Research On Fault Diagnosis And Section Location For Active Distribution Network Based On ?PMU

In order to improve the reliability of power supply,medium and low voltage distribution network always uses small current neural grounding in China.When the single phase to ground fault occurs,the fault current only has few amperes and the fault feature is weak and hard to detect.In active distribution system,the distributed generation will also contribute the fault current.More complex fault feature of active distribution network brings greater challenges.In the latest regulations,State Grid Corporation of China has changed 'work for 2 hours with fault' into 'determine section&trip' when permanent single-phase to ground fault occurs in small current grounding distribution network.Traditional measurements have problems of low precision and low sampling rate.The development of distribution system has been limited greatly.Micro Phasor Measurement Unit(?PMU)provides new opportunities for the issues.It is a high-precision power disturbance recorder adapted for making voltage phase angle or synchrophasor measurements,capable of storing,analyzing and communicating data live.Under this background,this paper presents a new method in fault detection and section location based on synchronous measurement information of multi ?PMUs.The details are as follows:(1)Analyze the current distribution,feature of zero sequence current,harmonic characteristics and the current provided by distributed generation when the single phase to ground fault in small current grounding system occurs.The fault stable characteristics show the disadvantages of fault detection by single feature.Furthermore,introduce the structure and data precision of ?PMU.Considering the communication pressure,synchronous measurement data and data processing ability,local-master station coordination architecture for fault detection and section location in distribution network is presented.(2)Under the coordination architecture,a new fault detection method based on multi synchrophasor measurements is presented.At the local side,extract auxiliary features(information entropy and wavelet absolute maximum value)in the data pane.All the local information will upload to the master station in lower frequencies.At the master station,the dimensions of data section for each moment will be reduced by principal component analysis.Then train the support vector machine with low-dimensional feature vectors of multi-scenarios and detect the status of distribution network.Finally,recall the high frequency data when detecting the fault.Verify and revise the result through continuous section analysis of rationality and auxiliary features.(3)To realize fast fault section location after detection,a new method based on the difference of two-terminal data is proposed.First,calculate the cluster center of different fault section by k-means clustering method.Next,determine the proportion of the fault section location using Bayesian Discriminant.The proposed method was tested on a 34 nodes distribution network in PSCAD/EMTDC and the results in different scenes justified the detection accuracy and effectiveness of the proposed methods.