Evaluation of Current, Future, and Beyond Li-Ion Batteries for the Electrification of Light Commercial Vehicles: Challenges and Opportunities

The first step involves getting an estimate of the power and energy requirements under different conditions of use. The power load of a vehicle can be determined with instantaneous velocity profile, v(t), over a specific trip distance. This data for LCVs and pickup trucks was obtained from a study by Liu et al.,²¹ who have employed a method of 'naturalistic drive cycle synthesis', to construct representative categories using real-world data from the Transportation Secure Data Center (TSDC) of the National Renewable Energy Laboratory.²² The driving conditions identified in their work are Congested-Urban (CU), Uncongested-Urban (UU), Urban-Suburban (US), Urban-Suburban-Highway (USH), and Highway (HW). As detailed in their study, Liu et al.,²¹ examine the naturalistic drive cycle data for pickup trucks from TSDC to procure small sections of drive cycles called 'micro-trips' which are categorized according to metrics like mean velocity and trip duration. After considering the probability of a vehicle transitioning from one micro-trip to another, these micro-trips are clustered together to form a large set of new drive cycles using Markov Chain theory. Based on well-defined criteria for representativeness of a driving condition like the standard deviations of velocity and acceleration, mean positive acceleration, number of stops, etc., 'representative drive cycles' for each driving condition are sifted out from the set of drive cycles by finding the ones which show the smallest error when compared to the defined criteria.

The instantaneous power requirement for each of these categories can be constructed using two relations representative of a dynamic vehicle model, Eq. 1, for the net force on the vehicle, and Eq. 2, which utilizes the drive cycles along with Eq. 1 to provide the different load profiles which the battery pack is subjected to.

where P(t) is the power load, v(t) is the drive cycle, ρ is the density of air, and g is the acceleration due to gravity. Vehicle design parameters are C_d, the coefficient of drag, A, the frontal area of the vehicle, C_rr, the coefficient of rolling resistance and W_T the total mass of the vehicle including the battery pack and the added payload.

In our study, we ignore towing load capacity since it would change each of the vehicle design parameters significantly. Further, this is a secondary performance parameter. Z is the gradient in road elevation, which represents the elevation and contours of the road, and it can be accounted for by increasing the mass of the vehicle as they act effectively as added weight. Under this assumption, to a first approximation, we can neglect the road gradient term. Also, regenerative braking with a recovery efficiency of 40% is used in regions of rapid deceleration.²³ The length of the drive cycle, v(t) for a given driving condition is based on the range in consideration.

Energy consumption estimation studies for electric vehicles typically do not account for the influence of pack weight on the consumption and pack requirements.²⁴ However, this is essential in order to obtain accurate estimates. The following set of Eqs. 3–8 comprise of our estimation framework which needs to be iteratively and self-consistently calculated and this provides accurate values of the required pack energy E_P, pack weight W_P, and the total number of cells n_cells, for a given battery chemistry. The total weight of the vehicle can be categorized into,

where W_V is the weight of the chassis and body and W_L is the added payload. W_V and W_P together specify the curb weight of the vehicle, where W_P, the pack weight is first provided an initial guess value.

where E_i is the estimated energy, t_T is the total trip duration, and P_i(t) is load profile obtained from Eq. 2. Trapezoidal numerical integration scheme is used for calculating the energy requirement. E_i provides the minimum energy required from the battery pack for completing the trip. The actual pack energy, E_P, is calculated after accounting for the battery discharge efficiency, η_b of 95%²⁵ and the battery-to-wheels efficiency, η_btow of 80%²⁵ as shown in Eq. 5.

where n_cells is the number of cells required to assemble a pack of energy E_P, and E_cell is the energy of a single cell, both of which are dictated by the battery chemistry selected and require inputs from Cell design and characterization section. The pack weight from the cells is given by

It is to be noted that the weight of each cell is determined by the battery chemistry.

where W^new_P is the new pack weight based on the packing burden factor f_burden, and Eq. 9 assigns a new value to the weight of the pack which is used in Eq. 3, and the system is solved self-consistently to determine the weight of the pack within a specified tolerance weight of 1g.

The packing burden factor, f_burden, accounts for the weight of the various inactive materials in the battery pack like the coolant/thermal management system, module hardware, and cell interconnects. This factor is particularly important since the theoretical specific energy is reduced significantly at the pack level, which signifies a practical system. Such factors have been studied by researchers^6,26 who have used f_burden factors of (0.25–0.36),⁶ 0.36,²⁶ and 0.4²⁷ for Li-ion systems. Eroglu et al.,²⁸ have generalized a factor equivalent to f_burden where they study the variation of the factor with open-circuit voltage (OCV) for different theoretical specific energies, and observe that for theoretical specific energies below 500 Wh/kg within an OCV range of (3–4.5)V the factor plateaus at value of ∼0.45, a space which specifies most Li-ion systems available currently. Hence, for the purpose of this study, we assume a constant packing burden factor of 0.45 using the cell weight in place of the theoretical weight in order to account for the weight of the oversizing of the battery pack as well.

This framework is first used to estimate the number of cells and pack size for a number of battery chemistries, for the base case of electrifying a Ford F-150 with a range of 200-miles with an added load of 1000 kg. The payload capacity was estimated based on other studies which mention an average payload of 1170 kg²⁹ for similar vehicles. Other base case parameters are mentioned in Table I.

The cell design is carried out such that all the chemistries in consideration are constructed with the same cell geometry and dimensions. The choice of cell geometry and the various issues that need to considered while assembling a battery pack have been extensively explored by other studies^30,31 and do not form the focus of the present work. Automotive battery packs generally use the prismatic cell geometry,^32,33 since it is known that prismatic cells have a more uniform heating, thereby making thermal management easier.³⁴ Further, for a given pack energy the weight of a battery pack assembled using cylindrical cells is about 11% more than a pack assembled using small prismatic or pouch cells.³⁰ Based on these considerations, we expect the use of prismatic stacked electrode design (SED) format as the most likely form factor for electric LCVs.

There has been tremendous progress in the mathematical modeling of Li-ion batteries for determining their performance under various conditions.³⁵ The basis of the governing equations for today's models of intercalation Li-ion systems can be found in the work of Newman and co-workers¹⁴ which resulted in a simulation platform¹⁵ for a Li-ion cell. These models were validated for experimental data^17,20 proving their effectiveness. Parallel to this, thermal models of Li-ion batteries have been developed^18,36 along with approaches to couple the electrochemical and thermal behavior.³⁷ A more detailed discussion of the various models and the development of robust solvers can be found elsewhere.^35,38

In our study, the platform used for the design and characterization process was AutoLion-1D,³⁹ a solver that couples the electrochemical response with changes in temperature using a physics-based Thermally Coupled Battery (TCB). The materials database in the software is experimentally validated and has been used for a range of temperatures and discharge rates^39–41 which makes it ideal for use in this study.

The governing equations of the TCB model are based on the Doyle-Fuller-Newman model,¹⁴ the physics-based approach of Wang et al.,^42,43 and Gu et al.,³⁷ These equations are summarized by Fang et al.,⁴⁰ and their implementation within the AutoLion is shown by Kalupson et al.,³⁹ using Eqs. 10–14 which elucidate a 1-D transport model coupled with a lumped thermal model. Eqs. 10,11 together with Eqs. 12,13 describe the electrochemical response, providing the concentration of ions and the potentials across the cross section of the battery, both in the solid electrodes and the liquid-phase electrolyte. This system of equations when coupled with Eq. 14, which represents the lumped thermal model, introduces the changes in properties such as diffusivity and conductivity which are functions of temperature, thus completing the TCB model. The solid phase charge conservation is given by

where σ^eff is the effective electronic conductivity and Φ_s is the potential of the solid phase. The electrolyte phase charge conservation equation is given by

where k^eff is the effective ionic conductivity, Φ_e is the potential of the electrolyte phase, k^eff_D is the diffusional conductivity obtained from concentrated solution theory, and c_e is volume averaged Li concentration in the electrolyte phase. The species conservation in the electrolyte phase is given by

where ε_e is the volume fraction, D^eff is the effective diffusion coefficient, t⁰₊ is the transference number, and F is the Faraday constant. In each of the equations, the effective transport properties are evaluated using the Bruggeman relation. The species conservation in solid phase is given by

where c_s is the concentration of Li in the solid phase, D_s is the diffusion coefficient in the solid phase, and r represents the radius of the particles of active material. This equation uses co-ordinates in the radial direction, r to represent the microscopic solid-state diffusion.⁴⁰ The energy balance is represented as a lumped thermal model given by

which accounts for q_r reaction heat, q_j the joule heating, q_c the heating due to contact resistance between the current collector and electrode materials, and q_e the entropic heating. The equation also accounts for a heat dissipation term where h_conv is the coefficient of heat dissipation and A_s is the cell external surface area.

In terms of cathode materials, we consider NCA, NMC, LMO, and LFP, and for anodes we considered graphite. LCO is avoided typically due to known issues with safety and cost,³⁴ and LTO anode based systems have a low specific energy,⁴⁴ thereby limiting their use to hybrid electric vehicles. Designing a cell and the specification of the various electrode properties is an extremely difficult task due to the large number of design parameters and the influence each parameter has on the other as well as the performance of the cell, these challenges have been documented in several studies.^45,51 The cell design process is also specific to application for which the cell is employed, and generally the design parameters are specified such that either the specific energy-power, rate capability or calender and cycling life is maximized. For the scope of this study, in order to facilitate a comparison between different cathode chemistries, we fix the total volume of all cells at ∼ 56 cm³, which translates to an assumption that all the cells are prepared within the same prismatic cell architecture. The graphite anode has a fixed porosity of 30% and the thickness of the anode is determined by the anode to cathode coating capacity ratio (N/P) which is fixed at 1.15.

In the context of the LFP cathode, several studies report the influence of properties like optimal electrode thickness,^45–47 active material and inactive material weight fractions,^46–49 and porosity^48,50 on the performance of the full cell. Fongy et al.,⁵⁰ use a parameter which defines the rate of loss in capacity with increase in discharge current to find a critical porosity window between 30% and 40%. They conclude that a porosity of 30% or below would be highly detrimental to the performance due to restricted diffusion of Li-ions in the pores and above 40% the performance would be affected by reduced electronic conductivity. Zheng et al.,⁴⁵ have studied the effect of electrode thickness of LFP and NMC along with a specification of the active and inactive material fractions for high electrochemical performance, and found that increasing the thickness of the electrodes would significantly reduce the rate capability due to restricted diffusion of Li-ions across the electrode, but higher electrode thickness would increase the energy density of the cell due to a reduction in the required amount of inactive cell materials like the current collectors, separator, and cell enclosure per unit energy. They also observed a reduction in capacity retention over cycles for cells with thicker electrodes in the case of LFP and NMC. Other studies on the effect of porosity and thickness on the energy density or specific energy and the rate capability of NMC and/ or LFP cells have made similar conclusions.⁵¹ In another study, Zheng et al., ⁴⁹ studied the effect of the ratio of binder to the conductive additive in NCA cathodes and found that at a PVDF-to-CB ratio of 5:3 the rate capability is not influenced by the total amount of inactive material in the NCA cathode. But other studies⁵² have used a 1:1 ratio in their analysis. For LMO cathodes, certain studies use active material weight fractions of as low as 73.5%⁵³ and others use values as high as 90%.⁵⁴ All electrodes in our study are designed with Polyvinylidene difluoride (PVDF) as the binder and Carbon black (CB) as the conductive additive, the active material fraction of the cathode is maintained at approximately 85% by weight, and the binder and conductive additive ratio is decided on the basis of the values used by other studies.^{45,49,52,54,55} The graphite anode active material fraction is fixed at 89% and the PVDF and CB fractions are 8% and 3% respectively.⁴⁵

Since the focus of our analysis is to compare the use of different cathode materials and the effective pack weight and the performance of the vehicle, in order to facilitate a high specific energy, we use a high cathode thickness of 80 μm for all cathodes since studies⁴⁵ show that, for this cathode thickness, at a discharge rate under 10 C the reduction in rate capability (loss in capacity) is not significant. LFP cathodes are designed with a porosity of 32% based on the results of Fongy et al., and in order to facilitate a fair comparison, the porosity of other cathodes (NMC, NCA, and LMO) is fixed at 30%, although studies⁵⁵ studies show that a much lower porosity is required to achieve higher specific energy. The loading capacity per unit area (mAh/cm²) of the electrode is varied to meet these specifications. The weight fraction of the active material, binder, and conductive additive used in the cell design process have been summarized in Table II. Each single cell is characterized using the built-in simulation tool in AutoLion-1D³⁹ at discharge rates of C/5, C, and 5 C, at temperatures of −10^oC, 25^oC, and 45^oC.

We select one possible battery chemistry based on Estimation framework section, and the required pack energy.³⁹ The (x)S(y)P series-parallel configuration is considered such that the total pack voltage is about 400 V. The pack is assumed to be placed in the Ford F-150 and is subjected to the load profiles of all the drive cycles considered. The pack size corresponding to the most energy intensive drive cycle is used for all the simulations, performed in a Matlab-Simulink platform which uses AutoLion-ST in a Software-in-the-loop approach. The pack is simulated for 1000 cycles and after every discharge cycle, the charging process follows the CC-CV protocol which the most common charging protocol for electric vehicles,^56,57 in this study, the constant current leg was maintained at C/2 and a constant voltage leg of around 400 V. Our analysis does not account for cell-to-cell variation in terms of temperature changes but does account for the temperature variation within a cell due to current. The end of life for the battery pack is generally assumed to be the point at which the fractional capacity remaining is about 0.8 or 80% of the initial capacity.⁵⁸

Modeling the capacity fade mechanisms within Li-ion cells has evolved following the pioneering work of Darling and Newman.¹⁹ The degradation model within the AutoLion platform is based on two processes, film growth on the electrodes and active material isolation at the respective electrodes. The growth of the Solid Electrolyte Interphase (SEI) layer on the anode is modeled similar to other well-known studies^59,60 and the cathode film growth is calculated in a similar manner. The SEI forms as a reduction product of the electrolyte solvent, ethylene carbonate (EC), and the side reaction current density, i_s is cast in the Tafel-form in the following manner

where i_os is the exchange current density, α_s is the electron transfer symmetry factor of 0.5, Φ_s and Φ_e are the solid phase and liquid phase potentials, U_SEI is the equilibrium potential for SEI formation, R_SEI is the SEI layer resistance, and i_t is the total current density which includes the intercalation and the side reaction current densities. i_os requires the concentration of EC at the reaction surface which is obtained using a transport model based on Fick's law. The growth rate of the SEI layer is calculated by

where M_SEI is the molecular weight of the SEI layer and ρ_SEI is the density. The resistance of the SEI layer is calculated using the effective conductivity of the electrolyte through the SEI, by

The loss of active material due to changes in the structural properties of electrodes is accounted for by using an isolation rate model within the 'active material isolation' sub-model³⁹ based on other studies.⁶¹ The submodel is formulated such that the isolation rate depends on the current density as well as an empirical parameter which accounts for the temperature effect. These submodels together quantify the damage to the battery cell and thereby the pack. The degradation sub-models on AutoLion have been validated for different discharge rates and temperatures.³⁹

In order to assess the cost of each of the battery pack, we use BatPaC^62,63 which is an open-source cost estimation tool developed by Argonne National Laboratory.²⁸ It has been employed in several cost modeling studies^32,64–66 to assess the cost of battery systems for electric vehicles. BatPaC is a process-based cost model which uses a bottom-up design approach starting with the required raw materials and proceeds to the modeling of the various manufacturing costs of Li-ion batteries for a specified type of electric vehicle. Once the battery design process is complete based on the inputs for chemistry, pack energy, number of cells, and power, BatPaC provides the cost to Original Equipment Manufacturer (OEM) per pack.⁶³

The cells designed and characterized on AutoLion-1D are then translated into the cell design within BatPaC, maintaining the same active material fractions, porosity, and other specifications. The number of cells calculated from the framework in Estimation framework section along with the pack energy, and the maximum required power are used as inputs for the pack cost estimation. BatPaC calculates the cell capacity for each chemistry based on its cell design routine, and a comparison with the values for AutoLion-1D is in Table IV. The salient governing equations of the cell design routine within BatPaC⁶² are summarized below

where I is the current density, P_max is the maximum rated power which is one of the given inputs, A_pos is the area of the positive electrode, and V_cell is the cell voltage at which the rated power is achieved.

where the energy of the pack is calculated using the capacity of the cells C_cell, or alternatively, as implemented in this study, the capacity of each cell is calculated based on the energy of the pack. Eq. 19 takes into account the discharge losses that occur at a discharge rate of C/3, based on the ASI, is the area specific impedance for the energy.

determines A_pos based on the ASI_power, where V_OCV is the open circuit voltage of the cell.

determines the ASI, where α and β are constant parameters and L_pos is the thickness of the positive electrode. The process for determining ASI can be found elsewhere.^62,67

The first type of cost considered are the materials costs, purchased items, and pack integration cost.⁶² The active materials costs of the cathode are calculated using Eq. 22,

where C ($/kg) is the cost of the lithiated oxide and C_o is the baseline cost. C_i is the cost of the lithium and transition metal, considered on the basis of their molar stoichiometry x_i, and molecular weight MW_i. MW is the molecular weight of the final product. Graphite is the only anode material considered in a ($/kg) basis. The electrolyte, solvent, and current collectors are also accounted for based on the material used, and the consistency of inputs between AutoLion and BatPaC is ensured. The purchased items including the pack and module hardware along with pack integration costs for the battery and thermal management systems are assigned ($/module) or ($/kW) units depending on the type of application. The baseline manufacturing plant within BatPaC uses a manufacturing rate of 100,000 packs per year, and considers the various operational, materials handling, and assembly costs, as well as the 'plant investment costs'. Other details of the cost model can be found in the relevant literature.⁶²

The final price per battery pack is obtained with uncertainty bounds based on 5% error for 'electrode thickness and capacity limits' and a 10% error for 'materials cost and processing limits'.⁶³ Using these cost estimates, we obtain cost inputs in terms of ($/kWh) and ($/cell) for each battery chemistry and then incorporate them into our estimation framework.

Another key parameter that determines the technical feasibility is the pack-to-curb weight ratio given by,

A higher ratio indicates a heavier battery pack which would consume a significant fraction of its own energy to travel with the vehicle. Furthermore, as the weight of the pack increases at a fixed empty vehicle weight, the acceleration and handling performance of the vehicle begins to deteriorate, as noted by electric vehicle manufacturers^68,69 and the amount of energy required to achieve the same acceleration would be much higher for a higher PTC. Furthermore, PTC also provides information on the wells-to-wheels efficiency of the vehicle, which would be significantly reduced if the PTC increases, since, for a fixed vehicle weight, a higher PTC implies that the fraction of energy used to transport the same load would be much greater. In our study, a PTC of over 0.5 is considered a technically unfeasible system.

Each of the drive cycles considered for this study is shown in Fig. 1. For the base case, the energy consumption (Wh/mile) for each drive cycle for the different chemistries shown in Table III. On examining these values along with Fig. 1, we find that the highway based cases (USH, HW) are naturally more energy intensive, due to higher average velocities as compared to the urban-based cases (CU, UU, US). The energy consumption for an NCA-based battery pack is in the range of [395,605] Wh/mile, and [432,639] Wh/mile for LFP. In comparison, other studies⁷⁰ have estimated an energy requirement of 460 Wh/mile for the vehicle class of Pickup trucks.

Zoom In Zoom Out Reset image size

We can see the metrics for the battery pack as estimated by the framework for the base case, comparing the n_cells, W_P, E_P, and the energy consumption in Fig. 2. Due to the lower number of cells required, and lower pack weight, NCA has the lowest required pack energy, while LFP has the highest. LMO and LFP also result in systems with a PTC over 0.5 (Table III), which exceeds the technical feasibility limit. These results confirm the need for higher specific energy in order to achieve high-performance LCVs. NCA and NMC seem to be the most suitable, with NCA performing marginally better than NMC in each of the metrics (10% higher specific energy, and 7% lower PTC for the base case) resulting in a 10 Wh/mile lower energy consumption. The comparative analysis reveals that the use of low specific energy chemistries like LFP and LMO is impractical for LCV applications. To conduct cycling analysis, we pick the system with the NCA-based battery pack which has the lowest energy demand of 121 kWh for the base case, with a required pack size of ∼152 kWh. The estimates for other chemistries can be found in Table III.

Zoom In Zoom Out Reset image size

Aging performance analysis conducted for the base case scenario for all the drive cycles, at a fixed ambient temperature is shown in Fig. 3a, where we see the change in the State-of-Health (SoH) of the pack, which is the percentage of initial capacity left after completing a fixed number of cycles. The NCA based battery pack was used for this analysis, sized at 152 kWh or 421.2 Ah, in a 100S78P configuration with a nominal voltage of ∼365 V. On examining Table III and Fig. 3a together, we find that the Congested-Urban case, the least energy intensive drive cycle, is the most detrimental to the SoH of the pack. A lower average velocity of the CU case indicates that the vehicle would cumulatively take a longer time to complete the same number of 200-mile trips (cycles), resulting in the battery pack being engaged for a much longer duration thereby causing higher degradation. Within the set of drive cycles CU, UU, USH, and HW, a lower average velocity seems to result in greater damage to the pack. However, it is observed that the Urban-Suburban (US) drive cycle does not fit this generalization when compared to the USH drive cycle, since the US case has an intermediate average velocity of 15.77 m/s compared to 9.41, 12.42, 21.27, 22.33 for CU, UU, USH, and HW respectively which does not increase the time required for traveling 200,000-miles or the energy consumption for the same. A combination of these factors causes the pack degradation in the US case to be comparable to the USH case. From these results, we can see that for a fixed ambient temperature, if two vehicles travel at significantly different average velocities, for the same distance traveled, we would see a lower SoH for the vehicle with the lower average velocity. This goes to show that miles driven or energy consumption alone might not be accurate metrics to study battery health, and there is a need for a better metric which couples the miles driven and the cumulative time of battery pack usage.

Zoom In Zoom Out Reset image size

Fig. 3a also tells us that the total projected miles covered over the lifetime of the battery pack is much higher than 200,000 miles since the SoH is more than 90% after 1000 cycles. Changes in ambient temperature would affect this since higher temperatures are known to affect SoH drastically,⁷¹ this can be seen in Fig. 3b which shows the change in SoH at an ambient temperature of 35^oC for a set of drive cycles, and at 45^oC for the CU case where the end of life is reached at 800 cycles or 160,000 miles. In the overall analysis, the cycle life observed is acceptable for electric LCVs since studies⁷² suggest that electric vehicles are competitive when they are driven over 180,000 miles, with a vehicle age of ∼ 15 years. It should be noted that a similar analysis for longer range or a different payload capacity would significantly change the estimated cycle life.

The cost estimation process described in Cost estimation section, is followed to analyze the base case battery packs. The results expressed in ($/pack) and ($/cell) values shown in Fig. 4. NCA has the lowest cost per pack, in spite of having the highest cost per cell. As seen in Fig. 2, the number of NCA cells required to assemble a pack is around 7800 cells, compared to 8600, 11500, and 11800, for NMC, LMO, LFP respectively. And owing to the high specific energy of NCA, the required pack energy is also lower. These two factors together reduce the overall cost of the NCA-based battery pack. Within the range of chemistries considered here, it is evident that specific energy of the battery pack can lead to reductions in pack size and hence the pack-level cost. In terms of commercial viability, NCA and NMC seem to be placed at a better position compared to LMO and LFP. The resultant pack-level mean-cost per unit energy ($/kWh) of these chemistries was 280 for LMO, 257 for LFP, 224 for NCA, and 241 for NMC. Studies⁷³ on prismatic cells estimate the cell cost ($/kWh) to be slightly over 200 for LMO, NCA, and NMC, and over 280 for LFP. Another study³² reports cell costs of 290 $/kWh for LMO, 310 for LFP, 260 for NCA, and 250 for NMC, for slightly different electrode design parameters with cathode thickness of 80 μm. The difference between the estimates could be due to the difference in thickness of the electrode coating considered or, in the context of the latter study due to different electrode design parameters like material weight fractions and porosity.

Zoom In Zoom Out Reset image size

Now we tackle the key question of what trade-off exists between the various vehicle design parameters and how they might influence near-term and longer-term electric LCVs. In order to address this, we have carried out a detailed analysis, based on possible changes in vehicle design parameters (Table V). We consider two of these parameters at a time, within a bivariate analysis setup, and study the pack cost and PTC as functions of the two chosen parameters, while the remaining are fixed at the base case values. The changes in pack cost and PTC observed are a result of the changes in pack energy and pack weight requirements obtained from the estimation framework. Each of the following figures can be examined using a cross-hair approach where we fix the cross-hair at certain values for the two parameters considered and we examine the resultant pack cost and PTC for different chemistries.

Fig. 5 sheds light on the pack cost (Fig. 5a) and PTC (Fig. 5b) as functions of range and the drag coefficient of the vehicle, C_d. Fig. 5a shows that, for the base case range of 200 miles, a C_d reduction of 0.2 changes the cost by $10,000 for NCA and NMC packs, and about $15,000 for LFP. On comparing the cost of a pack needed for 200-miles to that of 500-miles, at a C_d of 0.4, we find that LFP-based systems see an increase of ∼$90,000, and LMO of $70,000, as compared to ∼$60,000 for NMC and ∼$50,000 for NCA-based systems. Switching to the technical implications of this, in Fig. 5b, we see that the pack weight for LMO and LFP systems remains extremely high, and a range of over 300 miles is unattainable at low C_d's as well. At a C_d of 0.2, a range of over 350-miles would be technically feasible with NCA and NMC systems. With LMO and LFP systems, at a threshold PTC of 0.5, the maximum technically feasible range is 300-miles which is achieved at a C_d of 0.2.

Zoom In Zoom Out Reset image size

If we hold the C_d fixed at 0.4 and take up the payload capacity as a parameter, we obtain Fig. 6, where we see cost and PTC as functions of range and payload capacity. Fig. 6a shows that the reduction in cost seen by changing the payload capacity from 2000 kg to 500 kg is around $5,000 for LFP and LMO systems while it is close to $2,000 for NCA and NMC. Clearly, for systems designed with high specific energy battery packs, a reduction in payload capacity does not change the cost of pack significantly. Fig. 6b shows that, at a fixed range, the PTC of NCA and NMC systems changes by about 0.05, for a change in payload of 1500 kg. LMO and LFP-based systems are unlikely to support any significant payload since PTC is over 0.5 even at small payloads.

Zoom In Zoom Out Reset image size

Given the challenges in pack cost reduction through significant re-design required to decrease drag coefficient to about  0.2–0.3, a reduction in vehicle weight could become crucial to reducing the energy consumption and required pack energy and thereby a reduction in pack weight. We look at trade-offs considering the range and vehicle weight as parameters in Fig. 7. Examining Fig. 7a first, we see that reducing the vehicle weight would reduce pack size requirements and thereby reduce the pack cost, while the magnitude of cost reduction is comparable to that seen in Fig. 6a since both the vehicle weight and payload capacity affect the energy demands in the same manner. Although it should be noted that the vehicle weight affects the curb weight which is constant throughout the lifetime of the vehicle, whereas, payload capacity is a parameter the vehicle is designed for and does not necessarily remain the same. This fact plays into the changes in PTC, seen in Fig. 6b, where a lower vehicle weight for a fixed range results in higher PTC's. These results show the limitation of using PTC alone as a metric for technical feasibility, specifically with parameters like the vehicle weight, since a lower PTC achieved through an increase in vehicle weight would also result in higher pack energy requirement, a higher energy consumption, and a higher pack cost as seen in the Fig. 7a. The effect of increasing the vehicle weight on the energy consumption and range capability of the electric vehicle has been noted in other studies.²⁵

Zoom In Zoom Out Reset image size

Finally, Fig. 8 shows cost and PTC studied as functions of vehicle weight and C_d, where we observe a few more interesting results. We see that the cost of the pack in (8a) shows a higher sensitivity to changes in C_d than to changes in the vehicle weight, which implies that the aerodynamic losses form a significant part of the energy consumption since the cost of the pack is directly related to the required pack energy. As explained previously, examining PTC as a metric for technical feasibility when vehicle weight is involved does not provide straightforward results. A conclusion similar to Fig. 7b can be seen in Fig. 8b, where a lower PTC is achieved with higher vehicle weight, but the pack cost increases due to increased energy consumption. Another interesting observation is that at low vehicle weights, the amount of change in PTC achieved by halving the C_d is much higher than 0.1 while it is much lower than 0.1 at high vehicle weights. This is due to the compound effect of lowering the vehicle weight which has an effect on the pack weight as well due to reduced energy consumption at lower vehicle weights. The effect of changing the vehicle weight on the PTC is similar for all examined chemistries, as seen in (Fig. 8b).

Zoom In Zoom Out Reset image size

The trade-offs for changes in the rolling resistance, C_rr were also studied, and it was found that a lower C_rr naturally resulted in lower pack size requirements and hence lower costs and PTC, but the magnitude of improvement for both PTC and cost was considerably small. In all the trade-offs studied here, the changes in the pack-level cost and PTC brought about by increasing the driving range is the most drastic, followed by the C_d, while vehicle weight, added load, and C_rr result in marginal changes. Attempting to reduce the PTC should be performed such that the other metrics like the energy consumption or cost do not increase. A vehicle design effort undertaken for any fully electric LCV should take these effects into consideration in order to select a suitable chemistry for a set of specified performance parameters. Our analysis suggests that a first-principles based vehicle re-design such as that carried out for the Model S by Tesla Inc. would be essential to develop a cost-competitive electric LCV.

Our analysis suggests that energy density plays a crucial role in determining the sizing and cost of the battery pack. In this part of the study, we intend to quantify the potential improvements that could be obtained at a pack-level with beyond Li-ion systems. All these systems are assumed to be built with a target cost of 150 $/kWh based on estimates by other studies.²⁶ The specific energy values used in our analysis are practical pack-level values, based on estimates from prior work,^6,26,74 where we use current target and potential specific energies of [200,300] Wh/kg for future Li-ion batteries, and [350,600] and [600,1000] Wh/kg for Li-S and Li-air systems respectively. We examine the impact of specific energy on cost, PTC, and energy consumption, for different drag coefficients and driving range. The other parameters are fixed at the base case scenario, (Table I). With a cost of 150 $/kWh we observe that the cost of the base case battery pack drops well below $30,000 for current Li-ion batteries. In Fig. 9a, at a fixed C_d, we observe improvements in terms of cost reduction of the order of about $2,000 with future Li-ion and Li-S systems through the increase in specific energy to about 350 Wh/kg. The PTC at this specific energy is reduced from 0.4 of current systems to a value as low as 0.15, or in other words, the weight of the battery pack is only 15% of the total vehicle weight, thereby implying that the system efficiency, as well as the wells-to-wheels efficiency, is significantly higher. The energy consumption per mile also reduces with increasing specific energy as seen in Fig. 9c. For specific energies greater than 350 Wh/kg, the improvements obtained by increasing specific energy in terms of cost and energy consumption saturate out. For a fixed C_d, the energy consumption of a Li-S system of 350 Wh/kg is only marginally higher than a Li-air system with twice the specific energy. The PTC for the Li-air system is reduced by about 0.05, Fig. 9b, while both systems would cost approximately the same, as seen in Fig. 9a.

Zoom In Zoom Out Reset image size

With the perspective of accelerating the electrification of the transportation sector in the LCV segment, once the specific energy improves such that the PTC of the system is about 0.15, the focus will have to shift toward vehicle design and improvements on battery manufacturing process, since system-level gains will be minimal. This calls for a need to pay attention to the materials used in fabricating battery systems of the future. The changes in system-level cost and PTC studied for increasing values of driving range can be seen in Fig. 10, where we observe similar trends of a level off that occurs after a certain specific energy. While higher specific energies would enable longer range with similar battery pack metrics, the improvements seen in driving range with increasing specific energy saturate beyond a specific energy of 350 Wh/kg. Based on Figs. 9 and 10, for the near future, if we assume a cost cap of approx. $35,000 per pack (with $150/kWh), with a C_d of 0.4, we can achieve a range of 300 miles with a 350 Wh/kg battery pack with a PTC of under 0.3 and decreasing the C_d would extend the range further. If we assume no cost cap, a practical specific energy of close to 200 Wh/kg, with advanced Li-ion batteries and Li-S systems, would enable a technically feasible system with a driving range of 500-miles and a PTC of under 0.5, but the pack would cost over $60,000. To re-iterate these targets in terms of cell-level specific energies, 200 Wh/kg translates to a 440 Wh/kg cell, and 350 Wh/kg translates to about 770 Wh/kg, assuming an f_burden factor of 0.45.

Zoom In Zoom Out Reset image size

Each of these beyond Li-ion systems currently entail their own challenges to assemble practical systems.^5,75,76 For example, future Li-ion systems with practical specific energies of greater than 200 Wh/kg would be achieved via high capacity anodes based on Li metal which has issues with dendrite growth and cycling²⁶ or with Si which currently suffers from volumetric expansion and rapid degradation.⁷⁷ The Li-S system suffers from challenges with polysulphide formation-dissolution, volumetric changes at the cathode, and stability of the electrolyte, and current targets of Li-S systems are still only comparable to Li-ion systems with high capacity anodes.⁷ The Li-air battery has issues with specific power,⁷⁵ rechargeability,⁵ and other challenges with finding stable electrolyte and cathode materials,⁷⁵ and currently the target is to achieve a specific energy of 600 Wh/kg,²⁶ the value which was used in our trade-off analysis in this study. Researchers are also attempting to develop other metal-air batteries such as Na-air,^80,81 where the state of understanding and research is at a very early stage,⁵ and hence such systems have not been included in this study.