Supervised classification and fault detection in grid-connected PV systems using 1D-CNN: Simulation and real-time validation

Photovoltaic (PV) systems are prone to various faults, including short-circuit, open-circuit, partial shading, and inverter bypass diode issues, which reduce power output and can damage components. This study presents an innovative fault detection and online monitoring technique for grid-connected PV (GCPV) systems, combining Internet of Things (IoT) technology with a one-dimensional convolutional neural network (1D-CNN) deep learning approach. The method involves developing a temperature-dependent PV system model using series resistance and ideality factor, capturing real-time data from a 15kWp GCPV system with optimally placed sensors to minimize sensor count while maintaining data accuracy, and validating the model through MATLAB/Simulink simulations and real-time experiments under various fault scenarios. The collected data is used to train the 1D-CNN model to classify different fault types. The model is then implemented on an IoT platform for real-time monitoring and fault detection, displaying system status and alerts via a dashboard. The proposed system achieves a high fault detection accuracy of 98.15 % and 93.12 % during cyberattacks, with an uncertainty of ±4 %, significantly enhancing fault detection reliability and efficiency compared to existing methods. The IoT dashboard provides an effective tool for monitoring system performance and issuing alerts under abnormal conditions.