An advanced BPNN/RVEA coupled control strategy for novel immersed liquid cooling battery thermal management system

Abstract

This study introduces a novel distributed-inlet circulatory immersion liquid cooling (DIC-IC) battery thermal management system (BTMS), optimised using computational fluid dynamics (CFD), backpropagation neural networks (BPNN), and the reference vector-guided evolutionary algorithm (RVEA). The BPNN model accurately predicted maximum temperature T<inf>max</inf>, temperature difference ΔT<inf>max</inf>, and input power. Multi-objective optimisation (MOO) achieved a 24 % reduction in T<inf>max</inf> to 30.13 °C and a 70 % reduction in ΔT<inf>max</inf> to 3.58 °C with power consumption of only 0.7 m3/s3, significantly reducing energy usage by up to 95 % compared to initial designs. Dynamic cooling strategies, including static, equilibrium, and boost modes, were proposed to address various operational conditions. The boost mode reduced T<inf>max</inf> and ΔT<inf>max</inf> by an additional 5.9 % and 4.3 % respectively under 5C discharge, demonstrating superior thermal safety. These findings provide a practical framework for efficient and low-energy BTMSs that enhance the performance of lithium-ion batteries in high-power density and high-temperature applications.