Research On Methods Of High Impedance Fault Detection In Distribution Networks

Compared with the transmission networks,distribution networks have wide cover range and high probability of fault.Currently,the scale of the distribution networks continues to expand,the importance of safety,reliability and economic operation of distribution networks is also increased.The requirements for safety and reliability of distribution networks feeder become higher and higher.In comparison with the single phase grounding fault,high impedance fault(HIF)generally is difficult detect by conventional protection devices,because they have high impedance at the fault point and can not cause an excessive change of current and voltage.Once the small fault current exists for a long time and can not be found,it may result in a fire hazard and damage of the electrical equipment.Moreover,when this type of fault happens,energized high voltage conductor may fall within reach of personnel.Therefore,it is of significance to research on methods of HIF detection in distribution networks.First of all,the research directions in field of HIF detection home and abroad are summarized in this paper.After comparing the advantages and disadvantages of multiple signal extraction methods and intelligent classifiers,it is shown that local characteristic scale decomposition(LCD)is effective in processing nonlinear and non-stationary signals.And the possibility of applying deep learning algorithm to detect HIF is proposed.The signal processing process of LCD bandpass filter is explained in detail.An example of decomposition shows that the LCD band-pass filtering algorithm has high reliability in signal decomposition and can reflect the time-frequency information of signal.Considering some transient disturbances phenomena in the distribution networks are similar with HIF,HIF,single-phase grounding fault and transient disturbances phenomena(capacitor switching,load switching and no-load line switching)are included in the data samples.Two methods of HIF detection in distribution networks are proposed in this paper.The one is a method based on LCD band-pass filter and multi-level SVM.First,the half waveform before the fault and the one and half waveform after the fault are extracted from the three phase and zero sequence voltage of bus bar.Then,the data are pretreated by LCD band-pass filter algorithm.The standard deviation of each frequency band are extracted from the reconstructed time-frequency matrix to be characteristics for training and testing SVM.The other one is a method based on block time-frequency spectrum and CNN.Reconstructed time-frequency matrixes are processed to get block time-frequency spectrums by calculating energy block.The normalized block time-frequency spectrums are input images for training and testing 7-layer CNN.Two models of 10kV distribution network are constructed to acquire samples for training and testing.The one is constructed by PSCAD/EMTDC.And another one is a model of physics experiment based on distribution network dynamic simulation system.The test results show that the two methods have high accuracy in the HIF detection in distribution networks.They have good adaptability in noise interference,sampling asynchrony,DG supply access and so on.The second method is no need to construct and extract characteristics artificially.And it is more excellent than another method in terms of robustness and adaptability.