Detecting and diagnosing faults in photovoltaic modules using Neural Networks and monitoring them via Internet of Things IOT technology
Keywords:
Fault diagnosis, photovoltaic panels, Artificial Neural Network (ANN), Deep Neural Network (DNN), Internet of Things (IOT), intelligent monitoring system, MATLAB.Abstract
The proposed system relies on Artificial Intelligence Techniques using Artificial Neural Networks ANN and Probability Neural Networks PNN in the process of fault detection and diagnosis and employs a comprehensive data acquisition system that continuously measures the voltage and current signals of each PV module. The implementation includes two interconnected platforms: a physical hardware module, representing the real photovoltaic behavior under practical conditions, and a software-simulated module designed to replicate the electrical and dynamic response characteristics of the real system. This hybrid design enables controlled fault simulation and dual-layer data analysis under safe conditions.
Main fault types were introduced )short circuit, open circuit, grounding( through dedicated control buttons that simultaneously affect both the hardware and the simulated module, allowing synchronized acquisition of actual and virtual signals. Neural networks were trained on 50 training epochs containing 500 samples (350 for training and 150 for testing). The predictive ANN model was used for operational state estimation, while the probabilistic PNN model classified fault types. The system achieved an overall diagnostic accuracy of 91.6%, with detailed accuracy rates of 96% for short circuit, 97% for normal, 88% for open circuit, and 90% for grounding, and a total response time of 2.3 seconds from data acquisition to alert transmission.
Real-time monitoring was done using Internet of Things (IoT) technology by connecting the designed system with Telegram's platform, which provides live visualization of voltage, current, temperature, and fault status, along with instant alert notifications in case of any malfunction.