The rapid expansion of the global electric vehicle (EV) market has brought the issues of driving range, charging time, and charging anxiety to the forefront of consumer concerns. As EV battery pack technologies evolve toward higher energy densities and faster charging capabilities, the demand for ultra-fast charging has become an urgent industry trend. However, high-rate fast charging generates enormous heat inside lithium-ion batteries, causing rapid temperature rises and increasing the risk of thermal runaway. In this context, a well-designed battery thermal management system (BTMS) is crucial to ensure the safe operation, lifetime, and performance of the EV battery pack. This study addresses the thermal management design optimization of an EV battery pack for ultra-fast charging, aiming at enhancing the cooling capability, temperature uniformity, and system-level efficiency. A hybrid approach combining theoretical analysis, experimental characterization, and computational fluid dynamics (CFD) simulations is employed throughout the research.
1. Heat Generation Characteristics of Lithium-Ion Batteries
The lithium-ion cell selected in this work is a 109 Ah prismatic nickel-cobalt-manganese (NCM) ternary cell, which is widely adopted in modern electric vehicles for its high energy density and relatively mature manufacturing process. The internal structure of a lithium-ion cell consists of positive and negative electrodes, current collectors, separators, electrolyte, and a metallic casing. During charge and discharge, lithium ions shuttle between the electrodes through the electrolyte, and electrons travel through the external circuit, resulting in electrochemical reactions that are accompanied by heat generation. The total heat generated inside a cell can be categorized into Joule heat, polarization heat, reaction heat, side-reaction heat, and mixing heat. Under normal operation, the mixing heat is usually negligible; thus, the total heat generation rate can be estimated with the Bernardi model:
$$q_{\text{gen}} = \frac{I}{V_{\text{cell}}} \left[ (U_{\text{OCV}} – U_t) + T \frac{dU_{\text{OCV}}}{dT} \right]$$
where \(q_{\text{gen}}\) is the volumetric heat generation rate (W/m³), \(I\) is the current (A), \(V_{\text{cell}}\) is the cell volume (m³), \(U_{\text{OCV}}\) is the open-circuit voltage (V), \(U_t\) is the terminal voltage (V), and \(T\) is the absolute temperature (K). The term \(dU_{\text{OCV}}/dT\) is the entropic heat coefficient. To obtain accurate values for the internal resistance and open-circuit voltage at various temperatures and states of charge (SOC), a series of Hybrid Pulse Power Characterization (HPPC) tests were conducted in a temperature-controlled chamber. The battery was conditioned at three ambient temperatures (283 K, 300 K, and 313 K), and the current pulses were applied at every 10% SOC interval. From the voltage response curves, the ohmic and polarization resistances were extracted according to the following relations:
$$R_{\text{ohm}}^{\text{dis}} = \frac{U_1 – U_2}{I_{\text{dis}}}, \quad R_{\text{polar}}^{\text{dis}} = \frac{U_2 – U_3}{I_{\text{dis}}}$$
$$R_{\text{tot}}^{\text{dis}} = R_{\text{ohm}}^{\text{dis}} + R_{\text{polar}}^{\text{dis}}$$
Similar equations were used for the charge pulses. The measurement results demonstrate that the internal resistance strongly depends on temperature and SOC. At lower temperatures, the ionic conductivity of the electrolyte decreases, leading to much higher internal resistances. For example, at 5% SOC, the total resistance is only 1.27 mΩ at 313 K but increases to 5.67 mΩ at 283 K. In contrast, the open-circuit voltage is only weakly influenced by temperature, but varies significantly with SOC, especially below 30% and above 95% SOC. Therefore, the recommended operating window for the EV battery pack is between 30% and 95% SOC, where the internal resistance remains relatively low and stable.
Thermophysical Property Measurements
The density of the cell was determined by measuring its mass and volume and was found to be approximately 2172 kg/m³. An Accelerating Rate Calorimeter (ARC) was used to measure the average specific heat capacity by applying a constant heating power to the cell under adiabatic conditions. The specific heat capacity was calculated as
$$C_p = \frac{1}{m} \frac{P}{\Delta T / \Delta t}$$
where \(P\) is the heating power, \(m\) is the cell mass, and \(\Delta T / \Delta t\) is the temperature rise rate. The resulting average heat capacity was 850 J/(kg·K). The anisotropic thermal conductivity of the cell was determined from the layered structure of the electrodes and separators. Thermal conductivities in the \(x\), \(y\) and \(z\) directions were set to 21 W/(m·K), 14 W/(m·K), and 4 W/(m·K), respectively. The density, heat capacity, and thermal conductivity values are listed in Table 1 along with other cell parameters.
| Parameter | Value |
|---|---|
| Nominal capacity | 109 Ah |
| Nominal voltage | 3.5 V |
| Voltage range | 2.75–4.4 V |
| Cell mass | 1.83 kg |
| Dimensions (h × w × t) | 177.5 × 108.7 × 43.65 mm |
| Average density | 2172 kg/m³ |
| Specific heat capacity | 850 J/(kg·K) |
| Thermal conductivity x | 21 W/(m·K) |
| Thermal conductivity y | 14 W/(m·K) |
| Thermal conductivity z | 4 W/(m·K) |
2. Heat Generation Model Development and Fast-Charging Experiments
To verify the theoretical heat generation model and calibrate its parameters, transient temperature measurements were performed on fresh cells under various charging rates. Each cell was instrumented with five thermocouples on its largest surface and placed inside a temperature-controlled environmental chamber. Before each test, the cell was fully charged to 100% SOC using a standard constant-current constant-voltage (CC-CV) protocol. Then, the cell was discharged to 10% SOC with a 1C rate, and after a 30-minute rest, it was charged at the desired C-rate until the cut-off voltage was reached. The experiments covered ambient temperatures of 283 K, 298 K, 300 K, and 313 K, and charging rates of 1C, 2C, 3C, and 5C. The measured temperature profiles provided essential data for validating the model.
Three-Dimensional Thermal Model
A three-dimensional finite-volume model of a single lithium-ion cell was built using STAR-CCM+. The cell was treated as a homogeneous solid with anisotropic thermal conductivity, while the internal heat generation was assumed to be uniform. The governing equation for heat conduction in the cell is
$$\rho_b C_{p,b} \frac{\partial T_b}{\partial t} = \nabla \cdot (\mathbf{\lambda}_b \nabla T_b) + q_b$$
where \(\mathbf{\lambda}_b\) is the thermal conductivity tensor, \(\rho_b\) is the density, \(C_{p,b}\) is the specific heat capacity, and \(q_b\) denotes the user-defined volumetric heat source obtained from the Bernardi model. The outer surfaces of the cell are exposed to the environment. Natural convection is considered only when the cell is placed outside the pack; for modules inside the pack, the enclosed air is not explicitly modeled, because the heat transfer between the cell and the ambient is dominated by conduction through the cold plates and the surrounding structure. The mesh independence study was performed by refining the base mesh size from 8 mm to 2 mm. The difference in the predicted maximum cell temperature at the end of a 1C charge was found to be less than 0.05 K for mesh counts above 29,452. Therefore, a mesh containing 48,953 cells was selected.
The simulated cell temperatures were compared with the experimental data for charging rates of 1C, 3C, and 5C. Figure 1 shows an example comparison for the 3C charge. The maximum deviation between the simulated average temperature and the averaged thermocouple readings was only 0.56 K, which confirms the high fidelity of the heat generation model. In addition, two identical cells were tested at room temperature, and the experimental curves bounded the simulation result, thereby accounting for small sample-to-sample variations. This detailed validation provides confidence for the subsequent battery-pack-level simulations.
| Time (s) | Simulation avg. (K) | Experimental avg. (K) | Error (K) |
|---|---|---|---|
| 100 | 300.18 | 300.35 | 0.17 |
| 200 | 301.05 | 300.99 | 0.06 |
| 400 | 303.09 | 302.78 | 0.31 |
| 600 | 305.41 | 305.02 | 0.39 |
| 800 | 308.01 | 307.53 | 0.48 |
| 958 | 310.87 | 310.31 | 0.56 |
3. Thermal Management Structure Design and Component Selection
A liquid-cooled thermal management system was selected because of its high heat transfer coefficient and mature engineering implementation. A conventional bottom cooling layout, where a single liquid cold plate is placed beneath the battery pack, was initially designed. The EV battery pack contains multiple modules, each consisting of nine or ten prismatic cells connected in series by aluminum busbars. Between adjacent cells, an MPP foam is inserted as a thermal barrier to mitigate thermal runaway propagation. A thermally conductive pad is placed between the cell bottom surface and the cold plate to reduce contact thermal resistance. The material properties of the main components, including the coolant (50% ethylene glycol-water mixture), aluminum cold plate, insulating plate, thermal pad, and sealing gasket, are summarized in Table 3.
| Component | Material | Density (kg/m³) | Specific heat (J/(kg·K)) | Thermal conductivity (W/(m·K)) |
|---|---|---|---|---|
| Coolant | 50% ethylene glycol | 1068 | 3396 | 0.4 |
| Flame-retardant pad | MPP foam | 80 | 920 | 0.04 |
| Insulating end plate | Polycarbonate | 1230 | 1260 | 0.2 |
| Liquid cold plate | AL3003-O | 2730 | 893 | 193 |
| Aluminum busbar | AL1060-O | 2700 | 900 | 243 |
| Pack housing | AL6061 | 2730 | 896 | 167 |
| Thermal pad | PU-8592-E | 1700 | 2240 | 0.8 |
| Sealing gasket | Silicone | 400 | 2400 | 0.08 |
CFD Governing Equations and Boundary Conditions
The fluid flow inside the cold plates and the heat transfer between the solid and fluid domains were simulated using the three conservation laws. The coolant is treated as incompressible and Newtonian. The governing equations are summarized as follows. The continuity equation reads
$$\frac{\partial \rho_l}{\partial t} + \nabla \cdot (\rho_l \mathbf{v}) = 0$$
The momentum equation for the fluid is
$$\frac{\partial (\rho_l \mathbf{v})}{\partial t} + \nabla \cdot (\rho_l \mathbf{v} \mathbf{v}) = -\nabla p + \nabla \cdot (\mu_l \nabla \mathbf{v}) + \mathbf{S}$$
and the energy equation for the fluid is
$$\frac{\partial (\rho_l C_{p,l} T_l)}{\partial t} + \nabla \cdot (\rho_l \mathbf{v} C_{p,l} T_l) = \nabla \cdot (\lambda_l \nabla T_l)$$
For the solid parts, such as the cell, the cold plate, and the busbars, the heat conduction equation was applied. The laminar or turbulent nature of the coolant was determined from the Reynolds number
$$Re = \frac{\rho_l u d_h}{\mu_l}$$
where \(d_h\) is the hydraulic diameter. For the serpentine channels in the cold plate, the characteristic Reynolds number under typical operation is below 2300, so a laminar flow model was used. In cases where the flow approaches transition, a low-Reynolds-number turbulence model was applied. At the interface between the cold plate and the cell, a no-slip and perfect-contact condition was assumed because the numerical model flattens the microscopic roughness; thus, no additional thermal resistance was introduced.
Comparison of Cold Plate Geometries
The design of the liquid cold plate was first studied at the module level. Four representative channel layouts were proposed and compared using steady-state CFD simulations. These included (a) a series serpentine channel, (b) a parallel serpentine channel with a single outlet, (c) a “organ-pipe” parallel channel, and (d) a parallel serpentine channel with dual outlets. The dimensions of the local cold plate were 520 mm × 160 mm × 10 mm, and the inlet mass flow rate was fixed at 0.0178 kg/s (1 L/min). The performance was evaluated in terms of pressure drop, maximum coolant velocity, heat transfer rate, surface temperature uniformity, and coolant temperature homogeneity. Table 4 summarizes the simulation results.
| Layout | Pressure drop (Pa) | Max velocity (m/s) | Heat transfer (W) | Surface ΔT (K) |
|---|---|---|---|---|
| (a) Series serpentine | 878.2 | 0.385 | 145 ± 15 | 3.0 |
| (b) Parallel serpentine, single outlet | 249.3 | 0.299 | 168 | 5.0 |
| (c) Organ-pipe | 79.8 | 0.247 | 159 | 5.0 |
| (d) Parallel serpentine, dual outlets | 210.1 | 0.232 | 168 | 5.0 |
The series serpentine layout exhibited the highest pressure drop and produced unsteady flow oscillations in the residual curves. In contrast, the dual-outlet parallel serpentine layout provided the highest heat transfer and the lowest maximum velocity among all four, while its pressure drop was 22% lower than that of the single-outlet parallel layout. Therefore, the dual-outlet serpentine structure was selected as the fundamental pattern for the bottom cold plate. A mesh independence check was also performed on this structure; an appropriate mesh size of 1 mm with a prism layer thickness of 0.4 mm was selected for the subsequent full-scale simulations.
4. Design Optimization and Performance Validation
Proposed Three-Sided Liquid Cooling Configuration Based on Charging Pile
To overcome the insufficient cooling capability of a single bottom plate during ultra-fast charging, this work proposes a novel three-sided liquid cooling architecture whose cooling loop is partly integrated into the charging pile. During fast charging, the charging pile connects not only the high-voltage electrical circuit but also an external coolant circuit to the EV battery pack. This additional circuit supplies cold liquid to a top cold plate and a side cold plate that are embedded in the battery pack. The bottom cold plate is still cooled by the vehicle’s own coolant loop. In this manner, three surfaces of each cell — the bottom, one side, and the top — can be simultaneously cooled during the charging process. This concept offers several advantages: (1) the cooling power is supplied by the charging station, which is less sensitive to parasitic power consumption; (2) the charging-pile-side heat exchanger can provide a lower coolant inlet temperature than the vehicle loop; and (3) the added cold plates serve as thermal masses that improve temperature uniformity during driving and parking. The system architecture is illustrated schematically in the next paragraph.
Because the top surface of the cells must now be available for cooling, the orientation of the prismatic cells was changed. The cell terminals, originally on top, are repositioned to one side. Two adjacent rows of cells are arranged with their terminal sides facing each other, leaving a space for electrical interconnects, while the opposite side is dedicated to the side cold plate. The top surface of the rows is covered by a top cooling plate. The cold plates are inserted between the cell top surface and an insulating cover. Each module is cooled by one bottom plate, one side plate, and one top plate. The complete EV battery pack structure is shown in the conceptual diagram.
Flow Channel Design and Topology Optimization
The top cold plate was designed as three identical parallel serpentine units with a single inlet and two outlets. The side cold plate was designed as an organ-pipe type to match the longer length of the row. Four cold plates (three top sections and one side plate) are supplied in parallel from an inlet header tank and then merged into an outlet header tank. To achieve uniform flow distribution among the different plates, topology optimization was applied to the header tanks. A density-based topology optimization approach was adopted, solving a modified Darcy-forchheimer equation in the design domain. The following governing equations were used:
$$\frac{\partial (\rho u_j)}{\partial x_j} = 0$$
$$\frac{\partial}{\partial x_j} (\rho u_i u_j) = \frac{\partial}{\partial x_j} \left[\mu \frac{\partial u_i}{\partial x_j}\right] – \frac{\partial p}{\partial x_i} – \mu \alpha(\gamma) u_i$$
$$\frac{\partial}{\partial x_j} (\rho C_p T u_j) = \frac{\partial}{\partial x_j} \left( \lambda \frac{\partial T}{\partial x_j} \right) + q(\gamma)$$
where \(\gamma\) is a design variable between 0 (solid) and 1 (fluid), and \(\alpha(\gamma)\) is the inverse permeability. After more than 1000 iterations, the optimized header geometry reduced the maximum flow-rate imbalance among the four cold plates from 40.9% to 26.0%. The bottom cold plate was also topologically optimized by smoothing the corners and eliminating recirculation zones. The final pressure drop was reduced by 8.43% compared with the initial layout. Figure 2 shows the simulation result of the coolant velocity distribution in the manifold and cold plates after optimization.
Simulation under Fast-Charging Conditions
The transient thermal behavior of a full battery pack containing cells, bottom cold plate, top cold plate, side cold plate, housing, and thermal pads was simulated. The reference configuration employed only the bottom cold plate at the same total coolant flow rate. The initial temperature of the cells was set to 30°C. Three constant-current charging rates were considered: 1C, 3C, and 5C, with the charging duration corresponding to a SOC swing from 10% to 90%: 2880 s, 960 s, and 576 s, respectively. The coolant inlet temperature for all plates was set to 22°C. Since the cell temperatures are high in the fast-charging scenario, the heat dissipation through the pack walls is small compared to the cooling by the liquid; therefore, convection losses were neglected in the first set of simulations. Figure 3 compares the maximum temperature and the maximum temperature difference among cells for both configurations.
| Charging rate | Config. | Max cell temp. (°C) | Max temp. difference (°C) |
|---|---|---|---|
| 1C | Three-sided | 31.77 | 7.81 |
| 1C | Conventional | 33.98 | 9.74 |
| 3C | Three-sided | 31.6 | 9.21 |
| 3C | Conventional | 39.3 | 15.82 |
| 5C | Three-sided | 43.0 | 20.49 |
| 5C | Conventional | 53.1 | 27.89 |
As shown in Table 5, the proposed three-sided cooling configuration significantly reduces the maximum cell temperature. Compared with the conventional bottom cooling, the maximum cell temperature is reduced by 2.21°C, 7.7°C, and 10.1°C for 1C, 3C, and 5C charging, respectively. The maximum temperature difference between cells is also improved, with reductions of 1.93°C, 6.61°C, and 7.4°C. At the 5C rate, the conventional configuration causes the maximum temperature to exceed 50°C, which is detrimental to battery lifetime and safety, whereas the three-sided cooling confines the maximum temperature below 45°C.
Another important metric is the through-plane temperature gradient. Since the bottom cooling causes strong heat extraction from the bottom only, the upper portion of the cell typically becomes significantly hotter. By cooling the top and one side simultaneously, the temperature field in the cell becomes more homogeneous. The average temperature was sampled on three horizontal cross-sections at heights of 35 mm, 55 mm, and 75 mm from the bottom. The results are presented in Table 6. At 5C charging, the conventional cooling case shows a difference of more than 12°C between the lowest and highest cross-section, while the three-sided cooling case reduces this gradient substantially.
| Charging rate | Config. | 35 mm (°C) | 55 mm (°C) | 75 mm (°C) |
|---|---|---|---|---|
| 1C | Three-sided | 31.10 | 31.52 | 31.44 |
| 1C | Conventional | 32.45 | 33.62 | 34.60 |
| 3C | Three-sided | 35.60 | 37.47 | 36.81 |
| 3C | Conventional | 38.50 | 43.25 | 45.15 |
| 5C | Three-sided | 40.91 | 44.76 | 43.03 |
| 5C | Conventional | 44.13 | 51.90 | 56.29 |
Influence of Coolant Inlet Temperature
Since the top and side cold plates are supplied by the charging pile, the inlet temperature can be adjusted independently of the vehicle loop. To quantify the effect of lower coolant temperature, three inlet temperatures (22°C, 15°C, and 10°C) were imposed on the top and side cold plates, while the bottom plate inlet was kept at 22°C. Figure 4 presents the maximum temperature and maximum cell-to-cell temperature difference during 1C, 3C, and 5C charging. The results show that each 5°C reduction in coolant temperature lowers the maximum cell temperature by about 2–3°C, with a larger effect at higher charging rates. However, the lower coolant temperature also increases the temperature difference among cells, especially at high C-rates. Therefore, the inlet temperature should be carefully selected to balance cooling performance and temperature uniformity.
Low-Temperature Auxiliary Heating during Charging
The proposed charging-pile-side circuit includes a positive temperature coefficient (PTC) heater. This enables preheating of the coolant before entering the battery pack during cold-weather fast charging. A charging scenario at an ambient temperature of −10°C was simulated. The coolant was initially heated to 25°C by a 4.5-kW heater. The heating was enabled when the minimum cell temperature was below 10°C and stopped once that temperature was reached. Liquid cooling was enabled when the maximum cell temperature exceeded 35°C. The simulation showed that the battery can be charged while maintaining the maximum temperature at 37.8°C and the maximum temperature difference at 6.5°C. Thus, the battery pack can be warmed up without consuming the battery’s own energy during charging, which is beneficial for cold climates.
Driving and Parking Conditions
The performance of the proposed EV battery pack configuration was also assessed under realistic driving cycles and during long-term parking. A one-dimensional vehicle dynamics model was built in Amesim to calculate the cell heat generation rate during three consecutive China Light-duty Vehicle Test Cycle (CLTC) cycles. This heat generation profile was then imported into the three-dimensional CFD model, where only the bottom cold plate is active (top and side plates are filled with stagnant coolant). The results indicate that the additional thermal mass of the new configuration delays the initial temperature rise, after which the average temperature closely follows that of the reference case. The maximum temperature difference is slightly reduced, from 4.49°C in the conventional case to 3.41°C in the new configuration, demonstrating better temperature uniformity under daily driving conditions.
In the parking scenario, the EV is left outside for 10 hours at an ambient temperature of −20°C, starting from an initial battery temperature of 30°C. Natural convection on the external surfaces was modeled with a heat transfer coefficient of 5 W/(m²·K). The new three-sided cooling structure, owing to its larger thermal mass and the additional insulating effect from the stagnant coolant in the top and side plates, retains a higher cell temperature after 10 hours (−9.45°C) compared with the conventional bottom-cooled pack (−17.54°C). This advantage improves the next-day cold-start performance and reduces the energy required for battery heating.
Impact on Fast Charging Strategy
Finally, the benefit of the proposed three-sided cooling approach on the actual charging time was evaluated using a multi-stage constant-current (MCC) charging protocol. A one-dimensional battery/charging model was constructed in Simulink, incorporating the temperature-dependent resistance data and the cooling performance curves obtained from CFD simulations. The battery temperature was maintained below 45°C, and the charging process started from 10% SOC and continued until 95% SOC. In the conventional cooling case, the charging current had to be reduced at an early stage to prevent overheating, leading to a total charging time of 2718 s. In contrast, the three-sided cooling configuration allowed a higher current for a longer period, resulting in a charging time of only 2194 s. Thus, the total charging duration was shortened by 19.28% under the same thermal constraint. Moreover, the time required to charge from 10% to 80% SOC was reduced by 23.54%, a substantial improvement that directly addresses the charging pain points of electric vehicle users.
5. Conclusions
This paper reports a comprehensive thermal management design optimization study for an EV battery pack under ultra-fast charging conditions. The main contributions and findings are summarized as follows.
(1) The HPPC tests and ARC experiments provided essential thermal-physical properties of the 109 Ah NCM cell. The cell heat generation rate was determined using the Bernardi model, and a three-dimensional thermal model was constructed. The simulated temperature profiles agreed with the experimental measurements within 0.56 K, validating the model’s accuracy.
(2) A dual-outlet parallel serpentine channel was identified as the preferred bottom cold plate design after comparing four channel layouts. The cold plate layout was further refined using topology optimization, reducing the pressure drop by 8.43% and improving the flow uniformity among multiple cold plates.
(3) A novel three-sided liquid cooling architecture was proposed. The top and one side of each cell are cooled by coolant supplied from an external loop integrated into the charging pile. Under fast-charging conditions, this architecture outperforms conventional single-surface bottom cooling by reducing the maximum cell temperature by up to 10.1°C and the cell-to-cell temperature difference by up to 7.4°C. In addition, the temperature gradient within individual cells is greatly mitigated.
(4) The performance of the proposed EV battery pack thermal management configuration was also verified under driving and parking conditions. The additional cold plates act as a thermal buffer that improves temperature uniformity and reduces temperature drop in cold weather. With the same MCC charging strategy, the three-sided cooling layout allows 19.28% shorter 10%–95% SOC charging time under a 45°C maximum temperature constraint.
The proposed design approach offers a practical pathway for enabling large-rate fast charging in modern EVs without sacrificing safety or battery life. As future work, the battery’s electrochemical aging behavior and more sophisticated electro-thermal coupling will be investigated to further refine the charging protocol and thermal management control strategy. The findings of this research are expected to contribute to the next-generation thermal management solutions for high-performance EV battery packs.