考虑高斯有色噪声的FOMC-HTLS-Adaline算法在低频振荡模式辨识中的研究

针对广域测量系统的实测信号受高斯色噪声的影响,提出一种利用FOMC-HTLSAdaline进行低频振荡在线辨识的新方法。首先,为抑制高斯色噪声的影响,利用四阶混合累积量的盲高斯性,将四阶混合累积量(FOMC)序列代替实测序列进行低频振荡的辨识。然后,利用HTLS和自适应神经网络算法(Adaline ANN)相结合,估计出低频振荡的频率、衰减因子、幅值和相位。Adaline神经网络的引入解决了四阶混合累积处理后,模式幅值和相位不易确定的难点,同时减少矩阵处理引入的误差累积,提高检测精度。四机两区域系统仿真算例和实测相量测量单元(PMU)算例共同表明,FOMC-HTLS-Adaline算法可以在高斯色噪声环境下,精确地在线辨识系统振荡模式。

Research on Low Frequency Oscillation Mode Identification Based on FOMC-HTLS-Adaline Algorithm Considering Colored Gaussian Noises

The colored Gaussian noises will affect the measured signals in wide area monitoring systems. Thus, a new identification method for low frequency oscillation modes based on FOMC-HTLS-Adaline was proposed. Firstly, with the advantages of blindness to Gaussian noise, four order mixed cumulants (FOMC) sequence replaced measure signals to identify low frequency oscillation modes. Secondly, Hankel total least squares (HTLS) and Adaline artificial neural network (ANN) estimated the frequency, attenuation factor, amplitude and phase of low frequency oscillation. The introduction of Adaline ANN solves the problem that amplitude and phase of modes are difficult to estimate after FOMC process, and reduces error accumulation of matrix calculation and improves the identification accuracy. The 4-machine two-area power system and measured phasor measurements units (PMU) both indicate that FOMC-HTLS-Adaline method accurately identifies low frequency oscillation modes under the circumstances with colored Gaussian noises.