Intelligent control of air conditioning system for an electric vehicle: Based on reinforcement learning

The air conditioning (AC) systems of electric vehicle usually consumes 30% of total energy, resulting in a decrease in driving range. An AC-cabin intelligent control strategy based on dynamic thermal modeling and reinforcement learning (RL) is proposed. The thermal model is validated by road test cycle (RTC) experimental results. The intelligent control strategy is trying to find the mathematical relationship between the environmental states and the temperature of the passenger compartment and the dynamic responses of different control methods are compared. The results show that the dynamic response of the compressor speed to environmental changes is better and the temperature fluctuation is smaller under the RL-based control strategy. Compared with the Logic Threshold (LT) control strategy, the RL control strategy saved 6.7%, 9.6%, and 8.6% of compressor energy consumption, respectively, under three operating conditions (NEDC, WLTC, and RTC), while the PID control strategy saved 4.6%, 5.3%, and 4.5%, respectively.