Fault Detection of Photovoltaic Systems using Pre-Train CNN-VGG16

Fault detection in photovoltaic systems is crucial for boosting the output power and extending the lifespan of a PV system. However, conditions such as partial shading, high impedance faults, and the presence of Maximum Power Point Tracking make fault detection challenging in weather conditions. Most research in this area has only identified and classified faults in a few scenarios. This study identifies and classifies faults in the PV system using two-dimensional Deep Convolutional Neural Networks (2-D CNN) and features extracted from two-dimensional scalograms generated from PV system data. Unlike previous methods for fault classification, our study considers different faults by combining MPPT. This study shows that the proposed method, including a pre-trained CNN, performs better than existing methods and achieves 98.76 fault detection accuracy.