Research On State Estimation Method For Systems With Complex Uncertain Noise

The estimation theory of stochastic systems,also known as the filter theory,is a very important branch of Modern Control Theory as well as a key subject of Signal and System.The so-called "estimation" is to extract the practical information from the observation data containing random noise,so as to estimate the parameters or the state of the system.With the rapid development of space simulation,geological exploration,building engineering,financial analysis,voice processing,biological engineering,marine science and so on,the real-time,robustness and accuracy of estimation algorithms are becoming more important.This paper mainly analyzes and discusses the two most prominent problems of the state estimation algorithm in practical application: 1.The dynamic characteristics of the practical system are uncertain,so the dynamic characteristics of the system process model can not be accurately obtained.2.The practical system contains complex uncertain noise,which seriously affects the accuracy of system state estimation.First of all,we give the following two solutions to the first problem: 1.The idea of datadriven model is introduced in this paper.Based on the basic idea of ’data contains the model and the model is integrated into the data’,the model parameters are updated in real time.2.The statistical distribution of the state of the practical system is more consistent with the Rayleigh distribution than the uniform distribution.In this paper,the concept of the current model is applied to make the algorithm more convergent.Secondly,based on the complex uncertain noise contained in the system,we propose the following two solutions: 1.In order to deal with the colored process noise contained in the practical system,the colored process noise is regarded as a first-order Markov process by using the Wiener theorem,and the colored process noise is whitened.2.In order to deal with the complex and uncertain measurement noise contained in the practical system,this paper uses the principle of covariance matching and the noise innovation to update the measurement noise variance in real time,and deals with the colored measurement noise of the system.Finally,the proposed state estimation algorithm is applied to the practical application of large-scale civil structure health monitoring.The experimental results show that the proposed method is robust and real-time.The accuracy can be well applied to the state estimation problem of systems with complex uncertain noise.