Backtracking Location Of Fault First Arrival Time Based On System Smoothing And Filtering

The state estimation method of the system is divided into smoothing,filtering and prediction according to different requirements in the time series.They have a broad application background in the fields of fault detection,navigation,audio and video processing.Filtering is an estimation of the real-time state,while smoothing is an estimation of the historical state.At present,the commonly used smoothing algorithms all need to record the forward filtering data in the iterative process to update the estimated value of the historical point,and there is no clear definition for the selection of the window.This paper provides a clear and effective window selection method for this type of smoothing problem,and designs a type of fixed hysteresis smoother with a limited storage structure.Finally,for the system under both nonlinear and nonGaussian settings,a combination of feature function filtering based on unscented transform and RTS smoother is proposed.The main content of this article can be summarized as follows:1)A fixed lag smoother is established that optimizes window size and estimation accuracy.Through the analysis of the state reconstruction of the linear system,it is concluded that the estimation accuracy of the fixed-lag smoothing method is positively correlated with the size of the lag window,and the discrimination threshold is increased by setting the accuracy,and the window is minimized while maintaining appropriate accuracy.Size,and finally applied it in the simulation experiment of fault retrospective location.2)Design a fixed-lag smoother with a limited storage structure.This kind of fixed lag smoother with limited storage is different from the traditional fixed lag smoother.It does not need to record a large amount of historical information of forward filtering,thereby saving storage space and reducing computational complexity,and can maintain good estimation performance.This is also in line with the influence of the window size on the estimation accuracy in the idea of state reconstruction.Finally,a fault disturbance location method is set up through the residual generator.3)Aiming at the nonlinear system,design a method based on the feature function filtering without trace change and the smooth combination of RTS.It retains the advantage that feature function filtering based on unscented transform can be applied to nonlinear non-Gaussian systems,and the convenient RTS smoother can further improve the estimation accuracy.Finally,the conjecture was verified by comparison simulation experiments of several nonlinear filters and smoothers.