Adaptive nonlinear model predictive control for a class of multivariable chemical processes

An adaptive nonlinear model predictive control (ANMPC) algorithm was proposed for a class of multivariable nonlinear systems in chemical processes.By on-line identifying the parameters of the predictive model with recursive least-squares algorithm, a BP (back propagation) neural network steady-state model was developed to represent the static nonlinearity of the systems, and the parameters of the ARX predictive model were also modulated according to the dynamic gain calculated from the steady-state model.The design procedure of ANMPC controller was elaborated.The simulation results in a multivariable pH neutralization process demonstrated the feasibility and effectiveness of the ANMPC.