Time series data cleaning method based on correlation analysis and principal component analysis

Abstract

The invention discloses a time series data cleaning method based on correlation analysis and principal component analysis, comprising the following steps: finding out hidden correlation between transformer faults and other power data by using a Pearson's correlation coefficient (PCC); reducing the dimension and noise of all relevant time series through principal component analysis (PCA); and inputting part of cleaned data as a training set into a BP neural network (BPNN) for training and learning, and taking the remaining data as a test set to verify a model. Compared with the traditional technology, the accuracy of transformer fault diagnosis is significantly improved, the accuracy of classification is improved, and the operation time is shorter for high-dimensional data.

Description

technical field [0001] The invention relates to a time series data cleaning method based on correlation analysis and principal component analysis. Background technique [0002] As the lifeblood of the national economy and security, the safe operation of the power grid has always been the core and essential requirement of the power grid. Oil-immersed transformers are crucial equipment in the power system. Their working status directly affects the safety of the entire power system, and any failure may cause serious economic losses. Therefore, it is necessary to predict potential failures in order to take appropriate solutions to repair equipment in time. Dissolved gas analysis (DGA) is an effective method for fault diagnosis of power transformers. According to the experience of the power industry, the working state of the transformer is mainly related to the content of dissolved gases in several oils, mainly including: hydrogen (H 2 ), methane (CH 4 ), acetylene (C 2 h 2...

AI Technical Summary

Problems solved by technology

The three-ratio method is more suitable for some common serious faults, but it is not very sensitive to some low-frequency faults, because some codes do not correspond to the correct fault types

They can all be used to improve accuracy when dealing with fault data similar to electrical equipment, but because they only focus on the training process of machine learning and lack of preprocessing of collected data, there are obvious limitations in efficiency: that is, when the data size is getting larger and larger When , there is a lot of redundancy and noise in the training data, the training efficiency is low and the final accuracy is affected to a certain extent

Images