Anti-outlier adaptive Kalman filtering method for frequency scale output hopping detection

Abstract

The invention belongs to the technical field of frequency standard fault detection, and discloses an anti-outlier adaptive Kalman filtering method for frequency scale output hopping detection. The anti-outlier adaptive Kalman filtering method for the frequency scale output jump detection comprises the steps of enabling the outlier adaptive Kalman filter to perform recursive filteringby utilizing the observation data, and meanwhile, obtaining the ratio of the actual variance to the theoretical variance of measurement information according to facts; and obtaining a posterior frequency deviation matrix according to the priori frequency deviation matrix, obtaining a frequency deviation gain matrix according to the priori mean square error of the priori frequency deviation, and obtaining a posterior frequency deviation mean square error matrix according to the gain matrix. According to the present invention, the sample data with frequency hopping or outliers often enables a Kalman filter to perform error correction in state prediction of a system, so that a filtering result is deviated and even diverged, and the filtered data can be closer to a real state after the innovation is weighted.

Description

technical field [0001] The invention belongs to the technical field of frequency standard fault detection, and in particular relates to an anti-outlier self-adaptive Kalman filter method for frequency standard output jump detection. Background technique [0002] The most widely used detection methods for frequency standard jumps are the least squares method, block average, continuous average, maximum likelihood estimation, Kalman filter algorithm and dynamic Allen variance algorithm. When using the least squares method, it is necessary to automatically ignore the variable error and then fit the corresponding points on the curve, and when using this method for frequency hopping detection, the detection accuracy will not be high enough, and when more data is detected, it will be Use a longer detection time; block averaging is a method of comparing the average value in adjacent windows with a selected threshold. This detection method is only suitable for processing a small amou...

AI Technical Summary

Problems solved by technology

[0004] To sum up, the problem existing in the existing technology is: when the Kalman filter cannot determine the precise mathematical model and the precise noise statistical characteristics of the research object, the filtering accuracy will be greatly weakened, and even cause the filtering divergence in severe cases.

Difficulty in solving the above technical problems: In reality, the Kalman filter model cannot be very accurate, and the noise statistical characteristics are mostly based on estimation calculations, so when using the Kalman filter, the problem of insufficient filtering accuracy may occur. And there may be misjudgments, so how to automatically adjust the data, identify the existence of outliers, and adaptively adjust the data accuracy is an existing problem

Images