A Study Of Electric Power Transient Signal Detection And Calssification Based On Wavelet Entropy Measure

This paper is a part of the project supported by National Science Foundation of China.The detection and classification of transient signal is applied widely in many fields of power system, such as transient protection, power quality analysis, fault location and distance computation, apparatus status supervision and transient stability analysis. In order to deal with the fault and prevent cataclysm in power system, it's necessary for analyzing electric power transient signal, detecting kinds of faults instantly, and classifying transient signals quickly and accurately. So it's important to build a suit of efficient methodology used in detection and classification.In this paper, the present situation of detecting and classifying electric power transient signal, processing transient signal by wavelet and wavelet entropy's application are analyzed, and the existent problems are pointed out.Based on reach of wavelet transform, the definitions of wavelet entropy measure and their meanings, the potential that wavelet entropy is applied in detecting and classifying electric power transient signal are analyzed, which show that there is a wide prospect for wavelet entropy used in power system.A 500kV transmission line model is built, and the six kinds of transient signals including breaker switching, capacitor switching, single phase to ground, primary arc, lightning interference and faulty lightning strike are simulated. The contrast between wavelet transform maximal modulus and wavelet entropy shows that the latter is a better approach to detect transient signal accurately and timely.Aiming at the problem of transient signal's classification, the methods of classifying six familiar transient signals above, selecting short circuit fault phase and identifying power quality disturb transient are studied. In research,wavelet entropy weight is put forward. The simulation results show that the classification method based on wavelet energy spectrum can discriminate the disturbance transient from faulty transient, the feature extraction based on wavelet time-frequency entropy and wavelet entropy weight combined with BP neural network can differentiate the six transient signals delicately, and the approach to select fault phase is quick, accurate and not affected by resistance, instant and location of the fault.A software prototype of electric power transient signal wavelet analysis instrument based on Lab VIEW is designed in the last part, and it possesses the function including waveform show, wavelet transform, post-analysis and detection and classification. The test shows that the software can analyze and process data accurately, and run normally.