Automated optimization of generalized model predictive control for DC-DC converters

Simon Effler, Anthony Kelly, Mark Halton, Karl Rinne

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Generalized Predictive Control (GPC) offers a method of designing digital compensators directly in the discrete-time domain. In this paper, an automatic design process based on the optimization of a few GPC parameters is presented. The application to DC-DC converters offers real benefits because of its clearly defined design process, time-domain performance criteria, simple tuning technique and guarantee of stability. For practical applications, the guarantee of stability may not be sufficient, certain performance criteria must also be achieved. In this design process, a performance index is used in the optimization routine to quantify specific performance objectives. A novel performance index is presented which weights performance and robustness for a more optimized compensator design. For illustration purposes an optimal GPC compensator is designed and tested for a buck converter. The resulting compensator is critically assessed in simulation and validated with experimental hardware.

Original languageEnglish
Title of host publicationProceedings - CIS Workshops 2007, 2007 International Conference on Computational Intelligence and Security Workshops, CISW 2007
Pages134-139
Number of pages6
DOIs
Publication statusPublished - 2008
EventPESC '08 - 39th IEEE Annual Power Electronics Specialists Conference - Rhodes, Greece
Duration: 15 Jun 200819 Jun 2008

Publication series

NamePESC Record - IEEE Annual Power Electronics Specialists Conference
ISSN (Print)0275-9306

Conference

ConferencePESC '08 - 39th IEEE Annual Power Electronics Specialists Conference
Country/TerritoryGreece
CityRhodes
Period15/06/0819/06/08

Fingerprint

Dive into the research topics of 'Automated optimization of generalized model predictive control for DC-DC converters'. Together they form a unique fingerprint.

Cite this