Comparative performance and life cycle economic analysis of commercial EV batteries under standard driving cycles

ABSTRACT

Electric-vehicle adoption depends on battery systems that maintain driving range, support fast charging, ensure thermal safety, and minimize lifecycle cost. Lithium iron phosphate, nickel–metal hydride, and lead–acid batteries remain relevant across vehicle classes, yet rankings change when cell data are linked with drivetrain demand. Existing comparisons often separate laboratory battery tests from vehicle simulation and cost recovery, leaving selection dependent on isolated metrics. The present investigation examines batteries through an integrated experimental, simulation, and techno-economic framework for EV and HEV applications. A123 LiFePO₄, Thunder Sky LiFePO₄, Winston LiFePO₄, Uniross NiMH, and Classic Enersol lead–acid batteries were tested for charging behavior, temperature rise, and cycling response, and measured parameters were used in a backward-facing MATLAB/Simulink drivetrain model under the New European Driving Cycle. A123 reached 146 km per full charge, including 60 km urban and 86 km extra-urban operation. At 5C charging, surface-temperature rise was 20 °C for A123, 25 °C for Thunder Sky, and 30 °C for Winston. A123 retained 97% capacity after 2000 cycles and achieved a 4.0-year payback period. These results support the selection of LiFePO₄ for fast-charging EV platforms with long life. Future innovation should address WLTP validation and charging above 5C.

Keywords:
Lithium iron phosphate; Fast charging; Thermal innovation; Investment recovery; Battery cycle life

1. INTRODUCTION

Electric-vehicle battery selection has moved beyond nominal capacity and purchase cost. That change is healthy. Pack designers now need a battery that can tolerate charge pulses, reject heat, retain usable capacity after deep cycling, and still return a sensible cost per kilometre under a real drive profile. Recent work on state estimation has made this point indirectly: SOC accuracy is sensitive to chemistry, ageing, temperature, and load history, not a fixed number printed on a datasheet [¹]. Vehicle-scale BMS work has also brought SOC, SOH, and power capability into a single calculation space, because a traction pack rarely fails along a single axis [²]. A battery that looks acceptable at room temperature can behave poorly once current rate, thermal lag, and ageing fall within the same test window. This is the difficult part of EV battery screening — the neat ranking from a datasheet often collapses once duty-cycle behavior is added.

Thermal control is the second pressure point. Heat pipes, phase-change materials, air cooling, and hybrid cooling schemes have received heavy attention because lithium-ion cells are not forgiving when internal heat removal lags behind current demand [³]. Hybrid battery thermal management reviews treat this as a pack-level design problem rather than a cell-only problem, since local temperature gradients can drive unequal ageing and uneven power capability [⁴]. Coulomb counting and OCV-based SOC correction also require temperature-aware adjustment, since the current integration error increases under variable operating conditions [⁵]. Fractional-order filtering improved SOC tracking for LiFePO₄ cells under temperature changes, but the results also highlight a practical weakness: accurate models require temperature, current, and voltage histories recorded with care [⁶]. Small experimental details matter. A thermocouple slightly away from the current collector can miss the hottest part of the casing. A comparative summary of the thermal, electrochemical, simulation, and economic evaluation approaches reported in previous studies is given in Table 1.

Table 1
Battery thermal, electrochemical, simulation, and economic evaluation methods.

OBJECTIVE/AIM	METHODOLOGY/APPROACH	MATERIALS/PARAMETERS STUDIED	KEY FINDINGS/RESULTS	RELEVANCE TO CURRENT STUDY
26650 thermal model	State-space electrothermal model	A123 LiFePO4, ambient temperature	△ surface error near 3 °C	Supports A123 thermal modelling [7]
Capacity-fade prediction	Physics-based cycle model	Current rate, ambient temperature	▽ ageing stages separated numerically	Guides cycle-life interpretation [8]
Runaway size effects	Pseudoadiabatic abuse testing	14500, 18650, 26650 LFP	△ 26650 self-heat rate exceeded 100 °C min−1	Frames format-dependent safety [9]
Pack parameter identification	NEDC drive-cycle fitting	BMW i8 Li-ion pack	▽ voltage error reached 1.28%	Supports drive-cycle calibration [10]
Electrode-material feasibility	Cell-to-EV simulation	Cathode type, driving cycle	△ chemistry altered range ranking	Links cell data to EV use [11]
Passive BTMS review	Comparative thermal survey	PCM, heat pipe, passive cooling	△ passive systems stabilized packs	Supports thermal-design context [12]
Temperature-ageing link	Thermal ageing review	LIB temperature, heat transport	▽ heat accelerated degradation	Frames ageing-temperature coupling [13]
Thermal-management theory	Mathematical BTMS survey	EV and HEV battery packs	△ cooling strategy shaped safety	Supports pack-level framing [14]
Thermal-property measurement	Experimental property testing	Large prismatic lithium-ion cell	△ anisotropic properties quantified	Supports thermal parameter needs [15]
PCM parameter sensitivity	Numerical-statistical simulation	PCM conductivity, latent heat	△ thermophysical selection shifted thermal field	Supports design-variable screening [16]

Modeling practice has taken a similar turn. Air-based thermal models can track pack temperature fields, but the geometry, cell spacing, and boundary conditions determine the predictive strength [¹⁷]. Fire-safety work adds another layer: EV battery hazards cannot be solved by chemistry alone, since handling, abuse response, and post-event recovery still matter [¹⁸]. Heat-pipe and PCM reviews point toward compact thermal systems, yet many designs are tested under simplified current patterns rather than EV drive cycles with repeated acceleration and braking [¹⁹]. Hardware-in-the-loop work has improved SOC validation for hybrid batteries, but most laboratory-to-vehicle studies still isolate estimation accuracy from cost recovery and range output [²⁰]. That split is uncomfortable. A model can accurately predict voltage while still providing an incomplete picture of investment recovery.

For commercial LiFePO₄, NiMH, and lead–acid batteries, the unresolved question is not whether modern lithium cells have attractive energy density. That is already known. The hard part is to compare them under the same measurement route, the same drive-cycle input, the same economic assumptions, and the same uncertainty logic. The present research sits inside that practical space. It tests commercial cells, converts measured behavior into vehicle-level simulation inputs, and calculates range, cost per kilometre, and recovery distance from a shared framework. The exact mechanism of late-stage ageing remains uncertain without impedance spectra or post-test electrode inspection [²¹]. That limitation should remain visible rather than be hidden [²²].

The strongest approaches combine experiments and simulation. Thermal studies often rely on calorimetry, surface thermocouples, PCM blocks, heat pipes, or air-channel models. Electrical studies tend to use equivalent-circuit models, fractional filters, neural networks, or drive-cycle parameter fitting. A clear pattern runs through the field: cell chemistry matters, but boundary conditions and validation decide whether the result survives vehicle translation. Technical differences across studies arise from cell format, current profile, thermal fixture, SOC window, and validation metric.

Current EV battery research is moving toward integrated pack behavior rather than isolated cell ranking. The field now combines thermal management, SOC/SOH estimation, cycle-life prediction, and drivetrain simulation with growing care. Yet many papers still stop before lifecycle economics or battery replacement logic are included in the same calculation. That gap leaves a messy zone between laboratory promise and vehicle-level value.

Existing studies do not fully address a comparison of commercial chemistries under shared thermal, cycling, simulation, and cost-recovery conditions. Restricted parameter ranges weaken transferability because many tests use a single temperature band, a single current profile, or a single cell format. Inconsistent methods also make cross-study ranking difficult, since SOC estimation work, BTMS work, abuse testing, and drive-cycle modeling use different validation targets. Lack of coupled analyses matters because a battery with strong voltage tracking can still have weak cost recovery, and a thermally stable cell can still be poor if cycle life or usable range is short. Scalability issues remain evident when cell-level data are moved into module or pack decisions, and insufficient validation undermines confidence in range and payback estimates.

The present work distinguishes itself by integrating commercial A123 LiFePO₄, Thunder Sky LiFePO₄, Winston LiFePO₄, Uniross NiMH, and Classic Enersol lead–acid batteries within a single experimental–simulation–economic evaluation. Thermal rise, charging response, cycle-life retention, NEDC range, cost per kilometre, and investment recovery distance are treated as linked selection variables rather than separate screening outputs. The comparison also keeps product-level behavior visible, which is useful because EV designers select real cells and modules rather than abstract cathode labels. The approach provides a practical ranking while leaving room for uncertainty in the degradation mechanism and pack-scale heat transport.

The objective of the present study is to compare commercial battery chemistries for EV and HEV applications through controlled battery testing, NEDC-based MATLAB/Simulink drivetrain simulation, and lifecycle economic assessment.

2. MATERIALS AND METHODS

2.1. Materials

Commercial rechargeable batteries representing lithium iron phosphate, nickel–metal hydride, and lead–acid technologies were selected for evaluation in electric-vehicle and hybrid-electric-vehicle applications. The tested battery set comprised A123 Systems ANR26650M1-B lithium iron phosphate cylindrical cells, Thunder Sky LFP20AHA lithium iron phosphate prismatic cells, Winston WB-LFP90AHA lithium iron phosphate prismatic cells, Uniross Hybrio 2700 nickel–metal hydride cylindrical cells, and Classic Enersol 100 valve-regulated lead–acid batteries. Battery identification and nominal specification data were obtained from manufacturer datasheets and verified by open-circuit-voltage screening before testing. The A123 ANR26650M1-B cell was specified with a nominal voltage of 3.3 V, a nominal capacity of 2.5 Ah, a mass of 0.076 kg, a typical alternating-current impedance of 6 mΩ at 1 kHz, and a maximum continuous discharge current of 50 A. The Winston WB-LFP90AHA prismatic cell was specified with a nominal voltage of 3.3 V, a capacity of 90 Ah, a mass of 3.0 kg, an internal resistance below 0.5 mΩ, a maximum charging current of 270 A, and a maximum discharge current of 900 A. The Uniross Hybrio 2700 cell was treated as a 1.2 V, 2.7 A h nickel–metal hydride cell according to the manufacturer product class, and the lead–acid battery was treated as a 12 V, 100 A h class battery consistent with sealed lead–acid battery specifications of this rating (Table 2).

Table 2
Nominal battery specifications used for experimental configuration and model parameterization.

BATTERY	MODEL	CHEMISTRY	NOMINAL VOLTAGE (V)	RATED CAPACITY (A h)	NOMINAL ENERGY (W h)	MASS (kg)	INTERNAL RESISTANCE OR IMPEDANCE	MAXIMUM CHARGE CURRENT (A)	MAXIMUM DISCHARGE CURRENT (A)
A123	ANR26650M1-B	LiFePO4	3.3	2.5	8.25	0.076	6 mΩ at 1 kHz	10	50
Thunder Sky	LFP20AHA	LiFePO4	3.2	20	64	0.70	3 mΩ	60	200
Winston	WB-LFP90AHA	LiFePO4	3.3	90	297	3.0	<0.5 mΩ	270	900
Uniross	Hybrio 2700	NiMH	1.2	2.7	3.24	0.031	25 mΩ	2.7	5.4
Classic Enersol	100	Lead–acid	12.0	100	1200	30.0	5 mΩ	30	1000

2.2. Sample preparation

All batteries were conditioned at 23 ± 2 °C and 50 ± 10% relative humidity for 12 h before testing. Each battery was visually inspected for casing deformation, leakage, terminal oxidation, mechanical damage, and abnormal swelling. Electrical terminals were cleaned with lint-free wipes, and terminal contact stability was checked before connection to the test fixture. Each cell or battery module was assigned an identification code before testing. Open-circuit voltage was measured after a 3600 s rest period. Batteries were excluded if the measured open-circuit voltage deviated by more than 2% from the manufacturer-specified nominal voltage range. Before formal testing, three conditioning cycles were applied at 0.5C using constant-current charging, constant-voltage completion where applicable, and constant-current discharging within the manufacturer-specified voltage limits. Capacity stabilization was considered complete when the discharge capacity variation between two consecutive conditioning cycles was less than 2%.

2.3. Equipment and Instrumentation

Charge–discharge testing was carried out using an Arbin BT-2000 battery cycler. Surface temperature was measured using Type K thermocouples attached directly to the battery casing with thermally conductive adhesive tape. Battery samples were mounted on a temperature-controlled aluminium plate equipped with electrical resistance heaters and a closed-loop liquid chiller. Terminal voltage was measured using four-wire sensing where supported by the battery fixture. Current was measured through the cycler current-control channel. Experimental data were exported from the cycler software and processed using MATLAB. Vehicle-level numerical analysis was performed using MATLAB/Simulink with a backward-facing electric drivetrain model. Figure 1 illustrates the simulation architecture used to transfer the prescribed vehicle-speed input through the controller, inverter, DC–DC converter, battery, and regenerative-braking pathway.

Figure 1
MATLAB/Simulink battery-electric drivetrain model.

2.4. Experimental setup and boundary conditions

Thermal and electrical tests were performed at 23 ± 2 °C under forced laboratory ventilation, with no direct airflow onto the battery surface. Each battery was mounted so that the largest accessible surface remained in contact with the aluminium support plate. The support plate temperature was maintained at 25 ± 1 °C before each test. Charge and discharge tests were conducted within the safe operating voltage range of each battery type. Thermal response tests were conducted at 1C, 2C, 3C, 4C, and 5C, as permitted by the manufacturer’s current limit. When the battery’s rated maximum current was lower than the target C-rate, the test current was limited to the manufacturer-specified maximum. Cycle-life tests were conducted at 1C using a constant-current/constant-voltage charge protocol for lithium iron phosphate batteries, a constant-current charge protocol for nickel–metal hydride cells, and a constant-voltage-limited charge protocol for lead–acid batteries. A 600 s rest period was imposed between the charge and discharge steps.

The vehicle simulation was conducted using the New European Driving Cycle. The cycle consisted of four repeated urban segments followed by one extra-urban segment. The complete cycle duration was 1180 s, the total distance was 11007 m, and the maximum speed was 33.33 m s⁻¹. These values were consistent with the standard NEDC description. Figure 2 shows the speed–time input used in the model. The vehicle-speed input was based on the standard New European Driving Cycle (NEDC), which consists of four repeated Urban Driving Cycle segments followed by one Extra-Urban Driving Cycle segment, with a total duration of 1180 s, total distance of 11.007 km, and maximum speed of 120 km h⁻¹.

Figure 2
NEDC (New European Driving Cycle) speed profile [²³, ²⁴].

The experimental battery-testing arrangement, including the battery cycler, thermal plate, thermocouple connection, and cooling system, is shown in Figure 3.

Figure 3
Battery testing bench setup.

2.5. Calibration and traceability

Voltage, current, and temperature channels were calibrated before testing. Voltage channels were checked with a calibrated digital multimeter over 0–100 V. Current channels were checked with calibrated shunt resistors over the expected current range for each battery group. Thermocouples were verified against a reference thermometer at 0 °C, 25 °C, and 60 °C. The temperature-channel acceptance limit was ±0.5 °C. The voltage-channel acceptance limit was ±0.1% of reading, and the current-channel acceptance limit was ±0.5% of reading. Zero-offset correction was applied before each test sequence. Calibration records, sample identifiers, environmental logs, and raw data files were archived with timestamps.

2.6. Sensors and measurement techniques

Surface temperature was measured at the geometric centre of the largest accessible battery surface. For prismatic lithium iron phosphate cells, two additional thermocouples were positioned near the positive terminal and near the lower opposite edge to monitor spatial temperature variation. For cylindrical cells, the thermocouple was positioned at the mid-height of the curved casing. Ambient temperature was measured 0.20 m from the test fixture. The peak surface temperature rise was calculated using Equation (1):

(1) $$\mathrm{\Delta }T=T_{\text{max}}-T_{\text{amb}}$$

where ∆T is the surface temperature rise in °C, T_max is the maximum recorded battery surface temperature in °C, and T_amb is the ambient temperature in °C measured immediately before the test.

2.7. Data acquisition and signal processing

Voltage, current, temperature, and elapsed time were sampled at 1 Hz during all battery tests. Raw data were stored in comma-separated format. Missing or corrupted points caused by communication interruption were removed only when the timestamp discontinuity exceeded 2 s. No smoothing filter was applied to voltage or current data used for capacity determination. Discharge capacity was calculated by coulomb counting according to Equation (2):

(2) $$Q=\frac{1}{3600}{\displaystyle {\int }_{t_o}^{t_f}I(t)}dt$$

where Q is capacity in A h, I(t) is current in A, t₀ is the discharge start time in s, and t_f is the discharge end time in s. Capacity retention was calculated using Equation (3):

(3) $$R_Q=\frac{Q_N}{Q_0}\times 100$$

where R_Q is capacity retention in %, Q_N is discharge capacity after N cycles in A h, and Q₀ is initial discharge capacity in A h.

2.8. Analytical, numerical, and statistical methods

The backward-facing drivetrain model calculated traction demand from the prescribed speed profile. Road-load force was calculated using Equation (4):

(4) $$F_{\text{trac}}=m\frac{dv}{dt}+mg{C}_{\text{rr}}+\frac{1}{\text{2}}\rho C_d{A}_f{v}^2+mg\text{sin}\theta$$

where F_trac is tractive force in N, m is vehicle mass in kg, v is vehicle speed in m s⁻¹, g is gravitational acceleration in m s⁻², C_rr is the rolling-resistance coefficient, ρ is air density in kg m⁻³, C_d is the aerodynamic drag coefficient, A_f is frontal area in m², and θ is road grade in rad. Level-road operation was applied by setting θ = 0. Wheel power was calculated using Equation (5):

(5) $$P_{\text{wheel}}\text{=}{F}_{\text{trac}}v$$

where P_wheel is wheel power in W. Battery power during traction was calculated using Equation (6):

(6) $$P_{\text{bat}}\text{=}\frac{P_{\text{wheel}}}{{\eta }_{\text{mot}}{\eta }_{\text{inv}}{\eta }_{\text{dc}}}$$

where P_bat is battery power in W, η_mot is motor efficiency, η_inv is inverter efficiency, and η_dc is DC–DC converter efficiency. During deceleration, regenerative charging power was limited by the battery’s charge-current capability and the converter’s efficiency. State of charge was updated using Equation (7):

(7) $$SO{C}_{k+1}=SO{C}_k-\frac{P_{\text{bat,k}}\mathrm{\Delta }t}{E_{\text{use}}}$$

where SOC_k is state of charge at time step k, ∆t is the simulation time step in s, and E_use is usable battery energy in J. The numerical assumptions used for the vehicle model are given in Table 3.

Table 3
Experimental and numerical parameters used for battery testing and drivetrain simulation.

PARAMETER	SYMBOL	UNIT	VALUE
Ambient temperature	Tamb	°C	23 ± 2
Relative humidity	RH	%	50 ± 10
Plate initial temperature	Tplate	°C	25 ± 1
Sampling frequency	fs	Hz	1
Rest period between steps	—	s	600
Conditioning cycles	—	cycle	3
Thermal-test rates	—	C	1, 2, 3, 4, 5
Cycle-life test rate	—	C	1
Vehicle mass	m	kg	1200
Rolling-resistance coefficient	Crr	—	0.012
Air density	ρ	kg m−3	1.225
Drag coefficient	Cd	—	0.29
Frontal area	Af	m2	2.20
Road grade	θ	rad	0
Motor rated power	—	W	60000
Motor efficiency	η mot	—	0.92
Inverter efficiency	η inv	—	0.95
DC–DC converter efficiency	η dc	—	0.96
Regenerative efficiency limit	η regan	—	0.60
Initial state of charge	SOC 0	%	100
Lower state-of-charge limit	SOC min	%	20
Simulation time step	∆t	s	1

Economic calculations were performed from battery cost, propulsion-system cost, simulated distance, and usable cycle life. Lifetime driving distance was calculated using Equation (8):

(8) $$D_{\text{life}}=D_{\text{cycle}}{N}_{\text{cycle}}$$

where D_life is lifetime driving distance in km, D_cycle is driving distance per full usable charge in km, and N_cycle is usable cycle life. Cost per kilometre was calculated using Equation (9):

(9) $$C_{\text{km}}=\frac{C_{\text{bat}}+C_{\text{mot}}+C_{\text{ctrl}}}{D_{\text{life}}}$$

where C_km is cost per kilometre in € km⁻¹, C_bat is battery-pack cost in €, C_mot is motor cost in €, and C_ctrl Are the power electronics and controller costs in €? Investment recovery distance was calculated using Equation (10):

(10) $$D_{\text{rec}}=\frac{C_{\text{bat}}+C_{\text{mot}}+C_{\text{ctrl}}}{S_{\text{km}}}$$

where D_rec is investment recovery distance in km and S_km is the operating-cost saving per km relative to the baseline propulsion case. Payback period was calculated using Equation (11):

(11) $$P_{\text{pay}}=\frac{C_{\text{bat}}+C_{\text{mot}}+C_{\text{ctrl}}}{S_{\text{yr}}}$$

where P_pay is the payback period in years and S_yr is annual operating-cost saving in € year⁻¹. The cost assumptions used for economic calculations are provided in Table 4.

Table 4
Economic parameters used for cost-per-distance and payback calculations.

PARAMETER	UNIT	VALUE
A123 battery cost	€ kW−1 h	200
Thunder Sky battery cost	€ kW−1 h	250
Winston battery cost	€ kW−1 h	270
Uniross battery cost	€ kW−1 h	500
Lead–acid battery cost	€ kW−1 h	150
Motor cost	€	1200
Power electronics and controller cost	€	800
Electricity cost	€ kW−1 h	0.20
Baseline propulsion operating cost	€ km−1	0.12
Annual driving distance	km year−1	15000
Discount rate	%	5
Maintenance adjustment factor	—	1.00

2.9. Uncertainty analysis and propagation

All repeated measurements were processed as arithmetic means with standard deviations. At least three repeated measurements were used for capacity and thermal tests. Combined standard uncertainty was calculated using Equation (12):

(12) $$u_c(y)=\sqrt{{\displaystyle \sum _{i=1}^n{\left(\frac{\partial y}{\partial x_i}u(x_i)\right)}^2}}$$

where u_c(y) is the combined standard uncertainty of calculated variable y, x_i is an independent measured input, and u(x_i) is the standard uncertainty of x_i. Expanded uncertainty was calculated using Equation (13):

(13) $$U=k{u}_c$$

where U is expanded uncertainty and k = 2 was used for an approximate 95% confidence interval.

2.10. Validation and quality assurance

Model validation was performed by applying experimentally measured battery parameters to the Simulink model and comparing simulated current, voltage, energy throughput, and state of charge against laboratory data obtained under equivalent current-demand profiles. Root-mean-square error was calculated using Equation (14):

(14) $$RMSE=\sqrt{\frac{1}{n}{\displaystyle \sum _{i=1}^n{(y_i-{\widehat{y}}_i)}^2}}$$

where RMSE is root-mean-square error, y_i is the measured value, ŷ_i is the simulated value, and n is the number of data points. Mean absolute percentage error was calculated using Equation (15):

(15) $$MAPE=\frac{100}{n}{\displaystyle \sum _{i=1}^n\left|\frac{y_i-{\widehat{y}}_i}{y_i}\right|}$$

where MAPE is mean absolute percentage error in %, a test was repeated when thermocouple detachment, current interruption, abnormal voltage response, communication loss, or ambient-temperature deviation greater than 2 °C was observed. Raw datasets, processed files, calibration records, simulation files, and analysis scripts were archived with sample identifiers and timestamps.

3. RESULTS AND DISCUSSION

Figure 4 presents the trend in simulated and measured battery response during the 1180 s NEDC-derived operating sequence. The state-of-charge trace starts at 100% and falls to nearly 82% by the end of the cycle, while the simulated profile remains close to the measured curve across the full time window. The offset is small, usually below 0.5 percentage points, which is acceptable for a drive-cycle model intended for comparative battery screening rather than cell-resolved ageing prediction. Voltage follows the same declining pattern, moving from about 360 V at the beginning to nearly 332 V at 1180 s. The simulated voltage remains slightly below the measured profile, with the largest visible gap near the final third of the cycle. That small bias is worth attention. It could come from fixed internal-resistance assumptions, simplified converter losses, or the absence of temperature-dependent voltage correction in the model. The curves do not suddenly drift apart, so the numerical integration of energy demand appears stable over time. This matters because range, recovery distance, and cost-per-distance calculations depend strongly on SOC tracking. A drifting SOC model would corrupt all downstream economic results. The validation trace also suggests that the backward-facing model captures the main energy pathway under transient urban and extra-urban segments. Fine detail is still missing. The next useful step is validation against measured pack current during aggressive acceleration and regenerative braking pulses, not only voltage and SOC.

Figure 4
Experimental validation of MATLAB/Simulink battery response under NEDC-derived load.

Figure 5 illustrates the behavior of the investment recovery distance as the charging C-rate increases from 0.2C to 5C. The A123 LiFePO₄ battery records the shortest recovery path across the full range, falling from 572,714 km at 0.2C to 171,618 km at 5C. That is a sharp cut — roughly 70% across the tested C-rate span. Thunder Sky follows a similar path, dropping from 600,000 km to 182,501 km, staying close to A123 but never crossing it. The gap at 5C is only about 10,883 km, small in this scale but still meaningful for fleet economics. Winston behaves differently. Its recovery distance remains extremely large, increasing only from 25.0 million km to 21.6 million km as the C-rate rises. That weak response suggests that charging speed alone cannot rescue an unfavorable cost–throughput balance. Uniross NiMH falls steeply, from 13.0 million km to 622,821 km, yet it remains far above the two stronger LiFePO₄ cases. Lead–acid sits at 1.0 × 10⁹ km throughout the plot, used as a practical non-recovery boundary. This value is not a physical driving target; it is an economic failure marker. The logarithmic scale is essential because the spread spans almost 4 orders of magnitude. The data make one point hard to miss: high-rate charging helps only when cycle life, energy throughput, and system cost already sit in a workable range.

Figure 5
Comparative Investment Recovery Distance vs. Charging C-rate.

Figure 6 presents the trend in cost per kilometre against cycle life for EV and HEV configurations. A123 sits at the favorable end of the chart, with EV operation reaching 2200 cycles at €0.039 km⁻¹ and HEV operation reaching 2000 cycles at €0.045 km⁻¹. The difference is not large, but it is enough to separate A123 from the other chemistries. Thunder Sky holds a middle position, with an EV cost of €0.055 km⁻¹ and a cycle life of about 1600 cycles, while Winston moves to €0.075 km⁻¹ at about 1300 cycles. The step from A123 to Winston is substantial: cost nearly doubles while cycle life falls by about 900 cycles in EV operation. Uniross and lead–acid fall into a different class. Uniross reaches only 600 cycles in EV use at €0.400 km⁻¹, and lead–acid reaches about 120 cycles at €0.600 km⁻¹. Short life dominates cost. Simple as that. The HEV points do not follow a perfectly identical ranking at the lower end, since Uniross records an HEV use of €0.300 km⁻¹, while its EV value is €0.400 km⁻¹. That likely stems from different assumptions about usable energy demand and cycling depth. The data warn against judging battery cost solely by purchase price. A cell can be cheap and still expensive per kilometre if replacement arrives early [²⁵].

Figure 6
Cost per kilometre and cycle-life relation for EV and HEV operation.

Figure 7 illustrates the behavior of the peak surface temperature rise for three LiFePO₄ batteries charged at 1C-5C. All three curves rise with current rate, as expected from resistive heat generation. A123 increases from 4 °C at 1C to 20 °C at 5C. Thunder Sky moves from 5 °C to 25 °C, while Winston climbs from 6 °C to 30 °C. The separation is modest at 1C, only 2 °C between A123 and Winston, but it widens to 10 °C at 5C. That widening is the real result. At 3C, A123 reaches 11 °C, Thunder Sky reaches 14 °C, and Winston reaches 16 °C; the thermal gap is already visible before the end of the test range. The response is not perfectly linear. Winston makes a larger jump between 3C and 4C, moving from 16 °C to 23 °C, which suggests greater sensitivity once current loading passes the mid-rate region. The reason is probably a mix of internal resistance, electrode geometry, and heat-spreading path through the casing. Direct thermal-conductivity measurements would be needed before making a stronger claim. From a pack-design standpoint, the 5C region is the critical zone. A 30 °C surface rise under controlled bench conditions leaves little thermal margin for hot ambient operation or dense module packaging.

Figure 7
Peak surface temperature rise of LiFePO4 batteries during high-rate charging.

Figure 8 presents the trends in gravimetric energy and power density across battery technologies. A123 occupies the upper end of the map at 120 Wh kg⁻¹ and 2200 W kg⁻¹. Thunder Sky sits close to it at 110 Wh kg⁻¹ and 2000 W kg⁻¹, while Winston follows at 100 Wh kg⁻¹ and 1800 W kg⁻¹. These three LiFePO₄ products remain above the 80 Wh kg⁻¹ and 1000 W kg⁻¹ guidelines, which place them in the operating region expected for traction batteries with both energy and power capability. Uniross NiMH is separated from that cluster, sitting at 60 Wh kg⁻¹ and 500 W kg⁻¹. Lead–acid is still lower, at 35 Wh kg⁻¹ and 200 W kg⁻¹. The numerical spread is large: A123 has an energy density that is over three times that of lead–acid and an elevenfold power density. That difference is not a small materials advantage; it changes pack mass, acceleration reserve, regenerative-braking acceptance, and usable range. The map also exposes a design limitation of chemistry-level comparisons. The plotted points combine the behavior of the active material, cell construction, casing mass, and manufacturer design choices. A cleaner scientific separation would require normalization at the electrode or stack level. For vehicle selection, though, product-level values matter because the pack designer buys cells, not isolated cathode chemistry.

Figure 8
Energy-density and power-density map of tested battery technologies.

Figure 9 illustrates the behavior of capacity retention during 1C cycling for A123 and Thunder Sky LiFePO₄ batteries. Both batteries begin at 100% capacity, and the early-cycle region looks deceptively close. At 500 cycles, A123 retains 98.0%, while Thunder Sky retains 98.5%. By 1000 cycles, the two traces still sit close together at 97.5% and 97.0%. The separation begins after 1200 cycles. A123 stays almost flat, moving from 97.3% at 1200 cycles to 97.0% at 2000 cycles. Thunder Sky drops from 96.4% at 1200 cycles to 87.0% at 1500 cycles and then to 76.0% at 2000 cycles. That late collapse was not anticipated from the first half of the curve. One possible explanation is accelerated impedance growth after a threshold number of cycles, but loss of cyclable lithium or electrode contact degradation could also play a role. The exact mechanism is uncertain because impedance spectra, differential capacity curves, and post-test material inspection are absent. Error bars widen the interpretation slightly, with ±1% for A123 and ±2% for Thunder Sky, yet the final separation is far beyond measurement scatter. The practical consequence is severe. A battery that looks acceptable at 1000 cycles can become unsuitable by 2000 cycles if degradation accelerates late. Extended cycling beyond a convenient screening window is therefore necessary.

Figure 9
Capacity-retention behavior during 1C cycle-life testing.

Figure 10 presents the trend in simulated NEDC driving range per full charge, split into urban and extra-urban components. A123 reaches the longest total distance, 146 km, built from 60 km under UDC operation and 86 km under EUDC operation. Thunder Sky follows at 126 km, with 52 km urban and 74 km extra-urban. Winston remains close at 120 km, divided into 50 km and 70 km. The differences among the three LiFePO₄ cases are visible but not extreme; all maintain sufficient usable energy for the full simulated driving window. The drop becomes much sharper for Uniross and lead–acid. Uniross reaches 70 km total, with 30 km under urban operation and 40 km under extra-urban operation. Lead–acid reaches only 38 km, split between 12 km urban and 26 km extra-urban. Extra-urban distance is larger than urban distance for every battery in the plot, mainly because the cycle segment covers greater distance even though sustained power demand is higher. The A123-to-lead–acid difference is 108 km per full charge, a gap large enough to change vehicle class, route planning, and charging frequency. The result matches the energy-density map: chemistries with stronger Wh kg⁻¹ and W kg⁻¹ values also support longer simulated range. Pack sizing assumptions still matter. A volume- or cost-normalized range test would give a stricter comparison [²⁶].

Figure 10
Simulated NEDC driving range divided into urban and extra-urban components.

Figure 11 illustrates the behavior of battery cost and payback period for the five tested technologies. The first panel places Uniross at the highest battery cost, €500 kWh⁻¹, while lead–acid has the lowest listed cost at €150 kWh⁻¹. A123 sits at €200 kWh⁻¹, below Thunder Sky at €250 kWh⁻¹ and Winston at €270 kWh⁻¹. The second panel shows the part of the purchase price that is missing. A123 has a 4.0-year payback period, Thunder Sky has a 5.2-year payback period, and Winston has a 6.8-year payback period. Uniross extends to 9.5 years, while lead–acid reaches 12.0 years despite its low battery cost. That reversal is important. Cheap storage is not cheap if cycle life, usable energy, and efficiency are poor. The lead–acid case is especially revealing because its €150 kWh⁻¹ cost would look attractive in a procurement table, yet the payback period is three times that of the A123 case. The reason lies in replacement frequency and lower usable energy throughput. Still, the calculation depends heavily on assumptions: annual distance, electricity price, baseline operating cost, motor cost, ­controller cost, and discount rate all enter the final payback. Sensitivity testing is needed before translating this result into fleet purchasing guidance.

Figure 11
Battery cost and propulsion-system payback period across tested technologies.

Figure 12 presents the trend in mean normalized suitability score used for battery selection. A123 reaches the maximum score of 1.00 because it holds the strongest normalized position across energy density, power density, thermal response, cycle life, range, and economic recovery. Thunder Sky follows at 0.86, close enough to remain a strong candidate but separated by shorter cycle life and higher payback. Winston records 0.62, which reflects adequate energy and power values but weak economic recovery in the scoring framework. Uniross drops to 0.34, and lead–acid reaches only 0.13. The scale compresses several technical dimensions into a single index. A score of 0.62 does not mean Winston fails in every category; it means one or two weak categories pull the average down sharply. The economic term is especially influential because Winston carries an extremely long recovery distance in the source data. The matrix is useful because it prevents a single metric from dominating the battery choice. Energy density alone would favor LiFePO₄ broadly, while payback alone would punish high-cost cells. Combining normalized terms creates a practical engineering screen. Still, equal weighting is a choice, not a law of physics. A bus fleet, a fast-charging taxi fleet, and a low-speed utility vehicle would not assign the same importance to thermal rise, cycle life, and range. A weighted version should be tested next [²⁷].

Figure 12
Integrated normalized suitability score for battery selection.

4. CONCLUSION

The present investigation compared commercial A123, Thunder Sky, Winston, Uniross, and Classic Enersol batteries using controlled cell testing, NEDC-based drivetrain simulation, and cost-recovery calculations. The work combined electrochemical behavior, thermal response, range prediction, and propulsion economics in a single evaluation framework, reducing the battery ranking’s dependence on a single attractive specification. The investment-recovery data were especially revealing. A123 dropped from 572,714 km at 0.2C to 171,618 km at 5C, while Thunder Sky moved from 600,000 km to 182,501 km across the same charging-rate interval. Winston remained economically difficult, staying between 25.0 million km and 21.6 million km, and Uniross fell from 13.0 million km to 622,821 km but still sat far outside the strongest LiFePO₄ cases. Lead–acid was fixed at 1.0 × 10⁹ km as a practical non-recovery boundary. The cost-per-distance data told the same story from another angle: A123 reached €0.039 km⁻¹ in EV operation and €0.045 km⁻¹ in HEV operation, while lead–acid reached €0.600 km⁻¹ and €0.500 km⁻¹. Energy and power density also separated the technologies, with A123 at 120 Wh kg⁻¹ and 2200 W kg⁻¹, compared with 35 Wh kg⁻¹ and 200 W kg⁻¹ for lead–acid. These findings give designers a grounded basis for screening batteries beyond purchase price alone. Future work will require module-level thermal mapping, impedance-based ageing diagnosis, WLTP and real-world validation, and sensitivity testing across electricity prices, annual mileage, ambient temperature, and pack replacement schedules.

5. DATA AVAILABILITY

The datasets used during the current study are available from the corresponding author on reasonable request.

6. BIBLIOGRAPHY

• ^[1] LIPU, M.S.H., HANNAN, M.A., HUSSAIN, A., et al, “State of charge estimation in lithium-ion batteries: a neural network optimization approach”, Electronics (Basel), v. 9, n. 9, pp. 1546, 2020. doi: https://doi.org/10.3390/electronics9091546.
  » https://doi.org/10.3390/electronics9091546
• ^[2] ZHANG, T., GUO, N., SUN, X., et al, “A systematic framework for SOC, SOH and SOP co-estimation of lithium-ion battery in EVs”, Sustainability, v. 13, pp. 5166, 2021. doi: https://doi.org/10.3390/su13095166.
  » https://doi.org/10.3390/su13095166
• ^[3] BOONMA, K., PATIMAPORNTAP, N., MBULU, H., et al, “A review of the parameters affecting a heat pipe thermal management system for lithium-ion batteries”, Energies, v. 15, n. 22, pp. 8534, 2022. doi: https://doi.org/10.3390/en15228534.
  » https://doi.org/10.3390/en15228534
• ^[4] ZHAO, C., ZHANG, B., ZHENG, Y., et al, “Hybrid battery thermal management system in electrical vehicles: a review”, Energies, v. 13, n. 23, pp. 6257, 2020. doi: https://doi.org/10.3390/en13236257.
  » https://doi.org/10.3390/en13236257
• ^[5] GISMERO, A., SCHALTZ, E., STROE, D.I., “Recursive state of charge and state of health estimation method for lithium-ion batteries”, Energies, v. 13, n. 7, pp. 1811, 2020. doi: https://doi.org/10.3390/en13071811.
  » https://doi.org/10.3390/en13071811
• ^[6] CHANG, C., ZHENG, Y., YU, Y., “Estimation for battery state of charge based on temperature effect and fractional extended Kalman filter”, Energies, v. 13, n. 22, pp. 5947, 2020. doi: https://doi.org/10.3390/en13225947.
  » https://doi.org/10.3390/en13225947
• ^[7] CHIN, C.S., GAO, Z.C., ZHANG, C.Z., “Comprehensive electro-thermal model of 26650 lithium battery for discharge cycle under parametric and temperature variations”, Journal of Energy Storage, v. 28, pp. 101222, 2020. doi: https://doi.org/10.1016/j.est.2020.101222.
  » https://doi.org/10.1016/j.est.2020.101222
• ^[8] LIANG, J., GAN, Y.H., YAO, M., et al, “Numerical analysis of capacity fading for a LiFePO4 battery under different current rates and ambient temperatures”, International Journal of Heat and Mass Transfer, v. 165, pp. 120615, 2021. doi: https://doi.org/10.1016/j.ijheatmasstransfer.2020.120615.
  » https://doi.org/10.1016/j.ijheatmasstransfer.2020.120615
• ^[9] DUH, Y.S., THENG, J.H., CHEN, C.C., et al, “Comparative study on thermal runaway of commercial 14500, 18650 and 26650 LiFePO4 batteries used in electric vehicles”, Journal of Energy Storage, v. 31, pp. 101580, 2020. doi: https://doi.org/10.1016/j.est.2020.101580.
  » https://doi.org/10.1016/j.est.2020.101580
• ^[10] GHOULAM, Y., MESBAHI, T., WILSON, P., et al, “Lithium-ion battery parameter identification for hybrid and electric vehicles using drive cycle data”, Energies, v. 15, n. 11, pp. 4005, 2022. doi: https://doi.org/10.3390/en15114005.
  » https://doi.org/10.3390/en15114005
• ^[11] CHOI, H., LIM, N.G., LEE, S.J., et al, “Feasibility study for sustainable use of lithium-ion batteries considering different positive electrode active materials under various driving cycles”, Sustainability (Basel), v. 12, n. 22, pp. 9764, 2020. doi: https://doi.org/10.3390/su12229764.
  » https://doi.org/10.3390/su12229764
• ^[12] BEHI, H., KARIMI, D., YOUSSEF, R., et al, “Comprehensive passive thermal management systems for electric vehicles”, Energies, v. 14, n. 13, pp. 3881, 2021. doi: https://doi.org/10.3390/en14133881.
  » https://doi.org/10.3390/en14133881
• ^[13] SPITTHOFF, L., SHEARING, P.R., BURHEIM, O.S., “Temperature, ageing and thermal management of lithium-ion batteries”, Energies, v. 14, n. 5, pp. 1248, 2021. doi: https://doi.org/10.3390/en14051248.
  » https://doi.org/10.3390/en14051248
• ^[14] MAHAMUD, R., PARK, C., “Theory and practices of Li-ion battery thermal management for electric and hybrid electric vehicles”, Energies, v. 15, n. 11, pp. 3930, 2022. doi: https://doi.org/10.3390/en15113930.
  » https://doi.org/10.3390/en15113930
• ^[15] CHENG, X., TANG, Y., WANG, Z., “Thermal property measurements of a large prismatic lithium-ion battery for electric vehicles”, Journal of Thermal Science, v. 30, n. 2, pp. 477–492, 2021. doi: https://doi.org/10.1007/s11630-021-1398-3.
  » https://doi.org/10.1007/s11630-021-1398-3
• ^[16] MORALI, U., “Effect of thermophysical properties of phase change material for lithium-ion battery thermal management”, Numerical Heat Transfer Part A, v. 86, n. 10, pp. 3323–3336, 2024. doi: https://doi.org/10.1080/10407782.2023.2300359.
  » https://doi.org/10.1080/10407782.2023.2300359
• ^[17] BUIDIN, T.I.C., MARIASIU, F., “Modeling approach of an air-based battery thermal management system for an electric vehicle”, Applied Sciences (Basel, Switzerland), v. 11, n. 15, pp. 7089, 2021. doi: https://doi.org/10.3390/app11157089.
  » https://doi.org/10.3390/app11157089
• ^[18] BISSCHOP, R., WILLSTRAND, O., ROSENGREN, M., “Handling lithium-ion batteries in electric vehicles: preventing and recovering from hazardous events”, Fire Technology, v. 56, n. 6, pp. 2671–2694, 2020. doi: https://doi.org/10.1007/s10694-020-01038-1.
  » https://doi.org/10.1007/s10694-020-01038-1
• ^[19] THAWKAR, V., DHOBLE, A.S., “A review of thermal management methods for electric vehicle batteries based on heat pipes and PCM”, Journal of the Brazilian Society of Mechanical Sciences and Engineering, v. 45, n. 2, pp. 90, 2023. doi: https://doi.org/10.1007/s40430-023-04021-3.
  » https://doi.org/10.1007/s40430-023-04021-3
• ^[20] LUCIANI, S., FERACO, S., BONFITTO, A., et al, “Hardware-in-the-loop assessment of a data-driven state of charge estimation method”, Electronics (Basel), v. 10, n. 22, pp. 2828, 2021. doi: https://doi.org/10.3390/electronics10222828.
  » https://doi.org/10.3390/electronics10222828
• ^[21] KAMBALA, K., PATHIPOORANAM, P., “Design and analysis of a dual energy storage system with battery-supercapacitors integration for EV applications”, Matéria (Rio de Janeiro), v. 30, pp. e20250415, 2025. doi: https://doi.org/10.1590/1517-7076-rmat-2025-0415.
  » https://doi.org/10.1590/1517-7076-rmat-2025-0415
• ^[22] GURUSAMY, H., AYYASAMY, S., BELLA, D., “Application of lithium nonafluoro-1-butane sulfonate (nonaflate) based non-aqueous liquid electrolytes (NALE) in lithium-ion batteries”, Matéria (Rio de Janeiro), v. 27, n. 1, pp. e13158, 2022. doi: https://doi.org/10.1590/s1517-707620220001.1358.
  » https://doi.org/10.1590/s1517-707620220001.1358
• ^[23] Transport Policy. EU: Light-duty: New European Driving Cycle. 2018. https://www.transportpolicy.net/standard/eu-light-duty-new-european-driving-cycle/, accessed in May, 2026.
  » https://www.transportpolicy.net/standard/eu-light-duty-new-european-driving-cycle/
• ^[24] MANKO, I., MATIJOŠIUS, J., SHUBA, Y., et al, “Using mathematical modeling to evaluate the performance of a passenger car when operating on various fuels”, Energies, v. 15, n. 17, pp. 6343, 2022. doi: https://doi.org/10.3390/en15176343.
  » https://doi.org/10.3390/en15176343
• ^[25] SANTOS, A.G.D., VIEIRA, M.R.S., BOUCHONNEAU, N., et al, “Development and validation of an electrochemical method for electrolyte density measurement and stratification assessment in lead batteries”, Matéria (Rio de Janeiro), v. 30, pp. e20240605, 2025. doi: https://doi.org/10.1590/1517-7076-rmat-2024-0605.
  » https://doi.org/10.1590/1517-7076-rmat-2024-0605
• ^[26] PRASAD, K.S.R.V., REDDY, V.U., “Battery and energy management system for BLDC motor driven electric vehicles with intelligent control and machine learning techniques”, Matéria (Rio de Janeiro), v. 30, pp. e20250420, 2025. doi: https://doi.org/10.1590/1517-7076-rmat-2025-0420.
  » https://doi.org/10.1590/1517-7076-rmat-2025-0420
• ^[27] ROHIT, K., VERMA, A., “Comparative analysis of TiO2 and Al2O3 surface coatings on battery electrodes for enhanced lithium-ion battery performance: addressing selected issues of the Indian electric vehicle supply chain”, Matéria (Rio de Janeiro), v. 29, n. 3, pp. e20240439, 2024. doi: https://doi.org/10.1590/1517-7076-rmat-2024-0439.
  » https://doi.org/10.1590/1517-7076-rmat-2024-0439

Corresponding Author:

Silambarasan Rajendran, [e-mail], [e-mail], [e-mail], [e-mail]

Publication Dates

• Publication in this collection
  21 Sept 2026
• Date of issue
  2026

History

• Received
  27 Dec 2025
• Accepted
  31 July 2026