Machine Learning-Assisted Physics-Constrained Optimization for Thermal Management in Large-Format Batteries for Nuclear Applications

The synergy between nuclear power and lithium-ion energy storage is highly promising. However, the temperature uniformity control of large-format energy storage batteries, especially under extreme operating conditions, remains a significant engineering challenge. This paper presents an adaptive cooling strategy for bottom-liquid-cooled large-format lithium-ion battery thermal management systems, based on a machine learning-assisted, physics-constrained optimization framework. The framework integrates Gaussian Process Regression (GPR) for temperature prediction with Sequential Quadratic Programming (SQP) for cooling parameter optimization. The study shows that the coolant flow rate has a limited effect on both the maximum temperature (T_max) and temperature difference (ΔΤ) of large-format batteries, whereas the inlet coolant temperature (T_in) and ambient temperature (T_env) are the dominant factors affecting thermal management performance. Based on this, a coolant flow rate of 0.06 m/s is selected as the baseline, and the optimal inlet temperature variation under changing environmental temperatures (25-35°C) is further explored. The results demonstrate that even at the maximum ambient temperature of 35°C, the adaptive cooling strategy maintains a safe state (T_max ≤ 40°C); when the temperature is below 33.2°C, the system remains in the effective state (T_max ≤ 40°C, ΔΤ ≤ 5°C); and when the ambient temperature is below 28.3°C, the system achieves the optimal state (T_max ≤ 35°C, ΔΤ ≤ 5°C). The study indicates that as ambient temperature increases, raising the inlet temperature of the liquid cooling system helps improve the temperature uniformity of the battery pack but weakens the system's heat dissipation capability, causing an increase in the maximum temperature. This study demonstrates the potential application of machine learning and physics-based optimization methods in nuclear energy storage systems, providing a new perspective for enhancing the synergy between nuclear power and lithium-ion energy storage.