Research On Low Frequency Oscillation Pattern Identification Based On Wavelet Packet And SVM

With the rapid development of China's power industry,the scale of the power system has continued to expand and has been developed into a large-scale interconnected power grid.Interconnected power grids are conducive to improving the economic performance and reliability of power transmission and transmission,but the interconnection of multiple regional power grids may cause low-frequency oscillations.The low frequency oscillation process of the power system has strong random and nonlinear characteristics.Analyzing the characteristics of the low frequency oscillation mode of the power system can reveal the oscillation mode and propagation characteristics.Effectively tracking the oscillation mode of the system can accurately predict system instability and provide important reference for system analysis and emergency control.However,the low-frequency oscillation modes in the low-frequency oscillation process of the power system are varied,the oscillation process is complex and varied,and the response process is often accompanied by various external noise disturbances,resulting in the poor performance of some of the existing low-frequency oscillation dynamic characteristics analysis methods.Therefore,it is of great theoretical and practical value to study the new method of analyzing the characteristics of low-frequency oscillation modes in interconnected power grids.This paper combines modern signal processing methods,and proposes a principle that is more simple.The identification method applies the new method to the relevant role of low-frequency oscillation mode identification.This paper proposes a low-frequency oscillation mode identification method based on wavelet packet and support vector machine(SVM),and introduces the support vector machine for classification identification into low-frequency oscillation mode identification.This paper describes the application of wavelet packet to signal denoising analysis and the basic theory of signal energy feature vector extraction,and applies it to the extraction of low-frequency oscillation signals.Then it describes the basic principles of support vector machine,how to classify data,etc.A detailed analysis was performed.In this paper,according to the characteristics of low-frequency oscillation signals,the wide-area measurement signal is first denoised by wavelet packet,and the wavelet basis applicable to denoising is obtained through a large number of experiments.The denoised signal is obtained and compared with the original signal.Then,the denoised signal is extracted by wavelet packet energy feature vector to obtain the energy feature vector containing the pattern feature.Since the dimension is larger,principal component analysis is used to reduce the dimension by the contribution rate,and the dimension is reduced.After the energy feature vector,the SVM is trained using the energy feature vectors of a large number of data samples to obtain an identification classifier for low-frequency oscillation mode identification.After the signal is processed,the SVM pattern identifier is used to identify the low-frequency oscillation signal.Through example analysis and simulation analysis of the node model's measured signal,the results are compared with the results obtained from analysis of the Prony algorithm.It is verified that the method principle is simple and convenient,the classification results are accurate,the reliability is high,and the real-time performance is good.Oscillation performs a good pattern recognition.