A cryogenic air separation unit (ASU) plays a vital role in future oxygen-blown combustion/gasification-based power plant. It is subject to periods of significant changes in production demand due to which an effective control structure is needed to minimize the impact of transient operation on product purity. In a typical ASU, the column pressure, temperature and composition dynamics float in tune with the feed-air flowrate which, in turn, is feedforwarded/ratioed with gasifier oxygen demand. Since the amount of column reflux/reboil is tied-up and is a function of this flowrate, the system inherently entails significant non-linearity, even for operation close to nominal operating point. This paper extends previous studies1,2 using single linear model for predictive control over entire operating range to using multiple linear models, each corresponding to different part-load conditions. Each linear model is chosen to represent a discrete subspace of the overall nonlinear operating space. When all n models are combined, the resulting bank of linear models spans the entire nonlinear operating space. Hence, at any operating point of the ASU, the multiple model predictive control (MMPC) algorithm naturally and continuously adapts by promoting the dominant model(s), through model-weight calculation, based on plant measurements only, thus providing a better 'pseudo-nonlinear' model prediction.
This study has been done using a fully pressure-driven Aspen Plus Dynamics model of the ASU, acting as a surrogate 'plant-model', whereas multiple 'control-models' have been devised using a simultaneous PRBS multiple input excitation to the Aspen 'plant-model' at different operating load points. The MMPC approach has been shown to provide better composition control at high ramp-rates compared to MPC approach. This in-turn allows for higher ramp-rates of ASU, without violating the oxygen-purity constraints. Since the ASU poses a bottleneck in the entire IGCC plant operation by limiting the ramp-rate during load-following operation, the proposed control strategy provides a better response to the plantwide IGCC load-following problem compared to conventional PID and MPC approaches, preventing the need for larger liquid oxygen/air storage requirements. In addition, a linear optimization problem being solved at each time step requires much less computational resources compared to a nonlinear MPC approach.
References:
1. P. Mahapatra and B. Wayne Bequette. Oxygen Purity Control in the Air Separation Unit of an IGCC Power Generation System during rapid production fluctuation. In Annual International Pittsburgh Coal Conference 2009 CDROM Proceedings, 20-23 Sept, 2009.
2. P. Mahapatra and B. Wayne Bequette. Process design and control studies of Elevated-Pressure Air Separation Unit for IGCC power plants. In 2010 American Control Council Annual Meeting, Baltimore, DC, June, 2010.
See more of this Group/Topical: Computing and Systems Technology Division