Intelligent Algorithms Of Parameter Identification Of Wiener Nonlinear Systems

Due to the fact that almost all industrial processes are nonlinear systems, using linear model to describe nonlinear system can’t satisfy the demand of modern industrial production control. In other words, the linear system identification theories and methods are not applicable for nonlinear system identification. Besides, a precise and effective nonlinear model is the key and the basic premise to identify the nonlinear system and optimize the control process.There is no way to describe all the nonlinear process with a single mathematical model, so the nonlinear models based on modular attract widespread attentions. They are not only simple in structure, but also can describe nonlinear systems accurately, so it is universally significant and practically valuable to study them. The Wiener model is a nonlinear model, which consists of a dynamic linear module and a static nonlinear module. In view of the problem that the traditional optimization algorithm can’t meet the requirements of speed and accuracy in Wiener model parameter identification, this paper introduced the intelligent optimization algorithms.Firstly, this paper introduced particle swarm algorithm and proposed an improved fusion algorithm to solve the problem of slow convergence speed through the analysis of the search performance. This improved algorithm consisted of the key idea of bacterial foraging algorithm and catfish effect to add the chances of jumping out of local optimal value and the premature convergence. Several simulation experiments of the test functions, the classic Wiener model and a kind of fluid control valve identification, comparing with several other improved methods, verified the effectiveness and availability of the improved algorithm.Secondly, this paper made a detailed analysis of the control parameters of differential evolution algorithm in the identification of a classic Wiener model, and presented an improved method of adaptive parameter. Inspired by the firefly algorithm, a hybrid differential evolution algorithm is proposed to further optimize the performance of the algorithm. The improved algorithm showed the superiority of getting better parameter estimation than several other improved differential algorithms in experiments of parameter identification of classic Wiener model and saturating sensor model.Finally, in order to analysis capabilities of two improved algorithms in identifying complex nonlinear Wiener models, the improved algorithms were applied to identify the third order Wiener model and several kinds of Wiener models with dead zone characteristics. The results showed that both of the improved algorithms can successfully identify the parameters of the models. What’s more, the speed and accuracy of the improved algorithms is much better than the standard algorithms.