(710e) Risk-Constrained Two-Stage Demand Response Scheduling for Green Hydrogen Production

Hydrogen, which is used as an energy carrier or chemical feedstock, is increasingly being incorporated into global net-zero plans with a particular emphasis on electrolysis-based renewable-powered “green” hydrogen.¹ When integrated with the power grid, the demand response (i.e., electrical charge or discharge to satisfy demand) scheduling of a hydrogen production plant can be posed as a cost minimization problem that uses electricity price signals, or predictions, as inputs.² In reality, these price signals are often subject to considerable uncertainty and, especially when participating in several power markets (e.g., day-ahead, intraday), the performance of a deterministic scheduling solution can be significantly suboptimal.³

In this work, we present a stochastic risk-constrained formulation to determine optimal production schedules and unit capacities in an integrated green hydrogen process² with electrolysis, storage, a fuel cell, and a steel-production furnace. These units, which consume or produce energy, are subjected to uncertain multi-market electricity price signals. The integrated green hydrogen system is optimized on a year-long horizon, where the electrical dispatch can be adjusted as recourse once the true electricity price trajectories are observed. This is formulated as a two-stage stochastic program that allows for a scenario-based representation of possible price futures. Further, a risk-constrained⁴ formulation is posed by using conditional value-at-risk (CVaR)⁵ to explicitly bound distribution tail-risk (i.e., cases of extreme loss) to a user-specified level. The proposed approach also determines the optimal proportion of participation in each power market.

The two-stage risk-constrained formulation is shown to outperform the deterministic optimization problem and approach the performance of an oracle with perfect market knowledge. The time at which the true price trajectories are observed determines how closely the stochastic solution approaches this oracle. We further observe a risk-reward trade-off whereby tighter bounds on CVaR result in higher expected process costs (i.e., a narrower distribution with a higher mean). More risk-averse bounds are also shown to result in larger unit capacity decisions, which allow for larger storage to buffer against energy price fluctuations. Stochastic programming approaches, such as the one presented here, allow for optimal integration between energy markets and industrial systems, particularly when storage capacity is available.

[1] UK Department for Energy Security and Net Zero. (2023). Hydrogen Production Delivery Roadmap.

[2] Tsay, C., Qvist, S. (2023). Integrating process and power grid models for optimal design and demand response operation of giga-scale green hydrogen. AIChE J. 69(12), e18268.

[3] Schulze, T., McKinnon, K. (2016). The value of stochastic programming in day-ahead and intra-day generation unit commitment. Energy 101, 592–605.

[4] Herding, R., et al. (2024). Risk-aware microgrid operation and participation in the day-ahead electricity market. Adv. Appl. Energy 15, 100180.

[5] Rockafellar, R.T., Uryasev, S. (2000). Optimization of conditional value-at-risk. J. Risk 2, 21–42.-