Research On Methods Of Fault Line Selection And Location For Single-Phase Disconnection Fault In Distribution Network

With the high development of the national economy,the demand for electrical power supply is increasing day by day,and higher requirements are put forward for power supply reliability.The network structure of distribution system is becoming more and more complicated.With an increase in the proportion of overhead insulated conductors for distribution networks,disconnection fault occurs more and more frequently.There are many reasons for the disconnection fault.Once a fault occurs,it can directly affect the safe and stable operation of the grid even endanger the personal safe.To determine the fault line and fault location in a timely and accurate manner,this paper summarizes the existing fault diagnosis methods and their existing problems,and proposes methods of single-phase disconnection fault line selection and fault location.The following conclusions are demonstrated by analyzing single-phase disconnection faults in ungrounded distribution networks,and comprehensively considering the effects of fault type,grounding resistance,fracture position and load impedance.After a single-phase disconnection fault occurs,regardless of whether the fracture is grounded or not,the amplitude of the negative-sequence current in the faulty line is much greater than that in the healthy line.The amplitude of line voltages on power side remains unchanged.On load side,the line voltages related to fault phase decreases,while the other line voltage remains unchanged.The variation characteristics of these electrical quantities provide a theoretical basis for fault line selection and fault location.For overhead lines,a new fault line selection method for single-phase disconnection fault based on Hausdorff distance is proposed.Firstly,a fault feature extraction method based on diversity entropy(DE)is proposed.On the basis of DE,time-shift multiscale method is introduced to replace the traditional coarsening process,and the multiscale standard deviation is used as the weight of DE.Further,the refining method is used to avoid the fluctuation of DE values at larger scale factors.To improve the stability of entropy,a feature extraction method based on refined time-shit multiscale standard deviation diversity entropy(RTSMSDDE)is proposed.Then,according to the RTSMSDDE values of negative-sequence current signal for each feeder,the Hausdorff distance is calculated to measure the similarity of fault feature information.Finally,the fault line is selected by analyzing matrix elements.Through simulation and practicle application,it is verified that the RTSMSDDE algorithm can effectively extract the important constructing information of the original data,and the proposed fault line selection method can accurately identify fault line.Aiming at hybrid overhead and cable lines,a fault line selection method based on mixed kernel function support tensor machine(MKSTM)is proposed.Firstly,a new feature exreaction method of negative-sequence current signals based on refined time-shift multiscale diversity q-complexity entropy(RTSMDQC)is proposed by incorporating refined time-shift multiscale diversity entropy(RTSMDE)with complexity entropy theory,solving the problem of incomplete feature extraction using a single entropy method.Then,the MKSTM is established based on the tensor data composed of feature values of each feeder to achieve the selection of faulty line,and the improved salp swarm algorithm(ISSA)is applied for optimizing the parameters of MKSTM.The simulation and practicle application results show that the RTSMDQC has the ability of fault feature identification,and the MKSTM has higher classification precision.The method for locating single-phase disconnection faults in distribution networks is studied.A method based on adaptive Stacking(Ada Stacking)ensemble model is proposed to achieve fault location in distribution network.Combining the improved gamma transformation and gradual similarity ratio(GSR),a data augmentation and data screening method are proposed under inadequate samples.On the basis of Stacking,the Ada Stacking model is proposed to realize the adaptive ensemble and configuration optimization of the basic model.The experiment results show that the model can effectively learn the distribution characteristics of the original samples by using the data augmentation and data screening method,thus improving the precision of fault location in distribution network.