Considering Cloud Cover Using Deep Learning Techniques

Abstract

Numerous environmental elements, including sun irradiation, temperature, shade, wind speed, and others, can significantly influence the output of photovoltaic (PV) systems. Solar generation forecasts across various time ranges are required to meet demand and grid restrictions. Short-term predictions are essential for energy trading, real-time control, and grid balancing. This paper presents a concise overview of shading due to cloud cover on the generation of PV systems. Furthermore, the convolutional neural network (CNN) deep learning model predicts PV generation by considering cloud movement over a specific area. The results demonstrate that cloud cover significantly impacts the generation of PV systems. Consequently, PV power generation forecasting is essential for enhancing the stability, quality, and reliability of smart grids and modern power systems. The suggested deep learning model occupies 12% of the memory size of the state-of-the-art model. The suggested model surpasses these models in the MSE, RMSE, and MAE test set metrics, yielding values of 1 KW, 2.37 KW, and 1.35 KW, respectively.