Analysis and Comparison of PI and Reinforcement Learning Controllers for the Dual Active Bridge Converter

Currently, power electronic converters have a wide variety of applications, among these converters is the Dual Active Bridge (DAB) converter. Due to its high efficiency and power density, it is attractive and can be used in electric vehicle chargers. The objective of this paper is to present the reinforcement learning (RL) based Deep Deterministic Policy Gradient (DDPG) algorithm as an option to control the DAB converter with SPS modulation, replacing the traditional PI control in voltage mode. System simulations and training of the DDPG algorithm were developed in Matlab/Simulink software. Finally, the results show that the RL control achieves a better performance than the PI control, considering the response time to reach the steady-state and avoiding overshoot at different disturbances occurring in the system.