Low-cost green hydrogen from renewable energy curtailment: A techno-economic analysis with a real-time AI-driven approach

The increasing penetration of renewable energy has resulted in more frequent curtailment events, highlighting the potential for utilizing curtailment for cost-effective green hydrogen production. This study proposes a real-time hydrogen production framework based on artificial intelligence (AI) forecasting of renewable energy curtailment and electricity prices (local marginal price, LMP). Using historical data on curtailment, electricity prices, and meteorological data, three machine learning models—Long short-term memory (LSTM), Bidirectional LSTM (BiLSTM), and Extreme gradient boosting (XGBoost)—were developed to predict hourly curtailment and LMP. The predicted values were then used to compute the hourly Levelized Cost of Hydrogen (LCOH) for a 2 GW alkaline electrolyzer under realistic operational constraints. Results indicate that LSTM provided the most accurate forecasts for solar curtailment, while BiLSTM outperformed others for wind curtailment and LMP forecasting. The analysis revealed that LCOH significantly decreases during periods of high curtailment and low electricity prices. Sensitivity analyses on electrolyzer capacity and minimum curtailment thresholds showed that median LCOH values could be maintained below $3/kg under assumed conditions. This work presents an integrated, data-driven framework that links AI-based forecasting with techno-economic planning of hydrogen production, offering a viable pathway to enhance the efficiency and cost-effectiveness of renewable green hydrogen production.