Monitoring of PV Modules and Hotspot Detection using TensorFlow

In recent years, solar energy plants have increased dramatically as global awareness of environmental issues has grown. However, maintaining solar power plants is a difficult task, particularly in large-scale or remote power plants, detecting failure of photovoltaic (PV) cells is difficult. Consequently, finding these cells and replacing them before a serious event occurs has become increasingly important. This paper proposes an approach to develop a model in TensorFlow and train it with generated PV hotspots data and finally obtain classification and localization of hotspots generated in Photovoltaic modules with better accuracy. The proposed method enables the automatic classification of thermographic images from the system's convolutional neural network (CNN), with an accuracy of 98 percent in tests lasting only a few minutes. When compared to previous approaches, this approach has several advantages, including speed of execution, quickness of diagnosis, lower costs, and lower electricity production losses.