Detecting the faults of solar photovoltaic module due to the temperature and shading effect by convolutional neural network

Abstract

This study aims to use the convolutional neural network (CNN) to build a new model for detecting the faults of solar PV module performance under the temperature and shading effects. Two sets of data were used based on solar data from Mendeley, and experimental data from the University of Sharjah. The transfer learning model was also used in this study. Both CNN and transfer learning models were trained and tested for Mendeley and experimental data under different data and epoch numbers. The accuracy of both models was tested under different numbers of data and epochs. The accuracy was higher than 96 % for 50 epochs when half and full data were used for both models. In comparison, the transfer learning model still shows 90 % for all data when the number of epochs is reduced to 10 epochs. The transfer learning model in this study can be used to detect faults in other renewable energy applications, such as wind energy systems.

Keywords

• Convolutional neural network
• Solar PV
• Faults
• Shading effects
• Temperature effects