Adaptive implementation of Wiener model based nonlinear predictive control

This paper discusses issues of model adaptation in a Wiener model based nonlinear predictive control (WNPC). WNPC is an extension of state-space generalized predictive control with a Wiener model. It combines the advantages of predictive control with the ability to control highly nonlinear processes, while retaining a moderate level of computational demand. It is aimed at a class of nonlinear processes that can be described with a Wiener structure. Model adaptation is required when the process is affected by nonmeasurable disturbances that severely affect the nonlinear characteristics of the system. It is shown that better disturbance rejection performance can be achieved at the cost of an increased computational demand, and that robust tuning of the estimator is a difficult problem. The method is evaluated with a pH neutralisation process case study.