AI‐Enhanced PV Power Forecasting Using Cloud Thickness and Motion in Kayseri, Türkiye

ABSTRACT The incorporation of renewable energy in photovoltaic (PV) systems has made significant progress. The inherent intermittency nature of PV generation, nevertheless poses an obstacle to accurate energy forecasting. Historical PV production plus meteorological data such as temperature, humidity, and atmospheric pressure are largely utilized in present methods of forecasting. However, cloud thickness and dynamics‐integrated system, has not been investigated and tested in real‐world examples yet. This research seeks to fill this gap in research through the development of a new AI‐based PV forecasting model that incorporates cloud thickness, cloud motion, and solar position into the forecasting model. Cloud properties and their impact on solar radiation are computed through a deep learning‐based panel‐shadowing model. For cloud movement forecasting, a gated recurrent unit (GRU) is used, while multiple convolutional neural networks (CNNs) are used for estimating cloud thickness. These outcomes are then integrated with measurements from environmental sensors to improve the accuracy of the predictions. The system was implemented and tested at Abdullah Gül University and exhibited a remarkable improvement in forecasting accuracy compared to current models. The results prove that cloud motion and thickness improve the accuracy of PV predictions, which is important for energy market stability and power grid operations.