(418c) Minimizing the Utility Cost of An ASU Based On Multi-Scenario NMPC with Economic Objective Function

Cryogenic air separation unit (ASU) is currently the most efficient way to produce large quantities of oxygen, nitrogen and argon. Since the air feed needs to be brought to extremely low temperature (approximately -300 degrees Fahrenheit or -185 degrees Celsius), air separation is an energy intensive process. Roughly 80% of total power consumption corresponds to the liquefier power. On the other hand, the price of the power supplied by utility companies is often subject to high fluctuations. This is because electricity is not readily stored and must be used (or wasted) as it is produced. Energy providers often turn to complex pricing schemes to try to allocate resources to greatest effect. This creates an opportunity to minimize average utility cost by changing the operating conditions of the air separation unit. In a previous study [1], a scheduling method was proposed to switch an ASU among three different operation modes to minimize the average cost, and the future variability in energy price is modeled as a two stage stochastic programming problem. On the other hand, this work was based on steady state model and does not consider the transition period.

In this work, we propose to minimize the utility cost of the ASU by solving an economically oriented nonlinear model predictive control (NMPC) problem. In our previous work [2], we have developed a large-scale first principle dynamic model of an ASU and applied NMPC to drive the ASU between different operation conditions. In that study we applied a sensitivity based NMPC strategy, called advanced step NMPC, that performs most of the computation in background. Once the measurement is received, a fast update is used to determine the control to be injected into the plant, this minimizing computational delay. Here, we extend this NMPC strategy to consider economic objective functions. ARIMA models are generated to forecast the power price for real-time pricing strategy. The multi-scenario problem is solved on-line to take account different power pricing scenarios. This leads to a large-scale dynamic optimization problem which can be solved efficiently with simultaneous NLP formulations. Moreover, the advance step NMPC algorithm is implemented to reduce the on-line computational delay.

Reference: [1] M.G. Ierapertritou, D. Wu, J. Vin, P. Sweeney and M. Chigirinskiy. Cost minization in an energy-intensive plant using mathematical programming approaches. Ind. Eng. Chem. Res. 2002, 41: 5262. [2] R. Huang, V.M. Zavala and L. T. Biegler. Advanced step nonlinear model predictive control for air separation units. J. of Process Control 2009, 19: 678.