The rapid growth of the electric vehicle market has brought the safety and reliability of lithium-ion batteries into sharp focus. As the core energy storage component of an EV battery pack, the lithium-ion battery is highly sensitive to temperature. Both excessive heat and uneven temperature distribution can lead to performance degradation, accelerated aging, and even catastrophic thermal runaway. To ensure safe and efficient operation, an effective battery thermal management system (BTMS) is indispensable. In this work, I present a combined liquid cooling plate and vapor chamber (LCP‑VC) composite cooling system tailored for a soft‑pack ternary lithium‑ion cell. The proposed system aims to enhance the temperature control accuracy and temperature uniformity of an EV battery pack under high‑rate discharge conditions. Through a combination of numerical simulation and experimental validation, I systematically investigate the cooling performance, structural optimization, and low‑temperature preheating capability of the composite system.
Thermal management of an EV battery pack is challenging because the heat generation rate is not uniform within a cell. The electrochemical reactions, ionic transport, and ohmic losses create local hot spots, especially near the tabs. A conventional liquid cooling plate (LCP) can effectively reduce the average temperature, but it often fails to provide sufficient temperature uniformity. To address this issue, I introduce a vapor chamber (VC) as an intermediate heat spreader between the battery and the LCP. The VC rapidly spreads the heat in two dimensions, thereby reducing the temperature gradient across the cell surface. This composite architecture exploits the high heat capacity of liquid cooling and the excellent thermal conductivity of phase‑change heat transfer in the VC.
1. Battery Structure and Heat Generation Mechanism
A lithium‑ion battery consists of a positive electrode, a negative electrode, a separator, an electrolyte, and a casing. The positive electrode material used in this study is a nickel‑cobalt‑manganese oxide (NCM) ternary compound, while the negative electrode is graphite. During operation, lithium ions shuttle between the electrodes, generating heat through four distinct mechanisms: reaction heat \(Q_r\), polarization heat \(Q_p\), ohmic heat \(Q_j\), and side reaction heat \(Q_s\). The total heat generation rate can be expressed as:
$$Q_t = Q_r + Q_p + Q_j + Q_s$$
The reaction heat originates from the entropy change of the electrochemical reactions. It is reversible and can be calculated as:
$$Q_r = T \Delta S$$
where \(T\) is the absolute temperature and \(\Delta S\) is the entropy change. The polarization heat and ohmic heat are both irreversible and can be represented together by an equivalent internal resistance \(R\). Using Joule’s law:
$$Q_j = I^2 R t$$
where \(I\) is the discharge current and \(t\) is the time. The side reaction heat is usually negligible under normal operating conditions.
In this work, I adopt the Bernardi heat generation model, which is widely used for battery thermal simulation. The volumetric heat generation rate \(q_B\) is given by:
$$q_B = \frac{I}{V_B}\left( U – U_0 + T \frac{dU_0}{dT} \right) = \frac{I}{V_B}\left( IR + T \frac{dU_0}{dT} \right)$$
where \(V_B\) is the battery volume, \(U\) is the open‑circuit voltage, \(U_0\) is the terminal voltage, and \(R\) is the total internal resistance. The term \(T \, dU_0/dT\) represents the reversible heat, while \(IR\) represents the irreversible heat.
2. Thermal Properties and Model Validation
The battery under investigation is a soft‑pack cell with the dimensions 167 mm × 164 mm × 10 mm and a nominal capacity of 46 Ah. Since the internal structure of the cell is highly complex and consists of multiple layers of different materials, I simplify the cell as a homogeneous solid with equivalent thermal properties. The equivalent density \(\rho_B\), specific heat capacity \(c_{p,B}\), and thermal conductivities in the thickness direction (\(k_x\)) and in the planar directions (\(k_y, k_z\)) are calculated using the volume‑weighted and series‑weighted averages:
$$\rho_B = \frac{\sum_{i=1}^{n} \rho_i V_i}{\sum_{i=1}^{n} V_i}, \qquad c_{p,B} = \frac{\sum_{i=1}^{n} c_{p,i} m_i}{\sum_{i=1}^{n} m_i}$$
$$k_x = \frac{\sum_{i=1}^{n} L_i}{\sum_{i=1}^{n} L_i / k_i}, \qquad k_y = k_z = \frac{\sum_{i=1}^{n} k_i L_i}{\sum_{i=1}^{n} L_i}$$
The computed thermal parameters are listed in Table 1.
| Parameter | Value |
|---|---|
| Dimensions (mm) | 167 × 164 × 10 |
| Nominal capacity (Ah) | 46 |
| Positive electrode material | NCM |
| Negative electrode material | Graphite |
| Cut‑off discharge voltage (V) | 2.75 |
| Cut‑off charge voltage (V) | 4.2 |
| Density \(\rho_B\) (kg/m³) | 1933 |
| Specific heat \(c_{p,B}\) (J/(kg·K)) | 1080 |
| In‑plane thermal conductivity \(k_x\) (W/(m·K)) | 22.3 |
| Through‑plane thermal conductivity \(k_y, k_z\) (W/(m·K)) | 0.66 |
To obtain the internal resistance as a function of state of charge (SOC), I performed a hybrid pulse power characteristic (HPPC) test on a fresh cell at 25 °C. The test sequence consisted of discharging the cell in 10% SOC steps and applying a specific current pulse at each step. The ohmic resistance \(R_{ohm}\) and polarization resistance \(R_{pol}\) were extracted from the voltage response:
$$R_{ohm} = \frac{U_1 – U_2}{I}, \qquad R_{pol} = \frac{U_2 – U_3}{I}, \qquad R = R_{ohm} + R_{pol}$$
where \(U_1\), \(U_2\), and \(U_3\) are the voltages at the beginning, end of the pulse, and after the pulse, respectively. The measured total resistance versus SOC was fitted with a fourth‑order polynomial (R² = 0.997):
$$R(SOC) = 0.00272 – 0.00559 \, SOC + 0.00954 \, SOC^2 – 0.0059 \, SOC^3 + 0.000255342 \, SOC^4$$
The actual capacity of the cell was measured by a 1 C charge/discharge cycling test. The average measured discharge capacity was 46.057 Ah, very close to the nominal value. This verified that the cell was in a healthy state.
In order to validate the heat generation model, I placed thermocouples on the battery surface and performed a 2 C constant‑current discharge in a naturally convective environment at 25 °C. The measured temperature profiles were compared with the simulation results from ANSYS Fluent, where the Bernardi heat generation model was applied via a user‑defined function. The comparison showed excellent agreement, with a maximum relative error of less than 3.5%. Table 2 summarizes the temperature comparison at the end of discharge.
| Quantity | Simulation (°C) | Experiment (°C) | Error |
|---|---|---|---|
| Average temperature | 46.92 | 46.43 | 1.05% |
| Maximum temperature difference | 0.76 | 4.35 | — |
The simulation assumed a uniform heat source, which explains the under‑prediction of the maximum temperature difference. Nevertheless, the average temperature error was below 1.1%, confirming the reliability of the heat generation model for subsequent studies.
3. Design and Optimization of the Liquid Cooling Plate
The liquid cooling plate was designed to have dimensions identical to the battery. The plate material is aluminum (thermal conductivity 238 W/(m·K)), and the coolant is water. A straight‑channel configuration was selected for simplicity and manufacturing feasibility. The LCP is placed in direct contact with the battery surface; heat is conducted from the battery into the plate and then removed by the flowing coolant.
Four key geometric and operating parameters were considered for optimizing the LCP: channel width \(A\), channel height \(B\), number of channels \(C\), and coolant inlet velocity \(D\). Each parameter was assigned four levels, as shown in Table 3. An orthogonal experiment based on the \(L_{16}(4^4)\) array was designed, leading to 16 different LCP configurations. The performance indicators were the average battery temperature, the maximum temperature difference, and the pressure drop across the channel.
| Level | A (mm) | B (mm) | C (channels) | D (m/s) |
|---|---|---|---|---|
| 1 | 9 | 5 | 4 | 0.01 |
| 2 | 11 | 6 | 5 | 0.02 |
| 3 | 13 | 7 | 6 | 0.03 |
| 4 | 15 | 8 | 7 | 0.04 |
All simulations were performed in ANSYS Fluent with a 2 C discharge rate, ambient and coolant temperature of 25 °C, and a discharge duration of 1800 s. The flow was laminar (Reynolds number below 2300 for all cases). Mesh independence was verified with six mesh densities, and the fifth mesh (224,641 elements) was chosen as a compromise between accuracy and computational cost.
The orthogonal experimental results are summarized in Table 4.
| Case | A | B | C | D | Average temp. (°C) | Temp. diff. (°C) | Pressure drop (Pa) |
|---|---|---|---|---|---|---|---|
| 1 | 9 | 5 | 4 | 0.01 | 35.60 | 2.44 | 3.07 |
| 2 | 9 | 6 | 5 | 0.02 | 33.74 | 2.91 | 5.72 |
| 3 | 9 | 7 | 6 | 0.03 | 32.52 | 2.73 | 8.70 |
| 4 | 9 | 8 | 7 | 0.04 | 31.92 | 2.70 | 10.84 |
| 5 | 11 | 5 | 7 | 0.02 | 33.78 | 2.74 | 7.56 |
| 6 | 11 | 6 | 6 | 0.01 | 35.33 | 2.63 | 2.21 |
| 7 | 11 | 7 | 5 | 0.04 | 31.94 | 2.78 | 8.99 |
| 8 | 11 | 8 | 4 | 0.03 | 32.62 | 2.56 | 5.10 |
| 9 | 13 | 5 | 5 | 0.03 | 32.67 | 2.81 | 8.55 |
| 10 | 13 | 6 | 4 | 0.04 | 31.98 | 2.68 | 8.63 |
| 11 | 13 | 7 | 7 | 0.01 | 35.07 | 2.46 | 1.64 |
| 12 | 13 | 8 | 6 | 0.02 | 33.54 | 2.80 | 3.08 |
| 13 | 15 | 5 | 6 | 0.04 | 31.98 | 2.78 | 11.55 |
| 14 | 15 | 6 | 7 | 0.03 | 32.60 | 2.90 | 6.73 |
| 15 | 15 | 7 | 4 | 0.02 | 33.65 | 2.78 | 2.84 |
| 16 | 15 | 8 | 5 | 0.01 | 34.97 | 2.54 | 1.18 |
Range analysis and variance analysis were used to evaluate the influence of each factor. For each factor \(i\), the average value \(\bar{k}_{i,n}\) at each level \(n\) was calculated, and the range \(R_i\) was computed as:
$$R_i = \max(\bar{k}_{i,n}) – \min(\bar{k}_{i,n})$$
Similarly, the standard deviation \(S_i\) was obtained from:
$$S_i = \sqrt{\frac{1}{4}\sum_{n=1}^{4} \left(\bar{k}_{i,n} – \frac{1}{4}\sum_{n=1}^{4}\bar{k}_{i,n}\right)^2}$$
The results of the range and standard deviation analyses are presented in Table 5.
| Indicator | Factor | Range \(R_i\) | Std. dev. \(S_i\) | Rank |
|---|---|---|---|---|
| Average temperature | A | 0.150 | 0.065 | 3 |
| B | 0.244 | 0.097 | 2 | |
| C | 0.136 | 0.055 | 4 | |
| D | 3.288 | 1.245 | 1 | |
| Temperature difference | A | 0.123 | 0.044 | 3 |
| B | 0.078 | 0.028 | 4 | |
| C | 0.142 | 0.051 | 2 | |
| D | 0.336 | 0.130 | 1 | |
| Pressure drop | A | 1.60 | 0.52 | 4 |
| B | 2.14 | 0.79 | 2 | |
| C | 2.18 | 0.68 | 3 | |
| D | 5.61 | 4.07 | 1 |
For the average temperature, the coolant inlet velocity (factor D) had the dominant influence, followed by channel height, channel width, and number of channels. For the temperature difference, factor D still ranked first, while factor B had the least influence. Pressure drop was also dominated by the coolant velocity, which is consistent with the Darcy–Weisbach equation:
$$\Delta P = f \frac{L}{d} \frac{\rho v^2}{2}$$
where \(f\) is the friction factor, \(L\) is the channel length, \(d\) is the hydraulic diameter, \(\rho\) is the coolant density, and \(v\) is the flow velocity.
Based on the overall consideration of all three indicators, the optimal LCP configuration was determined as: channel width \(A = 15\) mm, channel height \(B = 8\) mm, number of channels \(C = 7\), and inlet velocity \(D = 0.04\) m/s. The simulation of this optimized configuration (designated as Model 17) yielded an average battery temperature of 31.89 °C and a maximum temperature difference of 2.88 °C. Compared with the original design, the average temperature was reduced by up to 3.71 °C, which demonstrates the effectiveness of the orthogonal optimization.
4. Composite LCP‑VC Cooling System
Although the optimized LCP achieved a low average temperature, the maximum temperature difference of 2.88 °C still left room for improvement. To further enhance the temperature uniformity, I inserted a vapor chamber (VC) between the battery and the LCP. The VC has an effective thermal conductivity of 5000 W/(m·K) in the planar direction, which is more than twenty times higher than that of aluminum. Heat generated by the battery is first absorbed by the VC and spread uniformly across its surface; then the LCP removes the heat through the coolant flow. This composite architecture significantly reduces the thermal resistance path and eliminates hot spots.
The simulation comparison between the LCP cooling mode and the LCP‑VC cooling mode for a single cell is presented in Table 6. The discharge condition was 2 C, with coolant and ambient temperature at 25 °C and inlet velocity at 0.04 m/s.
| Indicator | LCP mode | LCP‑VC mode | Improvement |
|---|---|---|---|
| Average temperature (°C) | 31.89 | 32.06 | +0.17 (slightly higher) |
| Maximum temperature (°C) | 32.82 | 32.65 | −0.17 |
| Maximum temperature difference (°C) | 2.88 | 1.42 | −50.27% |
The LCP‑VC mode reduced the maximum temperature difference by more than half, from 2.88 °C to 1.42 °C, while the average temperature increased by only 0.17 °C. This indicates that the composite structure dramatically improves thermal uniformity without sacrificing the overall cooling capacity.
4.1 Module‑Level Configurations
To evaluate the scalability of the composite cooling concept, I designed three module‑level configurations consisting of 3, 6, and 9 cells, respectively. Each configuration was simulated with both the LCP mode and the LCP‑VC mode. Table 7 lists the resulting average temperature, maximum temperature difference, and maximum temperature of the battery module for all six cases.
| Configuration | Cooling mode | Average temp. (°C) | Temp. diff. (°C) | Max. temp. (°C) |
|---|---|---|---|---|
| Case1 (3 cells) | LCP | 31.67 | 6.98 | — |
| Case1 (3 cells) | LCP‑VC | 31.90 | 5.85 | — |
| Case2 (6 cells) | LCP | 33.90 | 5.18 | — |
| Case2 (6 cells) | LCP‑VC | 33.23 | 2.42 | — |
| Case3 (9 cells) | LCP | 36.03 | 8.37 | 39.34 |
| Case3 (9 cells) | LCP‑VC | 36.41 | 4.16 | 38.75 |
For Case2, the LCP‑VC mode reduced the temperature difference by 53.3% compared with the LCP mode. For Case3, the reduction was 50.3%. Although the average temperature of the LCP‑VC mode was slightly higher in Case1 and Case3, the maximum temperature and temperature difference were consistently lower. Taking into account both thermal performance and structural cost, Case3 with the LCP‑VC cooling mode was chosen as the preferred module configuration for subsequent parametric studies because it satisfies the temperature requirements (25 °C – 40 °C and a cell‑to‑cell temperature difference below 5 °C) while using the fewest cooling components per cell.
5. Parametric Study of the Composite Cooling System
5.1 Effect of Coolant Flow Velocity
For the selected Case3 module with the LCP‑VC cooling mode, I varied the coolant inlet velocity from 0.01 to 0.04 m/s while keeping the inlet temperature at 25 °C. The results are presented in Table 8.
| Velocity (m/s) | Average temp. (°C) | Temp. diff. (°C) |
|---|---|---|
| 0.01 | 43.50 | 3.63 |
| 0.02 | 39.71 | 3.99 |
| 0.03 | 37.41 | 4.13 |
| 0.04 | 36.41 | 4.16 |
Increasing the velocity from 0.01 to 0.04 m/s reduced the average temperature by 7.09 °C, although the temperature difference increased slightly. The marginal benefit of increasing the velocity diminishes above 0.03 m/s; therefore, a velocity around 0.03–0.04 m/s offers a good balance between cooling performance and pump power consumption.
5.2 Effect of Coolant Inlet Temperature
The coolant inlet temperature was varied from 15 °C to 30 °C at a fixed velocity of 0.04 m/s. Figure shows the average temperature evolution and the temperature difference over time for the LCP‑VC mode. Table 9 summarizes the final values at the end of the 2 C discharge.
| Inlet temperature (°C) | Average temp. (°C) | Temp. diff. (°C) |
|---|---|---|
| 15 | — | 4.33 |
| 20 | — | 4.21 |
| 25 | 36.41 | 4.16 |
| 30 | — | 4.11 |
The average temperature increased almost linearly with the inlet temperature. However, the temperature difference decreased slightly as the inlet temperature rose, because the lower temperature difference between the coolant and the cell reduces the local heat flux gradient. In practice, a moderate inlet temperature around 20–25 °C is recommended for both cooling effectiveness and energy efficiency.
6. Low‑Temperature Preheating Performance
Batteries in an EV battery pack often face cold climates, where the electrolyte conductivity drops and the internal resistance increases. Preheating is essential to restore capacity and avoid lithium plating. I simulated a cold‑start scenario at −10 °C ambient temperature using the same module geometry. The heating medium (water) was circulated through the LCP at a velocity of 0.4 m/s, with inlet temperatures of 15, 20, 25, and 30 °C. The time required for the battery average temperature to rise from −10 °C to 10 °C was recorded for both the LCP heating mode and the LCP‑VC heating mode.
| Heating medium temperature (°C) | LCP mode (s) | LCP‑VC mode (s) | Time saving |
|---|---|---|---|
| 15 | 378 | 362 | 4.2% |
| 20 | 258 | 241 | 6.6% |
| 25 | 212 | 193 | 9.0% |
| 30 | 184 | 165 | 10.3% |
Table 10 clearly shows that the LCP‑VC heating mode always achieves a faster warm‑up than the LCP mode. The time saving becomes more significant at higher heating medium temperatures. Moreover, the temperature difference within the module was reduced by about 12.1% when using the LCP‑VC mode, demonstrating its advantage in providing uniform heating. The temperature uniformity is critical for preventing localized over‑discharge or over‑charge during cold starts.
7. Experimental Validation
To validate the simulation results, I constructed an experimental test bench consisting of a battery test system (Neware CT‑4004), a temperature acquisition unit (TOPRIE TP700), K‑type thermocouples, a constant temperature and humidity chamber, a peristaltic pump, and the LCP/VC components. Eight thermocouples were attached to both sides of the cell. The test procedure was as follows:
- Set the chamber temperature to 25 °C and charge the cell to 100% SOC using a constant current–constant voltage protocol.
- Soak the cell until thermal equilibrium is reached.
- Connect the peristaltic pump to the LCP and set the coolant flow rate to the desired value.
- Apply a 2 C constant current discharge until the voltage reaches 2.75 V.
- Measure the temperature histories at the eight locations.
- Repeat the test without the VC to obtain the LCP cooling mode data.
For the coolant flow rate of 0.04 m/s and an inlet temperature of 25 °C, the experimental results are listed in Table 11.
| Indicator | LCP mode | LCP‑VC mode | Improvement |
|---|---|---|---|
| Average temperature (°C) | 35.95 | 35.13 | 0.82 °C reduction |
| Maximum temperature difference (°C) | 3.98 | 3.69 | 7.2% reduction |
The experimental average temperature was higher than the simulation value because of the thermal contact resistance between the battery, VC, and LCP, which was idealized as perfect in the simulation. Nevertheless, the experimental results confirm the same trend as the numerical analysis: the LCP‑VC composite system provides a lower average temperature and better temperature uniformity compared with the conventional LCP system.
Figure 5 shows the temperature at the measurement points for the LCP‑VC mode under different coolant velocities. The highest temperature was consistently recorded at the positive tab side of the battery, while the lowest temperature was near the bottom away from the tabs. This is consistent with the electrochemical heat generation pattern.
8. Conclusion and Outlook
In this work, I have presented a comprehensive investigation of a composite liquid cooling system for an EV battery pack. The main conclusions are as follows:
- A simplified thermal model of a soft‑pack NCM lithium‑ion battery was established and validated against experiments. The simulated and measured temperatures differed by less than 3.5%, confirming the accuracy of the Berardi heat generation model.
- An orthogonal optimization of the liquid cooling plate showed that the coolant inlet velocity is the most influential parameter for both the average temperature and the temperature difference. The optimized LCP reduced the average temperature by up to 3.71 °C.
- Integrating a vapor chamber into the LCP cooling system reduced the maximum temperature difference by 50.27% (from 2.88 °C to 1.42 °C for a single cell) while maintaining a comparable average temperature. For the module with nine cells (Case3), the LCP‑VC mode achieved a maximum temperature difference of 4.16 °C, satisfying the common requirement of <5 °C.
- The parametric study revealed that increasing the coolant velocity improves cooling at the cost of a slightly larger temperature difference. A velocity of 0.03–0.04 m/s was found to be optimal. The coolant inlet temperature has a direct impact on the average temperature, but only a minor effect on the temperature difference.
- The low‑temperature preheating simulation indicated that the LCP‑VC heating mode reduces the warm‑up time by up to 10.3% and improves the temperature uniformity by 12.1% compared with the LCP heating mode.
- Experimental validation confirmed the superiority of the LCP‑VC composite cooling system. At a 2 C discharge rate, the average temperature was reduced by 0.82 °C and the temperature difference by 7.2% relative to the LCP system.
Future work will focus on optimizing the VC wick structure for even higher effective thermal conductivity, exploring the influence of different coolant fluids (e.g., nanofluids), and integrating the LCP‑VC system with phase change materials for passive cooling. The composite cooling technology proposed here offers a promising route toward safer and more reliable EV battery packs under high‑rate discharge and extreme environmental conditions.