A NONLINEAR PREDICTIVE CONTROL STRATEGY BASED ON RADIAL BASIS FUNCTION NETWORKS

An identification and predictive control strategy for nonlinear processes based on radial basis function networks is proposed. First, a radial basis function network is developed as a process model using an orthogonal least squares learning algorithm. This model is then used to train the nonlinear predictive controller, which is also implemented as a radial basis function network. This strategy allows the application of the very powerful nonlinear predictive control technique to processes for which accurate physical process models are not available. Since no optimization problems have to be solved on-line, this control strategy is very easy to implement. The proposed identification and control strategy is applied to a simulated pH neutralization process. It shows both excellent setpoint tracking performance and disturbance rejection when compared to alternative control strategies.