Inferential Active Disturbance Rejection Control of a Heat Integrated Distillation Column Using Dynamic Principal Component Regression Models

This paper presents a multivariable inferential active disturbance rejection control strategy for product composition control in a heat integrated distillation column (HIDiC) to overcome the large measurement delay associated with composition measurement. The inferential estimator uses multiple tray temperature measurements to estimate the product compositions. Dynamic principal component regression is used to overcome the strong co-linearity among tray temperatures and incorporate dynamics to build the inferential estimator model. The effectiveness of the proposed method is demonstrated on a simulated HIDiC based on mechanistic model.