Study On Fault Cause Comprehensive Identification For Transmission Lines

Transmission lines distribute and span widely, run under complex meteorological and geographical environment. Due to exposure in the wild for long time, line faults happen frequently because of external force damage or bad weather such as lightning, pollution, ice and snow. Trip of transmission lines will damage the electrical equipment, even cause serious outage, endanger the life and property safety. When fault happens, the process from searching for fault to power restoration will take a long time, the fault cause is often difficult to determine especially with unobvious fault trace. When outage occurs, identifying the cause of fault timely and exactly has great significance to expedite the procedure of eliminating fault, reduce the outage loss and improve the level of system operations. At present, the major research methods are based on the difference analysis of fault waveform characteristics, or statistical analysis of historical fault rule to implement fault cause identification. Since fault causes are complicated and various, the existing methods have single identification object, inadequate identification basis, low identification accuracy and other problems.This paper studies on six single-phase faults of transmission lines caused by lightning, insulator pollution, wildfires, bird damage, objects and crane contact. The in-depth analysis of fault mechanism and characteristics reveals that, various kinds of faults have differing meteorological conditions and numerical characteristics because of the differences of flashover mechanism. Therefore, we can extract effective fault features from the fault recording data and meteorological information, then regard the correspondence law between fault features and causes as the identification basis. However, the corresponding relationship is highly complex and nonlinear, it can’t be calculated directly, BP neural network is fit for dealing with this problem for its good ability of nonlinear mapping. The input of BP network is fault comprehensive features vector, which consist of fault weather, time, season, transition resistance properties, the harmonics and DC component of zero sequence current and reclosing. The output of BP network is fault cause type vector. By using actual fault samples to train BP network,we can establish the nonlinear mapping between fault features and causes. When the BP neural network training is completed, it can be used for fault cause identification.In this paper, the division standards of fault numerical features are set by the calculation and analysis of actual fault recording datas. In order to compare the effects of different algorithms for fault cause identification, establish several BP neural network models that considering fault comprehensive features, considering only meteorological features or considering only fault numerical features. Then train and test these BP models with a large number of actual fault samples. Test results for different objects show that, the identification accuracy of the BP algorithm based on fault comprehensive features is higher than other two methods, and this algorithm is more effective and stable with the increase of fault cause types.In conclusion, according to the comprehensive analysis of transmission line fault characteristics, the paper proposes a fault cause identification method based on BP neural network combing fault meteorological features with numerical features, that can effectively identify various fault cause types. On the basis of calculation and analysis of actual fault datas, the proposed algorithm has abundant evidence and high accuracy for fault cause identification, as well as the needed fault information can be obtained timely and accurately, that can meet practical engineering requirements.