High Impedance Fault Detection Method Based On Limited Synchronous Measurement Data For Distribution Networks

High-impedance faults（HIFs）often occur in distribution network systems,and the fault current is generally less than 10%of the load current.Although the detective level of high-impedance fault can reach 1kΩ-2kΩin practical applications,but the grounding impedance value is often greater than 2kΩ-3kΩ,or even more than 10kΩ.Thereby,40%-50%of HIF still cannot be detected by the protective device.Although the low current generated by the high-impedance fault will not immediately affect the normal operation of the distribution network,the HIF is often discovered after tens of minutes or even more than a few hours,which will destroy conductor insulation and cause phase-phase faults,and may also cause casualties and severe economic losses.Therefore,it is essential to propose a reliable and accurate real-time high-impedance fault detection method.The high reliability,low latency,and high precision real-time measurement characteristics of the micro-phasor measurement unit（μPMU）provide the possibility for real-time detection of high-impedance faults.However,it is constrained by economic factors such as equipment costs,operation and maintenance costs,etc.Therefore,it is an urgent problem to solve how to use the synchronous data measured by the limitedμPMU to complete the detection and identification of high-impedance faults.This paper focus on this issue,and the main contents are as follows:Firstly,this paper introduces the high-impedance fault characteristics,theμPMU device and its data characteristics based on the application ofμPMU in the distribution network,the optimal configuration of the device,and the research status of high-impedance fault detection.It then gives the high impedance fault model and the synchronization data structure used in this study.Secondly,aiming at"high impedance"problem at the station end,the improved l_2,1principal component analysis method is used to identify the nodes sensitive to the system state change in the distribution network.So,the global optimal placement are determined.The fault monitoring domain,based on the optimal placement results,are constructed to realize the division of the fault monitoring section,which lays a foundation for the distributed fault detection and location of the distribution network.Then,the section location of the"high impedance"fault at the station end was completed through the mode time difference method.The simulation verification shows that the fault monitoring domains,constructed based on the measurement device optimization placement,can effectively solve the"high impedance"problem at the station end through the distributed fault location method.Finally,theμPMU optimal placement is realized based on the above optimal placement method.On this basis,a novel HIF detection method is proposed based on an improved capsule network.The method first introduces the inception structure and the Layer Normalization（LN）into the capsule network to change the original network structure.Besides,the proposed packet routing algorithm is applied to improve the generalization performance of the capsule network.Then,the improved capsule network is trained with limited synchronous measurement data.The single-phase grounding high-impedance faults are detected and identified using the mature improved capsule network.The simulation and experimental results show that the proposed method can accurately detect high-impedance faults with limited synchronous data.The proposed improved capsule neural network algorithm has robust to the transition impedance and fault location changes,switching actions,and noise.