Solar Radiation Prediction in PV Power Systems: a Comparison of Deep Learning Models Using Big Data

Abstract

Photovoltaic (PV) energy systems are one of the most significant renewable resources, requiring efficient solutions for solar power generation and maintenance. Accurately predicting PV energy generation and solar radiation is essential for managing grid maintenance and making energy market decisions. This study proposes prediction models based on the use of deep learning, including convolutional neural networks (CNN) and long short-term memory (LSTM) networks, to fill the gap in big data analysis as more renewable energy data is collected. These models will be compared with traditional machine learning methods including Artificial Neural Networks (ANN) and Support Vector Regression (SVR). Data from the GAPYENEV Centre at the University of Harran is used to implement predictive models. Several prediction error metrics such as MAE, MSE, RMSE, R² and accuracy are used to evaluate the predictive ability of the models.