Effect of Disturbance Model Selection upon Performance, Robustness and Control Effort of Model Predictive Controllers
Date
2014-01-29
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Unmeasured disturbances are encountered in the chemical process industries but modelling them accurately can be a difficult task. Instead, commercial Model Predictive Control packages assume various disturbance models a priori. In an attempt to identify the best default model, this research evaluates closed-loop properties such as best achievable servo and regulatory performances, robustness and control effort associated with selected disturbance models. Results show that poles of assumed disturbance model should match those of the actual disturbance as closely as possible. Further, numerator polynomial of the model, or “T-filter”, represents a strong handle for affecting compromise between robustness and achievable regulatory performance. Rather than adopting a default T-filter (e.g. z = 0.8) it is recommended that selection of the order and pole of the T-filter be made on a case-by-case basis. Detailed guidelines for choosing these parameters are provided for processes whose open-loop dynamics can be well-approximated as first-order-plus-deadtime.
Description
Keywords
Engineering--Chemical
Citation
Liu, W. (2014). Effect of Disturbance Model Selection upon Performance, Robustness and Control Effort of Model Predictive Controllers (Master's thesis, University of Calgary, Calgary, Canada). Retrieved from https://prism.ucalgary.ca. doi:10.11575/PRISM/25796