A combined reinforcement learning and sliding mode control scheme for grid integration of a PV system

Abstract

The paper presents development of a reinforcement learning (RL) and sliding mode control (SMC) algorithm for a 3-phase PV system integrated to a grid. The PV system is integrated to grid through a voltage source inverter (VSI), in which PV-VSI combination supplies active power and compensates reactive power of the local non-linear load connected to the point of common coupling (PCC). For extraction of maximum power from the PV panel, we develop a RL based maximum power point tracking (MPPT) algorithm. The instantaneous power theory (IPT) is adopted for generation reference inverter current (RIC). An SMC algorithm has been developed for injecting current to the local non-linear load at a reference value. The RL-SMC scheme is implemented in both simulation using MATLAB/SIMULINK software and on a prototype PV experimental. The performance of the proposed RL-SMC scheme is compared with that of fuzzy logic-sliding mode control (FL-SMC) and incremental conductance-sliding mode control (IC-SMC) algorithms. From the obtained results, it is observed that the proposed RL-SMC scheme provides better maximum power extraction and active power control than the FL-SMC and IC-SMC schemes.

Keywords

• fuzzy logic
• incremental conductance
• instantaneous power theory
• reinforcement learning
• sliding mode controller