Abstract

Wind power forecasting provides valuable insights for demand response, energy market operation, and plant control optimization. Probabilistic approaches—which capture uncertainty in these complex systems by representing predictions as probability distributions—represent a growing area of research in wind energy science. However, their applications for wind-based hybrid systems, particularly those focused on sustainable hydrogen production, remain underexplored. This thesis illustrates the applicability of probabilistic methods for predicting and analyzing the performance of hybrid systems under real conditions, presenting a robust, modular strategy that integrates short-term wind speed and direction forecasts with power curve prediction. It applies these models in a case study using Supervisory Control and Data Acquisition (SCADA) readings from the Kelmarsh wind farm and an open-source Proton Exchange Membrane (PEM) hydrogen electrolyzer model to project hour-ahead hydrogen production from wind power. The study results demonstrate effective hydrogen forecasts alongside quantified uncertainty in both wind and power forecasts. The proposed framework, and its accompanying open-source software library, REStats, establishes an extensible foundation that accommodates various forecasting strategies and horizons, while highlighting potential future avenues for the design, monitoring, and analysis of hybrid systems.