Power transformer fault detection method based on 2dpca and svm

Abstract

The invention relates to a substation power transformer fault detection system and detection method based on 2DPCA and SVM, including a first audio collection system, a second audio collection system, an audio sample library, a first preprocessing system, a second preprocessing system, and an audio training system , data center, detection system, real-time collection of unknown audio data from transformers, converting audio samples and unknown audio data into data types with amplitude-frequency characteristics; converting the processed data into feature vector sequences representing audio; detection system calls The time series of feature vectors representing various audios in the data center are subjected to SVM calculations, and unknown audios are detected based on the calculation results to determine whether the power transformer is faulty and the type of fault occurred. Through the detection system of the invention, safe real-time and uninterrupted online substation power transformer working state monitoring can be achieved, faults can be found early and repaired, and major accidents can be prevented.

Description

technical field [0001] The invention relates to a substation power transformer fault detection system and detection method based on 2DPCA and SVM, and belongs to the technical field of audio discrimination. Background technique [0002] Substation power transformers are very important electrical equipment in the power system. When they fail, they can cause huge harm. Bring huge losses to production and life. [0003] The failure of a power transformer in a substation means that its working state is abnormal, some functions fail, or its performance index exceeds its rated range. Usually, in these cases, the power transformer has entered a fault state. The cause of failure mainly refers to the physical, chemical, biological or mechanical processes that lead to failure of the equipment under operating conditions, such as corrosion, creep, wear, heating, aging, etc. [0004] At present, in the maintenance process of power transformers in substations in my country, regular plan...

AI Technical Summary

Problems solved by technology

However, there are the following defects in this patent: 1. The transformer fault detection method based on dissolved gas in oil is adopted: the measurement of gas in oil is not very convenient, which will inevitably cause the complexity of the fault detection system; 2. Adopt dissolved gas in oil as Sample feature extraction, because various gases are mixed together, this will inevitably lead to the complexity of feature extraction and the appearance of errors, because the appearance of faulty gas is a slow process, only when the faulty gas reaches a certain level, the faulty gas characteristics will be dominant , at this time the gas features can be well extracted during extraction

3. Use the Euclidean distance to compare the characteristics of the data to be tested with the sample characteristics. Because the Euclidean distance is simple in structure and the probability of error is very high, it is the most basic measurement method in classification and recognition. Therefore, if the Euclidean distance is used as the first detection method step, if an error occurs, it will inevitably lead to the occurrence of subsequent detection errors

Images