Study On The Identification And Classification Of Lightning Over-voltage

It is suggested that the lightning stroke is the main factor responsible for transmission line faults from domestic and international operating experience. In order to maintain the safe operation of transmission lines, some lightning protection methods are introduced, such as the implement of ground wires and lightning rod, reducing the tower grounding resistance, increasing the line insulation and installing arrestors, etc. But even with these protective methods, the lightning outage rate remains high, due to the complexity of the process of lightning and the various reasons for the lightning accident. Now many systems for lightning location and over-voltage monitoring have been used to analyze lightning characteristics and development process. But the huge over-voltage data need special expert field analysis, which greatly limits the practical of the lightning online monitoring device in the failure analysis. Therefore extracting the lightning over-voltage characteristics and classification online, not only can find the point of failure quickly, reducing maintenance time and achieve rapid restoration of electricity, but also can take lightning protection pertinently and improve the reliability of power supply.In this paper, a 110kV electromagnetic transient model is set up to simulate the lightning over-voltage signal, used wavelet transform and mathematical morphology to extract the voltage features and classify the type of the over-voltage with support vector machine (SVM). The major works are:(1) A 110kV electromagnetic transient model is set up with the use of ATP/EMTP and five kinds of transient signals such as transmission line grounding fault, lightning strikes on conductor with no fault, back flash, shielding failure and induced lightning over-voltage are simulated. Combination of voltage generating mechanism and influence factors, analyze the characteristics of voltage waveform as a basis for future study.(2) Combining wavelet transforms with entropy theory put forward the best basis algorithm and find out the most appropriate wavelet for analyzing lightning voltage. According to these characteristics, by introducing the wavelet energy spectrum matrix similarity degree coefficient, over-voltages arose from induced lightning and direct lightning stroke are identified by the coefficient magnitude. After that, wavelet transforms are performed on the initial traveling waves. Then the back flashover and shielding failure over-voltages are distinguished from each other by the polarity of maxima modulus and the sudden change starting time of voltage. This method is simple and easy to understand and its reliability and validity for identifying the over-voltages under different fault conditions have been proved by a large number of the transient simulation results.(3) According to the mathematical morphological basic theory and characteristics of lightning over-voltage, firstly, determine the shape, length and amplitude of the structure element when it is used to analyze the lightning voltage. Secondly, construct a cascade of multiresolution morphology gradient MMG, compare the calculation accuracy and speed of MMG, MG and wavelet transform when they are applied in extracting the feature of lightning voltage front wave and structural F2 and T2 to identify lightning types. Finally extract the lightning pattern spectrum, compare its frequency characteristics with wavelet and identify the lightning types based on P6, P7, P8.(4) Study the theoretical basis of support vector machines and classification mechanism, use the wavelet transform and mathematical morphology feature extraction volume of the lightning to training SVM classifier and achieved good recognition, finally use the measured waveforms to verify the accuracy of the classification system.