Research On Time-frequency Analysis Of Non-stationary Signals With Single Component And Its Application In Oscillations In Power System

In the field of signal analysis,compared to classic spectrum analysis,time-frequency analysis methods can map one-dimensional time-domain signals to two-dimensional time-frequency planes,which can get joint time-frequency characteristics of non-stationary signals.The existing non-stationary signal time-frequency analysis methods are based on some single-component signals with specific time-frequency characteristics.The shortcomings of these methods are that some are difficult to obtain both high time resolution and frequency resolution,some are only applicable to specific types of signals which can’t represent general situations.In order to study a more general single-component signal form and the corresponding time-frequency analysis method that satisfies both high time resolution and high frequency resolution,the paper conducts research from the following aspects:(1)From a conceptual perspective,a new mathematical concept of "pan periodicity" was proposed and based on this,a general mathematical form of single-component non-stationary signals was given,covering the traditional concepts like "narrow-band signals" and "intrinsic mode function".Thus,a more rigorous theoretical basis for the time-frequency analysis of single-component non-stationary signals was established.(2)The existing typical methods(Fast Fourier transform,Hilbert Transform,Empirical AM-FM decomposition method,etc.)suitable for single-component non-stationary signal time-frequency analysis were discussed.Their deficiencies and the error sources in the time-frequency analysis of single-component non-stationary signals were analyzed.(3)Based on the characteristics of the single-component non-stationary signal,four new methods was proposed: Instantaneous Frequency Estimation by Midpoints of the Subintervals(IFEMS),Instantaneous Frequency Estimation by Zeros of the Function(IFEZF),Logarithm Decomposition with EMD(LDEMD),Scale Transform with Adaptive Integration(STAI).Their advantages,disadvantages and error sources compared with existing methods were discussed.By performing time-frequency analysis on signals with various non-stationary characteristics,IFEZF,IFEMS,and LDEMD are compared with existing methods.The result shows that these new methods proposed in this paper are superior to existing methods.(4)Finally,new methods of time-frequency analysis proposed in this paper were used to analyze the oscillation signals in power systems.The result shows that these new methods proposed by this paper can get more accurate mode parameters of low-frequency oscillation and instantaneous frequency of sub-synchronous power oscillation signals.