Controlled Variables Adaptation to Improve Process Optimality Using Historical Operating Data

Abstract

The selection of controlled variables (CVs) plays an important role in the process optimality and is highlighted in the methodology of self-optimizing control. In general, the self-optimizing control deals with expected disturbances via controlling CVs selected, while the unknown disturbances encountered in practice are not accounted for. A recent two-layer control architecture integrating self-optimizing control and modifier adaptation is able to handle both types of disturbances, which is however not effective in cases when the unknown disturbance are frequent. The controlled variable adaptation strategy proposed in this paper utilizes information in the historical operating data, endowing the self-optimizing control layer an ability of handling either disturbance mentioned above, in the aid of the upper modifier adaptation. Such transformation is beneficial to improve the process optimality because the self-optimizing control works in a much faster time-scale than the modifier adaptation. The Williams-Otto reactor is investigated to show the proposed methodology.