Smart Energy Management of Hybrid EV–Hydrogen Stations Through LSTM Forecasting and Model Predictive Control

The increasing adoption of electric vehicles (EVs) and hydrogen-powered vehicles requires integrated energy infrastructures capable of simultaneously addressing electricity and hydrogen demands while accommodating the variability of renewable energy sources. This study evaluates the operational performance of a renewable-powered hybrid EV charging and hydrogen refueling station through predictive energy management. The station integrates photovoltaic and wind generation, grid connection, an electrolyzer, hydrogen storage, and vehicle charging/refueling infrastructure. A Long Short-Term Memory (LSTM) model was employed to jointly forecast renewable generation, EV charging demand, and hydrogen demand using a 24-h historical window and a 10-h prediction horizon. These forecasts were incorporated into a Model Predictive Control (MPC) framework to coordinate coupled electricity–hydrogen energy flows and were comparatively evaluated against a rule-based energy management system (RB-EMS) under the same operating conditions. Aggregate forecasting performance was evaluated using MAE, RMSE, and R2 metrics. Compared with RB-EMS, LSTM–MPC reduced total grid energy consumption by 10.6% and peak grid demand by 20.8%, while providing a modest increase in hydrogen demand served. The results indicate that forecast-assisted predictive management can improve grid-side performance and support coordinated operation of renewable-powered EV–H2 infrastructures.