Evaluation of a simplified multi model predictive controller for a three stage power plant superheater based on projection of controller output onto the admissible set

Research focused on model predictive control (MPC) with low computational demands and good real time predictability usually exploits the properties of the explicit model predictive control. However, if the control problem is complex and the number of constraints and free control moves is high, explicit MPC can be even worse in the terms of computation time and real time predictability than standard MPC using on-line optimization. The purpose of this paper is to evaluate the application potential of a different approach to MPC with low computational demands: constraining the controller output by projection. This method was proposed by Tatjewski several years ago. However, it has found relatively little application. In this paper this method is extended to the case of switched multi linear MPC and evaluated using a case study of MPC control of a three stage power plant superheater. It is shown that even in the case of this relatively complex controlled plant, this simple method is able to achieve fairly good results. That is why it can be considered a suitable candidate approach for designing MPC controllers with low computational demands in applications with constraints on process control inputs and their rates of change.