Enhancing hydropower generation Predictions: A comprehensive study of XGBoost and Support Vector Regression models with advanced optimization techniques

Abstract

Hydropower plays a crucial role in electricity generation, contributing over 60% of total renewable energy output. Its ability to stabilize energy fluctuations makes it essential in green energy initiatives. Accurate prediction of hydropower production is vital, considering its dependence on various factors like weather, water storage, and electricity generation. Traditional methods struggle with the complexities involved. This study utilized Support Vector Regression (SVR) and eXtreme Gradient Boosting (XGBoost) algorithms, both individually and in hybrid models enhanced by optimization techniques like Slime Mould Algorithm (SMA), Aquila Optimizer (AO), and Grey Wolf Optimization (GWO). XGBoost outperformed SVR in single model predictions with an R² value of 0.8632 and RMSE of 40.90, and when optimized, the hybrid XGBoost models showed superior performance, with XGBoost-SMA achieving the highest accuracy. The results revealed that the XGBoost-SMA model achieved the most desired accuracy with an R² value of 0.9713 and a root mean square error of 18.73 for the test dataset. This research highlights machine learning’s applicability in hydropower prediction and suggests hybrid models as a promising approach for better accuracy, emphasizing XGBoost’s potential in efficient hydropower forecasting to meet global electricity demands.