Control-Relevant Prefiltering: A Systematic Design Approach and Case Study

System identification is the most demanding and time consuming step in the implementation of advanced control in the refining and petrochemical industries. As a result, control-relevant identification, which views the identification problem in terms of its impact on control system design, is a topic that possesses significant practical importance. In this paper, we specifically examine the use of control-relevant prefiltering applied to parameter estimation using prediction-error methods. The prefiltering step ensures that the estimated model retains those plant characteristics that are most significant with regards to the user’s control requirements. We describe how to systematically build the prefilter in terms of the estimated model struture, the desired closed-loop speed-of-response, and the setpoint/disturbance characteristics of the control problem. Two implementation algorithms are presented which are applied to the plant data obtained from a distillation column. The results show that substantial improvements are obtained from control-relevant prefiltering in output error and partial leastsquares estimation, while some caution must be exercised when applied to FIR and low-order ARX estimation.