Prototype-based Implementation of AI for Enhanced Solar Photovoltaic Performance

This paper presents the design and implementation of a prototype based solar photovoltaic (PV) system enhanced by, the application of Artificial Intelligence (AI) methods. The intention of the work described was to increase the reliability of solar power forecasting and to provide an automated fault detection system in PV installations. The authors created an experimental apparatus using a microcontroller-based data acquisition system, environmental sensors and a solar panel. The experimental apparatus collected RT data including solar irradiance, temperature, humidity, voltage and current over 30 days. The authors then trained a predictive power output machine-learning model using a Random Forest Regressor and their collected dataset. The Random forest Regressor model produced a prediction accuracy of 91%, indicating this model is reliable for estimating solar energy production given the changeability in climate condition. In addition, the authors implemented a fault detection module using a Convolutional Neural Network (CNN) that analyzed the panel images for problems like dust and physical damage. Their CNN reached 94.7% precision demonstrating its ability to classify faults in real time. The prototype system was preconfigured with wireless communication technology to allow for remote monitoring via mobile dashboard. The experimental findings validated the point of plausible AI integrated PV systems can be used to optimize energy efficiency, enable prognostic maintenance, and decrease operational losses. All in all, this is a contribution to the emerging field of intelligent, cost-effective and scalable solar energy management applications that can be deployed in decentralized and rural areas.