Fault detection and diagnosis in grid-connected PV systems under irradiance variations

Abstract

Nowadays, photovoltaic (PV) energy is considered as one of the most encouraging renewable energy sources. Nevertheless, the power delivered by a PV field is strongly attached to irradiance which undergoes rapid variations depending on the climatic conditions. Accordingly, it becomes extremely difficult to distinguish if it refers to faulty status in the system or healthy status under the irradiance variation (IV). Therefore, PV monitoring considering IV condition is fundamental in ensuring high reliability as well as improving power production of PV systems.In fault detection and diagnosis (FDD) field, researchers have considered the variation of irradiance (especially under low irradiance level) as faulty operating mode while others have considered it as fixed parameter during detecting faults.In this paper, therefore, firstly, the IV is introduced in the dynamic model of the grid connected PV (GCPV) system in different operating conditions. Then, an efficient and robust FDD approach based on machine learning and deep learning techniques is proposed in order to identify the healthy and faulty operating conditions.The obtained results through simulated data of a 12 kW PV module are extremely encouraging with a high accuracy under different studied cases.