An open-loop identification method for industrial process multi-order inertial systems based on deep learning

Abstract

The invention discloses an open-loop identification method for a multi-order inertial system of an industrial process based on deep learning. The present invention is based on a deep learning random inactivation neural network, and the first-order inertial filtering link with an inertial time constant of 30 seconds and 60 seconds and the second-order inertial filter with an inertial time constant of 30 seconds, 60 seconds and 160 seconds are set at the input end of the network After the filtering link, an open-loop object recognizer based on deep learning is formed; when identifying the open-loop object model, a forward and reverse step input is added to the input of the open-loop object model, and the open-loop object model outputs the corresponding data , and then input the input data and output data into the open-loop object recognizer at the same time. After the deep learning random deactivation neural network is trained offline, it can effectively identify the characteristics of the open-loop object model. The invention can easily and accurately identify the open-loop object model of the multi-order inertial system, and effectively improves the control quality of this type of system.

Description

technical field [0001] The invention relates to the field of industrial process control, in particular to an open-loop identification method for industrial process multi-order inertial systems based on deep learning. Background technique [0002] A common type of controlled object in industrial processes is the multi-order inertial system, which has the characteristics of large delay and large inertia. In order to obtain good control performance, it is often necessary to identify the characteristics of the object model. [0003] At present, the identification of the object model characteristics of multi-order inertial systems usually adopts the conventional least squares identification algorithm, but the conventional least squares identification algorithm needs to add more complex excitation signals, and the accuracy of the identified object model is not high. Contents of the invention [0004] The technical problem to be solved by the present invention is to overcome the ...

AI Technical Summary

Problems solved by technology

[0003] At present, the identification of the object model characteristics of the multi-order inertial system usually adopts the conventional least squares identification algorithm, but the conventional least squares identification algorithm needs to add more complex excitation signals, and the accuracy of the identified object model is not high.

Images