An intelligent active equalization control strategy based on deep reinforcement learning for Lithium-ion battery pack

The inconsistency in large-scale series-connected lithium battery pack significantly impacts the usable capacity of the battery pack and raises the likelihood of safety risks. In this paper, an equalizer based on Buck–Boost converter is utilized. This equalizer comprises a pulse width modulation (PWM) controlled Buck–Boost equalization circuit and a switch array. In order to maximize the performance of the equalizer, this paper proposes an equalization strategy based on reinforcement learning algorithm to reduce the equalization time. Under the simulation environment of MATLAB/Simulink, the equalization experiments under different working conditions are designed. Compared to the mean-difference algorithm, the proposed algorithm based on deep reinforcement learning has up to 46% improvements on balancing speed, demonstrating the feasibility and effectiveness of the balancing scheme. •A new equalization topology using Buck–Boost circuit and switch array is utilized.•Deep reinforcement learning algorithm is employed to optimize the equalization strategy.•Good performance of the proposed equalization strategy is verified by various experiments.