Improved Image Processing Technique Based Internet of Things and Convolutional Neural Network for Fault Classification of Solar Cells

Abstract

In the clean, renewable electricity generation, the solar photovoltaic (PV) classification structure became the most appealing. Furthermore, due to varied characteristics and ambient temperature, performance varies. To analyze its performance, a real-time and remote monitoring system is required. The use of the Internet of Things (IoT) in the solar cells classification and in the solar PV systems monitoring is dependent on image processing, and its effectiveness has been investigated. Data gathering, data gateway, and a Constitutional Neural (CNN) model for fault classification prediction in solar cells make up the enhanced proposed system. This research uses a 2,426 solar cells acquired datasets from high-resolution electroluminescence (EL) photographs for automated fault probability detection. The gathered images depict both faulty and functioning solar cells with varying degrees of degradation in polycrystalline and monocrystalline solar modules. Experts categorized the solar cells and labeled the photos basing on the fault possibility in each image. The tagged images could be used to develop machine-learning algorithms and computer vision for detecting and forecasting faults including cell quality, PID, fracture interconnects, and fractures, as well as anticipating power efficiency losses.