Principal Component Analysis-Based Shading Defect Identification and Categorization in Standalone PV Systems Using I-V Curves

Abstract

Photovoltaic (PV) system health monitoring and fault diagnosis are essential for optimizing power generation, enhancing reliability, and prolonging the lifespan of PV power plants. Shading, especially in PV systems, leads to unique voltage-current (I-V) characteristics, serving as indicators of system health. This paper presents a cost-effective and highly accurate method for detecting, diagnosing, and classifying shading faults based on real I-V data obtained through electrical measurements under both healthy and shaded conditions. The method leverages Principal Component Analysis (PCA) to separate classes, and a confusion matrix assesses classification accuracy. The results demonstrate a success rate exceeding 98% in various configurations, using experimental data from a 250 W PV module. Importantly, this method relies solely on existing electrical measurements, eliminating the need for additional sensors, making it both efficient and cost-effective.