Research Of Fault Diagnosis And Location Method Of Power Transformer Based On Deep Learning

With the rapid development of smart power grids,higher requirements are put forward for the safe and stable operation of power equipment.How to diagnose and locate power transformer faults on line quickly and accurately is an urgent problem in power field.When power transformer failure occurs,it can often be considered in two parts,which are electrical accident and mechanical breakdown.We can diagnose the fault by judging the type of electrical fault and locate the trouble unit and orientation by analyze mechanical vibration.For the frequently requirement of high timeliness of online diagnosis and positioning,we choose deep learning to address this issue.The strong learning ability of convolution neural network can effectively improve the accuracy of judgment,and the way of offline training and online diagnosis also greatly improve the efficiency of online classification and detection.In this thesis,the convolution neural network in deep learning and the improved convolution neural network algorithm are applied to the study of transformer faults.The main work is as follows:(1)Power transformer failure can often be considered from two aspects of electrical fault and mechanical fault,from the two aspects,this thesis respectively analyzes the principle and different characteristics of different failures.And a fault database generation method is proposed.(2)In order to obtain transformer failure diagnosis and location data online,the characteristics of various sensors are analyzed and compared in this thesis.What’s more,online monitoring method of dissolved gas in transformer oil and transformer vibration signal is proposed,which saves manpower and material resources and shortens the time of diagnosis and positioning.(3)In terms of fault diagnosis and positioning,a transformer fault diagnosis method based on convolutional neural network is proposed.This method can directly extract and select features from raw data,which is easy to implement.Then,an improved convolutional neural network is designed in this paper,which can detect and locate the cause of the fault,thus realizing the online positioning of the transformer fault.This method is proposed to solve the problem of multi-noise interference,under the operating conditions of transformers.In the transformer fault location,multi-sensor variable input condition is still suitable.(4)The validity and timeliness of the proposed fault diagnosis and location method are proved by simulation.