Optimal controlled variable selection using a nonlinear simulation-optimization framework

In feedback control, controlled variables are those process variables which are measured and fed back to controllers. Then in the presence of disturbances, controllers by the means of manipulating the inputs aim to maintain the controlled variables at their setpoints. The objectives for the selection of controlled variables can be conflicting and competing. These objectives include minimization of (1) economic losses, (2) input manipulations, (3) output variations and (4) changes in process states. This research aims to present a systematic framework for optimal selection of controlled variables. Each of the above-mentioned objectives is defined within a multi-objective function. In addition, the reasoning behind the selection of nonlinear steady state model is explained. The proposed methodology is benchmarked on an industrial distillation train. Optimization programming is presented and the paper discusses how the size of the optimization problem can be reduced by means of engineering insights and addressing the concerns regarding feasibility of the developed control structure. The methodology is scalable to large industrial problems, while maintaining its rigour. The results confirm that a very good trade-off is established between different objectives.