Advancing Solar Irradiation Prediction in Extreme Climates: A LASSO Regression Analysis in Tomsk

Abstract

This study presents a robust approach to predicting solar irradiation in the challenging climatic conditions of Tomsk using LASSO regression, with a particular emphasis on interpretability and climatic variability. Two distinct models were developed: Model 1, integrating specific humidity at 2 meters, and Model 2, excluding this variable to assess the impact of a wider range of meteorological factors. The comprehensive meteorological dataset from NASA’s POWER database underpinned the analysis. The models' efficacy was demonstrated by impressive R-squared values: 0.843 for Model 1 and 0.813 for Model 2, indicating a substantial proportion of variance in solar irradiation was captured. Notably, Model 1's RMSE of 0.0353 and Model 2's RMSE of 0.0386 affirm the precision of the predictions. The study advances the predictive modeling of solar power output, offering valuable contributions to renewable energy forecasting literature and operational practices by providing a methodological framework that is both accurate and comprehensible, even amidst the complexities of extreme weather patterns.

Keywords

• solar irradiation forecasting
• lasso regression analysis
• climatic variability in energy
• modelling renewable energy prediction in siberia
• meteorological data analytics in pv systems