Solar radiation prediction using boosted decision tree regression model: A case study in Malaysia.

Area of Science:

• Renewable energy systems
• Artificial intelligence in energy
• Climate modeling

Background:

• Malaysia possesses significant solar energy potential due to its equatorial location and climate.
• Current solar energy utilization in Malaysia remains low (2-4.6%) despite this potential.
• AI-based prediction models are increasingly used globally for solar radiation forecasting.

Purpose of the Study:

• To develop and evaluate an accurate solar radiation prediction model for Malaysia.
• To compare the performance of a boosted decision tree regression (BDTR) model against conventional algorithms.
• To investigate methods for enhancing model accuracy through normalization and parameter selection.

Main Methods:

• Application of the boosted decision tree regression (BDTR) algorithm for solar radiation prediction.
• Comparison with linear regression and neural network models.
• Investigation of Gaussian and binning normalizers, varying splitting sizes, and input parameters.
• Validation using sensitivity and uncertainty analyses.

Main Results:

• The BDTR model demonstrated superior performance and higher accuracy compared to linear regression and neural network models.
• Optimized normalization techniques and input parameters significantly improved prediction accuracy.
• Sensitivity and uncertainty analyses confirmed the robustness and reliability of the BDTR model.

Conclusions:

• The BDTR model is a reliable and accurate tool for predicting solar radiation in Malaysia.
• This AI-driven approach can significantly contribute to improving Malaysia's renewable energy sector.
• The findings support the development of sustainable energy resources and enhanced power generation.