Study Of Fault Location In Coal Mine Power Cable Based On Multiwavelet Neural Networks

The paper systematically analyzes the cause of coal mine cable fault and the method of domestic and international cable fault location, through the analysis and comparison of various methods and the analysis of the mine cable fault traveling wave, having a further study about the accuracy of the mine cable fault location in a series of in-depth analysis and research, and finally determine the fault location project of coal mine power cable based on the multiwavelet neural network.Firstly, the paper discusses the transmission characteristic of fault transient travelling wave, aiming at the problem of the wavelet analysis can’t accurately catch travelling wave front and the different location of the various scale after the wavelet transform, put forward the method of analyzing and detecting the fault transient traveling wave by using multiwavelet.What’s more,the paper describes the necessity of multiwavelet pretreatment and signal singularity detection principle of multiwavelet transform.Secondly, the models for fault location of the mine cable are simulated under the simulation environment of Matlab,and make use of signal singularity detection and modulus maxima of multiwavelet transform.The simulation results show that multiwavelet has a strong ability to signal analysis, the compact support can prevent energy leakage, calculates very fast of multiwavelets transform because of the orthogonality, various scales of singular point position deviation will not occur because of the symmetry, and the fault location can’t be affected by the position of fault, the fault resistance and fault initial phase angle and has higher positioning accuracy.Finally, aiming at the problem of the mine cable fault location accuracy is not high base on the multiwavelet in the nonideal state, the method of using multiwavelet neural network for the fault location is introduced.That is, multiwavelet transform modulus maxima matrix as the input of the BP neural network training samples for training, then using the BP neural network of nonlinear mapping ability and generalization ability to realize fault distance.Through theoretical analysis and simulation study concluded that the method has good self learning ability and nonlinear mapping ability, and learn with high efficiency, fast convergence speed, can realize the fault location effectively.