Research On Fault Identification Method For Distribution Network Based On Synchronous Waveform Measurement Data

The distribution network is an important hub that closely connects power systems and users.However,distribution lines often suffer from multiple factors that cause faults.With the integration of distributed energy,the anti-interference ability of distribution network gradually deteriorates.The transient ground faults are prone to deteriorate into serious interphase faults or permanent faults,leading to power outages.Therefore,accurate identification of transient grounding faults before they worsen is of great significance for ensuring the safe operation of distribution system.With the gradual development of synchronous waveform measurement devices,a large number of current and voltage waveforms with accurate synchronization have been collected,bringing a turning point for distribution network fault identification based on synchronous waveform measurement data.Aiming at the problems of complex fault mechanisms and difficult identification of fault causes in actual distribution network,this dissertation focuses on four aspects of fault detection,fault section location,fault cause feature extraction,and fault cause identification based on measured waveform data.The main research work and innovative achievements of this dissertation are as follows:(1)Aiming at the problem of difficulty in fault detection caused by weak fault characteristics,the distortion response characteristics of zero-sequence current waveform under different ground faults are revealed;A fault feature enhancement method based on mathematical morphology is proposed to amplify the fault response at the moment of occurrence;A fault moment detection method based on the Dirichlet process Gaussian mixture model is proposed.Through adaptively judging the enhanced fault features,the fast and accurate detection of grounding faults with different resistance values is achieved.(2)Aiming at the problem of the difficulty in locating sections of grounding faults with different resistance values in actual distribution network,the differences of waveform characteristics in time domain and frequency domain in different sections of the line is analyzed;A fault feature extraction method based on wavelet packet transform is proposed to construct fault differential features for different sections;A fault section location method based on gray correlation analysis is proposed,which calculates the correlation coefficients between different sections and achieves accurate location of grounding faults with different resistance values in the distribution network.(3)Aiming at the problem of difficult effective feature extraction of different fault cause waveforms in actual distribution network,a fault feature extraction method based on short-time Fourier transform is proposed.Starting from the mechanism level of various fault causes,the differences between time domain waveforms and frequency distributions of different fault causes are mined;A feature extraction method based on short-time Fourier transform is proposed to construct two-dimensional time-frequency images with clear physical significance,laying an effective feature foundation for fault cause identification.(4)Aiming at the problem of low accuracy of fault cause identification in actual distribution network,a fault cause identification method based on dual-channel convolutional neural network is proposed.The time-varying characteristics of various electrical quantities during the fault process are analyzed;A dual-channel convolutional neural network is proposed to extract high level features from various fault information and achieve cross channel multimodal information fusion;A fully connected layer based on Maxout unit is proposed to enhance the ability to extract hidden features of complex information and the classification accuracy of fault causes,achieving accurate identification of field fault causes.