Solar irradiation prediction using empirical and artificial intelligence methods: A comparative review.

Area of Science:

• Renewable Energy Systems
• Climate Science
• Data Science

Background:

• Solar irradiation data is crucial for the feasibility and effective utilization of solar energy projects, impacting both energy generation and agriculture.
• Accurate solar irradiation prediction is essential for site selection, project sizing, and optimizing crop choices.
• Physical measurement of solar irradiation is often impractical globally due to cost and technological limitations.

Purpose of the Study:

• To conduct a comprehensive review and comparison of empirical and artificial intelligence (AI) techniques for solar irradiation prediction.
• To identify the most accurate methods for forecasting solar irradiation based on recent scientific literature.

Main Methods:

• Systematic literature review of research articles published between 2017 and 2022 focusing on solar irradiation prediction.
• Comparison of the accuracy and applicability of empirical models (e.g., modified sunshine-based models) and AI methods (e.g., artificial neural networks, support vector machines).

Main Results:

• AI methods generally exhibit higher accuracy in solar irradiation prediction compared to empirical methods.
• Among empirical models, modified sunshine-based models (MSSM) showed the highest accuracy, followed by sunshine-based (SSM) and non-sunshine-based models (NSM).
• Artificial neural networks (ANN) and hybrid models achieved the highest accuracy among AI techniques, with key inputs including temperature, humidity, clearness index, and precipitation.

Conclusions:

• AI-based approaches, especially ANNs, are more effective for accurate solar irradiation forecasting.
• Empirical models, particularly NSM, can be valuable when sunshine data is unavailable.
• Simple empirical models predict accurately, and increasing polynomial order does not significantly improve results.