Stacked AutoEncoder based diagnosis applied on a Solar Photovoltaic System

Abstract

This paper deals with the modeling of a photovoltaic system connected to a grid for the simulation of normal and faulty operations and the generation of a data-set for learning a fault detection algorithm based on a Stacked Autoencoder. To evaluate the effectiveness of the proposed approach, a Mean Squared Error is used. This method enables early fault detection, enhancing system relability and efficiency while addressing the need for proactive fault management in the system under normal conditions. Obtained results under different radiation and temperature conditions highlight the relevance of the proposed model and the effectiveness of the fault detection algorithm.