Reinforcement learning-based control for the thermal management of the battery and occupant compartments of electric vehicles

Abstract

The complex coupling relationship in the cooling branch of direct-cooled battery thermal management systems leads to increased difficulty in controlling the temperature of the occupant compartment and the battery of electric vehicles as well as high energy consumption under high-temperature conditions in summer. In order to solve these problems, this study designed a secondary throttle (ST) orifice opening control at the refrigerant outlet of the battery branch and proposes a new cooling control strategy based on a deep reinforcement learning (RL) algorithm to control the compressor speed (N_compressor) and ST orifice opening. We also compared the performance of the RL control strategy and the proportion integration differentiation (PID) control under two different working conditions. The results show that the RL control strategy could better control the passenger compartment and battery cold plate at 22 °C and 20 °C, the system superheating was also more stable, and the compressor energy consumption (W_compressor) was 0.33 kW h⁻¹ and 0.38 kW h⁻¹ under a uniform climbing condition and NEDC condition, which were 5.7% and 7.3% lower than those of the PID control strategy, respectively.