Forecasting renewable energy consumption with hydrogen integration: A comprehensive regression approach

Abstract

Investments in hydrogen energy and its development and dissemination depend on the ability to produce longterm, high-accuracy forecasts. However, current forecasts for hydrogen energy remain insufficient. In T & uuml;rkiye, a prominent and influential country in terms of population and economy, solar and wind energy installations have increased significantly in the last two decades parallel to the world and have reached a certain saturation. In the coming years, similar to the growth observed in wind and solar energy, hydrogen energy consumption, considered the future energy source, is expected to increase nationwide. In this study, forecasts for renewable energy consumption, including hydrogen energy, are developed for T & uuml;rkiye using eight different regression models: autoregressive (AR), autoregressive with exogenous input (ARX), moving average (MA), autoregressive moving average (ARMA), nonlinear ARX (NLARX), autoregressive integrated moving average (ARIMA), seasonal ARIMA (SARIMA), and extended SARIMA (ESARIMA). These models are based on changes in T & uuml;rkiye's renewable energy consumption and global hydrogen energy consumption trends. Performance evaluation metrics are then employed to assess the models' effectiveness, and both scientific and economic insights are drawn to guide future hydrogen energy strategies in T & uuml;rkiye. The findings reveal that ARIMA-based approaches yield the lowest errors among the eight models, even at lower regression orders. Specifically, while the first-order ARIMA model offered faster computation times, the fourth-order ESARIMA model achieved the highest accuracy, with MSE and RMSE values of 0.035 and 0.187, respectively. Overall, the study concludes that ARIMA-based models provide the most stable long-term forecasts for hydrogen energy consumption.