Data-Driven Dynamic Modeling of the Acetylene Hydrogenation Process based on Nonlinear Slow Feature Analysis

The accuracy of process model has a significant impact on control and optimization. However, for complex time-variant systems, the traditional steady state modeling method is generally not effective on dynamic process. In this paper, a data-driven dynamic modeling method based on nonlinear slow feature analysis (SFA) is proposed to reduce the effect of deterioration with age, such as catalyst deactivation. Variables obtained by known mechanisms are used to consider nonlinear relationships between process variables, then, the SFA is executed to extract slowly changing characteristics, and finally, slow features predict model is constructed. The method is applied to the industrial acetylene hydrogenation reaction process, and it can be confirmed that the proposed method enables models to predict response accurately.