Research Of The Fault Detection In Transmission Line Based On Neural Network

Aerial cable is one important part of electricity system, detecting the fault exactly on time has very important significance for the electricity supply enterprise's economic and society benefits.The existing cable fault-detecting system is mainly used in detecting high voltage or super high voltage point to point capble fault, but it is not fit for cable whose voltage is below 35KV with complex structure.On the basis of analysis of the existing cable fault-detecting systems, a calbe fault detecting system based on neural net work is presented; the 10kv aerial cable is investigated under this system. The voltage, current, resistance, capacity and inductance data, catching by the device located on the aerial cable, is sent to the control center by carrier wave communication,where data interpolation is used to restore the parameter's real distribution;The basic distributed functions of every parameter, distributed function caused by work load, distributed funciont caused by environment and distributed function caused by impulse fault are got by curette transformation, and the noise is removed by strangeness-detecting.After segment sampling to each parameter, the sample data are given to the 3-layer neural network, the fault type and position is given by the statistical analysis. Before detecting the fault, the sample set is used to train the neural net work by LMBP algorithm.By EMTP simulating software, the simulating experiment is carried out. The results show that the accuracy rating increases greatly, the average relative error is under 2.1%, the maximum absolute error is under 2.91%, and the real working system's requirement is met greatly.