AI-Enhanced Framework for Predicting Shading and Soiling Impacts in Photovoltaic Systems

Photovoltaic systems offer cost-effective, sustainable energy, but partial shading and soiling significantly reduce efficiency, necessitating accurate shading prediction for optimized power output. Traditional energy management, reliant on human intuition, struggles with the complexity of modern energy optimization. This paper introduces an AI-driven energy management framework integrating computational intelligence, IoT monitoring, and deep learning to improve solar photovoltaic performance and energy storage. It addresses shading and soiling issues using EfficientNetB3, VGG16, ResNet50, and CNN models. Analysis of high-resolution images and performance metrics reveals nonlinear correlations affecting PV efficiency. The VGG16 model achieves 86.19% accuracy in anomaly detection, supporting predictive maintenance. A hardware-in-loop technique, using realtime feedback from the performance metrics of the most suited model, adjusts electrical parameters for optimal PV performance under soiling and shading conditions. This approach enhances energy efficiency thereby, contributing to smarter, sustainable cities.