Improving generalisation capability of artificial intelligence-based solar radiation estimator models using a...

Area of Science:

• Renewable Energy
• Environmental Science
• Data Science

Background:

• Reducing fossil fuel dependence is crucial for environmental protection.
• Solar radiation (Rs) is a key source of clean, renewable energy.
• Accurate estimation of Rs is vital for effective solar energy utilization.

Purpose of the Study:

• To estimate daily solar radiation (Rs) values in Iran using various models.
• To compare the performance of empirical models, support vector machine (SVM), and SVM coupled with cuckoo search algorithm (SVM-CSA).
• To evaluate two distinct data structuring approaches for model training and testing.

Main Methods:

• Daily Rs data from seven Iranian meteorological stations (2010-2019) were analyzed.
• Models employed include empirical methods, SVM, and SVM-CSA.
• Two data structures were used: station-specific training/testing and multi-station training/testing.

Main Results:

• Meteorological parameters improved SVM and SVM-CSA accuracy over geographical parameters.
• SVM-CSA reduced RMSE by up to 42.4% compared to SVM alone.
• Model performance varied across stations and radiation value intervals, with optimal results in specific thirds.

Conclusions:

• The SVM-CSA model offers a more accurate method for estimating solar radiation.
• Utilizing meteorological data enhances the predictive power of SVM-based models.
• Further research should explore climate change impacts and remote sensing data for Rs estimation.