A Multivariate Complementary Cross Validation Method for Partial Discharge Diagnosis Results

Abstract

The present invention relates to a method for performing multi-element complementary cross verification on a partial discharge diagnosis result. The method comprises following steps: performing partial discharge fault detection on power equipment using a plurality of partial discharge detection technologies; if a detection result of one of the partial discharge detecting technologies determines that the power equipment have a partial discharge fault, using the partial discharge detection technology as a leading detection technology, and giving a detection result; using several other partial discharge detection technologies as a set of auxiliary detection technologies, and performing multiple auxiliary detections on the power equipment from various aspects; selecting whether to retest the power equipment according to status of the multiple auxiliary detections, and meanwhile adjusting diagnostic confidence of the leading detection technology; and comparing diagnostic data of the leadingdetection technology with an expert collection to further confirm the diagnosis. The method solves the problem that one-sidedness of a single detection technology leads to unreliable analysis results, and avoids that some specific faults are ignored, thereby improving accuracy of fault diagnosis.

Description

technical field [0001] The invention belongs to the technical field of electric variable measurement, in particular to a method for multivariate complementary cross-validation on partial discharge diagnosis results. Background technique [0002] At present, the insulation performance of power equipment is mainly analyzed and judged by detecting the partial discharge of power equipment. Various physical quantities associated with partial discharge can be detected by different detection technologies, such as infrared temperature detection, chemical composition detection by micro-water or oil chromatography, electromagnetic wave intensity detection by UHF and high frequency, ultrasonic intensity detection by AE, etc. In practical applications, the above-mentioned detection technologies will be used and the detection data will be saved. However, due to the following problems, the diagnostic analysis of the insulation performance of power equipment still has inaccurate and low re...

AI Technical Summary

Problems solved by technology

In practical applications, the above-mentioned detection technologies will be used and the detection data will be saved. However, due to the following problems, the diagnostic analysis of the insulation performance of power equipment still has inaccurate and low reliability deficiencies:

[0003] 1. The above-mentioned detection technologies are all isolated when used and analyzed. The main reason is that the current detection equipment is designed according to different physical quantities, and there is no equipment for comprehensive detection of multiple physical quantities, which leads to different detection data of different physical quantities. no interaction with other data

Therefore, in the current analysis and processing of partial discharge conclusions, the analysis results will be one-sided, and the reliability of the analysis results given is not high.

For example, for electromagnetic detection technology, the detected data is easily affected by external electromagnetic interference, which leads to wrong analysis and judgment of partial discharge faults in power equipment in many cases.

[0004] 2. In the field application, due to the high and low business experience of the on-site inspection personnel, and the strong directionality, especially in the current situation that the workload of partial discharge has increased significantly but the number of experienced on-site personnel has not increased accordingly, The diagnostic analysis of on-site data is often overly dependent on the algorithm provided by the manufacturer of the detection device, but from the perspective of practical application, the accuracy of the algorithm of different manufacturers is different in different applications, which is related to the manufacturer's algorithm training method and data.

Images