Power System Short-circuit Fault Based On Wavelet Analysis And Neural Network Theory

The destroy is mostly due to short trouble when power system is running. Once the short trouble happened, the system varies severely from one state to another state following with complicated transient phenomenas. The tested signals contain lots of transient components. How to analyse this kind of signals and draw out their features to develop new protection device is an important research domain in the protection technology of power system all the time.The protection of power system is quick detection and localization of fault in order to act rightly and clear out the fault. At present, the analytic tools used for the power signals in microcomputer protection include FFT, Kalman filter and finitely shock response filter, etc. They are effective to analyze stationary signals but localizable for unstationary signals analysis. Especially it is difficult to identify nonlinear fault, e.g, the nonlinear high impedance short trouble is a long-term unsolved problem of power system.This paper puts forth a new method that using wavelet transform draws out the signals' details and identify fault incorporation with ANN for power system short trouble fault diagnosis. Firstly, make wavelet transform to the tested signals and draws out theirs properties. Secondly, take the properties as multilayer forward ANN's inputs, adopt different ANN for different outputs demands to judge the faults' phase, property and position. It collects the advantages of wavelet theory and ANN to realize the intelligent protection and promote the selectivity, sensitivity and reliability of protection, guarantee electric fence stability, improve the quality of power supply.The simulation results prove that the method is workable and effective and the fault identification is accurate. Especially it solves the identification problem of high impedance fault well.