In modern electric vehicles, the traction battery is the core energy storage component, and its performance is highly sensitive to operating temperature. Lithium-ion batteries are widely used in traction battery packs because of their high energy density and long cycle life. However, excessive temperature or non-uniform temperature distribution can lead to capacity degradation, internal short circuit, thermal runaway, and even fire. I therefore focused my research on the thermal management of a large-format lithium-ion traction battery pack, aiming to design a liquid cooling plate that can keep the maximum temperature below 35°C and the temperature difference under 5°C. To achieve this, I combined experimental characterization, computational fluid dynamics, single-factor analysis, surrogate modelling, and multi-objective optimization.
The traction battery pack investigated in this work has a nominal capacity of 139.2 kWh, and it is formed by a 2P88S configuration. In the pack, prismatic lithium-ion cells are placed in two layers, and each layer is independently cooled by liquid cold plates. Because simulating the entire pack is computationally expensive, I selected one layer of the battery pack as the research object. The cold plate designs studied in this thesis include conventional accordion channels, side-mounted plates, bottom-mounted plates, and finally rib-groove enhanced plates. The main objective was to improve the cooling performance and temperature uniformity of the traction battery without increasing the pressure drop and pumping power.
I first characterized the thermal properties of a 190 Ah prismatic ternary lithium-ion cell. The active cell has dimensions of 146 mm × 80 mm × 103 mm and a nominal voltage of 3.8 V. I used a hybrid pulse power characterization test to measure the internal resistances at different states of charge and different current rates. The heat generation rate was calculated according to the Bernardi equation, and the temperature rise was measured with thermocouples. Single-cell simulations were then performed in Fluent, and I compared the numerical temperature rise curves with experimental data. The errors between the simulation and the measurement were below 8%, which confirmed that the thermal model was reliable. After that, I established a battery pack model including the cells, cold plates, and thermal pads.
The heat transfer within the cell is governed by the energy conservation equation. For the fluid domain, the continuity and momentum equations are given by:
$$ \nabla \cdot \vec{U} = 0 $$
$$ \rho_{f} \left( \vec{U} \cdot \nabla \vec{U} \right) = -\nabla P + \nabla \cdot \left( \mu_{f} \nabla \vec{U} \right) $$
where $\rho_f$ is the fluid density, $\vec{U}$ is the velocity vector, $P$ is the pressure, and $\mu_f$ is the dynamic viscosity. The energy equation for the fluid is:
$$ \rho_f c_{p,f} \vec{U} \cdot \nabla T_f = \nabla \cdot \left( \lambda_f \nabla T_f \right) $$
while the solid domain satisfies:
$$ \nabla^2 T_s = 0 $$
I treated the heat source in the battery as a volumetric heat generation rate $q$, which is a fitted function of discharge time. The total heat generation inside the cell can be expressed as the sum of reaction heat, ohmic heat, polarization heat, and side-reaction heat. In normal operation, the side-reaction heat is negligible. Using the HPPC data, I derived the empirical formula for the heat generation rate at different C-rates:
$$ q = A_1 t^6 + A_2 t^5 + A_3 t^4 + A_4 t^3 + A_5 t^2 + A_6 t + A_7 $$
where $t$ is the discharge time in seconds, and $A_1$ to $A_7$ are polynomial coefficients. For the 2C discharge, the coefficients are presented below.
| C-rate | A1 | A2 | A3 | A4 | A5 | A6 | A7 |
|---|---|---|---|---|---|---|---|
| 0.5C | -4.76e-18 | 9.97e-14 | -7.86e-10 | 2.9e-6 | -4.87e-3 | 2.41 | 4001 |
| 1C | -3.5e-16 | 3.94e-12 | -1.62e-8 | 3.03e-5 | -2.37e-2 | 2.48 | 12004 |
| 2C | -2.02e-13 | 9.99e-10 | -1.77e-6 | 1.34e-3 | 0.382 | 9.23 | 40084 |
The thermophysical properties of the cell were obtained from the mass-weighted average of the components. The battery density is 2478.5 kg/m³, the specific heat capacity is 1100 J/(kg·K), and the thermal conductivities in the x, y, and z directions are 18.2, 2.0, and 18.2 W/(m·K), respectively. The transverse conductivity is much smaller than the in-plane conductivity because of the multilayer configuration of the electrode and separator. This anisotropy significantly influences the temperature gradient inside the traction battery cell. In the finite volume model, I used a poly-hexcore mesh and refined the boundary layers near the fluid solid interface. The cell surface heat transfer coefficient to the environment was set to 5 W/(m²·K).
For the battery pack simulation, I selected 50% water-ethylene-glycol as the coolant. The coolant density is 1071.1 kg/m³, its specific heat is 3300 J/(kg·K), the thermal conductivity is 0.384 W/(m·K), and the dynamic viscosity is 0.017 Pa·s at 25°C. The cold plate material is aluminum alloy 6061 with a thermal conductivity of 180 W/(m·K), and the thermal pad has a thermal conductivity of 2 W/(m·K). The coolant inlet temperature was fixed at 25°C, and the initial battery temperature was also 25°C. The total volume flow rate was initially set to 21.78 L/min for the whole battery pack, which corresponds to 2.42 L/min for each side-mounted cold plate.
In the first stage of my study, I compared two locations of the cold plates in the traction battery pack: bottom-mounted and side-mounted. The bottom-mounted configuration, commonly seen in vehicle battery packs, places the cold plates under the modules. The side-mounted configuration inserts the cold plates between the battery modules, which greatly increases the heat transfer area and utilizes the high in-plane thermal conductivity of the cell. Figure 1 shows the battery pack concept with the side-mounted liquid cold plates. I used the same accordion flow channel geometry for both positions so that the comparison was fair.
At the end of a 2C discharge process, the bottom-mounted cold plate battery pack reached a maximum temperature of 39.67°C and a maximum temperature difference of 13.27°C. The cell-level temperature difference was as high as 12.15°C, which greatly violates the design target. The side-mounted cold plate battery pack, on the other hand, reduced the maximum temperature to 35.6°C and the maximum pack temperature difference to 10.46°C. The cell-level temperature difference also decreased to 6.9°C. This improvement occurred because the side-mounted cold plate directly cools the large area of the cell surface where the in-plane thermal conductivity is high. The bottom-mounted cold plate mostly conducts heat through the small thickness direction of the cell, which has a low thermal conductivity of about 2 W/(m·K). As a result, a large vertical temperature gradient remains in the cell.
| Cooling layout | Tmax pack (°C) | ΔT cell (°C) | ΔT pack (°C) |
|---|---|---|---|
| Bottom accordion plate | 39.67 | 12.15 | 13.27 |
| Side accordion plate | 35.60 | 6.90 | 10.46 |
Since the maximum temperature of the side-mounted scheme was still close to 35°C and the temperature uniformity was not sufficient, I further optimized the coolant inlet and outlet positions. In the original side-mounted plate, all inlets were on one side and all outlets on the other side. In the improved design, the inlets and outlets were staggered in alternate cold plates. The inlet and outlet locations can significantly affect the coolant flow distribution and thus the temperature pattern. I compared the two configurations using the same accordion cold plate and the same boundary conditions. The staggered arrangement lowered the pack maximum temperature by 1.1°C and the pack temperature difference by 1.96°C. The cell-level temperature difference remained nearly the same, changing from 6.90°C to 6.93°C. The temperature distribution became more symmetric, and the high-temperature region was no longer concentrated downstream of the coolant flow.
| Inlet/outlet arrangement | Tmax pack (°C) | ΔT cell (°C) | ΔT pack (°C) |
|---|---|---|---|
| Same side | 35.60 | 6.90 | 10.46 |
| Staggered | 34.50 | 6.93 | 8.50 |
Although the staggered side-mounted cold plate produced better performance, the maximum temperature of 34.5°C was still very close to the 35°C limit, and the pack temperature difference remained above 8°C. In order to further increase the heat transfer coefficient without excessively increasing the pumping power, I introduced pin-fin-like convex ribs and concave grooves into the flow channels. These structures are common passive heat transfer enhancement elements. Convex ribs accelerate the fluid, break the boundary layer, and promote swirling motion. Concave grooves increase the heat transfer area and reduce the flow resistance compared with ribs. However, recessed cavities may cause local fluid stagnation inside the cavities, which increases the thermal boundary layer thickness.
I studied four different cross-sectional shapes for both ribs and grooves: circular, diamond, forward water-drop, and backward water-drop. The geometry parameters were kept the same: the distance between adjacent ribs/grooves along the flow direction was 25 mm, the diameter was 10 mm, and the height/depth was 2 mm. The volume flow rate was 2.42 L/min. I evaluated the cooling performance in terms of the maximum cold plate surface temperature, the standard deviation of the surface temperature, the pressure drop, the Nusselt number ratio, the friction factor ratio, and the PEC performance evaluation criterion. The PEC index is defined as:
$$ PEC = \frac{Nu / Nu_0}{(f/f_0)^{1/3}} $$
where $Nu_0$ and $f_0$ correspond to the original accordion channel without ribs or grooves. A PEC value larger than one indicates that the enhanced structure provides better overall thermohydraulic performance than the original channel.
The Nusselt number is calculated from the average heat transfer coefficient:
$$ Nu_{aver} = \frac{h_{aver} D_h}{\lambda_f} $$
and the hydraulic diameter is:
$$ D_h = \frac{2 H_{ch} W_{ch}}{H_{ch}+W_{ch}} $$
where $H_{ch}$ and $W_{ch}$ are the height and width of the rectangular channel. The friction factor is calculated as:
$$ f = \frac{2 \Delta p D_h}{\rho L u^2} $$
with $L$ being the channel length and $u$ the mean velocity.
For the shape comparison, I observed that all rib and groove structures gave lower surface temperatures than the original accordion plate. Circular ribs produced the lowest maximum temperature, followed by the two water-drop shapes, while the diamond ribs produced the highest temperature. The maximum temperature reduction was about 3°C for circular ribs, as can be seen in the following table. Concerning the temperature uniformity, the standard deviation of the cold plate surface temperature was generally lower for the original plate than for the ribbed or grooved plates. In general, concave grooves produced a smaller standard deviation than convex ribs. The best temperature uniformity among the enhanced structures was obtained with the diamond grooves.
| Structure | Tmax plate (°C) | Tsd (°C) | ΔP (Pa) | Nu/Nu0 | f/f0 | PEC |
|---|---|---|---|---|---|---|
| Original accordion | 33.80 | 0.980 | 293.3 | 1.00 | 1.00 | 1.00 |
| Circular rib | 31.64 | 1.340 | 394.7 | 1.59 | 1.22 | 1.49 |
| Diamond rib | 32.05 | 1.360 | 347.1 | 1.42 | 1.13 | 1.36 |
| Forward drop rib | 31.80 | 1.350 | 382.5 | 1.51 | 1.20 | 1.42 |
| Backward drop rib | 31.90 | 1.355 | 375.4 | 1.49 | 1.18 | 1.41 |
| Circular groove | 32.15 | 1.080 | 258.8 | 1.44 | 0.78 | 1.56 |
| Diamond groove | 32.60 | 1.020 | 242.6 | 1.30 | 0.72 | 1.45 |
| Forward drop groove | 32.40 | 1.050 | 248.9 | 1.36 | 0.75 | 1.50 |
| Backward drop groove | 32.50 | 1.040 | 251.2 | 1.34 | 0.76 | 1.47 |
Although the convex ribs produced stronger heat transfer enhancement than concave grooves, they also produced a much higher pressure drop. The circular rib plate had a pressure drop of 394.7 Pa, which was 34.6% higher than the original accordion plate. The circular groove plate, by contrast, had a pressure drop of 258.8 Pa, which was 11.8% lower than the original plate. Because the friction factor of the grooves was lower than one, the PEC index was consistently higher for grooves than for ribs. The circular groove obtained the best PEC of 1.56, while the circular rib obtained a PEC of 1.49. Considering both the cooling performance and the energy consumption, I decided to adopt the circular shape for both ribs and grooves. However, a single design with either pure ribs or pure grooves is not perfect. Clean ribs provide excellent heat transfer but consume extra pumping power; clean grooves save energy but have weaker heat transfer. To combine the advantages of both geometries, I proposed a hybrid liquid cold plate with staggered rib and groove structures. Along the flow direction, circular ribs and circular grooves are arranged alternately in the parallel channels. This staggered configuration allows the coolant to be alternately accelerated by the ribs and relaxed in the grooves, which may lead to a favorable balance between heat transfer and pressure drop.
Before performing the multi-objective optimization, I carried out a single-factor analysis of the volume flow rate, rib/groove spacing, and rib/groove diameter. First, I kept the geometry fixed (spacing = 25 mm and diameter = 10 mm) and varied the volume flow rate from 1.21 L/min to 3.62 L/min. The results showed that increasing the flow rate always reduces the maximum temperature. For instance, the maximum temperature of the circular rib plate dropped from 35.15°C at 1.21 L/min to 30.2°C at 3.62 L/min. However, the pressure drop increased superlinearly with the flow rate. The average Nusselt number ratio Nu/Nu0 increased slightly with the flow rate, while the friction factor ratio f/f0 remained almost stable. The PEC of both ribs and grooves decreased as the flow rate increased, which means that excessive flow is not beneficial from the viewpoint of overall efficiency. A reasonable operating point is 2.42 L/min, because further increasing the flow rate provides only a marginal temperature decrease while causing a large pressure drop penalty.
| Flow rate (L/min) | Tmax rib (°C) | Tmax groove (°C) | ΔP rib (Pa) | ΔP groove (Pa) | PEC rib | PEC groove |
|---|---|---|---|---|---|---|
| 1.21 | 35.15 | 35.32 | 174.5 | 112.2 | 1.58 | 1.72 |
| 1.81 | 33.40 | 33.58 | 255.3 | 164.9 | 1.52 | 1.64 |
| 2.42 | 31.64 | 32.15 | 394.7 | 258.8 | 1.49 | 1.56 |
| 3.02 | 30.85 | 31.05 | 542.1 | 367.5 | 1.43 | 1.49 |
| 3.62 | 30.20 | 30.35 | 701.3 | 482.6 | 1.40 | 1.45 |
Second, I varied the spacing between the pin centers from 35 mm to 25 mm. Smaller spacing means a larger number of ribs/grooves in each channel. For the circular rib plate, decreasing the spacing from 35 mm to 25 mm reduced the maximum surface temperature by 0.12°C and increased the Nu/Nu0 ratio from 1.46 to 1.60. The pressure drop increased slightly from about 376 Pa to 395 Pa. For the circular groove plate, the maximum temperature also decreased, but the pressure drop decreased unexpectedly because the enlarged cavities reduce the average flow velocity and wall shear. The PEC values increased with decreasing spacing for both ribs and grooves. Thus, the smallest spacing of 25 mm was selected.
Third, I fixed the spacing at 25 mm and changed the circular diameter from 6 mm to 10 mm. A larger diameter increases the projected area of the ribs/grooves. The maximum temperature decreased from 31.81°C at 6 mm to 31.65°C at 10 mm for the rib plate. The Nu/Nu0 ratio for the rib plate increased from 1.34 to 1.60 when the diameter increased from 6 mm to 10 mm. The pressure drop of the rib plate rose by nearly 30%, but the PEC also rose, showing that the heat transfer enhancement outweighed the extra pumping power. For the groove plate, the diameter had a weaker influence because the grooves produced lower disturbance. Larger grooves slightly increased the heat transfer while reducing the pressure drop. Based on these results, I selected a diameter of 10 mm for both ribs and grooves.
After the single-factor analysis, the following geometric parameters were fixed: circular shape, spacing = 25 mm, diameter = 10 mm, flow rate = 2.42 L/min. The remaining parameters to be optimized were the local heights of the convex ribs and the depths of the concave grooves. Since the coolant temperature increases along the flow direction, the heat transfer requirement is not uniform along the channel. It may be more effective to increase the rib height or groove depth near the outlet where the coolant is hotter. I therefore allowed the first and last ribs and grooves to have different heights and depths, while the intermediate values varied linearly from the inlet to the outlet. Specifically, I defined the first rib height $h_1$, the last rib height $h_{19}$, the first groove depth $d_1$, and the last groove depth $d_{18}$. The height of the $i$-th rib is given by:
$$ h_i = h_1 + (i-1)\delta, \quad i=1,2,\ldots,19 $$
where
$$ \delta = \frac{h_{19}-h_1}{18} $$
Similarly, the depth of the $j$-th groove is:
$$ d_j = d_1 + (j-1)\epsilon, \quad j=1,2,\ldots,18 $$
where
$$ \epsilon = \frac{d_{18}-d_1}{17} $$
The first rib height $h_1$ was allowed to vary from 0 mm to 2 mm, the last rib height $h_{19}$ from 0 mm to 2 mm, the first groove depth $d_1$ from -2 mm to 0 mm, and the last groove depth $d_{18}$ from -2 mm to 0 mm. Negative values mean that the grooves are cut from the channel wall. By varying these four variables, the heat transfer structure could be gradually strengthened from the inlet to the outlet, or vice versa, depending on the optimal solution. I used optimal Latin hypercube sampling to generate 84 design points. For each design point, I built a parametric CAD model, meshed it with poly-hexcore elements, and solved the conjugated heat transfer problem using Fluent. The response variables were the maximum temperature of the cold plate surface $T_{\max}$, the standard deviation of the surface temperature $T_{sd}$, and the pressure drop $\Delta p$.
To reduce the computational cost, I constructed a radial basis function surrogate model based on the 84 simulation samples. The accuracy of the surrogate was evaluated by the coefficient of determination $R^2$ between the predicted and simulated responses. The $R^2$ values were 0.90091 for $T_{\max}$, 0.92914 for $T_{sd}$, and 0.96244 for $\Delta p$. All $R^2$ values were greater than 0.9, indicating that the surrogate model had acceptable accuracy for optimization. Then I formulated a multi-objective optimization problem as follows:
$$
\begin{aligned}
\text{Find } & \mathbf{x} = (h_1, h_{19}, d_1, d_{18})^T \\
\text{Minimize } & T_{\max}(\mathbf{x}) \\
\text{Minimize } & T_{sd}(\mathbf{x}) \\
\text{Minimize } & \Delta p(\mathbf{x}) \\
\text{subject to } & 0\le h_1,h_{19}\le2~\text{mm}, \\
&-2 \le d_1,d_{18}\le 0~\text{mm}.
\end{aligned}
$$
I used the NSGA-II genetic algorithm to solve this multi-objective optimization problem. The population size was 20, the number of generations was 100, the crossover probability was 0.8, and the mutation probability was 0.1. The algorithm generated a Pareto front that showed the trade-off between the thermal objective and the hydraulic objective. Because $T_{\max}$, $T_{sd}$, and $\Delta p$ are conflicting, no single design can minimize all of them simultaneously. I selected a balanced solution from the Pareto front with $T_{\max}=31.726$°C, $T_{sd}=1.258$°C, and $\Delta p=276.7$ Pa. Rounding the design variables to practical manufacturing increments produced $h_1 = 0$ mm, $h_{19} = 2$ mm, $d_1 = -0.8$ mm, and $d_{18} = -2$ mm. This means that the ribs are absent near the inlet and grow toward the outlet, while the grooves are shallow near the inlet and become deeper near the outlet. Such a non-uniform structure enhances heat transfer exactly where the temperature is highest, without overly increasing the pressure drop near the inlet.
I then placed this optimal rib-groove cold plate back into the battery pack and performed full pack simulations under a 2C discharge condition. The results are summarized in the table below. Compared with the original accordion side-mounted cold plate with staggered inlet and outlet, the optimized rib-groove cold plate reduced the battery pack maximum temperature from 34.5°C to 33.46°C, a reduction of 3%. The cell-level temperature difference dropped from 6.93°C to 6.23°C, a reduction of 10.1%. The pack temperature difference dropped from 8.5°C to 7.74°C, a reduction of 8.94%. At the same time, the pressure drop of the whole cooling system decreased from 2703.6 Pa to 2460.6 Pa, which is an 8.99% reduction.
| Configuration | Tmax pack (°C) | ΔT cell (°C) | ΔT pack (°C) | ΔP (Pa) |
|---|---|---|---|---|
| Optimal rib-groove staggered plate | 33.46 | 6.23 | 7.74 | 2460.6 |
| Original accordion side plate | 34.50 | 6.93 | 8.50 | 2703.6 |
The optimized traction battery cooling system not only lowered the temperature level but also consumed less pumping power. This is particularly important for long-range electric vehicles because the cooling system consumes electrical energy from the traction battery itself. The simultaneous improvement in thermal performance and hydraulic performance can be explained by the combined effects of ribs and grooves. The ribs promote turbulence and boundary layer renewal in the high-temperature downstream region, whereas the grooves provide pressure recovery and enlarge the heat transfer area near regions where the temperature gradient is less severe. The optimal height/depth distribution also avoids over-disturbing the flow at the inlet where the coolant is still cold.
In addition to the continuous 2C discharge case, I evaluated the final cold plate design under three typical operating conditions: a fast-charging process, a low-temperature heating process, and a high-temperature highway cruise condition. These conditions represent realistic traction battery usage scenarios and present different thermal management challenges.
The fast-charging condition was modeled as a stepped current profile. From 0 to 0.6 state of charge, the cell was charged at 2C. From 0.6 to 0.9 SOC, the charging rate was reduced to 1C, and above 0.9 SOC, it was further reduced to 0.5C. The total charging time was 2880 seconds, and about 80% of the available capacity was restored within 30 minutes. The coolant inlet temperature was 25°C, and the flow rate per cold plate was 2.42 L/min. During the 2C charging stage, the maximum temperature of the battery pack peaked at 30.13°C and the maximum temperature difference reached 4.53°C, both below the recommended thresholds. At the end of the charging process, the maximum temperature declined to 26.45°C and the temperature difference declined to 1.3°C. Therefore, the cooling system can fulfill the thermal requirements during fast charging.
| Fast charging stage | Time (s) | Tmax pack (°C) | ΔT pack (°C) |
|---|---|---|---|
| 2C charge end | 1080 | 30.13 | 4.53 |
| 1C charge end | 1980 | 28.33 | 3.05 |
| Charge complete | 2880 | 26.45 | 1.30 |
For very cold regions, a traction battery must be warmed to at least 0°C before it can be charged at a low rate and to about 10°C before high-rate charging or discharging. I modeled a low-temperature scenario in which the battery, thermal pad, and cold plate were initially at -20°C. The coolant inlet temperature was set to 15°C, and the flow rate was set to 2 m/s in the channels. The cooling/heating system was used to warm the cells until the minimum battery temperature exceeded 10°C. The simulation showed that the minimum temperature of the battery pack rose to 0°C after about 620 s and to 10°C after about 1210 s. The average heating rate was about 1.5°C/min, which exceeds the common requirement of 0.3°C/min. At the end of heating, the minimum cell temperature was 10.23°C while the maximum temperature was 15.03°C. The temperature difference is not critical during heating, but the overall gradient remained within 5°C. This result proves that the rib-groove liquid plate can also function as a heater when warm coolant is supplied.
Finally, I considered a hot summer driving scenario. The vehicle was assumed to be exposed to high ambient temperature for one hour, and then the driver entered a highway and cruised at 120 km/h. The initial temperature of the battery and environment was 40°C. I calculated the required power to maintain 120 km/h based on the vehicle dynamic equation:
$$ P_{req} = \frac{1}{\eta} \left( \frac{G f u}{3600} + \frac{C_D A u^3}{76140} \right) $$
Using the vehicle parameters of an electric off-road SUV, the required power is approximately 65 kW. With a 2P battery configuration, each cell has to discharge at about 0.5C. The coolant inlet temperature was 25°C. The simulation showed that the battery pack temperature decreased quickly from 40°C to below 35°C within 390 s. The maximum temperature difference fell below 5°C within 770 s. After reaching a quasi-steady state, the maximum battery pack temperature was 25.77°C and the maximum temperature difference only 0.66°C. The traction battery was kept well within its optimal operating range. The optimized rib-groove liquid cooling system demonstrated excellent transient response and steady-state performance under the high-temperature highway cruise condition.
The results of the three typical operating condition analyses are summarized below.
| Operating condition | Initial T (°C) | Coolant T (°C) | Key result | Satisfied? |
|---|---|---|---|---|
| Fast charging | 25 | 25 | Tmax = 30.13°C, ΔT = 4.53°C at 2C stage | Yes |
| Low-temperature heating | -20 | 15 | Above 0°C in 620 s, above 10°C in 1210 s | Yes |
| High-temperature cruise | 40 | 25 | Reaches 35°C in 390 s, Tmax = 25.77°C | Yes |
This research has demonstrated that a well-designed rib-groove staggered liquid cold plate can significantly improve the thermal performance of a traction battery pack. The main novelty of this work lies in the spatially graded ribs and grooves, which are optimized to distribute the heat transfer enhancement according to the temperature profile of the traction battery. Traditional cold plates often use uniform flow channels or uniform pin fins over the entire surface. By allowing the rib height and groove depth to increase gradually from the coolant inlet to the outlet, the optimized cold plate effectively equalizes the driving temperature difference and avoids excessive pressure drop. The linear grading laws are simple enough to be manufactured by stamping or milling.
One important contribution of this study is the direct comparison between bottom-mounted and side-mounted liquid cooling plates. Many current electric vehicles use bottom cooling because of assembly simplicity. My simulation indicates that the temperature distribution in the traction battery with a side cold plate is much more uniform than that in the bottom-cooled battery because the large-area side contact exploits the high in-plane thermal conductivity of the electrode layers. The side-mounted arrangement also permits a more compact battery pack because the cold plate serves as both a heat exchanger and a structural spacer. When the cold plate is placed on the side of each cell or module, the thermal path is short and uniform, which suppresses the large vertical temperature gradients normally encountered in bottom-cooled traction batteries.
The numerical model I established can predict the temperature and flow field of the liquid cooling system under different operating conditions with acceptable accuracy. I treated the traction battery cell as a homogeneous anisotropic solid with a volumetric heat source. Such simplification greatly reduces the mesh count but still captures the main heat conduction behavior. The HPPC-derived heat generation function accounts for the variation of internal resistance with state of charge. In future work, an electrochemical-thermal coupled model could improve the prediction near the end of discharge, where the heat generation rises rapidly because of the increase in internal resistance. Additionally, the single battery simulation was validated by temperature-rise experiments on the cell surface, but the cold plate simulation itself was not directly tested in a hardware prototype. A future experimental study on the rib-groove cold plate would be desirable.
Another direction for further research is the system-level integration of the cold plate with the entire traction battery thermal management loop. In this thesis, each liquid cold plate was modeled as an independent component with the same inlet flow rate. However, in a real electric vehicle, multiple cold plates are connected in a network of pipes that may introduce unequal flow distribution. The inlet and outlet headers should be designed to ensure balanced flow among all branches. Future work may include one-dimensional flow network simulation or full three-dimensional simulation of all parallel cold plates.
In this study I focused mainly on the cooling mode, but the same liquid cold plate can be used for battery heating by circulating warm coolant. My low-temperature heating simulation proved the concept. More sophisticated control strategies could be developed to switch between heating and cooling automatically according to the battery temperature and charging state. For example, during cold weather, the traction battery may be heated by the motor waste heat via the coolant loop before fast charging. The rib-groove structure with enhanced heat transfer is also beneficial for heating because it increases the effective heat exchange area and promotes heat transfer from the coolant to the battery. Therefore, the proposed cold plate design is suitable for year-round traction battery thermal management.
In conclusion, the thesis contributes a systematic design methodology for a liquid cooling plate with staggered convex ribs and concave grooves for a large traction battery pack. I have verified the thermal model with experiments, selected the side-mounted and staggered inlet/outlet configuration, identified the circular rib and groove shapes, and optimized the height and depth distributions using surrogate-assisted NSGA-II. The final design outperforms the benchmark accordion cold plate by reducing the maximum battery pack temperature by 3%, the cell-level temperature difference by 10.1%, the pack-level temperature difference by 8.94%, and the pressure drop by 8.99%. The proposed cooling system satisfied the required thermal criteria under fast charging, low-temperature heating, and high-temperature highway cruise scenarios. The research demonstrates that passive heat transfer enhancement using staggered rib-groove structures can break the usual trade-off between cooling performance and pumping power in liquid-cooled traction battery systems.