A Deep Concurrent Learning-based Robust and Optimal Energy Management Strategy for Hybrid Battery Systems in Electrical Vehicles: A Thermal-And Health-Aware Design Approach

This study proposes a learning-based energy management strategy to reduce energy loss and improve battery-powered electrical vehicle’s (BEV) hybrid battery systems (HBS) with high power (HP) and high energy (HE) battery packs in the presence of external disturbances and modelling errors. A multi-constrained, least costly formulation is developed by considering over-temperature penalty and multi-stress-driven degradation costs of the lithium-ion battery (LIB) pack (HE and HP batteries). It also offers a reward term to test the HP battery pack’s optimal working range without limiting its charge, improving system efficiency. Compared to the current Deep Deterministic Policy Gradient (DDPG)-based method, the suggested method decreases energy loss by 14.3% and 16.4% during training and validation. It controls battery pack temperature within a threshold, outperforming the DDPG-based method in LIB safety.