【摘要】 由于间谐波具有在频域广泛分布、幅值相对于基波和谐波较小、周期难以确定等特点,因此间谐波的检测难度很高,要求能够尽量限制非同步采样对检测精度的影响。现代谱估计算法由于对同步采样没有要求,频率分辨率高等特点被广泛地用于电力系统谐波与间谐波检测中,由于其只能准确的估计信号的频率,不能很好的估计信号的幅值和相位,在实际应用中受到很大限制,而支持向量机如果要准确的估计信号的频率、幅值和相位,计算量很大。基于以上特点,将支持向量机(SVM)与总体最小二乘旋转不变(TLS-ESPRIT)算法结合起来,先用TLS-ESPRIT算法准确估计信号频率分量,再用支持向量机方法估计信号的幅值和相位,通过三个算例分别讨论了该方法在相邻信号、多信号等情况下的检测情况,证明该方法切实可行,能准确检测出间谐波的各种参数,并且比单独使用支持向量机计算量小。更多还原

【Abstract】 Because the interharmonics had such characteristics as wide distribution in the frequency domain,the amplitude of the interharmonics was less than that of the fundamental wave or harmonics and difficult to make sure the cycle of the signals including the interharmonics,the detection difficulty in interharmonics was higher than that in harmonics by far,and it is required to try to restrict the infection of the detection precision caused by asynchronous sampling.Modern spectral estimation method has been widely used in the electrical power system to estimate harmonics and interharmonics because it was not requested to the synchronized sampling,and had the high frequency resolution.Howeverr,it could not accurately estimate signal’s amplitude values and the phases but only accurately estimate signal’s frequencies.Therefore,the use of the modern spectral estimation method was limited.We proposed the union of the TLS-ESPRIT algorithm and support vector machine method.First,the TLS-ESPRIT algorithm was used to accuratly estimate signal’s frequency components,then,the SVM was used to estimate signal’s the amplitude values and the phases.Through three examples,the detection results of this method was discussed under different conditions.Results indicate this method is practical feasible,can estimate all parameters of the interharmonics and has smaller computation quantity than using SVM only.更多还原