Preprocessing of Raw Data for Developing Steady-State Data-Driven Models for Optimizing Compressor Stations

Compressors operate in parallel to increase the supply of a gas in many applications in process industries (e.g. an air separation process). To optimally distribute the load of the compressors in parallel, an optimization problem is formulated that takes into account the operational constraints of the compressors and the objective is to reduce operational costs, i.e. power consumption of the drivers. The optimization takes place when the system is in steady-state. The structure of the optimization employs steady-state data-driven models to represent the operation in steady-state. Many researchers reported that the identification of steady-states of the data plays a key role for accurate representation of the actual process by a data-driven model. However, to the best of the authors' knowledge, there is not much research on the quantification of the influence of the output of the steady-state detection methods on the data-driven models. For these reasons there is a need to examine this topic.