Factors affecting electric vehicle energy consumption

Abstract

Internal combustion (IC) engines waste a majority of the energy they consume, with only 20% actually going into moving the vehicle. The drivetrains of electric vehicles (EVs) can operate at over 80% efficiency which shows that they have great potential in reducing the transportation energy demand. This paper initially quantifies the energy needed to run an EV, having similar dimensions and performance to modern IC vehicles. Simple range and cost calculations were used to establish the advancements needed in battery technology to match the ranges of IC vehicles. Factors affecting EV energy consumption are then addressed, with the aid of MATLAB^® simulations, to ascertain what variations can be expected in real-world situations and the benefits of optimising vehicle parameters. The results are then compared with conventional and hybrid IC vehicles. It is shown that an optimised EV can achieve a 63% ‘tank-to-wheels’ energy reduction over the best conventional IC vehicles available, and 60% over hybrids. The effects of either a badly optimised EV, hard acceleration during the driving cycle, or constant large accessory power draws, such as heaters and demisters, are each shown to increase the EVs energy consumption by 70%/km. To achieve the performance and practicalities comparable with modern IC vehicles, new battery technologies with specific energies of >300 Wh/kg are required.

Keywords:

• drivetrain
• electric vehicles
• energy

Nomenclature

DOD	=	Depth of discharge
EV	=	Electric vehicle
FTP	=	Federal test procedure
GUI	=	Graphical user interface
IC	=	Internal combustion
NEDC	=	New European driving cycle

1. Introduction

The majority of the UK transport network is currently powered by fossil fuels and produces large quantities of emissions. These emissions present both direct and indirect health problems to the environment and humans. This, along with diminishing fuel reserves, tightening emissions regulations and rising fuel costs, necessitates the need to create a cleaner more sustainable transportation network.

It is critical that energy consumption is minimised if this is to be achieved. Most internal combustion (IC) vehicles waste up to 80% of the available energy before it even gets to the wheels, and then more through braking (Plotkin et al. 2001). For example, a car of mass 1300 kg travelling at 113 kph (70 mph) possesses approximately 640 kJ (177 Wh) of kinetic energy which, under braking, is lost as heat. Electric vehicles (EVs) lend themselves to regenerative braking, which can recoup some of this energy, and motors are far more efficient in terms of energy usage than IC engines.

Hydrogen fuel cells can be used to power EVs, but their efficiencies are around 40–55% and there are also considerable losses in producing hydrogen (Campanari et al. 2009). These disadvantages offset those gained by the motor, which is why the analysis in this paper focuses on battery-powered EVs.

The energy consumption of EVs is known to be dependent on many factors that include the following:

•	vehicle parameters, such as coefficient of drag (Marco et al. 2007);
•	driving style and conditions (Ye et al. 2008);
•	auxiliary power draws, most significantly demisters and cockpit heating, which is usually provided by the waste heat from IC engines and
•	operating conditions, noticeably cold weather, which reduces battery capacity (Doerffel and Sharkh 2006).

The greatest limitation of EVs, at present, is their energy storage. Their limited storage capacities have resulted in most vehicles being significantly lower powered than similar IC vehicles and their ranges being many fold less. This, coupled with higher initial costs and longer refuelling times, has led to limited numbers of EVs being sold and produced to date.

The current limited test data for EVs energy consumption have resulted in many reports either extrapolating what data are available or using previous estimates (Granovskii et al. 2006, Lane 2006, ARUP and CENEX 2008). Since most EVs are considerably smaller and slower than their IC counterparts, this can lead to inconsistent comparisons. The majority of test data is taken from vehicles keeping all the auxiliaries off, which often results in optimum values being quoted, especially by manufacturers (Mackay 2008), which are not realised in use.

This paper initially considers the minimum requirements needed for an EV to compete with the performance and practicalities that are expected of modern IC vehicles, specifically a small family-sized vehicle such as a VW Golf/Ford Focus. Hand calculations, quoted data and SIMULINK^® are then employed to estimate the energy requirements needed by such an EV (see Section 2.3). These initial results are then used to determine the requirements of the battery.

The second part of the paper deals with the effects of EV parameters and driving style on their ‘tank-to-wheels’ energy consumption. This was achieved by constructing a graphical user interface (GUI) using MATLAB^®, which could iterate these values, to identify the extent of their effects. The model used a ‘backwards-facing’ methodology controlled by velocity inputs from commonly used driving cycles. This approach allowed the program to remain relatively simple when compared with a command-based model such as Argonne National Laboratory's Powertrain System Analysis Toolkit vehicle model (Rousseau et al. 2004). The authors' model resembles the ‘backward-facing’ calculation path of the National Renewable Energy laboratory's ‘ADVISOR’ vehicle simulator (Wipke et al. 1999, Plotkin et al. 2001), but with the simplification that all the geartrain losses were lumped into one constant efficiency term. To assess the effect of this simplification, the authors' program was re-run with a variable efficiency gearbox model of the type given in Doucette and McCulloch (2011), which showed discrepancies of < 1% in the standard vehicles energy consumption over all the driving cycles.

This type of model has the restriction that it will always simulate the input velocity, regardless of whether the vehicle in question can feasibly achieve this (Markel et al. 2002). However, this was overcome by interrogating the motor output torque for each simulation to ensure that it was always within achievable limits and by using the separate SIMULINK^® model to evaluate the vehicle's maximum capabilities (see Section 3.2).

Finally, the values for EVs ‘tank-to-wheels’ energy requirements are compared with those of IC and hybrid vehicles.

2. Minimum energy requirements

2.1 EV performance requirements

To calculate the energy required by an EV, the minimum acceptable vehicle performance requirements need to be defined. Considering the requirements to keep up with modern traffic and those of similar IC vehicles, the authors defined the following minimums:

•	speed of >128.7 kph (80 mph) (to enable safe use on modern motorways) and
•	0–96.6 kph (0–60 mph) acceleration of < 15 s (the 2010 Citroen C3 1.1i takes 16 s and the Ford Focus 1.4 l takes 13.6 s (Parker's 2010)).

To meet the practicalities of modern vehicles, in terms of space and handling, there must also be a limit on the weight of batteries that can be used. The powertrain (excluding the transmission), for a lightweight IC vehicle, has been suggested to represent 30.7% of the vehicles mass (Burnham et al. 2006). For a 1300 kg vehicle, see Table 1, this gives an IC powertrain weight of 400 kg, which can be neglected in an EV. An electric motor and controller/inverter, of power similar to that needed to propel the vehicle in simulation, have been estimated to weigh about 100 kg (Burnham et al. 2006). This leaves a difference of 300 kg, which has been selected as the maximum permissible battery mass, to ensure that the simulated EV has a total mass comparable with that of similar IC vehicles.

Table 1 Default input parameters used for EV simulation.

Parameter	Default value
Vehicle mass (not including batteries)	1000 kg (2010 VW Golf weight 1314 kg, Volkswagen 2010) minus weight saved by swapping engine, exhaust and fuel tank for electric motor and possible light-weighting 314 kg)
Battery mass	300 kg
Coefficient of drag	0.25 (this is a low figure but has been achieved by the Toyota Prius/Audi A2 and Mercedes are aiming for < 0.2 within 5 years, Mercedes-Benz 2009)
Vehicle frontal area	1.8 m^2 (Ford Escort)
Percentage of energy recouped through braking	50% (this is thought to be the maximum that is practically achievable, Moeller 2002)
System efficiency	95% (allows for mechanical losses with single speed gearbox, Ehsani et al. 2005, Campanari et al. 2009)
Road incline	0°
Average power drawn by accessories	800 W (Rashid 2007)
Coefficient of rolling resistance	0.007 (this represents energy efficient tyres)
Density of air	1.25 kg/m^3 at an air temperature of ≈ 10°C
Tyres	175/65R14 having a radius of 0.3 m

To help produce accurate and realistic figures for EV energy consumption, the vehicle parameters were selected to provide optimum efficiency whilst still being attainable from a small family-sized EV, using currently available technologies. The selected values are given in Table 1 and were used in all the simulations quoted in this paper, unless stated otherwise.

2.2 Constant velocity power

When the car is at maximum speed, its acceleration will be zero. The force that must be overcome corresponding to this condition is given by Equation (1).

Using the figures given in Table 1 and the required velocity of 128.7 kph/35.76 m/s produced the following results which were subsequently converted into a power requirement using Equation (4):

This shows that a power of 16.1 kW is required at the wheels for the vehicle to cruise at 128.7 kph (80 mph) on a flat surface. Allowing for conversion, motor and mechanical losses of 12.5% (Campanari et al. 2009) gives a power requirement of 18.4 kW.

2.3 Power required for acceleration

To calculate the power required for acceleration, the output torque profile from the motor is required. For the purposes of this estimation, it was assumed that the vehicle was powered by a high speed AC induction motor having a maximum speed of 13,000 rpm similar to that used by General Motors' EV1 and the Tesla Roadster (Tesla Motors 2010). The constant volt/hertz method (Ehsani et al. 2005) was used to power the motor. This allowed the output torque to be modelled as a constant up to 7000 rpm and then assume the motor operated at constant power (Equation (5)) up to its maximum speed.

Using this model, the default values in Table 1, and Equations (1-3), a SIMULINK^® program was constructed that calculated the vehicle acceleration.

To compensate for the rotational acceleration forces, they were converted into their approximate equivalent translational effects. These translational effects were then applied to the vehicle acceleration force calculations, which resulted in a relatively small increase (Larminie and Lowry 2003, Miller 2004).

With the given wheel size and motor top speed, the maximum gear reduction ratio between the motor and wheels to achieve a speed of >128.7 kph was found. This was rounded to 11:1 which gave a top speed of 132.3 kph. Next, by iterating the maximum motor power, it was found that 57 kW was required to achieve an acceleration from 0 to 96.6 kph (0–60 mph) in 14.9 s. The motor efficiency was then calculated using the outputted speed/torque values and equations based on data of a similar motor. Section 4.2 describes the motor further and depicts the efficiency plot produced using these equations. Allowing for the motor inefficiencies indicated that a motor input power of 63.5 kW was necessary.

This showed that the power required for acceleration is the limiting factor in selecting a power supply/motor. Using the estimated peak power demand of 63.5 kW and maximum battery weight of 300 kg required the batteries to have a specific power of >212 W/kg.

2.4 Energy consumption

EVs having the requirements set out in Section 2.1 will be unable to achieve the ranges of IC vehicles using today's battery technologies. To aid in the following calculations, it was necessary to define a minimum requirement, for which 160 km (100 miles) was selected. This is a low figure compared to IC vehicles but in 2006 the average UK car journey was < 13.6 km and 93% of all journeys were < 40 km (ARUP and CENEX 2008). Nissan's ‘Leaf’, which is of similar size to that of the vehicle being modelled, is claimed to be able to cover up to 160 km on one charge from a 24 kWh Lithium-ion battery pack (Nissan-Leaf 2010). This indicates that the selected range should thus suffice for most journeys and still be achievable by today's battery technologies.

To estimate the battery requirements, the average energy consumption of the EV whilst in use is needed. Using the Nissan ‘Leaf’ data gives an approximate energy consumption of 0.15 kWh/km, assuming that the batteries are fully drained, or of 0.12 kWh/km if the batteries are cycled to 80% depth of discharge (DOD), which should be the case to prevent damage. This figure appears reasonable when compared with those quoted by other EV manufacturers, as is summarised in Table 2.

Table 2 Energy consumption data for selected conventional sized EVs on the market.

Vehicle	Energy consumption (kWh/km)
Peugeot Berlingo Electric	0.22 (Lane 2006)
Electric Smart fortwo	0.15 (Carroll 2010)
i-MiEV	0.12 (Mitsubishi Motors 2011)
Toyota RAV4 EV	0.15 (over SAE J1634 cycle; Idaho National Laboratory 1998)

Using the figure of 24 kWh as the minimum battery energy required to meet the 160 km range, and 300 kg for the maximum battery weight, shows that the batteries must have a specific energy of >80 Wh/kg.

3. Costs and range

3.1 Battery cost and selection

Currently, EVs cost significantly more than similar IC vehicles. For instance, the electric Mitsubishi i-MiEV was priced at £23,990 in 2011 (after deducting the £5K UK government subsidy; Mitsubishi Motors 2011), whereas the price of the larger IC Mitsubishi Colt starts at £9249. The price difference results largely from the cost of the batteries. This makes the mass, and chemistry, of the batteries selected critical to the economic viability as well as the practicalities of an EV. Table 3 summarises the performances and costs of various battery technologies.

Table 3 Average performance and estimated costs of various battery technologies (Ehsani et al. 2005, Westbrook 2007, ThermoAnalytics 2010).

Battery type	Specific energy (Wh/kg)	Specific power (W/kg)	Number of cycles down to 80% DOD	Approximate current cell cost (£/kWh)	Estimated large-scale cell production cost (£/kWh)
Lead-acid	35	150	1000	70	40
Valve regulated lead-acid	50	250	1500	130	130
Nickel cadmium	50	150	1000	255	190
Nickel-metal hydride	80	200	1200+	200	160
Lithium-ion	120	250	1000+	300 (BatterySpace.com 2011)	130
Zinc–air	200	120	200	200	65

From a cost perspective, lead-acid batteries are clearly the best. However, as shown in Section 2, for an EV having a range >160 km and other factors comparable with those of modern IC vehicles, the batteries will require a specific power >212 W/kg and a specific energy >80 Wh/kg. This leaves lithium-ion as the only current option.

It should be noted that the costs refer to the cell costs only. For lithium-ion batteries, the inclusion of all the necessary components/processes (e.g. the battery management system and assembly labour) to create a complete battery pack will increase the cell cost by approximately 60% (Arnold 2011).

3.2 Vehicle range

To estimate the range of EVs, a GUI was created using MATLAB^® which allowed the vehicle parameters to be changed. The program was based on Equations (1-4) and examples such as those shown in Larminie and Lowry (2003) and Ehsani et al. (2005). The default data, shown in Table 1, along with a gear ratio of 11:1, were used to model the vehicle parameters. The motor type was again assumed to be AC induction having its efficiency map based on a 102 kW version, similar to that used in General Motors' EV1 (see Section 4.2). The peak power of 102 kW is considerably higher than the 60 kW requirement predicted in Section 2.3. This is to ensure that the motor does not spend much time at maximum output, which could cause overheating. The energy source was modelled as 300 kg of lithium-ion batteries, with a specific energy of 120 Wh/kg, and the vehicle range was taken between fully charged down to 80% DOD, to prevent over-discharge. The DOD was based on the open circuit voltage and calculated by the program using Equation (6):

Equation (6) was derived from test data for a lithium-ion cell (Maxim 2008) using a polynomial fitting function in MATLAB^®. Over the 0–80% DOD range used in the model, this equation exhibited a maximum deviation of < 1% from the test data.

Accurate capacity modelling of lithium-ion batteries is complex and involves many variables relating to the precise battery chemistry and operating conditions (Sikha et al. 2004, Subramanian et al. 2007). Most battery capacities are severely affected by the discharge rate, but this effect is less pronounced for lithium-ion batteries. Some tests have shown that, if a lithium-ion battery's temperature is allowed to rise naturally, the capacity is unaffected, within limits, by discharge rate (Doerffel and Sharkh 2006). When modelling the lithium-ion batteries in this paper, their capacity was thus assumed to be unaffected by temperature.

The vehicle was simulated over the New European Driving Cycle (NEDC), which is the same cycle used to provide the quoted emission figures for IC vehicles in the UK.

Running the program with these parameters gave a predicted range of 196 km and an energy consumption of 0.142 kWh/km. Referring to the measured EV energy consumptions given in Table 2, and taking into account that the modelled lithium-ion batteries have a higher specific energy than was calculated necessary for the range of 160 km, this figure appears justifiable.

The program was then set to calculate the effects of varying the vehicle battery mass and the battery cost, based on the current values in Table 3. Modifications were then made to predict the effects of conventional lead-acid batteries, having a specific energy of 35 Wh/kg, and those of an advanced future battery, possibly zinc–air, having a specific energy of 200 Wh/kg (Equation (6) was used to estimate the advanced battery discharge profile). Figures 1 and 2 show the effects of these alterations on the EVs range and battery cost, with current zinc–air battery costs used for the advanced battery values.

Figure 1 Effect of battery mass and type on vehicle range and cost.

Figure 2 Effect of battery mass and specific energy on range.

From Figure 1, the clear range advantages and cost penalties of lithium-ion batteries can be seen over lead-acid batteries. The cost of a lithium-ion battery pack will place restrictions on the amount used in a vehicle, and consumers/manufacturers will need to judge at what point additional costs and mass outweigh the increased range.

Figure 2 shows the effects of employing extreme battery masses on range. This indicates that as the mass increases, the relative percentage increase in range decreases because more of the battery energy is consumed in transporting the increased battery mass. Increasing the specific energy of the batteries causes this effect to be less predominant at low battery weights. This produces slightly greater range increases than would be expected, for a certain percentage improvement in the batteries' specific energy.

The minimum distance that any current IC vehicle can cover on a full tank of liquid fuel is about 483 km (300 miles). For an EV to achieve this range, it would require one tonne of lithium-ion batteries (Figure 2). This weight is impractically high and will also increase the vehicle energy consumption by over 20%. Simulations showed that to achieve 483 km on a 300 kg battery pack, a battery with a specific energy of >300 Wh/kg would be needed, some two and a half times that of current lithium-ion batteries.

4. Parametric analysis of energy consumption

4.1 Vehicle design variables

The design of a vehicle has a significant impact on its energy consumption. With most companies chasing efficiency when designing EVs, rather than looks or space per se, EV parameters are often far better optimised than conventional IC vehicles. This results in smaller vehicles that use more advanced components, such as carbon fibre to minimise weight and low friction tyres to reduce the coefficient of rolling resistance. Such components typically act to increase the cost of the vehicle, which is already high due to the batteries.

In this section, the MATLAB^® GUI was used to iterate the values for the most significant vehicle parameters from those selected previously (see Table 1 and Section 3.2). Each parameter was altered by ± 50% to provide the result shown in Figure 3. The extremities of some of these predictions may be impractical in reality, but the study was conducted primarily to highlight the relative importance of various parameters and where best to focus development.

Figure 3 Effect of EV parameters on vehicle energy consumption.

Figure 3 shows that the relative effects of the vehicle's coefficient of drag, coefficient of rolling resistance, frontal area and average power draw by accessories are all surprisingly similar. Each of these showed an approximate 10% loss or gain in energy consumption for the ± 50% variation, which correlates with a range variation of about ± 19 km. It should be noted that the contributions from the coefficient of drag and vehicle frontal area are highly dependent on vehicle speed, and will, therefore, vary considerably with the driving cycle used in the simulation. The vehicle mass showed the highest variation with a 15% increase, which resulted in a range reduction of over 26 km. The results for regenerative braking showed a decrease in energy usage of 5% as the percentage efficiency was increased. This represented a maximum efficiency of 75%, which is unlikely to be achievable, but the original value of 50% had effectively already achieved an 8% decrease in energy consumption and a 15 km increase in range over no regenerative braking.

Demisting and cockpit heating are particular problems for EVs due to their high power requirements, consuming 2500 W (Rashid 2007) and 1800 W (Dhameja 2002), respectively. Therefore, the effects of having these systems switched on most of the time, a scenario likely to occur during winter for example, were judged to represent a 4 kW constant draw. Rerunning the simulation (over the NEDC) with the power draw by the accessories set to 4 kW indicated that over half of the vehicle batteries' energy was being used to power the accessories. This resulted in the energy consumption increasing by 70%, i.e. to 0.24 kWh/km, and the range decreasing by 40%, i.e. to 116 km.

4.2 Motors and gearing

Brushless DC motors are now becoming common for small EVs and it has been shown than they can reduce energy consumption by 10–20% (Kutz 2008), due to improved efficiencies and reduced sizes over induction motors. Induction motors are cheaper, better suited to larger, more powerful vehicles and they are more robust. In this simulation, it was, therefore, judged that an induction motor would be best to achieve the performance requirements because the vehicle being modelled was larger and more powerful than a typical EV available for sale in 2010/2011.

The energy consumption of an EV will vary depending on how efficient the motor is and the conditions at which it is operating. Figure 4 shows the motor efficiency map used for the simulation, along with dots showing the required torque and speed, from the motor every second, when the vehicle was simulated over the NEDC. The efficiency map, which includes the losses associated with the motor controller, was constructed using data from General Motors' EV1 (Larminie and Lowry 2003, Kutz 2008). The efficiency of a motor is usually different when being used as a generator rather than a driver, with most being optimised for one or the other; for simplicity, this model assumed both were identical.

Figure 4 Motor efficiency map with required motor outputs.

From the efficiency contours, it can be seen that the motor spends most of the time well below the optimum efficiency. This would suggest that altering the gear ratio could lower the energy requirements. To achieve this, and the required top speed, would necessitate the use of more than one gear ratio. The program was thus adjusted to include two gears: a top gear with a ratio of 11:1 (to maintain the top speed) and a second gear that would be used for speeds below 80 kph.

The optimal ratio is dependent on the driving cycle and motor but, for the purpose of establishing what merits a multispeed gearbox could offer, the simulation used the NEDC and motor described previously. The ratio of 7:1 was deduced, which gave a 4% energy reduction per km. Adding a second gear is likely to reduce the efficiency and increase the weight, so the program was run again with the mechanical efficiency dropped to 93% and 10 kg added to the vehicle weight. This halved the overall energy efficiency gains to 2%.

4.3 Driving cycles and vehicle range

One of the causes of variation encountered when modelling vehicle energy consumption is for which driving cycle the vehicle is modelled (Samuel et al. 2005, Marco et al. 2007, Ye et al. 2008). Many different cycles are in use, with varying levels of complexity to represent different conditions (Larminie and Lowry 2003, Ehsani et al. 2005). Previous studies have shown that this variation can easily exceed 35% for IC vehicles as shown in Table 4.

Table 4 Comparison of driving cycles (Ehsani et al. 2005, Kasseris and Heywood 2007).

	Driving cycle
	FTP urban	FTP highway	NEDC	US06
Average speed (kph)	27.9	79.3	33.5	77.5
Maximum speed (kph)	86.4	97.7	120	128.5
Fuel consumption for Toyota Camry 2.5 l (litre/100 km)	9.9	7.5	8.8	8.4

Extrapolation of the NEDC gave an acceleration from 0 to 96.6 kph in 48 s, which is well below that of the vehicle's capability. To provide a simple analysis of the effects of increased acceleration on the energy consumption, the European Urban Driving Cycle ECE-15 was used (due to the vehicle set-up, this cycle gave < 1% difference in energy usage compared with the NEDC). This cycle has a maximum speed of 50 kph, which allowed the velocity to be multiplied by 2.5 and still not exceed the EV maximum speed of 128.7 kph. Extrapolating the cycle's longest acceleration period gave a 0 to 96.6 kph time of 50 s, which was reduced to 20s when the velocity vector was multiplied by 2.5. The other two acceleration periods of the ECE-15 cycle have greater accelerations but never reached 96.6 kph. Increasing forces at higher speeds lead to reduced accelerations, so this greater acceleration rate was assumed acceptable up to the speeds involved. Running the simulations, using this ECE-15 cycle multiplied by up to 2.5, produced the results shown in Table 5 along with those for some constant velocity tests.

Table 5 Effect of driving cycle on energy consumption and range.

ECE-15 velocity multiple factor	Extrapolated 0–96.6 kph time (s)	Maximum acceleration (m/s)	Max. speed (kph)	Energy usage (kWh/km)	Range to 80% DOD (km)
0.5	100	0.49	25	0.172	163
0	50	0.97	50	0.142	196
1.5	33	1.46	75	0.159	174
2	25	1.94	100	0.195	140
2.5	20	2.43	125	0.241	111
Constant 48 kph	–	0	48	0.083	338
Constant 80 kph	–	0	80	0.120	232
Constant 113 kph	–	0	113	0.189	146

The effect of hard acceleration can be seen to give 50% decrease in vehicle range for a modest acceleration time of 20 s to 96.6 kph.

4.4 Energy consumption of different drivetrains

Table 6 compares the simulated EV energy consumptions with those of similarly sized current IC vehicles. The values are calculated using the lower heating value of petrol and diesel, taken as 32 and 38.6 MJ/l, respectively.

Table 6 Comparison of the energy consumption of different drivetrains (figures in parentheses indicate values for the most efficient vehicle in range).

Vehicles	Energy consumption per km (kWh)
Measured data
Average petrol Volkswagen Golf	0.53 (0.51)
Average petrol Ford Focus	0.60 (0.58)
Average diesel Volkswagen Golf	0.50 (0.38)
Average diesel Ford Focus	0.52 (0.43)
Toyota Prius Hybrid petrol/electric	0.35
Honda Civic Hybrid petrol/electric	0.41
Calculated data
EV using default values	0.14
EV with 4 kW accessory draw	0.24
EV over ECE-15 multiplied by 2.5	0.24
EV with combined worse effects of all parameters from Figure 3	0.23

5. Discussion

The results of the sensitivity study summarised in Figure 3 demonstrate the importance of a purpose-built, well designed, EV to maximise range and efficiency. For example, changing the coefficients of drag and rolling resistance to 0.4 and 0.013, respectively, and frontal area to 2.2 m², resulted in a 25% decrease in range.

It was shown in Section 4.3 that the driving cycle, which the EV was simulated over, has a dramatic effect on energy consumption, with a 70% increase experienced for harsh acceleration times. Driving cycles also affect fuel consumption for IC vehicles (Kasseris and Heywood 2007) but the effects appear less dramatic, with a 25% increase in fuel consumption associated with cycles containing similarly harsh accelerations (Samuel et al. 2005). However, because the ‘tank-to-wheels’ energy consumption of an IC vehicle is much greater than that of an EV, this percentage increase actually represents a 60% greater energy consumption increase for the IC vehicle over the 0.1 kWh/km found for the EV.

The energy consumption for the slowest driving cycle appears to be higher than for when the vehicle is travelling twice as fast, as shown in Table 5. This contradicts the trend of increasing energy consumption with increasing velocity, and was found to result from two factors. First, because the power draw by the accessories was set as a constant (because the majority of draws, i.e. lights and heaters, are needed for as long as the vehicle is in use regardless of speed), the longer the vehicle took to cover a certain distance the greater the amount of energy they consumed. The vehicle over the ECE-15 cycle multiplied by 0.5 took 40% longer to reach 80% DOD than the standard cycle. This resulted in over 50% of the energy being consumed by the 800 W accessory draw, compared with 30% for the standard cycle and only 7% for the ECE-15 multiplied by 2.5. Second, at low accelerations and velocities, the motor spent all of its time in the bottom left of the motor efficiency map (Figure 4), in which the efficiency is low. The gearing and motor could be switched to resolve this, but this would result in problems at other speeds. This fact makes it very easy to optimise a vehicle to one driving cycle and achieve excellent results, but this may result in disappointing figures when the vehicle is used in real situations.

The effects of temperature were neglected in the analysis, for simplicity, but cold conditions are known to reduce the capacity of lithium-ion batteries (Doerffel and Sharkh 2006). This effect will exaggerate the predicted 40% range reduction caused by additional power draws for heaters, lights and wipers, all of which are more likely to be on during cold weather. Battery heating could be used, but would in itself create another power draw. There are also many other minor effects of cold conditions which will further diminish the range, including increased mechanical friction and aerodynamic drag. For example, a decrease in temperature from 25 to 0°C results in about a 10% increase in drag (Cengel and Turner 2001, Dhameja 2002).

The energy consumption figures collected in Table 6 show that, even under adverse conditions, EVs can offer a 36% reduction over the most efficient conventional diesel vehicles and 31% over hybrids. All the energy consumption figures quoted in this paper refer to the vehicle consumption from the batteries (tank-to-wheels usage), but it should be noted that the energy required from the grid will be considerably higher. The energy efficiency (the ratio of the energy input to the energy extracted from a battery) of lithium-ion batteries is typically around 95% and chargers are only about 90% efficient. This, coupled with self-discharge of approximately 0.35% a day for lithium-ion batteries (Ehsani et al. 2005), makes the energy required from the grid to be about 20% higher. Taking these inefficiencies into consideration will reduce the EV advantage (in comparison with the most efficient IC hybrid vehicles using 0.35 kWh/km) to 18% for the EV consuming 0.24 kWh/km (a figure for an EV with large accessory draws) and 52% for the EV consuming 0.14 kWh/km (a figure for EVs using default values). These reductions are still significant but there are other losses involved in charging EVs. The average UK electricity grid transmission losses are 7.5% (DEFRA 2009) and current natural gas-fuelled stations are only around 40% efficient (Granovskii et al. 2006).

6. Conclusions

EVs can offer up to 75% ‘tank-to-wheels’ reductions in energy consumption over typical liquid-fuelled vehicles. This figure is shown to reduce to 58% when the vehicle was either not optimised, driven hard or had large auxiliary power draws. The power draw associated with EV heating and electrical accessories is predicted to lead to a decrease in range of 40%. This is a serious problem in cold climates where the battery capacity is also reduced, and some manufacturers have incorporated petrol-burning heaters to help alleviate the problem.

Investigation into vehicle driving style shows huge changes in range, varying from 338 km for cruising at 48 kph to only 111 km when the vehicle is used at approaching full throttle during the cycle. The addition of a multispeed gearbox showed a 2% energy reduction, but this was very dependent on the driving cycle and will require additional space and cost.

Lithium-ion is the only battery technology currently capable of fulfilling the performance requirements for an EV that can compete with modern IC vehicles in terms of size, weight and performance whilst achieving a range of over 160 km. The present cost of the batteries required can easily double that of the vehicles. This cost is expected to halve once mass production is fully underway, but this will still result in higher prices than similar IC vehicles. To achieve a range of >483 km for a mass of 300 kg of batteries, the simulations showed that advancements would be needed to create a chemistry capable of producing >300 Wh of energy per kg, some 2.5 times better than current lithium-ion variants.