MPPT for PV systems using deep reinforcement learning algorithms

Abstract

This work proposes the use of reinforcement learning (RL) techniques with deep-learning models for the control problem of the maximum power point tracking (MPPT) of a photovoltaic (PV) array. Particularly, we based on the methods of the deep deterministic policy gradient (DDPG) and with inverted gradient (IGDDPG) and also the method of the delayed twins (TD3) policies. In order to evaluate the performance of our proposal to address the MPPT control problem, several simulated tests were performed under different operating conditions in terms of temperature and solar irradiance. Results show a good performance of the methodologies with a fast response and a stable behavior that performs near to the maximum power point (MPP). Moreover, the algorithms do not require any previous knowledge about the dynamic behavior and characteristics of the photovoltaic (PV) array.