Design of model predictive control for industrial process with input time delay

Abstract

This paper focuses on designing model predictive control (MPC) for industrial process with input time delay. The process is typically represented by a first order plus dead time model which is obtained via a step test. The objective of design is to track reference input while control input has to satisfy the bound conditions. MPC utilizes two types of state observers, namely, input-delay observer and state-delay observer. For input-delay model, state consists of only the first order dynamic and the delay term is explicitly embedded in the input to the plant. For state-delay model, the state variable consists of the first order dynamic and the delay term. The formulations of MPC design of these observers are different. To determine optimal control inputs, MPC formulations are converted to quadratic programs with inequality constraints which can be efficiently solved by the active set method. Numerical examples employ the level control loop as a case study and compare the output tracking performance, control input, and the computational time of the proposed design.