基于神经网络自适应滤波的低频振荡Prony分析

针对传统Prony算法在分析低频振荡时对噪声非常敏感的缺点，提出一种基于神经网络自适应滤波和改进Prony算法相结合的电力系统低频振荡分析方法。该方法以广域测量信号作为输入，采用神经网络自适应滤波对低频振荡信号进行滤波预处理，调节性能指标阀值确定滤波效果，并通过改进Prony算法对滤波后的信号进行识别。仿真结果表明，该方法能有效滤除噪声，能较为准确地辨识低频振荡的主导模式。

Prony Analysis of Low Frequency Oscillations Based on Neural Network Adaptive Filtering

Because traditional Prony method have difficult to analyze low frequency oscillation signal with noise, a new low frequency oscillation analysis method is proposed ,which integrates neural network adaptive filtering with improved Prony method. Firstly, this method considers the wide area measurement signals as the inputs. Then,the neural network adaptive filter method is used for low fi＇equency oscillation signal pre-processing, adjusting the threshold of performance indicators to determine the filtering effect. Finally, the filtered signal is analyzed by improved Prony method. The simulation results show that the method can effectively filter out the noise, can be more accurately identify the dominant low frequency oscillation mode.