Forecast accuracy vs. Decision value in electricity price forecasting for power-to-X off-take agreements

Power-to-X (PtX) systems are emerging as key technologies in the transition to low-carbon fuels by converting renewable electricity into chemical energy carriers. Since electricity costs represent a major share of total production cost, accurate electricity price forecasts are essential for both the economic operation and contract design of PtX systems. However, forecasting models are typically evaluated using statistical error metrics that do not necessarily reflect their economic value in operational decision-making. This study proposes a decision-focused forecasting framework that integrates a Threshold-Aware Residual Network (TARN) with a Monte Carlo simulation based on a Mean-Reversion Jump-Diffusion (MRJD) process to improve the evaluation of PtX off-take agreements under electricity price uncertainty. The results demonstrate that superior statistical accuracy does not necessarily translate into superior economic performance, highlighting the importance of decision-focused forecasting for robust PtX contract assessment and investment decisions.