Predictive Control of a Reactive Distillation Column using Multi-rate DAE EKF

Control of product purity is of paramount importance in effective control of tightly integrated process such as reactive distillation. In practice, however, measurements of product concentrations may be unavailable or may be available at slower sampling rates when compared with other measurements. A cost effective approach to improve control of such systems is to develop a state estimator that can accommodate measurements at multiple rates and use it in controller development. In this work, DAE EKF formulation developed by Mandela et al. (2010) is modified to accommodate measurements available at multiple sampling rates. A successive linearization based nonlinear MPC scheme is then developed for controlling a system modeled as DAEs. Observer error feedback approach developed by Hunag et al. (2012) has been extended to achieve offset reduction and to improve regulatory control of a multi-rate sampled data system. The efficacy of the proposed approach is demonstrated by conducting simulation studies on an ideal reactive distillation system. Analysis of the simulation results reveals that the feedback introduced using the multi-rate concentration measurements reduces the offset significantly in the face of unmeasured disturbances of moderate magnitude.