CAJ | 학술논문

为研究干式变压器局部放电超声信号的模式识别，设计制作了针-板、沿面和气隙3种局部放电模型，用超声检测系统得到3种局部放电的超声信号波形数据。运用小波包理论对采集到的局部放电超声信号进行分解，用Shannon熵代价函数来确定最优小波包基，得到最优小波树，提取在最优基下包含绝大部分信息能量的小波包分解系数的统计量作为模式识别的输入特征量，并应用CPN网络分类器进行放电类型的识别，取得了很好的效果。
위연구간식변압기국부방전초성신호적모식식별，설계제작료침-판、연면화기극3충국부방전모형，용초성검측계통득도3충국부방전적초성신호파형수거。운용소파포이론대채집도적국부방전초성신호진행분해，용Shannon적대개함수래학정최우소파포기，득도최우소파수，제취재최우기하포함절대부분신식능량적소파포분해계수적통계량작위모식식별적수입특정량，병응용CPN망락분류기진행방전류형적식별，취득료흔호적효과。
To study the pattern recognition of partial discharge ultrasonic signal for dry-type transformer, we designed three kinds of partial discharge models, which are needle-plate, surface, and air gap, and we obtained three kinds of partial discharge ultrasonic signal waveforms by ultrasonic test system. The par-tial discharge ultrasonic signal was decomposed by wavelet packet theory, and the optimal wavelet packet basis and the optimal wavelet tree were determined by Shannon entropy cost function. The statistics of wavelet packet decomposition coefficients, which contains majority of information energy under the opti-mal wavelet packet basis, was extracted as an input characteristic parameter for pattern recognition, and the discharge type was identified by CPN network classifier. The recognition effect is excellent.