Soft sensor design for the optimisation of parallel debutaniser columns: an industrial case study

Abstract

This work demonstrates a practical implementation of a soft sensor to estimate the C5 hydrocarbon impurity in the butane product of a liquid petroleum gas (LPG) recovery system. Such a sensor can then subsequently be used to optimise the process. The process has two parallel debutaniser columns that feed a common LPG recovery system. The optimisation objective is to minimise the Reid vapour pressure (RVP) of the two debutaniser bottoms’ products. This optimisation problem can be solved with a simple advanced control implementation. However, the ability of the controller to minimise the process variation and drive the process to the optimal point is directly influenced by the quality of the constraining process variable. In this case, the key controlled variable (CV) is the debutaniser overheads C5 mass fraction. The designed soft sensor for this CV uses the general distillation shortcut (GDS) method, and is shown to represent the distillation column operation well. This work presents a derivation of the GDS method, and formulates a new approach for the feedback biasing of the two parallel debutaniser soft sensors.