Distributed modeling method of large chemical process based on cca-pls

Abstract

The invention discloses a distributed modeling method for a large-scale chemical process based on CCA-PLS. First, all process variables are collected for the large-scale chemical process, output variables are determined according to process quality indicators, and characteristic components of the process variables are extracted by a canonical correlation analysis method. Calculate the maximum correlation coefficient between each output variable and all input variables and the corresponding principal axis vector according to the characteristic components, select the independent input variables and interaction input variables of each subsystem according to the absolute value of each component of the principal axis vector, and realize the decomposition of the large system . After the subsystems are divided, the PLS algorithm is used to model each subsystem, and the component that maximizes the covariance between the input and output variables is extracted, and the subsystem model is obtained by using the regression modeling technology. The beneficial effects of the invention are that only the input and output data of the process are used, and the typical correlation analysis principle is used to screen the input variables of the subsystem, the model dimension is reduced, the model structure is simplified, and the PLS modeling algorithm is used for the subsystem modeling, which eliminates the need for practical The calculation difficulty caused by the collinearity of a large number of variables in the application, and the model accuracy is high.

Description

technical field [0001] The invention belongs to the technical field of chemical process modeling and relates to a large chemical process distributed modeling technology based on canonical correlation analysis (CCA) and partial least squares (PLS). Background technique [0002] With the advancement of technology and people's pursuit of economic interests, the scale of chemical production process is getting larger and more complex. A process is usually composed of many subsystems interacting through material, energy and information flow. There are complex correlations between subsystems, and there are variability and uncertainty in the external environment and internal disturbances. The traditional centralized predictive control (Centralized MPC) has encountered difficulties in object modeling and optimization calculations [1], and has been unable to meet the Industrial production development needs. The above-mentioned new characteristics of chemical process, coupled with the...

AI Technical Summary

Problems solved by technology

[0013] The purpose of the present invention is to provide a data-driven distributed modeling algorithm for large chemical process, which does not require knowledge of process mechanism, only based on production data analysis for subsystem division and dimensionality reduction, and further establishes a subsystem model to solve the current problem of large chemical process The problem that the complexity and accuracy of the model structure cannot be taken into account provides a model basis for distributed predictive control and a simple and feasible method for modeling the actual large-scale chemical process

Images