Research On Improved Algorithm For Nonlinear System Identification With Quantized Output

In this paper,the output-error(OE)model with quantized output and the nonlinear output-error(NOE)model with quantized output are studied.A recursive prediction error identification algorithm based on modularization and recursive idea is proposed.On this basis,in order to use more available data in each recursion,a weighted recursive prediction error identification algorithm is proposed.In the improved weighted algorithm,the information in the current step is used to update the estimate of the unknown parameter.The estimated value of the parameter at time t is updated by the correction term,and the estimated value of the parameter will be updated by the correction term at the current recursive step and the correction term at the previous moment.To ensure the accuracy of the estimate,the new correction term will be the weighted sum of the correction term at the current recursive step and the correction term at the previous moment.The main research work of this paper is as follows:1.A linear output error model is followed by a known quantizer,the output of the output error model cannot be directly measured,making it difficult to identify the parameters of the output error model.Therefore,for the output error model with quantized output,this paper uses the recursive prediction error algorithm to identify the parameters of the output error model.On this basis,a weighted recursive prediction error algorithm,a variable gradient weighted recursive prediction error algorithm and an improved weighted recursive prediction error algorithm are proposed.In these three weighting algorithms,the estimated value of the parameter is updated by the weighted sum of the correction term of the current recursive step and the correction term of the previous moment.From the simulation experiment results,the weighted algorithm can obtain better identification performance.2.This paper presents the recursive prediction error algorithm for the nonlinear output error model with quantized output.In the algorithm,the derivative of the quantizer is taken by the constant as the simplest approximation.Consider that the derivative of the quantizer is used to approximate a large error,a variable gradient recursive prediction error algorithm is proposed,and the derivative of the quantizer is approximated by a smoothing function.In the recursive prediction error algorithm and the variable gradient recursive prediction error algorithm,the parameter values used in the algorithm recursive process use the recursive initial value.In order to make each step more efficient,a modified recursive prediction erroralgorithm is proposed.In the improved recursive prediction error algorithm,the parameter values obtained in the recursion are used to correct the correlation calculation in the algorithm.