vehicles in Fig. 9 show that the relationship between vehicleelectricity consumption and motor efficiency is not perfectly linear(i.e. a slight curvature can be seen on the plots), however the R2 fit-ting parameters in Table 6 show that a linear equation of the formof Eq. (1), with x being the average motor efficiency (%), produces agood fit.

5. Conclusions

Given the ambitious targets for electric vehicle deployment inIndia under the National Electric Mobility Mission Plan whichwas announced by the Government of India, there are significantconcerns with the impact that EV charging will have on an alreadystrained Indian electricity grid. This study is part of a larger efforttowards estimating the impact on the Indian electricity grid fromsubstantial deployment of EVs on the Indian grid to subsequentlyplan the deployment of new generation and distributioninfrastructure.

This study used detailed vehicle powertrain models to estimatethe per kilometer electrical consumption of several types of EVs,including a scooter, a 3-wheeler, a low power 4-wheeler, and ahigh power 4-wheeler. Electrical consumption data for scooters,3-wheelers, and small 4-wheelers has previously been unavailablein the literature, particularly for the Indian context where drivingconditions will be different than in developed countries and airconditioning load will be a significant factor. The powertrain modelmethodology was validated against experimental measurementsfor a Nissan Leaf. The main conclusions from this study are asfollows:

1. Average electrical consumption: Vehicle size has the greatestimpact on per km electrical consumption, followed by thedriving characteristics (i.e. city vs. highway driving). In citydriving conditions average electrical consumption results were:33 Wh/km for the scooter, 61 Wh/km for the 3-wheeler, 84 Wh/km for the low power 4-wheeler, and 123 Wh/km for the highpower 4-wheeler. For highway driving conditions average elec-trical consumption results were: 133 Wh/km for the low power4-wheeler, and 165 Wh/km for the high power 4-wheeler. The

scooter and 3-wheeler were incapable of sustaining highwayspeeds. Readers are referred to Section 4.1 for a detailed break-down of electrical consumption for different driving conditions.

2. Impact of air conditioners and ancillary component loads on elec-trical consumption: Ancillary components have a significantimpact on electrical consumption, with per km electrical con-sumption increasing linearly with greater ancillary componentloads. The slope of increasing electrical consumption is largerfor lower speed driving conditions (i.e. in cities), but is not sen-sitive to vehicle type.

3. Impact of variations in vehicle mass on electrical consumption: Perkm electrical consumption also increases linearly with increas-ing vehicle mass (i.e. for more passengers or cargo). The slope ofincrease is fairly consistent across different driving conditionsand vehicle types.

4. Impact on variations in motor efficiency on electrical consumption:Per km electrical consumption decreases linearly with greatermotor operating efficiency, however the slope of this decreaseis highly sensitive to vehicle size. An important finding is thatfor smaller vehicles, like scooters, increasing motor efficiencyhas little impact on electrical consumption. As a result, theuse of inexpensive and less efficient motors to minimize thecost of electrical scooters will only have minimal impact onelectrical consumption and thus on EV range. For larger vehi-cles, however, motor efficiency has a significant impact, withmore efficient motors allowing significantly reduced electricalconsumption. For larger vehicles there is also an impact of driv-ing characteristics, with higher speed driving conditions show-ing greater variation of electrical consumption with changes inmotor operating efficiency.

Acknowledgements

This work was supported by the Assistant Secretary of Policyand International Affairs, Office of Policy and International Affairs,of the US Department of Energy and the Regulatory AssistanceProject through the US Department of Energy under Contract No.DE-AC02-05CH11231.

Appendix A.

This Appendix presents a brief description of the powertrainand component models that are used to model the four types ofelectric vehicles considered in this study. For a detailed descriptionof each model, readers are referred to the documentation associ-ated with the commercially available powertrain modeling soft-ware Autonomie, which was used in this study.

A.1. Overall powertrain architecture

The electric vehicle models in Autonomie include the compo-nent models shown in Fig. 10, as well as an overarching propulsionand brake control model.

The propulsion control model translates driver accelerationcommands, which are governed by the specified drive cycle, intomotor torque demands while simultaneously considering vehicleand motor speed, battery state of charge, maximum torque outputbefore wheel slip at a given speed, and loading from ancillarycomponents.

The braking control model performs a similar function of trans-lating driver braking commands, which are governed by the spec-ified drive cycle, into braking torque demands while consideringseveral factors and constraints. One further function of the braking

S. Saxena et al. / Applied Energy 115 (2014) 582–590 589

model is to specify the braking torque provided by the tractionmotor and the mechanical brakes. In general, braking torque isprovided entirely by the traction motor until the motor orbattery power limits are encountered. Beyond these limits,mechanical braking is used to absorb the remaining required brak-ing torque.

A.2. Battery model

The battery model calculates the state of an individual cell andassumes that all cells operate identically. Cell state of charge (SOC)is calculated according to the coulomb counting approach in Eq.(1):

SOC ¼�R

I3600 dt þ Ahinit

Ahmaxð1Þ

In Eq. (1), I is the charging or discharging current requested fromthe battery, Ahinit is the amount of energy stored in the batterywhen the model is initialized, and Ahmax is the maximum energystorage capacity of the battery as a function of cell operating tem-perature, as shown in Eq. (2):

Ahmax ¼ f ðTcellÞ ð2Þ

The values for Ahmax are specified in an initialization file using mea-surement data for the maximum capacity of a cell that is dischargedat a C/5 rate.

The open circuit voltage and the internal resistances of the cellon charging or discharging are determined as a function of SOC andcell temperature, as shown in Eq. (3) through Eq. (5) respectively:

VOC ¼ f ðSOC;TcellÞ ð3Þ

Rint;chg ¼ f ðSOC;TcellÞ ð4Þ

Rint;dis ¼ f ðSOC;TcellÞ ð5Þ

Open circuit voltage and internal resistance data on charging anddischarging is specified in an initialization file based on experimen-tally measured data. For lithium ion batteries, open circuit cell volt-age typically spans a range from 3.5 to 4.2 V.

The cell output voltage at any given operating condition (i.e. atthe battery terminals) is calculated using Eq. (6) for charging andEq. (7) for discharging:

Vout;chg ¼ VOC � gcoulIoutRint;chg ð6Þ

Vout;dis ¼ VOC � IoutRint;dis ð7Þ

gcoul is the coulombic efficiency, which for these models is simplyset to 1.0.

A simple thermal model is included as part of the battery modelto estimate the cell operating temperature. The rate of heat gener-ation in a cell is calculated according to Eq. (8) while charging andEq. (9) while discharging:

_Q gen;chg ¼ I2outRint;chg � VoutIoutð1� gcoulÞ ð8Þ

_Q gen;dis ¼ I2outRint;dis ð9Þ

Heat dissipation is calculated by assuming a fan flows cooling airacross the cells within a pack. Eq. (10) is activated when the celltemperature rises above a specified threshold to cause the batterymanagement system to turn the cooling fan on.

_Q cooling ¼Tmodule air � Tmodule

Thermal resistanceð10Þ

In situations where the cooling fan remains off, Eq. (10) is simply setto zero. The module air temperature is calculated using Eq. (11):

Tmodule air ¼ Tair � 1=2_Q cooling

_mcooling airCp;moduleð11Þ

Finally, the module temperature is calculated through the bal-ance of heat generation and heat dissipation rates in Eq. (12),and it is assumed that each cell within the module will have thesame temperature.

Tcell ¼ Tmodule ¼

R _Qgen � _Q cooling

� �dt

mmoduleCp;moduleð12Þ

A.3. Motor model

The motor model provides the torque demanded by the propul-sion controller, while taking into account the effects of losses androtor inertia. Motor temperature is used to determine the time thatthe motor can spend above the maximum continuous rated torquelevels.

The maximum continuous torque is specified in an initializationfile for a full range of motor speeds according to experimental data.The absolute maximum torque output is specified according toa predefined value for continuous to peak torque ratio. Themotor efficiency map is specified for a full range of torque andspeed points in the initialization file using experimentallymeasured data. The motor model inputs are the command to themotor (i.e. required propulsion torque), the input voltage, andthe motor speed.

The maximum propulsion and regenerative torque capabilitiesof the motor are determined as a function of motor speed. Themaximum torque map (as a function of speed) is specified in thevehicle initialization file. The specified torque map enables maxi-mum torque at low speeds (up to roughly 2000 RPM) and subse-quently decaying maximum torque up to the high speed limits ofthe motor.

Section 4.2.3 of this paper examines the impacts of differentlevels of motor efficiency for the vehicles that are modeled. Motorefficiency is scaled by multiplying the efficiency map specified inthe initialization file by a scaling factor. The scaling factor is de-fined as the ratio of desired maximum motor efficiency over themaximum efficiency specified in the map defined in the initializa-tion file.

A.4. Torque coupling, final drive and wheel model

The final drive and torque coupling models are functionallysimilar, and serve to apply a fixed gear reduction ratio to both tor-que and speed by taking into account the losses. The torque cou-pling and final drive are assumed to be 97% efficient across theentire torque/speed range.

The wheel model serves to transform rotational energy into lin-ear. Losses from mechanical braking and tire friction are calculatedwithin this model. Linear force exerted or absorbed by the tires iscalculated using Eq. (13):

F ¼ T=rwheels ð13Þ

In Eq. (13), T is the total input or output torque to the tires, andrwheels is the wheel radius. Torque input or output is calculatedusing Eq. (14):

T ¼ Tin � Tbraking � Tres ð14Þ

In Eq. (14), Tin is the torque input from the vehicle powertrain,Tbraking is the braking torque exerted by the mechanical brakes,and Tres is the resistive torque from tire rolling resistance which iscalculated using a third-order polynomial function of speed. The

590 S. Saxena et al. / Applied Energy 115 (2014) 582–590

coefficients for the polynomial are specified in an initialization filebased on experimentally measured data.

A.5. Chassis model

By balancing the total powertrain output against the totalopposing forces, the linear acceleration and vehicle speed is finallycalculated at the chassis model. Powertrain output (or regenerativeinput) is calculated at earlier sub-models based on the specifieddrive cycle and the specified component parameters. Opposingforces include factors such as hill climbing, aerodynamic losses,and tire rolling resistance.

Aerodynamic losses are calculated within the chassis modelusing Eq. (15):

Floss;aero ¼ 1=2qCdAV2 ð15Þ

In Eq. (15), q is the density of air, Cd is the vehicle drag coefficient, Ais the frontal area of the vehicle, and V is the vehicle speed.

Opposing force from hill climbing is calculated using Eq. (16):

Floss;hill ¼ mg sinðhÞ ð16Þ

In Eq. (16), m is the vehicle mass, g is the acceleration from gravity,and h is the hill grade.

The acceleration of the vehicle is subsequently calculated usingEq. (17):

a ¼ Fin � Floss

mstatic þmdynamicð17Þ

In Eq. (17), Fin is the input from the vehicle powertrain, Floss is thesum of all opposing forces, mstatic is the static mass of the vehicleand mdynamic is the dynamic mass of the vehicle from rotating com-ponents. Vehicle speed is calculated by integrating Eq. (17) overtime.

A.6. Ancillary components models

Power losses from ancillary components (such as air condition-ing and electronic in-vehicle equipment) are calculated as a spec-ified continuous power draw. The power that is flowed toancillary components is assumed to travel through a power con-verter which maintains its output voltage at the required voltageinput for ancillary components (i.e. 12 V). The power converter isassumed to have 95% conversion efficiency.

References

[1] Department of Heavy Industry, Government of India, National Electric MobilityMission Plan 2020, <http://dhi.nic.in/NEMMP2020.pdf>, [accessed19.03.2013].

[2] Foley A, Tyther B, Calnan P, Gallachoir BO. Impacts of Electric Vehicle chargingunder electricity market operations. Appl Energy 2013;101:93–102. http://dx.doi.org/10.1016/j.apenergy.2012.06.052.

[3] U.S. Department of Energy, EV Everywhere grand challenge blueprint, 2013,<http://www1.eere.energy.gov/vehiclesandfuels/electric_vehicles/pdfs/eveverywhere_blueprint.pdf>.

[4] The Central People’s Government of the People’s Republic of China, Energy-saving and new energy automotive industry development plan, 2012, <http://www.gov.cn/zwgk/2012-07/09/content_2179032.htm>.

[5] Romero JJ. Blackouts illuminate India’s power problems. IEEE Spectrum2012;49(10):11–2. http://dx.doi.org/10.1109/MSPEC.2012.6309237.

[6] Hidru MK, Parsons GR, Kempton W, Gardner MP. Willingness to pay for electricvehicles and their attributes. Resour Energy Econ 2011;33(3):686–705. http://dx.doi.org/10.1016/j.reseneeco.2011.02.002.

[7] Collins MM, Mader GH. The timing of EV recharging and its effects on utilities.IEEE Trans Vehicular Technol 1983;VT-32(1):90–7. http://dx.doi.org/10.1109/T-VT.1983.2394.

[8] Rahman S, Shrestha GB. An investigation into the impact of electric vehicleload on the electric utility distribution system. IEEE Trans Power Deliv1993;8(2):591–7. http://dx.doi.org/10.1109/61.216865.

[9] Farmer C, Hines P, Dowds J, Blumsack S. Modeling the impact of increasingPHEV loads on the distribution infrastructure, In: 43rd Hawaii internationalconference on system sciences; 2010, doi:10.1109/HICSS.2010.277.

[10] Denholm P, Kuss M, Margolis RM. Co-benefits of large scale plug-in hybridelectric vehicle and solar PV deployment. J Power Sources 2013;236:350–6.http://dx.doi.org/10.1016/j.jpowsour.2012.10.007.

[11] Qian K, Zhou C, Allen M, Yuan Y. Modeling of load demand to EV batterycharging in distribution systems. IEEE Trans Power Systems2011;26(2):802–10. http://dx.doi.org/10.1109/TPWRS.2010.2057456.

[12] Wang J, Liu C, Ton D, Zhou Y, Kim J, Vyas A. Impact of plug-in hybrid electricvehicles on power systems with demand response and wind power. EnergyPolicy 2011;39(7):4016–21. http://dx.doi.org/10.1016/j.enpol.2011.01.042.

[13] Wu D, Aliprantis DC, Gkritza K. Electric energy and power consumption bylight-duty plug-in electric vehicles. IEEE Trans Power Systems2011;26(2):738–46. http://dx.doi.org/10.1109/TPWRS.2010.2052375.

[14] Kelly JC, MacDonald JS, Keoleian GA. Time-dependent plug-in hybrid electricvehicle charging based on national driving patterns and demographics. ApplEnergy 2012;94:395–405. http://dx.doi.org/10.1016/j.apenergy.2012.02.001.

[15] Weiller C. Plug-in hybrid electric vehicle impacts on hourly electricity demandin the United States. Energy Policy 2011;39(6):3766–78. http://dx.doi.org/10.1016/j.enpol.2011.04.005.

[16] Peterson SB, Whitacre JF, Apt J. Net air emissions from electric vehicles: theeffect of carbon price and charging strategies. Environ Sci Technol2011;45(5):1792–7. http://dx.doi.org/10.1021/es102464y.

[17] Lawrence Berkeley National Laboratory, V2G-Sim - A platform forunderstanding vehicle-to-grid integration, <http://powertrains.lbl.gov/research-projects/v2g-sim>.

[18] U.S. Department of Transportation, National Household Travel Survey, <http://nhts.ornl.gov/download.shtml>, [accessed 19.03.2013].

[19] Saxena S, Phadke A, Gopal A, Srinivasan V. Understanding fuel savingsmechanisms from hybrid vehicles to inform optimal battery sizing for India.Int J Powertrains IJPT-44428 2013;1:2–3.

[20] Ericsson E. Independent driving pattern factors and their influence on fuel-useand exhaust emission factors. Transport Res Part D: Transport Environ2001;6(5):325–45. http://dx.doi.org/10.1016/S1361-9209(01)00003-7.

[21] Kudoh Y, Ishitani H, Matsuhashi R, Yoshida Y, Morita K, Katsuki S, et al.Environmental evaluation of introducing electric vehicles using a dynamictraffic-flow model. Appl Energy 2001;69(2):145–59. http://dx.doi.org/10.1016/S0306-2619(01)00005-8.

[22] Joumard R, André M, Vidon R, Tassel P, Pruvost C. Influence of driving cycles onunit emissions from passenger cars. Atmos Environ 2000;34(27):4621–8.http://dx.doi.org/10.1016/S1352-2310(00)00118-7.

[23] Hansen JQ, Winther M, Sorenson SC. The influence of driving patterns on petrolpassenger car emissions. Sci Total Environ 1995;169(1–3):129–39. http://dx.doi.org/10.1016/0048-9697(95)04641-D.

[24] Wang H, Fu L, Zhou Y, Li H. Modelling of the fuel consumption for passengercars regarding driving characteristics. Transport Res Part D: Transport Environ2008;13(7):479–82. http://dx.doi.org/10.1016/j.trd.2008.09.002.

[25] Saxena S, Gopal A, Phadke A. Understanding the Fuel Savings Potential fromdeploying hybrid cars in China. Appl Energy 2014;113:1127–33. http://dx.doi.org/10.1016/j.apenergy.2013.08.057.

[26] Chugh S, Kumar P, Muralidharan M. et al. Development of Delhi driving cycle:a tool for realistic assessment of exhaust emissions from passenger cars inDelhi. SAE technical paper 2012–01-0877; 2012. doi:10.4271/2012-01-0877.

[27] Kamble SH, Mathew TV, Sharma GK. Development of real-world drive cycle:case study of Pune, India. Transport Res Part D: Transport Environ2009;14(2):132–40. http://dx.doi.org/10.1016/j.trd.2008.11.008.

[28] Automotive Research Association of India, Testing type approval andconformity of production (COP) of vehicles for emission as per CMV rules115, 116 and 126, Chapter 3, Doc. No. MoRTH/CMVR/TAP-115/116, Issue 4, pp.914, <http://www.araiindia.com/CMVR_TAP_Documents/Part-14/Part-14_Chapter03.pdf>.

[29] Environmental Protection Agency, Detailed Test Information, <http://www.fueleconomy.gov/feg/fe_test_schedules.shtml>, [accessed 18.06.2013].

[30] Johnson VH. Fuel used for vehicle air conditioning: a state-by-state thermalcomfort-based approach, SAE technical paper 2002–01-1957; 2002.doi:10.4271/2002-01-1957.

[31] Lukic SM, Emadi A. Performance analysis of automotive power systems: effectsof power electronic intensive loads and electrically-assisted propulsionsystems. IEEE Vehicular Technol Conf 2002;56(3):1835–9. http://dx.doi.org/10.1109/VETECF.2002.1040534.

[32] Carlson RB, Lohse-Busch H, Diez J, Gibbs J. The measured impact of vehiclemass on road load forces and energy consumption for a BEV, HEV, and ICEVehicle. SAE Int J Alternat Powertrains 2013;2(1):105–14. http://dx.doi.org/10.4271/2013-01-1457.