Machine Learning Algorithms for Solar Irradiance Prediction: A Recent Comparative Study
Zaid Allal, Hassan Noura, Khaled Chahine
Abstract
Photovoltaic panels have emerged as a promising technology to generate renewable energy and reduce the dependence on fossil fuels by harnessing solar radiation and converting it into electricity. The efficiency of this conversion process relies on various factors, including solar panel quality and region-specific intake of solar radiation. Therefore, accurate solar irradiance forecasting is crucial for making informed decisions about designing and managing efficient solar power systems. One way to address this issue is through leveraging artificial intelligence algorithms that can predict precise amounts of irradiance in specific locations. To this end, this paper explores solar irradiance forecasting as a machine-learning problem. We utilize data from Izmir, Turkey, over a period spanning three years while testing several deep learning algorithms against traditional machine learning models, with results detailed within our comparative analysis. Based on these results, the multilayer perceptron stands out amongst other tested models regarding its effectiveness when applied to irradiance prediction.