Research On Incipient Arc Fault Detection And Identification Technology Of Underground Cable

In the process of accelerating urbanization construction,the underground power cable,with its advantages of short distance transmission,more economical and less space occupation,occupies an increasing proportion in the transmission and distribution system with a wider range of voltage levels.However,underground cables will have frequent partial discharge,flashover and other phenomena at their insulation defects,which will gradually evolve into special early intermittent arc faults,which will accelerate the deterioration of the insulation until it is completely carbonized and cause permanent faults.Such early faults are often concealed and difficult to be caught by detection equipment.Research how to quickly detect and identify such early intermittent arc faults with transient and recurring characteristics can avoid permanent faults in time and reduce troubleshooting time The investment of manpower and material resources is conducive to the economic operation of the power grid.Firstly,the occurrence and development mechanism of early intermittent arc fault of underground cable is analyzed,and a dynamic arc model combined with Cassie and Mayr model is proposed to simulate the early fault arc.On this basis,a simplified model of 35kV distribution network with cable feeder is built by PSCAD/EMTDC software,and two early fault forms of underground cable semi cycle arc and multi cycle arc are analyzed The simulation and brief analysis of the inrush current,capacitor switching,load change,motor start-up and other four kinds of over-current disturbance events in the power system are given to facilitate the subsequent detection work.Secondly,for the early fault detection research,on the basis that different transient events will present different overcurrent characteristics,the bilateral cumulative sum algorithm(CUSUM)is used to calculate the positive and negative offset cumulative values of the current sequence signals of each phase at the cable head end to determine the abnormal phase,and then the improved adaptive neural network(Adaline)is used to track and calculate the fundamental and harmonic amplitude content of the abnormal phase signal The early fault type and overcurrent disturbance type of cable are detected and identified by the defined harmonic exponential change curve.Simulation and case data verify the effectiveness of the proposed detection method under different working conditions.Finally,in order to solve the problem that the existing identification algorithm model depends too much on the establishment of absolute complete fault type database in actual engineering application,a method of early fault identification of unknown type based on Mahalanobis distance and improved support vector data description(SVDD)is proposed.For the feature vector set of early fault samples constructed in time-frequency domain,SVDD model is used to divide the feature space of training samples into three recognition regions of distance level according to Otsu method.In the process of classification and recognition of the samples to be tested,the sample points which are easy to cause misclassification between the two threshold regions are identified again,and the samples are divided into different types of sample sets according to the SVDD model The Mahalanobis distance is used to determine whether the sample is a known early fault type.The above method is used to classify and identify the four types of early fault signals obtained from the experimental platform.The combination of pairwise known and pairwise unknown verifies the effectiveness of the proposed method.Compared with other classification methods,The proposed method has higher predictive value and provides a new idea for the identification of unknown categories of early fault samples.