A Fault Classification Method Based on Improved Fuzzy Support Vector Machine

Abstract

The invention discloses a fault classification method based on an improved fuzzy support vector machine. The method comprises: obtaining sample data of a switch cabinet, establishing an optimization model for the sample data by introducing a relaxation factor, and calculating the hypersphere of the optimization model The minimum radius of the hypersphere; take the minimum radius of the hypersphere as the boundary of the hypersphere, and use the segmented semi-decreasing normal cloud model to calculate the degree of membership of the sample data inside and outside the boundary of the hypersphere; The degree of membership constitutes a sample set of fault types, and the sample set of the fault types is used to obtain the optimal segmentation function of the improved FSVM; the optimal segmentation function of the improved FSVM is used to establish a classifier of the improved FSVM, and the sample data are sequentially Through the classifier of the improved FSVM, the fault type of each sample data is obtained. This method achieves the effect of improving the fault classification of the switchgear.

Description

technical field [0001] The invention relates to the technical field of fault classification, in particular to a fault classification method based on an improved fuzzy support vector machine. Background technique [0002] At present, the operation status of cables, busbars and circuit breakers is monitored, fault characteristic parameters are extracted, and typical fault classification and identification can be realized by using fault characteristic data, which is conducive to the identification and diagnosis of various fault states and guides its operation and maintenance management. For the classification and identification of faults, there are currently decision tree classification, Bayesian classification, neural network classification and support vector machine classification, among which support vector machine (Support Vector Machine, SVM) can obtain training learning accuracy and test accuracy on small sample data. The optimal compromise between recognition capabilitie...

AI Technical Summary

Problems solved by technology

For the classification and identification of faults, there are currently decision tree classification, Bayesian classification, neural network classification and support vector machine classification, among which support vector machine (Support Vector Machine, SVM) can obtain training learning accuracy and test accuracy on small sample data. The optimal compromise between recognition capabilities has good adaptability, but SVM treats all effective samples and noise samples equally when constructing the optimal classification surface, and its noise samples will easily affect the classification results of SVM

Although Fuzzy Support Vector Machine (FSVM) can eliminate the influence of noise samples, it ignores the interaction and dispersion of similar samples, and the effect is not good. Therefore, an improved FSVM is proposed to classify switchgear faults. have a better effect

Images