CAJ | 학술논문

抑制现场噪声干扰、有效提取信号特征是局部放电信号检测和分析的关键。给出了利用Fisher可分离度选择具有最强类可分离度的双谱作为信号的特征参数，并利用特征参数训练径向基神经网络来判断信号的类型的算法。通过混有高斯白噪声的电磁波仿真软件得到的模拟不同局部放电源辐射的电磁波信号，利用该算法进行信号分离，验证了该算法的有效性。最后在变电站现场未知局部放电源的情况下，对采集到的局部放电辐射电磁波信号利用该算法进行处理得到信号类型数，并训练用于信号分离的径向基神经网络。基于现场实测信号分离结果，并结合基于时延序列的局部放电源定位结果验证了该算法在变电站现场干扰情况下分离多局部放电源的有效性。
억제현장조성간우、유효제취신호특정시국부방전신호검측화분석적관건。급출료이용Fisher가분리도선택구유최강류가분리도적쌍보작위신호적특정삼수，병이용특정삼수훈련경향기신경망락래판단신호적류형적산법。통과혼유고사백조성적전자파방진연건득도적모의불동국부방전원복사적전자파신호，이용해산법진행신호분리，험증료해산법적유효성。최후재변전참현장미지국부방전원적정황하，대채집도적국부방전복사전자파신호이용해산법진행처리득도신호류형수，병훈련용우신호분리적경향기신경망락。기우현장실측신호분리결과，병결합기우시연서렬적국부방전원정위결과험증료해산법재변전참현장간우정황하분리다국부방전원적유효성。
It is the key measures for the detection and analysis of partial discharge (PD) signals to suppress on-site noise interference and extract signal features effectively. An algorithm, which utilizes Fisher separable degree to choose the bi-spectrum with the strongest class separable degree as the characteristic parameters of PD signal and trains the radial basis function neural network (RBFNN) by characteristic parameters to judge the type of PD signal, is proposed. Using electromagnetic wave simulation software mixed with Gaussian noises, electromagnetic wave signals simulating the radiation of different PD sources are obtained and the signal separation is performed by the proposed algorithm to validate the effectiveness of the proposed algorithm. Finally, under the condition of on-site unknown PD sources in substation, the acquired electromagnetic wave signals radiated by PD are processed by the proposed algorithm to achieve the number of signal types, and the RBFNN for signal separation is trained. Based on the separation result of on-site measured signals and combining with the time delay sequence based location result of PD sources, the effectiveness of separating multi PD sources by the proposed algorithm under the condition of on-site interference in substation is validated.