Support vector machine power fault classification method and system based on multi-feature fuzzification processing

Abstract

The invention discloses a support vector machine power fault classification method based on multi-feature fuzzification processing. The method comprises a training step and a classification step. The training step comprises: collecting a sample characteristic signal characterizing a power equipment fault, and carrying out step membership function fuzzification processing to obtain fuzzified sample characteristic values; performing dimensionality reduction processing on the fuzzified sample characteristic values to reduce the quantity of the sample characteristic values; and constructing a support vector machine, optimizing parameters of the support vector machine by using a chaos optimization algorithm, and inputting the sample characteristic values into the support vector machine, wherein the output of the support vector machine is the type of the power equipment fault. The classification step comprises: collecting an actually measured characteristic signal of the power equipment and carrying out step membership function fuzzification processing to obtain fuzzified actually measured characteristic values; performing dimensionality reduction processing on the fuzzified actually measured characteristic values, and inputting the actually-measured characteristic values into a support vector machine, wherein the output of the support vector machine is the type of the power equipment fault.

Description

Technical field [0001] The present invention relates to a method and system for fault classification, particularly to a method and system power failure classification. Background technique [0002] In the power system, power equipment operating status will directly affect the security of the entire system, when a major power equipment failure may cause the power system to a standstill. Thus, in order to ensure that the power to run the system, after the event of power equipment failure, power failure is often necessary facilities for rapid diagnosis and classification, in order to repair the fault in the engineering personnel, to restore electricity to run the equipment. [0003] In this prior art, the oil dissolved gas analysis (the DGA) method has been recognized as one of the fault diagnosis apparatus and insulation life assessment most convenient and effective method of power. However, in practice the diagnostic process, using this data, fault diagnosis DGA three-ratio method...

AI Technical Summary

Problems solved by technology

However, in the practical diagnosis process, the three-ratio method or the improved three-ratio method using this DGA data for fault diagnosis has defects such as "lack of coding" and "too absolute coding boundaries".

[0004] According to the analysis of the mechanism of dissolved gas in oil when power equipment fails, it is found that there is no clear function mapping relationship between the gas content in oil and the type of power equipment failure, and the distribution characteristics of gas content are also difficult to speculate, and the actual field data Acquisition accuracy and quantity are also very limited

Therefore, the traditional diagnostic classifier based on the principle of empirical risk minimization cannot obtain enough knowledge for learning, its classification accuracy is not high, and the diagnostic error is large

Images