Reinforcement learning-based power sharing between batteries and supercapacitors in electric vehicles

Energy management of Battery/Supercapacitors (SCs) hybrid energy storage system (HESS) aims to reduce RMS battery current values and enhance the battery lifetime. This paper presents a reinforcement learning (RL) based energy management strategy for Electric Vehicles (EV). This approach allows for learning in real time the optimal power flow distribution between battery and supercapacitors starting from historic of the observation of RMS current of battery. The power management problem is presented with RL formulation verifying the electrical HESS constraints. The presented framework uses the RL technique to control the power flow distribution leading to the minimization of the RMS battery current. Particularly, we propose a methodology that generates optimal frequency sharing policy between battery and SCs taking into account the load variations of the EV dynamically in real time. Numerical simulations carried out on Matlab/Simulink confirmed the convergence of the RMS battery current to the optimal value without any prior knowledge of the driving conditions. The proposed framework aims to adapt automatically the power management policy to the optimal solution.