Data-Driven State-Wise Solar Energy Forecasting Using Advanced Machine Learning Models

Abstract: In this paper, we present a machine learning framework for forecasting state-wise solar power generation in India, leveraging historical solar output and meteorological data. The models employed Random Forest, Gradient Boosting, and XGBoost are designed to capture the complex, nonlinear relationships inherent in weather-influenced energy generation. To enhance predictive performance and model robustness, hyperparameter tuning techniques are applied. Among the approaches, XGBoost outperforms others, demonstrating superior accuracy and adaptability. These findings emphasize the potential of machine learning in advancing solar energy forecasting and enabling data-driven, sustainable energy planning.

Keywords: Solar forecasting, XGBoost, Hyperparameter tuning, Machine learning, Renewable energy