Opening and closing coil current analysis method combining Bayesian update and DS evidence theory

Abstract

The invention discloses an opening and closing coil current analysis method combining Bayesian update and a DS evidence theory. The Bayesian update compensates the influence of the environment on parameter estimation by performing feature residual extraction on the difference between an actual feature and an ideal feature and taking the difference as a new classification feature; and on the basis of a Bayesian parameter estimation result, an uncertainty probability and DS evidence reasoning method is introduced to realize DS data fusion. The method provided by the invention has relatively low knowledge dependency and relatively weak subjective degree, and trust distribution does not need to construct a complex membership function. According to the method, the method is low in degree of dependence on fault samples, and is suitable for occasions such as high-voltage circuit breakers where the reliability requirement is high and the fault samples are relatively difficult to obtain; and in practical application, the method can introduce characteristics of signals such as vibration, displacement and pressure in the fusion process, so that the circuit breaker state monitoring system which is more comprehensive and higher in fault diagnosis accuracy is realized.

Description

technical field [0001] The invention relates to a method for analyzing the current of opening and closing coils combining Bayesian updating and DS evidence theory, and belongs to the technical field of circuit breaker control. Background technique [0002] The operating mechanism is the most frequently faulted component in the circuit breaker system, and the opening and closing coil current analysis method is a method widely used in fault diagnosis of the circuit breaker operating mechanism. The obtained operating coil current is used as the basis for fault diagnosis to evaluate the reliability of the circuit breaker. Compared with other methods such as the vibration analysis of the operating mechanism body and the displacement analysis of the transmission spindle, the coil current acquisition does not require a major change in the mechanical structure, and has a better fault diagnosis performance at a relatively low modification cost. [0003] In terms of diagnostic algori...

AI Technical Summary

Problems solved by technology

[0003] In terms of diagnostic algorithms, Rao et al. used wavelet denoising and Euclidean space distance comparison methods to realize the identification of coil faults, but this method cannot judge the specific type of faults

On the basis of a large number of coil current observations, Razi-Kazemi analyzed the current characteristic changes under different faults, formulated a series of fault classification rules, and realized the classification of faults. However, this method can only be used after the waveform is obviously distorted. Analyze the cause of the failure

Strachan et al. adopted clustering method and decision tree rule extraction, and implemented a fault diagnosis method based on data mining, but this method relies on a large number of fault samples, and is easily constrained by the lack of router samples in application.

By introducing the DS theory, relaxation has carried out data fusion on the coil current characteristics, and realized the fusion reasoning of faults, but this method has the disadvantage of being too subjective in the process of assigning evidence measures.

In order to make up for this shortcoming, Li Haiying et al. proposed a diagnostic method based on fuzzy-evidence theory and achieved good diagnostic results. However, the choice of fuzzy membership function greatly affects the reliability of the method.

Ge Baohua started from the perspective of rough sets, simplified the attributes of the coil current characteristics, and established the minimum diagnosis rule decision table, but its discrete output results still cannot meet the needs of condition-based maintenance.

In order to solve the demand for digital evaluation indicators of condition-based maintenance, Guan et al. proposed to divide the characteristic probability distribution into deterioration, failure, normal and other areas, and calculated the probability of various types of failures based on the assumption of independent events, and realized the fault analysis. Risk assessment, but there are still disadvantages that the coil current characteristics are easily affected by voltage input, and the characteristic data is not a simple independent relationship, which may easily cause misjudgment of faults

Images