(340s) Moving Horizon Demand Response Scheduling Subject to Exogenous Uncertainties

Deregulation, abundance and affordability of natural gas, and the increase of renewable-based electricity generation from wind and solar photovoltaics have transformed U.S. energy markets. While beneficial to the environment, increased renewable generation rages pose a challenge for maintaining the stability of the power grid owing to the daily and seasonal fluctuations in generation rate. Managing demand, rather than generation, a concept referred to as “demand response” (DR), has emerged as an important means for mitigating demand variability. Electricity-intensive industrial processes are promising industrial DR candidates: production can be increased during off-peak electricity hours, and excess product stored for use during peak grid demand, when production rate is lowered. Engaging in DR also has the potential to significantly lower operating cost of the industrial entity.

DR participation involves appropriate scheduling of production, and scheduling horizons of several days should be considered to fully exploit the benefits of, e.g., fluctuations in electricity prices [1]. However, considering multi-day time horizons in DR scheduling presents the drawback of having to deal with the uncertainty associated with predicting electricity prices and product demand. Day-ahead electricity prices can be assumed to be known exactly for (the first) 24 hours, while product demand may fluctuate significantly during the day. To account for this uncertainty and generate robust DR operating schedules, we propose a moving horizon optimal scheduling framework, where we ``close the scheduling loop'' and periodically recompute the schedule based on updated information such as changes in electricity prices, ambient conditions, and chemical product demand.

Using an industrially relevant air separation unit capable of producing 50 tons of N2 a day as an example, we illustrate the implementation moving horizon scheduling with periodic updates for electricity prices and ambient temperature and present several methods of mitigating non-periodic disturbances related to fluctuations in product demand. We find that even simplistic methods of forecasting electricity prices and ambient temperatures yield economic gains relative to steady-state operation that is agnostic to changes in operating circumstances. Our results also reveal that conventional assumptions made in moving horizon control (e.g., persistent disturbances) can lead to infeasibilities for DR scheduling, and propose chance constrained problem formulations to deal with this issue.

[1] R. C. Pattison, C. R. Touretzky, T. Johansson, I. Harjunkoski, and M. Baldea, “Optimal Process Operations in Fast-Changing Electricity Markets: Framework for Scheduling with Low-Order Dynamic Models and an Air Separation Application,” Ind. Eng. Chem. Res., vol. 55, no. 16, pp. 4562–4584, Apr. 2016.