Stochastic sizing for off-grid green ammonia plants powered by wind energy

The plant architecture includes electrolysers, Haber-Bosch synthesis, air separation unit, and storage systems for hydrogen (HESS), nitrogen (NESS), and electricity (BESS). Electrolyser efficiency degradation (1.3% per year) and operational constraints such as minimum load and up/down times are incorporated.

Unlike deterministic approaches, which either become computationally intractable or fail to provide robust designs, the stochastic method balances investment decisions across multiple scenarios, ensuring resilience against prolonged wind droughts. The methodology is structured as a two-step sequential process to overcome the computational burden inherent in multi-year optimization.

Investment Phase: a stochastic optimization is performed over a two-year horizon using 24 climatic years of hourly wind data. Full chronological resolution (24 hours/day) is maintained to capture variability and extreme low-wind events and their likelihoods, which are critical for sizing hydrogen and ammonia storage, and to calculate LCOA on unbiased data. To keep the problem tractable, this phase uses Linear Programming (LP) rather than Mixed Integer

Programming (MIP), while enforcing technical constraints through penalty mechanisms. This approach ensures robust investment decisions under uncertainty without sacrificing realism. 1 Dispatch Phase: each candidate design from the investment phase is validated through detailed operational simulations using MIP. This phase applies true unit commitment logic, minimum up/down times, and realistic forecast horizons to stress-test flexibility and quantify operational performance. It computes actual ammonia production, shutdown frequency, and the final

LCOA, ensuring feasibility under real-world conditions.

Results show that stochastic optimization produces designs with slightly smaller process units

(4–5% reduction) but significantly larger storage capacities (30–40% increase) compared to the average of 24 deterministic solutions, improving reliability. Sensitivity analysis on flexibility penalties reveals that allowing controlled plant shutdowns during extreme wind shortages reduces hydrogen storage by over 60% and improves LCOA by 1.8%, while stricter continuity requirements increase costs by 2%. Dispatch-phase simulations using Mixed Integer

Programming validate operational feasibility under realistic forecast horizons and unit commitment constraints. The final designs achieve LCOA values between $810 and $825 per tonne of ammonia, with shutdown frequencies averaging 1–3 per year, confirming economic viability and robustness.

This study demonstrates that stochastic sizing is essential for off-grid ammonia projects relying solely on wind energy. It provides a systematic framework to navigate trade-offs between capital costs, operational flexibility, and reliability, offering insights for future integrated energy systems in the context of the energy transition.