Study On Fault Line Selection And Location Of Distribution Network

When distribution line occur single-phase grounded fault, its stedy current is weak, fault characteristic is not obvious and the fault arc is insable, resulting the difficulty in fault detection. Distribution network also has a "closed-loop structure, open-loop operation" feature; there are many branches, complex structure and topology changes with different system operation. These factors lead to that even accurately detect the fault and obtain the fault distance, it’s hardly to determine the exact location or the fault branch. In addition, high resistance faults caused by wire fallen to ground and tree s contacting the overhead feeders make the fault detection in the actual operation very difficult. Distribution feeder measurement points are only installed in the side of substation so that the protection and location mostly rely on single-end algorithm.With the decline in the cost of cable application and the transformation of urban and rural power grids, more and more cable lines are applied in distribution system. Thus, feeder capacitor charging current increases and arc extinction becomes more difficult. Travelling wave mentods are facing the effects of the large number and length variation of the feeders, also the load disturbance and such factors which make the first wave head is not easy to be calibrated reliably. There is also the problem of subsequent wave heads can’t be identified effectively caused by the increased camlexity in propagation process analisys. These factors determine the specificity, complexity and difficulty in fault detection and location of distribution system. How to find the exact fault positon quickly has long been a problem. As users’requirement of high power supply reliability and quality continue to increase, the importance of fault detection and rapid resoration becoming fast growing, a reliable and effective detection and location technology is needed in actual site.In this paper, a novel distribution system fault line selection method is proposed. It uses the wavelet analysis to extact wavelt coefficients in the highest frequency band of the zero-sequence currents of each feeder. And then accoeding to the opposite phase between fault feeder and sound feeders to build a new fault line selection method. When single-phase grounded fault occurs in straight line, avoid analysing the wave natures, using the powerful learning and generalization ability of ANN, inputing the polarity of first three wave heads and the time differences to achieve fault location. Because of the impact of node in multi-baranch distribution, identify travelling wave is more difficult, but the frequency spectra distribution can reflect the different fault position, whereby explore methods of location fault in radial distribution system.