Research On High Impedance Fault Detection Method Of Distribution Network Based On Statistical Distribution Characteristics Of Data

The high impedance fault usually happens in the form of a line coming into contact with a high impedance surface or tree in the distribution network,which will bring potential severe risks to system security.In recent years,fires and electrical shocks caused by high impedance faults have been frequently reported,causing massive casualties and property loss.Due to weak fault features,high impedance faults are challenging to detect by traditional overcurrent protection devices.Besides,under the influence of many factors such as grounding medium and grounding environment,the characteristics of high impedance faults appear strong nonlinearity and uncertainty,and the access of nonlinear elements such as distributed generations further aggravates the difficulty of high impedance faults detection and identification.The obvious differences in the probability distribution of equipment measurement data before the occurrence and during the duration of high impedance faults.Thereby,this paper proposes a distribution network high impedance fault detection method.The main work of this paper is as follows:(1)For the problem of weak high impedance fault currents and inconspicuous fault characteristics,this paper proposes a high impedance fault detection method based on distribution network data divergence.Using Wasserstein divergence to measure the difference of probability distribution between the historical first principal component and the online first principal component obtained by the principal component analysis method and combined with the proposed determination criteria,high impedance fault detection is achieved.Simulation and experimental results validate the effectiveness of the proposed method.(2)For the problem of high impedance fault detection under Gaussian sparse data,this paper transforms the problem of finding the difference of probability distribution into the problem of finding the difference of standard deviation between historical data and online data by improving the Wasserstein divergence method,and the proposed improved method is combined with sensitive sparse principal component analysis to achieve the detection of high impedance faults in distribution networks.The simulation and experimental results show that the proposed method is robust to factors such as fault inception angle,transition impedance,noise,and the layout of the measurement device and can effectively distinguish high impedance faults from switching events.(3)For the problem of high impedance fault detection under non-Gaussian data,this paper constructs a dynamic kernel independent component analysis by introducing hysteresis coefficients,and achieves high impedance fault detection according to the proposed new Wasserstein divergence statistic.The simulation results show that the proposed method can accurately identify high impedance faults in different fault locations,types,transition impedances,and inception angles.