Evaluating AI Models for Solar Irradiance Prediction: A Systematic Review of Strengths, Limitations, and Future Directions

Abstract

The growing need for accurate solar irradiance prediction to optimize solar energy generation has led to the exploration of Artificial Intelligence (AI) models. This study systematically evaluates the strengths, limitations, and future directions of AI models used in solar irradiance prediction. A total of 80 articles, sourced from databases such as Scopus, IEEE Xplore, and Google Scholar, were included based on relevance to key search terms including “solar irradiance prediction,” “Artificial Neural Networks (ANNs),” “Support Vector Machines (SVM),” “Random Forests (RF),” and “Deep Learning (DL).” The articles were selected using a comprehensive review process that ensured the inclusion of high-quality and relevant studies. The analysis and synthesis of the articles revealed that AI models, particularly ANNs, are widely used due to their ability to model complex, non-linear relationships and provide high prediction accuracy. However, limitations such as overfitting, the need for extensive computational resources, and challenges in data preparation were identified, especially with models like SVM and RF. Hybrid models that combine the strengths of different AI approaches were frequently recommended in the literature. Additionally, future directions for improving solar irradiance prediction included the integration of real-time data, satellite-based information, and the reduction of computational costs. This study highlights the substantial potential of AI in enhancing solar irradiance prediction while also pointing out key challenges. It concludes with recommendations for the development of hybrid models, better computational efficiency, and the use of real-time and satellite data to improve the scalability and accuracy of solar energy forecasting.