In the global transition toward renewable energy, the role of battery energy storage systems (BESS) has become increasingly critical for balancing supply and demand, ensuring grid stability, and enabling flexible energy dispatch. Among various storage technologies, lithium-ion batteries dominate due to their high energy density, long cycle life, and declining cost. However, the thermal behavior of lithium-ion batteries is highly sensitive to operating temperature. Elevated temperatures and large temperature gradients can accelerate capacity fade, increase internal resistance, and even trigger thermal runaway, posing severe safety risks. Therefore, efficient thermal management is essential for the reliable and safe operation of large-scale BESS.
Air cooling remains the most widely adopted thermal management approach for container-level BESS due to its simplicity, low cost, and ease of maintenance. Nevertheless, conventional air-cooled systems often suffer from poor airflow distribution, leading to localized hot spots and excessive temperature differences among cells. These issues become more pronounced as the scale of the storage system increases. Previous studies have focused on optimizing airflow paths within small battery packs or individual racks, but systematic investigations on the full container level—especially the influence of cold aisle configuration, air supply mode, and exhaust layout—are still limited.
In this work, we present a comprehensive numerical study on the thermal performance of a container-level BESS using a three-dimensional computational fluid dynamics (CFD) model. We focus on three key design parameters: the cold aisle structure (open versus closed), the air supply mode (underfloor supply versus inter-row air conditioning supply), and the exhaust outlet location (side versus rear). The objective is to identify the optimal configuration that minimizes both the maximum cell temperature and the temperature difference across the battery array. To ensure the reliability of our simulations, we first validate the CFD model against experimental data from a similar air-cooled system. We then systematically compare the temperature fields and flow patterns under different design scenarios. Our findings provide actionable guidelines for improving the thermal management of container-level BESS, enhancing both safety and operational efficiency.
Model Description and Methodology
Geometric Model
We modeled a standard 40-foot container-level BESS with internal dimensions of 4000 mm × 2440 mm × 2590 mm (length × width × height). Inside the container, two rows of battery racks are placed symmetrically, with a central cold aisle that extends from the floor to the top of the racks. The cold aisle has dimensions of 2650 mm × 1180 mm × 1960 mm. Each rack consists of multiple shelves, and each shelf houses a battery module containing 12 prismatic lithium-ion cells (LF50K, rated 50 Ah, 3.2 V). The cells are arranged with a 5 mm gap between them to allow airflow. Four air conditioning units are evenly distributed between the two rows for the inter-row supply cases. The container also features five underfloor air inlets (each 530 mm × 530 mm) and side/rear exhaust openings of size 1180 mm × 400 mm. Due to symmetry about the mid-plane along the length direction, we simulated only half of the container to reduce computational cost.
The computational domain includes the cold aisle, battery racks, air inlets, and exhaust openings. All walls are assumed adiabatic except for the inlet and outlet boundaries. The battery cells are treated as solid blocks with anisotropic thermal conductivity, and the heat generated during 3C discharge (a high-rate scenario) is applied as a uniform volumetric heat source.
Heat Generation Measurement
To accurately represent the heat generation rate of the battery cells, we performed experimental measurements using K-type thermocouples attached to three locations on the cell surface (positions T1, T2, and T3). The fully charged cells were placed in a temperature-controlled chamber (Jumbo, ECT-408-00-CP-SD) at an initial temperature of 25 °C. The cells were discharged at rates of 1C, 2C, and 3C, and the average temperature rise was recorded. At 3C discharge, the temperature rise reached 12.47 °C over the discharge period, yielding a volumetric heat generation rate of 37,959.2 W/m³. This value was calculated using the formula:
$$q = \frac{m C_p \Delta T}{t \cdot V}$$
where m is the cell mass (kg), Cp is the specific heat (J/(kg·K)), ΔT is the temperature rise (°C), t is the discharge time (s), and V is the cell volume (m³). This experimental heat source was then used in all subsequent CFD simulations.
Governing Equations and Boundary Conditions
The heat transfer within the battery cells is governed by the anisotropic heat conduction equation:
$$\rho_b C_p \frac{\partial T}{\partial t} = \frac{\partial}{\partial x}\left( \lambda_x \frac{\partial T}{\partial x}\right) + \frac{\partial}{\partial y}\left( \lambda_y \frac{\partial T}{\partial y}\right) + \frac{\partial}{\partial z}\left( \lambda_z \frac{\partial T}{\partial z}\right) + q$$
where ρb is the density of the battery material (1934 kg/m³), Cp is the specific heat (1399 J/(kg·K) for the main body), and λx, λy, λz are the thermal conductivities in the three directions (1.0, 10.0, 21.0 W/(m·K) respectively). The thermophysical properties are summarized in Table 1.
| Parameter | Value |
|---|---|
| Thermal conductivity λx (W·m⁻¹·K⁻¹) | 1.0 |
| Thermal conductivity λy (W·m⁻¹·K⁻¹) | 10.0 |
| Thermal conductivity λz (W·m⁻¹·K⁻¹) | 21.0 |
| Specific heat (body/positive/negative) (J·kg⁻¹·K⁻¹) | 1399/871/381 |
| Rated voltage (V) | 3.2 |
| Rated capacity (Ah) | 50 |
| Density (body/positive/negative) (kg·m⁻³) | 1934/2719/8978 |
| Internal resistance (mΩ) | ≤1.5 |
| Negative electrode material | Graphite |
| Positive electrode material | LiNiO₂+LiCoO₂+Li₂MnO₂ |
| Cell height (mm) | 185 |
| Cell width (mm) | 135 |
| Cell thickness (mm) | 30 |
| Terminal diameter (mm) | 12 |
| Terminal height (mm) | 5 |
The fluid flow and heat transfer in the air domain are described by the three-dimensional, transient Navier-Stokes equations with the k-ε turbulence model. The governing equations are:
Continuity:
$$\frac{\partial \rho}{\partial t} + \nabla \cdot (\rho \mathbf{u}) = 0$$
Momentum:
$$\frac{\partial}{\partial t}(\rho \mathbf{u}) + \nabla \cdot (\rho \mathbf{u u}) = -\nabla P + \rho \mathbf{g}$$
Energy:
$$\frac{\partial T}{\partial t} + \nabla \cdot (\mathbf{u} T) = \frac{\lambda}{\rho C_p} \nabla^2 T$$
where ρ is air density (1.2 kg/m³), u is velocity vector (m/s), P is pressure (Pa), g is gravitational acceleration (9.8 m/s²), λ is thermal conductivity of air (0.024 W/(m·K)), and Cp is specific heat of air (1006.4 J/(kg·K)). The air properties are given in Table 2.
| Property | Value |
|---|---|
| Density (kg·m⁻³) | 1.2 |
| Specific heat (J·kg⁻¹·K⁻¹) | 1006.4 |
| Thermal conductivity (W·m⁻¹·K⁻¹) | 2.4×10⁻² |
| Dynamic viscosity (kg·m⁻¹·s⁻¹) | 1.8×10⁻⁵ |
| Thermal expansion coefficient (K⁻¹) | 3.4×10⁻³ |
Boundary conditions: the inlet air temperature is 25 °C, the outlet is at ambient pressure (1 atm), all external walls are adiabatic, and a no-slip condition is applied at all solid surfaces. Gravity is considered along the y-direction (vertical). The initial temperature of the entire domain is 25 °C.
Validation of the CFD Model
We validated our simulation approach by reproducing the experimental configuration reported in a previous study on air-cooled data centers (which share similar airflow characteristics with BESS). The comparison between our simulation results and the measured temperature rise above the supply tile (T–Ttile) at the cabinet inlet and outlet is shown in Table 3. The maximum deviation is less than 1.5 °C, confirming that the k-ε model with the selected boundary conditions can accurately capture the thermal behavior of the air-cooled system.
| Height (m) | Inlet (Simulated) (°C) | Inlet (Experimental) (°C) | Outlet (Simulated) (°C) | Outlet (Experimental) (°C) |
|---|---|---|---|---|
| 0.2 | 0.5 | 0.6 | 3.2 | 3.0 |
| 0.5 | 1.8 | 2.0 | 6.8 | 6.5 |
| 0.8 | 3.5 | 3.7 | 10.5 | 10.8 |
| 1.1 | 6.2 | 6.0 | 14.2 | 14.0 |
| 1.4 | 9.8 | 10.1 | 16.8 | 16.5 |
Results and Discussion
Effect of Cold Aisle Structure
The cold aisle is the space between the two rows of battery racks. In an open cold aisle configuration, the top and sides of the aisle are open, allowing cold and hot air to mix freely. In a closed configuration, the top and both sides are sealed so that the cold air is confined to the aisle before entering the battery racks. We compared the airflow and temperature fields for both cases, keeping all other conditions identical (underfloor supply, side exhaust).
Table 4 presents the airflow velocity and temperature at two characteristic planes. In the closed configuration, the maximum air velocity in the cold aisle is higher (6.4 vs. 5.8 m/s at x=1820 mm, and 10.0 vs. 6.8 m/s at y=1000 mm), and the air temperature is lower (29.6 vs. 33.4 °C at y=1000 mm). This indicates that closing the cold aisle reduces mixing with warm return air, thereby delivering cooler and faster air to the battery racks.
| Cold Aisle Structure | Plane Location | Max Velocity (m/s) | Max Temperature (°C) | Max Temperature Difference (°C) |
|---|---|---|---|---|
| Open | x=1820 mm | 5.8 | 39.8 | 14.8 |
| Closed | x=1820 mm | 6.4 | 39.5 | 14.5 |
| Open | y=1000 mm | 6.8 | 33.4 | 8.4 |
| Closed | y=1000 mm | 10.0 | 29.6 | 4.6 |
The impact on cell surface temperatures is summarized in Table 5. The closed cold aisle reduces the maximum cell temperature from 44.6 °C to 43.7 °C, and the maximum temperature difference across a single cell drops from 6.1 °C to 5.2 °C—a reduction of 15%. The average cell temperature also decreases by 0.8 °C. The improvement is attributed to the elimination of hot air recirculation into the cold aisle, which enhances the uniformity of the inlet air temperature to the battery racks. Velocity vector plots reveal that under the open configuration, strong recirculation zones form at the top of the cold aisle, mixing warm exhaust air with the incoming cold air. Sealing the aisle suppresses these vortices, allowing more cold air to reach the lower racks.
| Cold Aisle Structure | Max Cell Temperature (°C) | Max Cell Temperature Difference (°C) | Average Cell Temperature (°C) |
|---|---|---|---|
| Open | 44.6 | 6.1 | 29.4 |
| Closed | 43.7 | 5.2 | 28.6 |
Based on these findings, all subsequent investigations were performed with the closed cold aisle configuration.
Effect of Air Supply Mode
We compared two common air supply schemes: underfloor air supply (UFS) and inter-row air conditioning (AC) supply. In the UFS mode, five floor inlets deliver cold air directly into the cold aisle from below. In the AC mode, four air conditioning units located between the racks blow cold air horizontally into the cold aisle. All other parameters (closed cold aisle, side exhaust) were kept constant.
Table 6 compares the airflow characteristics at the y=1000 mm plane. Although the AC mode produces a higher maximum velocity (10.6 vs. 10.0 m/s), it also results in a higher maximum air temperature (30.6 vs. 29.6 °C) and a larger temperature difference (5.6 vs. 4.6 °C). The reason is that the horizontal air jets from the AC units create strong recirculation zones near the inlet of the battery racks located adjacent to the AC units. These vortices impede the entry of cold air into certain racks, leading to uneven distribution of cooling flow.
| Supply Mode | Max Velocity (m/s) | Max Temperature (°C) | Max Temperature Difference (°C) |
|---|---|---|---|
| Inter-row AC | 10.6 | 30.6 | 5.6 |
| Underfloor supply | 10.0 | 29.6 | 4.6 |
The resulting cell temperatures are presented in Table 7. The underfloor supply yields a lower maximum cell temperature (43.7 vs. 44.1 °C), a lower average temperature (28.6 vs. 28.7 °C), and a substantially smaller cell temperature difference (5.2 vs. 5.6 °C). The improvement is 7.1% in terms of temperature uniformity. The underfloor supply provides a more uniform upward flow, which naturally aligns with the vertical arrangement of the racks, whereas the horizontal AC jets cause flow separation and non-uniform cooling.
| Supply Mode | Max Cell Temperature (°C) | Max Cell Temperature Difference (°C) | Average Cell Temperature (°C) |
|---|---|---|---|
| Inter-row AC | 44.1 | 5.6 | 28.7 |
| Underfloor supply | 43.7 | 5.2 | 28.6 |
Effect of Exhaust Outlet Location
Finally, we investigated the influence of the exhaust outlet position. Three configurations were considered: outlet on the left side (I), outlet on the right side (II), and outlet on the rear side (III) of the container. The dimensions of all outlets were identical (1180 mm × 400 mm). The cold aisle was closed, and underfloor supply was used in all cases.
Table 8 summarizes the airflow parameters at the x=1820 mm plane. The rear exhaust (III) yields the lowest maximum air temperature (39.1 °C) and the maximum temperature difference is comparable to the side outlets. However, the maximum velocity is much higher (23.8 m/s) because the rear exhaust creates a shorter flow path, reducing resistance and increasing local velocities.
| Outlet Location | Max Velocity (m/s) | Max Temperature (°C) | Max Temperature Difference (°C) |
|---|---|---|---|
| Left side (I) | 6.4 | 39.5 | 14.5 |
| Right side (II) | 6.5 | 39.6 | 14.6 |
| Rear side (III) | 23.8 | 39.1 | 14.6 |
The cell temperature results are given in Table 9. The rear exhaust achieves the best thermal performance: the maximum cell temperature is 42.9 °C, which is 0.8 °C lower than the left-side outlet and 1.1 °C lower than the right-side outlet. The maximum temperature difference across a single cell is only 4.9 °C, representing a 4.1% reduction compared to the left-side outlet (5.2 °C) and a 3.2% reduction compared to the right-side outlet (5.1 °C). The average cell temperature is also slightly lower (28.3 °C vs. 28.6/28.2 °C).
| Outlet Location | Max Cell Temperature (°C) | Max Cell Temperature Difference (°C) | Average Cell Temperature (°C) |
|---|---|---|---|
| Left side (I) | 43.7 | 5.2 | 28.6 |
| Right side (II) | 44.0 | 5.1 | 28.2 |
| Rear side (III) | 42.9 | 4.9 | 28.3 |
The superiority of the rear exhaust can be explained by the flow pattern. With side outlets, the hot air has to travel laterally across the entire width of the container before exiting, leading to accumulation of heat in the far corners. In contrast, the rear outlet is positioned directly behind the battery racks, allowing a straight, short path for the hot air to escape. This reduces the residence time of hot air in the rack region and enhances convective heat transfer from the cells. Although the rear exhaust causes a higher air velocity near the outlet, which might increase pressure drop, the overall thermal benefit is clear. In practice, the pressure drop can be managed by adjusting the fan speed.
Discussion
Our study systematically examined three critical design parameters for air-cooled container-level BESS. The results consistently show that a closed cold aisle combined with underfloor air supply and a rear exhaust outlet provides the best thermal management, limiting the maximum cell temperature to 42.9 °C and the cell-to-cell temperature difference to 4.9 °C under a 3C discharge condition. This configuration improves temperature uniformity by over 15% compared to the worst-case scenario (open aisle, inter-row supply, side exhaust).
The benefits of the closed cold aisle are twofold: (1) it isolates the cold supply air from the hot return air, maintaining a low and uniform inlet temperature; (2) it increases the velocity of the air entering the racks by preventing leakage through the sides. Underfloor supply is preferred over inter-row AC because the vertical buoyancy-driven flow works in harmony with the upward movement of heated air, whereas horizontal jets create disruptive vortices. The rear exhaust minimizes the flow path length, reducing heat buildup and improving heat transfer.
It is worth noting that our simulations assumed a homogeneous heat generation rate of 37,959.2 W/m³, which corresponds to a 3C discharge. In real applications, the heat generation varies with state of charge, depth of discharge, and ambient conditions. However, the comparative trends observed here are expected to hold for other operating conditions. Additionally, we did not consider the effect of fan curves or pressure drops; a future study could incorporate fan performance to optimize the system energy consumption.
Conclusion
In this work, we developed a validated CFD model to investigate the thermal performance of a container-level battery energy storage system. The main conclusions are:
- Closing the cold aisle significantly improves airflow uniformity and reduces the maximum cell temperature difference by 15% (from 6.1 °C to 5.2 °C) compared to an open cold aisle.
- Underfloor air supply outperforms inter-row air conditioning supply, yielding a 7.1% smaller cell temperature difference (5.2 °C vs. 5.6 °C) and a lower average temperature.
- Positioning the exhaust outlet on the rear side of the container provides the best cooling performance, achieving a maximum cell temperature of 42.9 °C and a cell temperature difference of 4.9 °C—improvements of 4.1% and 3.2% over side outlets.
The optimal configuration (closed cold aisle + underfloor supply + rear exhaust) demonstrates superior thermal management capability, which is crucial for the safe and efficient operation of large-scale lithium-ion battery energy storage systems. These findings offer practical guidance for the design and retrofit of container-level BESS, contributing to enhanced reliability and longer service life.