Policy-Relevant Forecasting of Green Hydrogen Viability: A Comparative Techno-Economic and Machine Learning Analysis of Costa Rica and the United Kingdom

Green hydrogen is increasingly viewed as a cornerstone of global decarbonization, yet its economic viability varies widely across national contexts. This study presents an integrated modeling framework that combines high-resolution spatial resource mapping, deterministic techno-economic forecasting, Monte Carlo simulation, and interpretable machine learning to assess green hydrogen development prospects in Costa Rica and the United Kingdom. The framework synthesizes established tools—Random Forest, XGBoost, and LightGBM—to forecast Levelized Cost of Hydrogen (LCOH) and Net Present Value (NPV) under policy and resource uncertainty.Model results show that in Costa Rica, 3 MW pilot-scale systems using Solid Oxide Electrolysis Cells (SOEC) reach LCOH values of $8.81/kg in 2025 and NPVs of up to +$3 million over 20 years. In contrast, 50 MW UK projects achieve LCOH as low as $2.07/kg by 2050, with SOEC configurations yielding NPVs exceeding +$895 million. SHAP-based sensitivity analysis highlights electricity price, CAPEX, and capacity factor as dominant cost drivers in both contexts.By providing a transferable, explainable scenario analysis tool, this study supports strategic planning and policy design in countries with divergent infrastructure maturity and development goals. It also identifies opportunities for bilateral cooperation, such as UK-supported offshore wind–hydrogen hubs in Costa Rica. The integrated framework offers a replicable pathway for evaluating green hydrogen investments and informing context-sensitive policy instruments under deep uncertainty.