(599b) Demand Response Operation of Air Separation Units Utilizing an Efficient MILP Modeling Framework

Kelley, M., The University of Texas at Austin
Pattison, R., The University of Texas at Austin
Baldea, M., The University of Texas at Austin
Baldick, R., The University of Texas at Austin
Deregulation, abundance and affordability of natural gas, and the increase of renewable-based electricity generation from wind and solar power has transformed U.S. energy markets. While highly beneficial in terms of the environment, the variability of solar and wind power coupled with the variable demand of consumers creates a challenge for grid operators. Consequently, increased reliance on renewables carries serious energy security risks, manifest in brown-outs or black-outs when supply cannot meet demand. Electricity storage is a natural way to synchronize renewable electricity generation with demand but remains cost-prohibitive. Managing demand rather than generation, referred to as demand response (DR), has emerged as the approach of choice for mitigating demand variability. Electricity-intensive air separation units (ASUs) are a promising industrial DR candidate—production can be increased during off-peak hours, and excess product stored (in either liquid or gas form) for use during peak demand when production rate is lowered.

DR operation of ASUs requires that the process dynamic characteristics and performance be included in production scheduling calculations to ensure that production rate changes (typically scheduled at hourly intervals) are feasible and do not violate product quality and operating safety constraints. However, the (first-principles) mathematical models that are needed to describe and optimize the DR behavior of industrial facilities cover multiple time scales and are inevitably large-scale, nonlinear, and stiff, posing computational challenges for simulation and optimization.

In this work, we address this problem by developing a computationally efficient optimal scheduling framework for ASUs operating under DR conditions. In particular, we propose using (nonlinear) low-order data-driven models of the scheduling-relevant dynamics of the plant [1], and present a set of reformulation and (exact) linearization techniques that allow us to cast the overall DR scheduling problem as Mixed Integer Linear Program (MILP).

We illustrate these concepts using an Air Separation Unit (ASU) capable of producing 50 tons of nitrogen per day. We show that DR operation can save over 3% in operating costs compared to a scenario where the production rate is kept constant during the day, a considerable savings in the context of an established industrial system that generates a commoditized product. Further, we demonstrate the computational efficiency of this framework and its potential for real-time implementation.

[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.