Intelligent Modeling Method For A Class Of Nonlinear Systems Based On SCNs

Aiming at the problem that it is difficult to establish an accurate mathematical model for a class of nonlinear discrete dynamic systems,a hybrid intelligent modeling method for nonlinear systems based on stochastic configuration networks(SCNs)is proposed.The forgetting factor recursive least square(FFRLS)is used to identify the unknown parameters of the lower-order linear model,and SCNs is used to estimate the linear partial error in this algorithm.In the estimation algorithm,the input weights and deviations of hidden layer nodes can be randomly allocated by SCNs according to the monitoring mechanism,and the output weights are automatically corrected,and the hidden layer nodes are added independently until the preset estimation accuracy is achieved,which improves the performance modeling accuracy of nonlinear system.The contributions is as follows:(1)The existing linear identification methods and the nonlinear identification methods are briefly reviewed and analyzed,and some aspects worthy of improvement are explored;(2)For a class of nonlinear discrete-time dynamic systems,firstly,the Taylor expansion of the system is carried out at the working point to express the model as the sum of lower-order linear part and the unmodeled dynamic part.On this basis,a hybrid modeling method for the nonlinear systems based on SCNs is proposed.The proposed algorithm combines SCNs with FFRLS,and the alternate identification strategy is adopted.At the end of each sampling,the linear model parameters are first identified by FFRLS algorithm,then the unmodeled dynamics are estimated by SCNS algorithm,and finally the hybrid intelligent model of the nonlinear system is established.(3)In order to illustrate the effectiveness of the proposed modeling algorithm,four groups of simulation experiments are carried out,and the experimental results show the effectiveness of the proposed algorithm.