A new method for locating faults on transmission lines based on rough set and FNN

摘要

Traditional ANN method for locating faults on transmission lines uses the voltage and current data after the faults, which causes the training samples great and brings the ANN to an ill condition. So, the convergence speed is slow or ANN cannot converge. In this paper, a new FNN faults location method based on rough set is presented. The method uses rough set before FNN to analyze the indiscernibility relation of data, to remove the redundant data and extract the key attribute of electrical component after faults. In order to use the method effectively, a FNN fault location model with layered and distributed structure system, is proposed, which first filters the input data and gets the three-phase power frequency voltage and current, then input the voltage and current to a sub-FNN to classify the fault type, finally according to the fault type uses different fault location sub-FNN (including three-phase fault location sub-FNN, two-phase fault location sub-FNN, two-phase grounded fault location sub-FNN and single-phase grounded fault location sub-FNN) to locate faults and output the fault position. The simulations show that FNN can converge easily and training time is reduced largely when using proposed method.