AI-Driven Hydropower Forecasting: A Multivariate Machine Learning Approach

Accurate forecasting of renewable energy production is fundamental for operational planning and grid management, especially in systems influenced by complex environmental and infrastructure-dependent factors. Building on previous research, this study extends a multivariate machine learning approach for hydropower forecasting in Water Distribution System-integrated plants. In contrast to our earlier work, which was limited to a shorter time frame and internal validation, the current study evaluates model performance over a longer period, characterized by new production patterns. Additionally, we introduce a lagged hydropower production variable as a proxy for internal system memory. The inclusion of this feature improves the model's ability to adapt to seasonal dynamics and results in significantly enhanced forecasting accuracy.