Research On Recursive Identification Methods For Time-varying Output Error Systems

Time-varying systems have always been present in complex industrial areas,physical processes and natural phenomena.Despite decades of development,the study of time-varying systems is still a popular research topic due to its wide range of applications.In this paper,we address the problem of parameter identification for timevarying output error-like systems,based on generalized expressions for time-varying parameters,using recursive class identification methods incorporating multi-innovation identification theories and online interaction estimation,etc.The main research elements are as follows:(1)A generalized time-varying system identification model based on time-varying parameter approximation relations(the product of the coefficient matrix and the measurable disturbance vector)is derived for the sliding average model in the output error class system;and an extended stochastic gradient algorithm based on the auxiliary model is derived using the auxiliary model idea;at the same time,a multi-innovation extended stochastic gradient algorithm based on the auxiliary model is derived in conjunction with the multi-innovation identification theory.(2)For the autoregressive sliding average model in the output error class system,the hierarchical technique is used to decompose the system into two low-order subsystems,while the online interaction estimation is combined with the auxiliary model idea to solve the problem of more parameters and the problem of containing unknown parameters;the generalized extended stochastic gradient method based on the auxiliary model and the generalized extended stochastic gradient method based on the hierarchical auxiliary model are derived,and the analysis of the computational effort of the two algorithms is given.(3)For the time-varying output error autoregressive sliding average model under colored noise interference,the data filtering technique is used to deal with the interference generated by the noise interference term on the time-varying parameters of the discrimination,while combining the online interactive estimation method and the measurable input and output data to interactively estimate the unknown parameters between the filtering subsystem and the noise subsystem,and derive a generalized extended stochastic gradient algorithm based on the data filtering;in order to further improve the identification accuracy,the multi-innovation generalized extended stochastic gradient algorithm based on the data filtering is derived by combining the multi-innovation identification theory.The paper derives several recursive class online estimation algorithms for the identification of time-varying parameters using generalized expressions for timevarying parameters for time-varying output error class systems,and performs numerical simulations and analysis.The feasibility and effectiveness of the proposed algorithms are demonstrated in simulation experiments.