Fast Prediction of Thermal Behaviour of Lithium-ion Battery Energy Storage Systems Based on Meshless Surrogate Model

Accurate and efficient temperature monitoring is crucial for the rational control and safe operation of battery energy storage systems. Due to the limited number of temperature collection sensors in the energy storage system, it is not possible to quickly obtain the temperature distribution in the whole domain, and it is difficult to evaluate the heat production behavior of the battery in real time. Numerical simulation has the ability to obtain the temperature distribution in the whole domain, but the computational efficiency is low, and the arithmetic power demand is large, which can not be directly applied to the online temperature monitoring and heat production evaluation of the system. In this paper, based on the CFD simulation of a battery module, combined with machine learning algorithms, a meshless surrogate model construction method is proposed to obtain the surrogate model of wind speed, time and node temperature. The analysis of different wind speed test cases shows that the surrogate model is able to complete the evolution of the whole field temperature distribution of the battery module during the discharge process of 1 C times in 2.6 s. Compared with the CFD simulation, the computational speed is increased by 1229 times, and the absolute error of the temperature is less than 0.023 K. The surrogate model is able to realize the application of complex numerical models in the online simulation of actual engineering, which provides a new idea for the monitoring and control of the temperature of the energy storage system