CAJ | 학술논문

　　针对变压器内部常见绝缘缺陷，设计了尖刺、悬浮微粒、沿面、固体气隙等4种缺陷模型，利用超高频法获取了放电谱图。通过统计理论计算了PD超高频信号三维谱图的统计特征，以BP神经网络作为分类器，对所获得的PD信号进行了识别，取得了良好的效果，为开发变压器局部放电在线或离线分析系统奠定计算基础。
　　침대변압기내부상견절연결함，설계료첨자、현부미립、연면、고체기극등4충결함모형，이용초고빈법획취료방전보도。통과통계이론계산료PD초고빈신호삼유보도적통계특정，이BP신경망락작위분류기，대소획득적PD신호진행료식별，취득료량호적효과，위개발변압기국부방전재선혹리선분석계통전정계산기출。
Partial discharge is an important feature of the characterization of transformer insulating state, but it also causes further deterioration of insulation. According to typical internal insulation defects, four kinds of defect modeles are designed, including Corona, suspended particulates, surface discharge and air gap in solid insulating material, the discharge spectra is obtained by UHF method. Features of discharge spectra are calculated by statistical theory, and with a BP neural network classifier to recognize PD signal, good result is achieved. .