ESRT-based parameter identification method for sub- / ultra-synchronous oscillation of wind farm

Abstract

The ESRT-based wind farm sub- / ultra-synchronous oscillation parameter identification method of the application has the following steps: in step S1, the oscillation recording data of the voltage or current of the key node of the power grid is obtained based on the PMU measuring device, and the oscillation recording data is preprocessed to obtain the measurement signal; in step S2, the measurement signal is processed based on the ESRT decomposition method, the STFT time-frequency coefficient matrix is obtained, the pseudo modal is removed, and the frequency is redistributed, and the sparse time-frequency matrix containing only the sub- / ultra-synchronous oscillation mode is obtained; in step S3, the multi-modal coupled oscillation signal is reconstructed based on the sparse time-frequency matrix to obtain the single modal component; in step S4, the single modal component is subjected to Hilbert transformation to extract the characteristic parameters of each oscillation mode; by proposing the enhanced synchronous redistribution transformation decomposition method (ESRT), the characteristic parameters of each oscillation mode can be accurately extracted, the oscillation source can be quickly located, the basis for oscillation suppression can be provided, and the safe and stable operation of the power system can be ensured.

Description

Technical Field

[0001] This invention relates to the field of power system operation and maintenance technology, and in particular to a method for identifying wind farm sub- / supersynchronous oscillation parameters based on ESRT. Background Technology

[0002] With the continuous increase in the penetration rate of renewable energy in the power system, a large number of power electronic devices have been widely connected and integrated into the power grid. These devices have multi-timescale characteristics, and their complex interaction with the power grid makes the subsynchronous / supersynchronous oscillation problem increasingly prominent. Judging from the oscillation accidents that have occurred in recent years, this type of oscillation has shown the characteristic of wide-area propagation from local areas to the global scope. This undoubtedly brings great challenges to the source analysis of accidents and the effective suppression of oscillations. Therefore, timely and accurate identification of the relevant parameters of subsynchronous / supersynchronous oscillations in the power system can not only provide key support for the location of oscillation sources, but also provide an important basis for the scientific formulation of suppression strategies, which is of great significance to ensuring the safe and stable operation of the power system.

[0003] Currently, common methods for identifying subsynchronous / supersynchronous signal parameters mainly include the Hilbert-Huang transform (HHT) and the short-time Fourier transform (STFT). The HHT method can handle nonlinear and non-stationary signals, but it suffers from mode aliasing when decomposing modes with similar frequencies. The STFT method can extract subsynchronous / supersynchronous oscillation parameters, but it is susceptible to noise and difficult to obtain accurate identification results in noisy environments. The Synchro-Reassigning Transform (SRT) method is an effective analysis method for nonlinear and non-stationary signals, but its application in power system oscillation parameter identification has not yet been reported. Summary of the Invention

[0004] In view of this, the present invention proposes a wind farm sub / supersynchronous oscillation parameter identification method based on ESRT, which solves the problems of mode mixing and susceptibility to noise interference in traditional sub / supersynchronous oscillation parameter identification methods, and improves the accuracy of parameter identification.

[0005] The technical solution of this invention is implemented as follows: The method for identifying wind farm / supersynchronous oscillation parameters based on ESRT includes the following steps: Step S1: Obtain voltage or current oscillation waveform data of key nodes in the power grid based on the PMU measurement device, and preprocess the oscillation waveform data to obtain the measurement signal. Step S2: Process the measurement signal based on the ESRT decomposition method to obtain the STFT time-frequency coefficient matrix, perform pseudo-mode elimination and frequency redistribution, and obtain a sparse time-frequency matrix containing only sub / supersynchronous oscillation modes. Step S3: Reconstruct the multi-mode coupled oscillation signal based on the sparse time-frequency matrix to obtain the single-mode component; Step S4: Perform Hilbert transform on the single-mode components to extract the characteristic parameters of each oscillation mode.

[0006] Preferably, the preprocessing in step S1 includes detrending the component and removing the DC component.

[0007] Preferably, step S2 includes the following specific steps: Step S2-1: Perform STFT transformation on the measurement signal to obtain the time-frequency coefficient matrix; Step S2-2: Introduce average mode energy and mode energy weights to remove pseudo-modes from the time-frequency coefficient matrix; Step S2-3: Extract the sparse time-frequency matrix containing only sub / supersynchronous oscillation modes from the time-frequency coefficient matrix after pseudo-mode removal using a three-step selection rule.

[0008] Preferably, the specific steps of step S2-1 are as follows: Preprocessed measurement signals The expression for the time-frequency coefficient matrix obtained by performing STFT transformation is as follows:

[0009]

[0010] ; in For measurement signal At time t, frequency The STFT time-frequency coefficient matrix at point u, where g(u) is the window function. , u∈[t-△u, t+△u], where △u is half the window length. For the complex exponential term of the Fourier transform, For measurement signal At any moment ,frequency The STFT coefficient at that location.

[0011] Preferably, the specific steps of step S2-2 are as follows: Average mode energy is constructed based on the time-frequency coefficient matrix. : ; in Time-frequency coefficient matrix Average mode energy at frequency band k Time-frequency coefficient matrix STFT coefficients at frequency band k; Model energy weights are constructed based on average model energy. : ; Time-frequency coefficient matrix The relative mode energy weight at frequency band k, where n is the total number of frequency bands; When mode energy weight Less than the threshold At that time, the STFT coefficients corresponding to its frequency band are set to zero, and the pseudo-modes are separated and eliminated. The time-frequency coefficient matrix after eliminating pseudo-modes is: ; in To remove pseudo-modes, the time-frequency coefficient matrix in time... ,frequency STFT coefficient at the location, To remove pseudo-modes, the time-frequency coefficient matrix in time... ,frequency The STFT coefficient at that location.

[0012] Preferably, the specific steps of steps S2-3 are as follows: Obtaining discrete time-frequency points The local maximum value within its domain is taken as the amplitude, and the amplitude satisfies the formula: ; Discrete time frequency points First derivative in the frequency direction The following first condition must be met: ; Based on satisfying the first condition, discrete time-frequency points The first derivative in the frequency direction exhibits a monotonically decreasing trend and satisfies the second condition: ; in Discrete time and frequency, respectively This is the error amount; Using the amplitude satisfying the formula, the first condition, and the second condition, a sparse time-frequency matrix containing only subsynchronous / supersynchronous oscillation modes is obtained: ; in It is a sparse time-frequency matrix.

[0013] Preferably, the reconstructed expression in step S3 is: ; in It is a multi-mode coupled oscillation signal. For the i-th single-mode component, Let be the instantaneous frequency of the i-th single-mode component. This represents the peak value of the window function at the center of the window.

[0014] Preferably, the i-th single-mode component The corresponding sparse time-frequency matrix is: ; in For multi-mode coupled oscillation signals At the i-th frequency point and the l-th time point coefficient.

[0015] Preferably, the specific steps of step S4 are as follows: For single-mode components Performing the Hilbert transform, we get: ; in for The result after performing the Hilbert transform. Let Cauchy's principal integral be the value. Step size, >0; Single-mode components Analyzed signal for: ; in and They are respectively The instantaneous amplitude and instantaneous phase; A signal with fixed frequency components can be represented as: ; in, Let be the initial amplitude of the i-th sub-mode. Let be the attenuation factor of the i-th submode. Let be the oscillation frequency of the i-th submode. Let be the phase angle of the i-th submode; From analytical signals The expression and the signal representation with fixed frequency components can be obtained as follows: ; By performing a linear fit on the above equation using the least squares method, the initial amplitude under the oscillation mode can be obtained. Attenuation factor and oscillation frequency .

[0016] Compared with the prior art, the beneficial effects of the present invention are: The wind farm subsynchronous / supersynchronous oscillation parameter identification method based on ESRT of the present invention first performs short-time Fourier transform on the wide-area measurement signal to obtain the corresponding time-frequency coefficient matrix. Then, it filters and removes noise pseudo-modes by mode energy weighting. Further, it extracts the instantaneous frequency trajectory of the system subsynchronous / supersynchronous oscillation mode from the time-frequency coefficient matrix after pseudo-mode removal through a three-step selection rule, and achieves accurate separation and time-domain reconstruction of each mode. Finally, it accurately extracts the characteristic parameters of each oscillation mode by combining Hilbert transform, and identifies its oscillation frequency, attenuation factor, amplitude and other parameters. This method solves the problems of mode aliasing and susceptibility to noise interference in traditional subsynchronous / supersynchronous oscillation parameter identification methods, provides key support for the location of oscillation sources, and ensures the safe and stable operation of the power system. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only preferred embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of the wind farm / supersynchronous oscillation parameter identification method based on ESRT of the present invention; Figure 2 This is an electromagnetic transient simulation model diagram of the wind farm sub- / supersynchronous oscillation parameter identification method based on ESRT of the present invention. Figure 3 The waveform diagram of sub / supersynchronous oscillation under 40% series complement is shown in the specific implementation of the wind farm sub / supersynchronous oscillation parameter identification method based on ESRT of the present invention. Figure 4 The WFT spectrum of the sub / supersynchronous oscillation waveform is shown in the specific implementation of the wind farm sub / supersynchronous oscillation parameter identification method based on ESRT of the present invention. Figure 5 The time-spectrum diagram of the sub- / supersynchronous oscillation waveform is shown in the specific implementation of the wind farm sub- / supersynchronous oscillation parameter identification method based on ESRT of the present invention. Figure 6 This is an IMF decomposition diagram of the sub- / supersynchronous oscillation waveform of a specific implementation of the wind farm sub- / supersynchronous oscillation parameter identification method based on ESRT of the present invention. Detailed Implementation

[0019] To better understand the technical content of this invention, a specific embodiment is provided below, and the invention will be further described in conjunction with the accompanying drawings.

[0020] See Figure 1 The wind farm / supersynchronous oscillation parameter identification method based on ESRT provided by this invention includes the following steps: Step S1: Obtain voltage or current oscillation waveform data of key nodes in the power grid based on the PMU measurement device, and preprocess the oscillation waveform data to obtain the measurement signal. Step S2: Process the measurement signal based on the improved synchronous redistribution transform decomposition method to obtain the STFT time-frequency coefficient matrix and perform pseudo-mode elimination and frequency redistribution to obtain a sparse time-frequency matrix containing only sub / supersynchronous oscillation modes. Step S3: Reconstruct the multi-mode coupled oscillation signal based on the sparse time-frequency matrix to obtain the single-mode component; Step S4: Perform Hilbert transform on the single-mode components to extract the characteristic parameters of each oscillation mode.

[0021] This invention discloses a wind farm subsynchronous / supersynchronous oscillation parameter identification method based on ESRT, used for subsynchronous / supersynchronous oscillation parameter identification in wind farm power grids. It acquires voltage or current oscillation waveform data of key nodes in the power grid using a pre-set PMU device. The acquired oscillation waveform data is then preprocessed to remove trend components and DC components to obtain measurement signals. These signals are then subjected to time-frequency analysis based on the Enhanced Synchronous Redistribution Transform (ESRT) method. The time-frequency analysis includes obtaining the time-frequency coefficient matrix through Short-Time Fourier Transform (STFT). After removing pseudo-modes from the time-frequency coefficient matrix, frequency redistribution is performed using a three-step selection rule to obtain a sparse time-frequency matrix containing only sub / supersynchronous oscillation modes. Then, based on the sparse time-frequency matrix, multi-mode coupled oscillation signals can be accurately separated and reconstructed in the time domain to obtain single-mode components. Finally, the characteristic parameters of each oscillation mode, including oscillation frequency, attenuation factor, and amplitude, can be accurately extracted using the Hilbert transform. These characteristic parameters can be used to quickly locate the oscillation source, providing a basis for oscillation suppression and ensuring the safe and stable operation of the power system.

[0022] Preferably, step S2 includes the following specific steps: Step S2-1: Perform STFT transformation on the measurement signal to obtain the time-frequency coefficient matrix. The specific steps are as follows: Preprocessed measurement signals The expression for the time-frequency coefficient matrix obtained by performing STFT transformation is as follows:

[0023]

[0024] ; in For measurement signal At time t, frequency The STFT time-frequency coefficient matrix at point u, where g(u) is the window function. , u∈[t-△u, t+△u], where △u is half the window length. For the complex exponential term of the Fourier transform, For measurement signal At any moment ,frequency The STFT coefficient at that location.

[0025] When using STFT for time-frequency analysis, the choice of window function is crucial to the time-frequency resolution. The Gaussian window function can achieve the optimal balance between time resolution and frequency resolution. Therefore, this invention selects the Gaussian window function as the window function. .

[0026] Step S2-2: Introduce average mode energy and mode energy weights to remove pseudo-modes from the time-frequency coefficient matrix; Since the STFT time-frequency coefficient matrix reflects the energy distribution characteristics of the signal in each frequency band to a certain extent, this invention introduces average mode energy and mode energy weights to quantify the mode energy contribution in each frequency band, thereby effectively identifying and eliminating the coefficients corresponding to pseudo-modes in the time-frequency coefficient matrix. The specific steps are as follows: Average mode energy is constructed based on the time-frequency coefficient matrix. : ; in Time-frequency coefficient matrix Average mode energy at frequency band k Time-frequency coefficient matrix STFT coefficients at frequency band k; Model energy weights are constructed based on average model energy. : ; Time-frequency coefficient matrix The relative mode energy weight at frequency band k, where n is the total number of frequency bands; When mode energy weight Less than the threshold At that time, the STFT coefficients corresponding to its frequency band are set to zero, and the pseudo-modes are separated and eliminated. The time-frequency coefficient matrix after eliminating pseudo-modes is: ; in To remove pseudo-modes, the time-frequency coefficient matrix in time... ,frequency STFT coefficient at the location, To remove pseudo-modes, the time-frequency coefficient matrix in time... ,frequency The STFT coefficient at that location.

[0027] By setting an energy weight threshold, effective elimination can be achieved. The pseudo-modes in the model are then used to determine the mode energy weights. relative mode energy weight threshold Comparison, when it is less than the relative mode energy weight threshold When a spurious mode is identified, its corresponding STFT coefficient is set to zero, thereby separating and eliminating spurious modes from the signal. This invention uses a relative mode energy weighting threshold. Set to 0.02.

[0028] Step S2-3: Extract the sparse time-frequency matrix containing only sub / supersynchronous oscillation modes from the time-frequency coefficient matrix after pseudo-mode removal using a three-step selection rule. The specific steps are as follows: Obtaining discrete time-frequency points The local maximum value within its domain is taken as the amplitude, and the amplitude satisfies the formula: ; Discrete time frequency points First derivative in the frequency direction It should be equal to zero. However, considering the numerical errors in actual analysis, it only needs to meet the condition that it is approximately zero, that is, the following first condition must be met: ; Based on satisfying the first condition, discrete time-frequency points The first derivative in the frequency direction exhibits a monotonically decreasing trend and satisfies the second condition, ensuring that the spectral function has local convexity in this direction: ; in Discrete time and frequency, respectively This is the error amount; Using the amplitude satisfying the formula, the first condition, and the second condition, a sparse time-frequency matrix containing only subsynchronous / supersynchronous oscillation modes is obtained: ; in It is a sparse time-frequency matrix.

[0029] Preferably, the reconstructed expression in step S3 is: ; in It is a multi-mode coupled oscillation signal. For the i-th single-mode component, Let be the instantaneous frequency of the i-th single-mode component. This represents the peak value of the window function at the center of the window.

[0030] Preferably, the i-th single-mode component The corresponding sparse time-frequency matrix is: ; in For multi-mode coupled oscillation signals At the i-th frequency point and the l-th time point coefficient.

[0031] Preferably, the specific steps of step S4 are as follows: For single-mode components Performing the Hilbert transform, we get: ; in for The result after performing the Hilbert transform. Let Cauchy's principal integral be the value. Step size, >0; Single-mode components Analyzed signal for: ; in and They are respectively The instantaneous amplitude and instantaneous phase; A signal with fixed frequency components can be represented as: ; in, Let be the initial amplitude of the i-th sub-mode. Let be the attenuation factor of the i-th submode. Let be the oscillation frequency of the i-th submode. Let be the phase angle of the i-th submode; From analytical signals The expression and the signal representation with fixed frequency components can be obtained as follows: ; By performing a linear fit on the above equation using the least squares method, the initial amplitude under the oscillation mode can be obtained. Attenuation factor and oscillation frequency .

[0032] This invention employs, as follows Figure 2 The electromagnetic transient simulation model shown serves as an example to verify the effectiveness and feasibility of the method proposed in this invention. The model consists of a doubly-fed induction generator (DFIG), a transformer, a transmission line, and an infinite bus system. The transmission line is connected to the infinite bus system via series capacitor compensation. By changing the value of the series capacitor, the series compensation degree is altered, resulting in subsynchronous / supersynchronous oscillation waveforms.

[0033] This invention selects the A-phase current as the input signal, keeps the wind speed constant at 11 m / s, and sets the initial series complement to 10%. At 2s, the series complement increases from 10% to 40%, thereby triggering subsynchronous / supersynchronous oscillations. Figure 3 The waveforms of the A-phase current under 40% series complement in the electromagnetic transient simulation model are shown. Taking the A-phase current under 40% series complement at 2-4 seconds as an example, WFT spectrum analysis is performed on it. Figure 4 The WFT spectrum of the A-phase current with 40% series complement in 2-4 seconds shows that, by observing the spectrum, it can be determined that the A-phase current contains 3 modes, including one subsynchronous mode and one supersynchronous mode. Figure 5 The time-frequency spectrum of the A-phase current with 40% series complement over 2-4 seconds shows that it contains three modes: one subsynchronous mode and one supersynchronous mode. Figure 4 The results are consistent with the WFT spectrogram analysis. Figure 6 The IMF mode decomposition diagram of the 40% series complement waveform is used. Hilbert transform is performed on IMF1 and IMF3 to calculate the characteristic parameters. The parameters identified by the algorithm of this invention are compared with those identified by the VMD method, and the results are shown in Table 1. Table 1 shows that the characteristic parameters calculated by the two methods are very close, verifying the effectiveness of the proposed method in identifying the characteristic parameters of subsynchronous / supersynchronous oscillations.

[0034] Table 1 Feature parameters identified by ESRT and VMD methods

[0035] The algorithm of this invention was compared with the parameters of the WFT spectrogram, and the results are shown in Table 2. Table 2 shows that the feature parameters obtained by the two methods are basically the same, thus further verifying the effectiveness of the algorithm of this invention.

[0036] Table 2 Characteristic parameters identified by ESRT method and WFT spectrogram

[0037] The feature parameters obtained by the algorithm of this invention are compared with those obtained by the FSST, SET, and MSST algorithms, and the results are shown in Table 3. Comparison of Table 3 with the WFT results shows that the algorithm of this invention has better recognition accuracy compared to the other three algorithms, verifying the superiority of the algorithm of this invention over the others.

[0038] Table 3 Feature parameters identified by ESRT method and other methods

[0039] exist Figure 3 Based on the subsynchronous / supersynchronous oscillation waveforms, 30dB and 20dB of Gaussian white noise were added respectively. The characteristic parameters identified by the algorithm of this invention are shown in Table 4. As can be seen from Table 4, the characteristic parameters identified after adding noise are very close to those identified in Table 1, verifying that the method proposed in this invention has good noise resistance in the identification of subsynchronous / supersynchronous oscillation characteristic parameters.

[0040] Table 4. Characteristic parameters identified by the ESRT method at different noise levels.

[0041] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for identifying wind farm generation / supersynchronous oscillation parameters based on ESRT, characterized in that, Includes the following steps: Step S1: Obtain voltage or current oscillation waveform data of key nodes in the power grid based on the PMU measurement device, and preprocess the oscillation waveform data to obtain the measurement signal. Step S2: Process the measurement signal based on the ESRT decomposition method to obtain the STFT time-frequency coefficient matrix, perform pseudo-mode elimination and frequency redistribution, and obtain a sparse time-frequency matrix containing only sub / supersynchronous oscillation modes. Step S3: Reconstruct the multi-mode coupled oscillation signal based on the sparse time-frequency matrix to obtain the single-mode component; Step S4: Perform Hilbert transform on the single-mode components to extract the characteristic parameters of each oscillation mode; The specific steps of step S2 include: Step S2-1: Perform STFT transformation on the measurement signal to obtain the time-frequency coefficient matrix; Step S2-2: Introduce average mode energy and mode energy weights to remove pseudo-modes from the time-frequency coefficient matrix; Step S2-3: Extract the sparse time-frequency matrix containing only sub / supersynchronous oscillation modes from the time-frequency coefficient matrix after pseudo-mode removal using a three-step selection rule; The specific steps of step S2-1 are as follows: Preprocessed measurement signals The expression for the time-frequency coefficient matrix obtained by performing STFT transformation is as follows: ； in For measurement signal At time t, frequency The STFT time-frequency coefficient matrix at point u, where g(u) is the window function. , u∈[t-△u, t+△u], where △u is half the window length. For the complex exponential term of the Fourier transform, For measurement signal At any moment ,frequency STFT coefficient at the location; The specific steps of step S2-2 are as follows: Average mode energy is constructed based on the time-frequency coefficient matrix. : ； in Time-frequency coefficient matrix Average mode energy at frequency band k Time-frequency coefficient matrix STFT coefficients at frequency band k; Model energy weights are constructed based on average model energy. : ； Time-frequency coefficient matrix The relative mode energy weight at frequency band k, where n is the total number of frequency bands; When mode energy weight Less than the threshold At that time, the STFT coefficients corresponding to its frequency band are set to zero, and the pseudo-modes are separated and eliminated. The time-frequency coefficient matrix after eliminating pseudo-modes is: ； in To remove pseudo-modes, the time-frequency coefficient matrix in time... ,frequency STFT coefficient at the location, To remove pseudo-modes, the time-frequency coefficient matrix in time... ,frequency STFT coefficient at the location; The specific steps of step S2-3 are as follows: Obtaining discrete time-frequency points The local maximum value within its domain is taken as the amplitude, and the amplitude satisfies the formula: ； Discrete time frequency points First derivative in the frequency direction The following first condition must be met: ； Based on satisfying the first condition, discrete time-frequency points The first derivative in the frequency direction exhibits a monotonically decreasing trend and satisfies the second condition: ； in Discrete time and frequency, respectively This is the error amount; A sparse time-frequency matrix containing only subsynchronous / supersynchronous oscillation modes is obtained using a three-step selection rule that includes amplitudes satisfying the formula, the first condition, and the second condition: ； in It is a sparse time-frequency matrix.

2. The method for identifying wind farm generation / supersynchronous oscillation parameters based on ESRT according to claim 1, characterized in that, The preprocessing in step S1 includes detrending the component and removing the DC component.

3. The method for identifying wind farm number / supersynchronous oscillation parameters based on ESRT according to claim 1, characterized in that, The reconstructed expression in step S3 is: ； in It is a multi-mode coupled oscillation signal. For the i-th single-mode component, Let be the instantaneous frequency of the i-th single-mode component. This represents the peak value of the window function at the center of the window.

4. The method for identifying wind farm generation / supersynchronous oscillation parameters based on ESRT according to claim 3, characterized in that, The i-th single-mode component The corresponding sparse time-frequency matrix is: ； in For multi-mode coupled oscillation signals At the i-th frequency point and the l-th time point coefficient.

5. The method for identifying wind farm sub-segments / supersynchronous oscillation parameters based on ESRT according to claim 1, characterized in that, The specific steps of step S4 are as follows: For single-mode components Performing the Hilbert transform, we get: ； in for The result after performing the Hilbert transform. Let Cauchy's principal integral be the value. Step size, >0; Single-mode components Analyzed signal for: ； in and They are respectively The instantaneous amplitude and instantaneous phase; A signal with fixed frequency components can be represented as: ； in, Let be the initial amplitude of the i-th sub-mode. Let be the attenuation factor of the i-th submode. Let be the oscillation frequency of the i-th submode. Let be the phase angle of the i-th submode; From analytical signals The expression and the signal representation with fixed frequency components can be obtained as follows: ； By performing a linear fit on the above equation using the least squares method, the initial amplitude under the oscillation mode can be obtained. Attenuation factor and oscillation frequency .

Images