Data-driven forecasting of hydrogen production and demand via weather-clustered and day-weighted optimization for advance scheduling

Effective hydrogen transport propositions rely on accurate production and demand forecasts. In order to provide different managers tailored hydrogen transport propositions, data-driven forecasting of hydrogen production and demand via weather-clustered and day-weighted optimization for advance scheduling is presented. First, the historical weather data are preprocessed through anomalous data detection and normalization to ensure robustness in subsequent modelling. Then Centre-Radial means (CR-means) is presented for classifying weather data into six distinct weather-relevant clusters. Next, a forecasting model for hydrogen production based on the clusters divided by CR-means is presented. The comparative results verify the feasibility and advantages of CR-means with lower root mean squared error and absolute error of each hour between the predicted and actual values of hydrogen production by each means. Moreover, a forecasting model for hydrogen demand is presented, and the efficacy of the model demonstrates that it provides highly credible hydrogen demand predictions with a more precise temporal scale. Finally, based on the foregoing models, effective hydrogen transport propositions for helping managers schedule ahead response are presented.