In the era of rapid energy transition, energy storage battery systems have become indispensable for grid stability, renewable integration, and peak shaving. Among various battery chemistries, lithium-ion energy storage battery modules dominate due to their high energy density, long cycle life, and flexible discharge characteristics. However, their thermal safety remains a critical challenge, as improper thermal management can lead to thermal runaway, fire, or explosion. To address this, we have developed and optimized a liquid cooling system specifically designed for energy storage battery modules operating under both constant current and real-world peak regulation and frequency regulation profiles. This paper presents our comprehensive study involving geometric design, parameter optimization, and performance validation of the liquid cooling system for energy storage battery applications.
Our research began with the construction of a three-dimensional numerical model of a lithium-ion energy storage battery module using COMSOL Multiphysics. The module consisted of 18 prismatic cells arranged in a 3-series, 6-parallel configuration, each with a nominal capacity of 20 Ah and dimensions 148 mm × 26 mm × 92 mm. We proposed two distinct liquid cooling plate layouts: Design 1, where cooling plates were placed longitudinally (vertical orientation) between cell columns, and Design 2, where cooling plates were placed transversely (horizontal orientation) between cell rows. Both designs utilized aluminum cold plates with internal microchannels through which water flowed as the coolant. Detailed geometric parameters are summarized in Table 1.
| Parameter | Design 1 (Longitudinal) | Design 2 (Transverse) |
|---|---|---|
| Number of cold plates | 4 | 7 |
| Cold plate dimensions (mm) | 221 × 6 × 92 | 221 × 6 × 92 |
| Channels per plate | 7 | 7 |
| Channel cross-section (mm) | 5 × 5 | 5 × 5 |
| Channel spacing (mm) | 15 | 15 |
| Manifold type | Rear inlet, front outlet | Right inlet, left outlet |
The heat generation of each energy storage battery cell was modeled using the Bernardi equation:
$$ q = \frac{1}{V} \left[ (U_0 – U) – T \frac{dU_0}{dT} \right] $$
where \(q\) is the volumetric heat generation rate, \(V\) the cell volume, \(U_0\) the open-circuit voltage, \(U\) the terminal voltage, \(T\) the temperature, and \(dU_0/dT\) the entropy coefficient. The thermal behavior of the battery was governed by the anisotropic heat conduction equation:
$$ \rho_k C_{p,k} \frac{\partial T}{\partial t} = \nabla \cdot (\lambda_k \nabla T) + q $$
with \(\rho_k\), \(C_{p,k}\), and \(\lambda_k\) representing density, specific heat, and thermal conductivity of each material. The coolant flow in the microchannels was described by the incompressible Navier-Stokes and energy equations:
$$ \rho_w (\mathbf{u}_w \cdot \nabla) \mathbf{u}_w = -\nabla p + \mu_w \nabla^2 \mathbf{u}_w $$
$$ \nabla \cdot \mathbf{u}_w = 0 $$
$$ \rho_w c_w \frac{\partial T}{\partial t} + \rho_w c_w \mathbf{u}_w \cdot \nabla T = \nabla \cdot (k_w \nabla T) $$
where subscripts \(w\) denote water properties. The aluminum cold plate itself was modeled as a passive solid with no internal heat generation:
$$ \rho_c c_c \frac{\partial T}{\partial t} = \nabla \cdot (k_c \nabla T) $$
These coupled equations were solved using the finite element method, after validating the battery thermal model against experimental data. Our validation at 0.5 C constant current charging showed a maximum error of only 0.22 °C and a root mean square error (RMSE) of 0.15 °C, confirming the accuracy of the simulation approach.
We first compared the thermal performance of Design 1 and Design 2 under constant current (CC) charging at 1 C and 3 C, with coolant flow rate fixed at 1 cm³/s, coolant temperature 25 °C, and seven channels per plate. Table 2 summarizes the key temperature metrics at the end of charging (SOC = 100%).
| Charge rate | Metric | Design 1 | Design 2 |
|---|---|---|---|
| 1 C | Maximum temperature (°C) | 34.93 | 37.91 |
| Maximum temperature difference (°C) | 2.44 | 4.10 | |
| 3 C | Maximum temperature (°C) | 52.58 | 58.82 |
| Maximum temperature difference (°C) | 5.01 | 8.54 |
Design 1 consistently outperformed Design 2, with up to 11.86% lower maximum temperature and 70.46% lower temperature difference at 3 C. The reason is the anisotropic thermal conductivity of the energy storage battery cells: the Y-direction (through-plane) conductivity is much lower than in the X and Z directions, making transverse cooling (Design 2) inefficient as heat must traverse the poorly conductive direction. Thus, we selected Design 1 for all subsequent optimization studies.
Next, we investigated the influence of four key operational and structural parameters on the cooling performance of Design 1 under constant current charging at 1, 2, and 3 C. The parameters included coolant flow rate, coolant inlet temperature, number of cooling channels per plate, and inlet orientation (same-side vs. staggered). The baseline condition for parametric sweeps was: flow rate 2 cm³/s, coolant temperature 20 °C, 7 channels, and staggered inlets (unless a specific parameter was being varied). We present the results through a series of tables.
| Charge rate | Flow rate (cm³/s) | Max temp (°C) | Max ΔT (°C) |
|---|---|---|---|
| 1 C | 1 | 33.05 | 2.65 |
| 2 | 31.27 | 2.74 | |
| 3 | 31.00 | 2.78 | |
| 2 C | 1 | 42.79 | 4.35 |
| 2 | 40.86 | 4.48 | |
| 3 | 40.31 | 4.53 | |
| 3 C | 1 | 49.66 | 5.55 |
| 2 | 46.90 | 5.72 | |
| 3 | 46.11 | 5.79 |
Increasing flow rate from 1 to 2 cm³/s reduced the maximum temperature by ~5% at all rates, but further increasing to 3 cm³/s yielded only ~1-2% additional reduction. The temperature difference slightly increased with flow rate due to more pronounced cooling near the inlet. Considering energy consumption of the pump, we selected 2 cm³/s as the optimal flow rate.
| Charge rate | Coolant temp (°C) | Max temp (°C) | Max ΔT (°C) |
|---|---|---|---|
| 1 C | 20 | 29.58 | 2.22 |
| 25 | 33.05 | 2.65 | |
| 30 | 36.49 | 3.09 | |
| 2 C | 20 | 39.28 | 3.60 |
| 25 | 42.79 | 4.35 | |
| 30 | 46.39 | 5.11 | |
| 3 C | 20 | 46.11 | 4.56 |
| 25 | 49.66 | 5.55 | |
| 30 | 53.34 | 6.54 |
Lowering coolant temperature by 5 °C reduced the maximum temperature by approximately 3.5 °C, independent of charge rate. However, the effect on temperature uniformity was modest. We chose 20 °C as the optimal coolant temperature, balancing cooling performance against chiller power requirements.
| Charge rate | Channels | Max temp (°C) | Max ΔT (°C) |
|---|---|---|---|
| 1 C | 5 | 32.59 | 2.10 |
| 7 | 29.58 | 2.22 | |
| 9 | 28.69 | 2.31 | |
| 2 C | 5 | 43.81 | 3.38 |
| 7 | 39.28 | 3.60 | |
| 9 | 37.96 | 3.74 | |
| 3 C | 5 | 51.82 | 4.18 |
| 7 | 46.11 | 4.56 | |
| 9 | 44.54 | 4.72 |
Increasing channel count from 5 to 7 provided significant cooling improvement (9-11% reduction in maximum temperature), but further increasing to 9 gave diminishing returns (~3% reduction). At the same time, the temperature difference increased slightly. Considering manufacturing complexity, 7 channels was selected.
| Charge rate | Inlet orientation | Max temp (°C) | Max ΔT (°C) |
|---|---|---|---|
| 1 C | Same-side | 29.68 | 2.31 |
| Staggered | 29.58 | 2.22 | |
| 2 C | Same-side | 39.41 | 3.89 |
| Staggered | 39.28 | 3.60 | |
| 3 C | Same-side | 46.29 | 4.98 |
| Staggered | 46.11 | 4.56 |
Staggered inlet arrangement (alternating inlet direction for adjacent channels) had negligible effect on maximum temperature but significantly reduced the temperature difference, especially at high discharge rates. The improvement became more pronounced at 3 C (0.42 °C reduction in ΔT). Therefore, staggered inlets were adopted as part of the optimal design.
Combining all findings, the optimal design combination for constant current operation was: coolant flow rate 2 cm³/s, coolant temperature 20 °C, 7 channels per plate, and staggered inlet configuration. Under this set, the battery module achieved maximum temperatures of 29.58 °C, 39.28 °C, and 46.11 °C at 1 C, 2 C, and 3 C respectively, with corresponding temperature differences of 2.22 °C, 3.60 °C, and 4.56 °C.
We then evaluated the performance of this optimized liquid cooling system under two real-world energy storage battery operating profiles: frequency regulation (rapid charge/discharge switching every ~60 s) and peak shaving (prolonged charging/discharging periods). The profiles were adopted from literature and actual storage station data. The ambient and coolant initial temperature were set to 25 °C. Figure 1 shows the current profile used. (We provide a conceptual link to the actual figure representing the typical duty cycles.)
During the frequency regulation test, the liquid cooling system reduced the maximum battery module temperature by 5.28 °C (10.74%) compared to natural convection, and the temperature difference decreased by 2.79 °C (44.1%). In the peak shaving test, the improvements were 5.13 °C (10.86%) in maximum temperature and 1.91 °C (37.45%) in temperature difference. The detailed temperature evolution showed that the liquid cooling system maintained battery temperature within the safe operating range throughout both cycles.
Recognizing that constant coolant flow leads to unnecessary energy consumption during low-load periods, we proposed a variable flow rate strategy for the peak shaving profile. The strategy adjusted the coolant flow rate based on the instantaneous charge/discharge rate: when the rate was low (<0.5 C), we reduced the flow to 0.5 cm³/s; when the rate was high (>1.5 C), we increased the flow to 2.5 cm³/s; otherwise, the flow remained at the baseline 2 cm³/s. Table 7 compares the performance of constant flow (2 cm³/s) versus the variable flow strategy.
| Parameter | Constant flow (2 cm³/s) | Variable flow |
|---|---|---|
| Maximum temperature (°C) | 42.11 | 39.84 |
| Maximum temperature difference (°C) | 3.19 | 3.27 |
| Total coolant volume consumed (cm³) | 49,846.15 | 44,707.52 |
The variable flow strategy actually reduced the maximum temperature by 2.27 °C (5.39%) compared to constant flow, because it supplied higher cooling power exactly when needed. The temperature uniformity was slightly degraded (ΔT increased by 0.08 °C, or 2.51%), but still well within acceptable limits. More importantly, the total coolant consumption was reduced by 5,138.64 cm³, a saving of 10.3%, leading to lower pumping energy and thermal management system cost.
In summary, we have systematically optimized a liquid cooling system for energy storage battery modules operating under both constant-current and realistic grid-service conditions. The key findings are:
- Longitudinally placed cooling plates (Design 1) significantly outperform transversely placed ones due to directional thermal conductivity of prismatic cells.
- Optimal parameters for constant current operation: 2 cm³/s flow rate, 20 °C coolant, 7 channels, staggered inlets.
- The optimized system reduces maximum temperature by over 10% and temperature difference by over 35% compared to natural convection in real peak shaving and frequency regulation applications.
- A variable flow rate strategy, adjusting cooling based on instantaneous load, further improves cooling performance while reducing coolant consumption by 10.3%.
These results provide a robust design guideline for efficient and safe thermal management of large-scale energy storage battery systems, contributing to the wider adoption of lithium-ion energy storage battery in grid applications.