Reliability Analysis of Thermal Management Structure of Power Battery Based on Deep Learning Methods

During the charging and discharging process of lithium batteries, heat is generated. If the heat is not effectively dissipated, it may lead to excessive battery temperature, which can adversely affect battery performance, shorten its lifespan, and even pose safety risks. This paper establishes a battery pack model and employs a cooling system primarily based on liquid cooling, with air cooling as a supplementary method, aiming to improve heat dissipation efficiency and ensure that the battery pack operates within the optimal temperature range. The study analyzes the effects of different parameter settings for the coolant and air, as well as the advantages and necessity of combined cooling. The results indicate that the combined cooling method is more effective, with the temperature and flow rate of the cooling medium having a significant impact on the battery pack's maximum temperature and temperature field distribution. Based on reliability theory, a Kriging surrogate model is used to establish the objective function, followed by reliability analysis using the improved first-order second-moment (AFOSM) and Monte Carlo simulation (MCS) methods. The deep learning network shows good fitting and training capabilities. By inputting the extracted sample array data into the network for training, the predicted values closely match the real values, with minimal error.