Generalised predictive control : a study and application

This dissertation investigates the Generalised Predictive Control (GPC) method derived by Clarke, Mohtadi and Tuffs in the mid eighties. GPC is an adaptive predictive control algorithm, of which there are number, and has proven to be one of the more popular variants having been applied to various control applications. The theory of the GPC method is studied in detail and a computer simulation program has been written for analyses. While the theory is perhaps not straightforward the actual use of GPC is quite easy. There are 'tuning knobs' available which provide a simple means of tuning the response to match specifications.The GPC method is extended to an overall Long-Range Predictive Control (LRPC) method using the Long-Range Predictive Identification (LRPI) developed by Shook, Mohtadi and Shah in the early nineties. The new identification algorithm replaces the recursive least squares (RLS) used in the original GPC and is essentially a duplicate of the control law. The effect of using LRPI as opposed to RLS is investigated theoretically by use of the simulator. In the simulations carried out comparison of the GPC and LRPC responses depended on the GPC parameters employed. On the whole the LRPI scheme improved the responses though, especially when disturbances and model changes were investigated.

Grantor dc:publisher.institution

Department of Electrical Engineering

Year dc:date.issued

1996

Author dc:creator

• Prince, Karl John

Advisor dc:contributor.advisor

• Braae, Martin

Language dc:language.iso

eng

Handle dc:identifier.uri

http://hdl.handle.net/11427/9227

OAI identifier oai:identifier

oai:open.uct.ac.za:11427/9227