The economic viability of offshore wind-to-hydrogen systems depends heavily on the accuracy of wind-speed forecasting. We evaluated three probabilistic scenario generators (historical bootstrapping

parametric Weibull fitting

and calibrated long short-term memory (LSTM) sequence model) using a decision-coupled framework. From 61 years of ERA5

we created 1000 synthetic 23-year hourly scenarios per method

propagated them through a techno-economic model

and scored the continuous ranked probability score (CRPS) on economic distributions (levelized cost of hydrogen (LCOH)

net present value (NPV)

and interest rate of return (IRR)). We report the decision bandwidth and its elasticity to skill

and run a financing price sweep to meet probability targets. Relative to bootstrapping and Weibull

the LSTM lowers CRPS by 30% for LCOH and NPV (25% for IRR)

narrowing risk bands at similar medians. The hydrogen price required to reach more than 90% probability is €7.76–7.78/kg for Pr(NPV > 0) and €9.16-9.18/kg for Pr(NPV > 0 and IRR > 10%). A cross-method spread of less than €0.02/kg indicates a threshold-saturated regime

in which better skills mainly contract risk. Robustness tests shift medians

but preserve method ranking and thresholds. This framework translates probabilistic forecasting skills into finance ability metrics for wind-to-hydrogen screening and policy design.