Q-Learning-Based Power Allocation Strategy for Battery-Supercapacitor Hybrid Systems in Electric Vehicles

Abstract

The main objective of this paper is to propose an intelligent power management approach for a hybrid system designed for electric vehicles (EV) comprising a battery and a supercapacitor. The approach is aimed at optimizing the power sharing between the two energy storages for improving battery life and overall energy efficiency of the vehicle. To this end, a reinforcement learning technique i.e., classical Q-Learning (QL) is integrated into the energy management system. Through this technique, the approximation of the power demand of the vehicle and the optimal calculation of the energy share to be delivered by the battery and/or the supercapacitor in consumption phases are enabled. It also enables efficient energy storage in regenerative braking phases. The results illustrate the effectiveness of the Q-Learning approach in maximizing the durability and energy efficiency of the hybrid system.