PCA-based Method of Identification of Dominant Variables for Partial Control

Abstract

Since the early use of automatic control, the Partial Control strategy has frequently been adopted in complex chemical processes having more process variables than manipulated variables. The key idea of Partial Control is to find the dominant variables which can be controlled to constant setpoints and in turn leads to acceptable variations in the operating objectives in the face of external disturbances occurrence. Although the idea seems simple to understand, the identification of the dominant variables can be a daunting task where presently this is largely done based on extensive process knowledge and experience. In this paper, we present a novel methodology to identify the dominant variables based on Principal Component Analysis. The method can greatly facilitate the implementation of Partial Control strategy because it does not require extensive process experience and knowledge. The effectiveness of the methodology is demonstrated based on its application to a complex extractive fermentation process.