Improved neural network based on rough set and application in fault line detection for distribution network
Qingle Pang, Xinyun Liu, Min Zhang · 2010 3rd International Congress on Image and Signal Processing · 2010
To overcome the shortcomings of the longtime training and the complicated structure in the neural network based fault line detection method for distribution network, the fault line detection method based on neural network and rough set is presented. All kinds of steady state and transient fault features are extracted from zero sequence current signals through the wavelet transform and Fourier transform. These fault features are regarded as the condition attributes of an information system and fault states as the decision attribute of the information system, then the information system is constructed. By use of attribute reduction and value reduction based on rough set theory, the information system is reduced. The reduced fault features are regarded as inputs a neural network and the reduced samples as the training samples of the neural network. Then the training samples are normalized using rough set theory and the normalized samples are used to train neural network. The trained neural network model can realize fault line detection. The simulation and field verification results show that the method reaches higher training speed and lower error rate.