Analysis Of Low Frequency Oscillation Based On Modal Separation And Stochastic Resonance Inversion Recognition

With the expansion of renewable energy interconnection structure,the large-scale construction of flexible AC/DC transmission and the vigorous development of DC distribution network technology,the global power grid has entered a period of a new form of power system with regional networking and overall optimal allocation of resources.The use of high magnification fast excitation system can improve the transient stability and voltage quality of power system,but it also leads to the increase of weak network in the system,which intensifies the occurrence of underdamped low frequency oscillation.when monitoring the low-frequency oscillation signal through a wide-area measurement system,we find that the oscillation signal is close to the shape of the random fluctuation of the system and often confuses the noise,which makes it difficult to distinguish the oscillation signal from the noise and the random fluctuation.At the same time,the low frequency oscillation signal of power system is a typical multi-component noise mixing signal,which brings great difficulty to the feature extraction of low frequency oscillation signal.Therefore,it is of great significance to accurately extract the information of the dominant modes of low frequency oscillation in the background of noise to improve the monitoring ftmction of power system and to ensure the safe and stable operation of power network.In this paper,the variational modal decomposition algorithm and the empirical wavelet transform algorithm with strong non-stationary signal processing ability are proposed in recent years as the preprocessing algorithm of the acquisition signal,and the feature online identification of the low frequency oscillation signal of the power system is realized by combining the stochastic resonance theory.Firstly,the definition,generation mechanism,analysis methods,research status and conditions for on-line identification of low-frequency oscillation in power system are briefly described,For the characteristics of low frequency oscillation signal,the principle of empirical mode decomposition method,which is a traditional pre-processing algorithm for low frequency oscillation signal,is introduced.In view of the phenomenon of modal aliasing,loss and distortion when dealing with low-frequency oscillatory signals,and the poor anti-noise performance of this method,variational modal decomposition algorithm and empirical wavelet transform algorithm are introduced in this paper.These two methods have strong theoretical basis and non-stationary signal processing ability.in order to make these two algorithms more suitable for low-frequency oscillation analysis in power system,the variational mode decomposition method based on bandwidth summation limitation and the empirical wavelet transform based on flat-top envelope mirror extension are proposed in this paper to realize the adaptive and effective extraction of the dominant modes of low-frequency oscillation signals.After that,combined with the analysis of the basic classical theory of stochastic resonance,a nonlinear system processing model of Duffing oscillator generalized parametric stochastic resonance is established,and the identification technique of stochastic resonance inversion is proposed.The noise energy of low frequency oscillation signal in power system is transferred to the signal to be tested,so as to strengthen the characteristics of the signal to be tested,and the effective extraction of the characteristic information of each dominant mode of low frequency oscillation is completed.Different from the idea of filtering noise by other methods,this method can realize the transfer of noise energy to the signal to be tested and highlight the characteristics of the signal to be tested when the parameters and nonlinear systems reach a reasonable match.The successful application of this method in low frequency oscillation signal processing provides a new idea for low frequency oscillation mode identification analysis in power system.Lastly,the method proposed in this paper is applied to the simulation system oflEEE16 68 nodes and the practical example analysis of power grid,and the effectiveness of this method is verified.At the same time,by comparing the traditional methods Prony method,HHT method and VMD-Hilbert method,it shows that the proposed method has good noise robustness and extraction accuracy,and meets the requirements of online recognition,it has certain engineering application value.