Photovoltaic Array Fault Diagnosis and Localization Method Based on Modulated Photocurrent and Machine Learning

Abstract

Photovoltaic arrays are exposed to outdoor conditions year-round, leading to degradation, cracks, open circuits, and other faults. Hence, the establishment of an effective fault diagnosis system for photovoltaic arrays is of paramount importance. However, existing fault diagnosis methods often trade off between high accuracy and localization. To address this concern, this paper proposes a fault identification and localization approach for photovoltaic arrays based on modulated photocurrent and machine learning. By irradiating different frequency-modulated light, this method separates photocurrent and directly measures the photoelectric conversion efficiency of each panel, achieving both high accuracy and localization. Through machine learning classification algorithms, the current amplitude and frequency of each photovoltaic panel are identified to achieve fault identification and localization. Compared to other methods, the strengths of this method lie in its ability to achieve high-speed and high-accuracy fault identification and localization by measuring only the short-circuit current. Additionally, the equipment cost is low. The feasibility of the proposed method is demonstrated through practical experimentation. It is determined that when utilizing a neural network algorithm, the fault identification speed meets measurement requirements (5800 obs/s), and the fault diagnosis accuracy is optimal (97.8%).