Deep Reinforcement Learning-Based Approach for Fault-Tolerant Control of PV Systems in Smart Grids

Photovoltaic systems are susceptible to a range of component-related failures and environmental disturbances, which have an impact on their regular operation and cause significant energy loss. To solve such issues, fault-tolerant control (FTC) techniques have been applied to control and supervise the PV system state and notify the user when any unexpected change in the systems normal operation occurs. In addition, finding the optimal operating point of a PV system in the presence of challenging environmental circumstances has been made possible through a variety of maximum power point tracking (MPPT) techniques. The goal of this work is to propose an FTC method for setting up the MPPT controller to seek for the sub-optimal point of operation. The deep reinforcement learning approach is the basis of the MMPT controller. The proposed FTC approach has good performance in the presence of diverse fault types, according to simulation results on Matlab/Simulink.