Photovoltaic Fault Detection and Classification Using Common Vector Approach Based on I-V Curve

Abstract

Photovoltaic systems are subject to various faults. These faults result in significant degradation in photovoltaic system performances. These degradations result in a significant reduction in system efficiency as a result of increase in operating costs and times of exclusion of the photovoltaic system. Therefore, health monitoring and high accuracy error detection are extremely important for a photovoltaic system. In this study, fault detection and classification in photovoltaic systems are performed by using common vector approach which is first proposed in the literature for speech recognition and then is applied for face recognition. The faulty conditions discussed in the study are partial shading and series resistance degradation. In both fault detection situations, the common vector approach method is used to determine the type of fault using simulation data obtained under healthy operation and faulty operation conditions and very high fault detection rates are obtained. In order to evaluate the accuracy of the proposed method, the data obtained are also evaluated with the principal component analysis method, which is previously presented for photovoltaic system fault detection in the literature. According to the results obtained, principal component analysis method completely fails in case of serial resistance degradation fault. However, by using common vector approach method proposed in this study, a very high fault detection rate such as 97.5% can be obtained in serial resistance degradation fault. Likewise, in the case of partial shading, higher fault detection rate is achieved in common vector approach method (99.6%) compared to principal component analysis method (95.4%).