基于ICA的多通道低频振荡模式识别方法

Multi-Channel Low-Frequency Oscillation Pattern Recognition Method Based on ICA

针对电力系统中多通道低频振荡模式辨识过程中计算复杂、辨识精度不高等问题,提出将独立分量算法(Independent Component Algorithm,ICA)引入低频振荡多通道信号分离中,结合Prony法实现对低频振荡的模式辨识。首先将待处理的多通道信号利用ICA算法进行分离(噪声可视作某一通道信号进行分离)预处理,进而利用Prony法对分离后的信号进行模式辨识。结果表明:利用ICA+Prony法进行多通道信号低频振荡模式辨识时,具备较好的抗噪性,辨识结果更加接近于理论值,相对误差均小于4%,且能够解决直接Prony方法在多通道信号辨识时的模式遗漏问题。在多通道背景下,该方法可以较大程度地保留原始信号特征,克服Prony方法的缺陷,提高多通道低频振荡模式辨识精度与准确度,更加能够满足实际电网的应用需求。

In view of the problems of complex calculation and low recognition accuracy in the identification process of multi-channel low-frequency oscillation modes in power systems,it is proposed to introduce the independent component algorithm(ICA)into low-frequency oscillation multi-channel signal separation mode identification of low frequency oscillations.First,the multi-channel signals to be processed are separated by ICA algorithm(noise can be regarded as a channel signal for separation)preprocessing,and then the Prony method is used to perform pattern recognition on the separated signals.The results show that the ICA+Prony method for multi-channel signal identification of low-frequency oscillation mode has better noise immunity,the identification results are closer to the theoretical value,the relative error is less than 4%,and can solve the problem of missing patterns during recognition by direct Prony method in multi-channel signal.In the multi-channel background,this method can retain the original signal characteristics to a greater extent,overcome the defects of the Prony method,improve the recognition accuracy and accuracy of the multi-channel low-frequency oscillation mode,and can better meet the needs of actual power grid applications.