<p>Driven by the urgent need to combat climate change and ensure a sustainable energy future, this study explores green hydrogen production using a hybrid machine learning (ML) model applied to a photovoltaic-thermal (PVT)-Rankine system in Shenyang, China. Green hydrogen, produced by electrolysis of water using forms of renewable energy, is a key part of the modern energy transition. The associated data were subjected to preprocessing and divided into 80 % training and 20 % testing subsets. The proposed hybrid ML model demonstrated outstanding performance in predicting power generation, achieving R² = 0.999 and low RMSE values (0.0204 for PV and 0.0139 for Rankine cycle). Nevertheless, the predictive performance for hydrogen production rate (HPR) showed a decreased accuracy as is reflected in the increased RMSE. Multi-objective optimization utilizing the MOPSO algorithm was employed to maximize photovoltaic (P_PV) and Rankine cycle (P_RC) power outputs. The optimum values for P_PV and P_RC were 292.99 kW and 233.73 kW, respectively, guaranteeing effective system performance. The major contributions of this investigation include: (i) construction of a hybrid ML model with a high accuracy specifically designed for a new PVT-Rankine-based hydrogen generation system; (ii) combination of the ML algorithms with the renewable energy management strategies for advancing the system performance; (iii) the implementation of an entire optimization routine under a set of realistic meteorological conditions, enabling the establishment of a robust decision support mechanism ensuring efficient and sustainable hydrogen and power production.</p>