Improved PSO-BPNN Multi-Parameter Identification Method and Its Application in Battery Thermal Network Model

Traditional battery thermal models either lack accuracy or have high computational costs, making it difficult for battery management systems to monitor battery temperature online. In this study, a three-dimensional thermal network model was developed specifically for a 120Ah square lithium-ion battery, significantly reducing the computational burden while ensuring relatively accurate outputs at multiple temperature points. Firstly, the heat generation rate of the entire battery during charging and discharging processes was obtained using an accelerated adiabatic calorimeter (ARC). Then, based on the positions of the electrode tabs and temperature measurement points, the battery was divided, and a three-dimensional thermal network model architecture was constructed. Special particle swarm algorithms and neural networks were used to fit the thermal parameters of the thermal network model. Finally, discharge temperature experiments were conducted to validate the feasibility of the thermal network model.