Study On The Simulation And Identification Of Lightning Strike On Transmission Line

High-voltage transmission system is an important part of the power system.Lightning stroke is the main factor responsible for transmission line insulation faults. Toprevent lightning disturbances occur frequently, the measures taken on site must betargeted. That is to say, according to different lightning stroke, different measures shouldbe take. Through analyzes the type of lightning is a prerequisite for fault clearing andimprovement of lightning protection. In this paper, a electromagnetic transient model isset up to simulate the lightning over-voltage signal, the voltage features are extracted,and intelligent classification method is used to realize the recognition of lightning stroke.A220kV transmission system simulation model is established by usingPSCAD/EMTDC transient simulation software. The modeling analysis was carried outon the transmission line lightning considering about segmented wave impedance modelof tower to gain five kinds of simulation waves as follows: common short circuit troubletraveling wave, inductive lightning strike traveling wave, direct lightning strike travelingwave, back-strike traveling wave and shield-failure strike traveling wave. Voltagewaveform of simulation is analyzed preliminarily combined with the overvoltagegeneration mechanism, and the paper analyzes the differences waveform of differenttypes of lightning in the time domain.The paper analyzes overvoltage signals by time domain, frequency domain and jointtime-frequency analysis method to reflect the characteristics of signal. Three-phasesimilarity W, voltage integral eigenvalue ω and wave head volatility B are extractedbased on the time domain waveform method. Q1、Q2、Q3and Q4are extracted based on theHilbert spectrum to reflect the correlation characteristics between frequency domain andtime-frequency. Wave front characteristics Tgand the characteristics of gradient Tmareextracted based on teager energy operator.Considering the characteristics of different disturbance and the physical meaning ofeach feature, the paper selects characteristics for each type of disturbance identification,in order to improve the recognition accuracy. The difference between lightning strike andcommon short circuit trouble is mainly embodied in the wave front volatility B and theratio of high/low spectral energy Q2. Lightning wave has a bigger undulatory propertyand higher content of high frequency components. Induced lightning strikes at the sametime appears in the three-phase of the transmission line, so three-phase waveformsimilarity W and the Hilbert spectrum matrix similarity Q3are higher to distinguishinduced lightning strike from direct lightning strike. Differences between back flashwaveform and shielding failure waveform are reflected in wave front. Calculating Teagerenergy reflected wave front steepness, of which the maximum is Tmand Tgthat reflectswhether the maximum in the first extreme value point can effectively distinguish back flash and shielding failure from the others.This paper studies the basic concepts and algorithm of relevance vector machine andthe selection of kernel function. Transmission line lightning type recognition model isestablished by using multiple binary classification relevance vector machine. Puttingextracted the characteristic into classifier chooses corresponding eigenvector for eachRVM. The identification model is trained and tested by using data obtained from thesimulation to realize the intelligent identification of transmission line lightning typeeventually. A large number of simulation analysis verified the effectiveness of the builtrecognition model.