Convolution model based predictive controller for a nonlinear process

The model predictive control (MPC) of a distributed parameter nonlinear laboratory heating system is studied. A nonlinear convolution model consisting of a linear dynamic and a nonlinear steady-state part is applied as the model of the process in the MPC algorithm. The dynamic part is represented by a relative impulse response model (IRM). The steady-state gain is derived from the first principle model of the system. The application of this special convolution model is as simple as the use of the transfer function model; however, it is valid on the whole operating range. MPC algorithms employing different models of the process are compared by simulation and physical tests.