Fractionating tower benzene content soft measurement method based on stack Elman neural network

Abstract

The invention discloses a fractionating tower benzene content soft measurement method based on a stacked Elman neural network. According to the method, by referring to the composition structure of thestack-type auto-encoder, a stack-type Elman neural network model formed by connecting multiple stages of Elman neural networks in series is built, and therefore layer-by-layer nonlinear feature extraction of fractionating tower process data is achieved. The benzene content is further regressed and predicted by fully using output estimation values of the Elman neural networks at all levels, so that a fractionating tower benzene content soft measurement model with higher precision is established. According to the method, the time-varying characteristic adapting capacity of the Elman neural network is utilized, and nonlinear characteristics for soft measurement of the benzene content can be extracted layer by layer. In addition, according to the method, nonlinear characteristics extracted byElman neural networks at all levels are fully utilized, and least square regression is used for further reducing the difference between a soft measurement value and an actual value.

Description

technical field [0001] The invention relates to a soft sensing technology, in particular to a soft sensing method for benzene content in fractionating towers based on stacked Elman neural networks. Background technique [0002] In the ethylene production process, the catalytic cracking fractionation tower is a kind of chemical equipment for fractionating mixed volatile liquids, and is an important device for secondary processing of petroleum. The basic principle of the catalytic cracking fractionation tower is to use the difference in the relative volatility of each component in the gas-liquid phase to separate. In a column, vapor rises from the bottom to the top, and liquid descends from the top to the bottom. When the two phases of gas and liquid are in contact with each other on each layer, the gas phase is partially condensed, and the liquid phase is partially vaporized. Due to the partial vaporization of the liquid, the light components in the liquid phase diffuse to ...

AI Technical Summary

Problems solved by technology

However, the traditional Elman neural network cannot achieve layer-by-layer progressive extraction of nonlinear features

Images