Combined optimal sensor network design and self-optimizing control with application in a typical sugarcane mill

The concept of self-optimizing control is applied in the context of sensor network design to support the simultaneous selection of optimal sensors and controlled variables. Optimality is defined as minimizing the capital cost of the sensor network as well as the revenue loss due to disturbances and measurement errors under a constant set-point policy. The method is applied to a sugarcane mill case study, where 41 controlled variables (CVs) are selected as linear combinations of the measured variables, and the optimal supporting sensor network is identified. The economically optimal sensor placement found in the present study reduced the average revenue loss due to disturbances and measurement errors by 39% compared to a mill operating with a typical sensor network. The substantial reduction in revenue loss for the optimal sensor placement is attributed to an improved precision in 19 CVs compared to the base case network. The case study demonstrates the natural extension of self-optimizing control to optimal sensor network design, as well as the potential benefits of the combined approach.