Journal article Open Access

Identification of a PWA Model of a Batch Reactor for Model Predictive Control

Gorazd Karer; Igor Skrjanc; Borut Zupancic

The complex hybrid and nonlinear nature of many processes that are met in practice causes problems with both structure modelling and parameter identification; therefore, obtaining a model that is suitable for MPC is often a difficult task. The basic idea of this paper is to present an identification method for a piecewise affine (PWA) model based on a fuzzy clustering algorithm. First we introduce the PWA model. Next, we tackle the identification method. We treat the fuzzy clustering algorithm, deal with the projections of the fuzzy clusters into the input space of the PWA model and explain the estimation of the parameters of the PWA model by means of a modified least-squares method. Furthermore, we verify the usability of the proposed identification approach on a hybrid nonlinear batch reactor example. The result suggest that the batch reactor can be efficiently identified and thus formulated as a PWA model, which can eventually be used for model predictive control purposes.

Files (210.3 kB)
Name Size
659.pdf
md5:bbe583f47e054f0ff1460a2b9850c2e5
210.3 kB Download
5
4
views
downloads
All versions This version
Views 55
Downloads 44
Data volume 841.1 kB841.1 kB
Unique views 55
Unique downloads 33

Share

Cite as