An innovative adaptive hybrid prediction model based on deep learning methods (CEEMDAN-GRU) for green hydrogen production, In case study: Turkey

Abstract

Due to global warming, countries have turned to renewable and clean energy sources. Among these sources, they have started to prefer hydrogen energy and green hydrogen production as the cleanest type of energy in producing hydrogen energy. In this study, an innovative adaptive hybrid forecasting model based on deep learning was proposed to be used in green hydrogen production forecasting. The energy required for the electrolysis method used in hydrogen production was planned to be provided by wind energy. A new perspective was introduced in determining hydrogen production with the proposed model. In particular, the wind regime in 9 provinces selected for the case study was determined and the most suitable wind turbine power for the system was calculated according to this regime by applying the decomposition method to wind speeds, the most important input of wind energy. Depending on the determined turbine powers, the amount of hydrogen that can be produced was estimated. In the prediction process, 70 % of the data was used as training data (24528), 5 % as validation data (1752) and 25 % as test data (8760). The highest accuracy rate obtained in hydrogen energy forecasting among the selected provinces is 92.045 %. The lowest error and the highest regression rates (The root mean square error (RMSE): 0.600, Mean absolute percentage error (MAPE): 0.100 and the coefficient of determination (R2): 0.990) in wind speed prediction were realized in İzmir province. The highest wind turbine power was calculated as 225 kW.