Neural fuzzy Wiener-Hammerstein model identification method based on combined signal source

Abstract

The invention discloses a neural fuzzy Wiener-Hammerstein model identification method based on a combined signal source. The method comprises the steps of: constructing a single-input single-output Wiener model which is formed by connecting an input dynamic linear link, an output dynamic linear link and a static non-linear link in series; combining Gaussian signals and binary signals to form a multi-signal source as the input of the model; separating and identifying the dynamic linear link and the static nonlinear link in series by multi-signal sources. The neural-fuzzy system is used to approximate the static nonlinear link, wherein, if the input is a Gaussian signal, the Wiener-Hammerstein series module is separated effectively according to an assigned theorem, and the input product model of the input dynamic linear link and the output product model of the output dynamic linear link are separated correctly according to the non-excitation characteristic of the binary signal, so as toobtain the constituent variable parameters of the static sub-linear link. The identification method of the invention greatly simplifies the identification process of the model and has high predictionaccuracy of the model.

Description

technical field [0001] The invention belongs to the technical field of Wiener-Hammerstein (Wiener-Hammerstein system) identification, in particular to a neuro-fuzzy Wiener-Hammerstein model identification method based on a combined signal source. Background technique [0002] The Wiener-Hammerstein model is an important class of nonlinear systems applicable to a variety of industrial processes, such as multi-sensor data fusion, electrical stimulation of skeletal muscle models, heat exchangers, etc. The unmeasurable intermediate variables of the Wiener-Hammerstein model make it impossible to directly use the input and output data to use some existing simple system identification methods to directly identify the linear part and the nonlinear part. Separation identification of the linear part requires indirect estimation of intermediate unmeasured variables. If only approximators such as neural networks or fuzzy systems are used to approximate the model, the dynamics of the mo...

AI Technical Summary

Problems solved by technology

[0004] The main purpose of the present invention is to provide a neuro-fuzzy Wiener-Hammerstein model identification method based on a combined signal source, which is used to solve the problems in the prior art that the algorithm is relatively complicated and the number of iterations cannot be effectively guaranteed. The specific technical solution is as follows :

Images