A Machine Learning Approach for Hydroelectric Power Forecasting

Sustainable energy solutions are essential for a cleaner and greener future, and hydroelectric power responds to this need. Recently, hydroelectric power plants have been integrated into some Water Distribution Systems (WDSs) to harness hydraulic jumps for electricity production. Optimizing the performance of these plants requires accurate forecasting of hydropower generation. Such forecasts are crucial for various purposes, including capacity planning, energy management, decision-making, and anomaly detection. This paper aims to develop a one-step-ahead monthly forecasting model for hydroelectric plants integrated into WDSs using machine learning algorithms not previously utilized in this context. In contrast to existing literature that primarily focuses on general run-of-river or storage-reservoir-based systems, the proposed approach is applicable to all WDSs-integrated plants, whose primary function is water delivery to consumers. After testing more than 500 models, the Temporal Convolutional Network architecture, with specific hyperparameters, achieved the best performance.