Power system low-frequency oscillation mode identification method and terminal

Abstract

The invention discloses a power system low-frequency oscillation modal identification method and a terminal, and the method comprises the steps: obtaining a low-frequency oscillation noisy signal of a power system, and carrying out the preprocessing of the noisy signal through an improved adaptive wavelet threshold algorithm; wherein the adaptive wavelet threshold algorithm adopts an improved wavelet threshold dynamically adjusted based on a noise standard deviation and a data length; constructing a Hankle matrix and performing singular value decomposition, calculating a singular value cumulative contribution rate based on PCA, and determining an optimal modal order of a TLS-ESPRIT algorithm; subspace division is carried out on a signal space by using TLS-ESPRIT, and oscillation modal parameters are solved and determined through twiddle factor eigenvalues; compared with a traditional method, the anti-interference performance and the identification accuracy are improved, accurate order determination can be achieved, and the method has a good application prospect in power system low-frequency oscillation early warning.

Description

Technical Field

[0001] The present invention relates to the field of electrical engineering, and particularly to a method and a terminal for identifying low-frequency oscillation modes of a power system. Background Art

[0002] Currently, the scale of power system interconnection is increasing day by day, and the long-distance transmission of electric energy supply across regions has highlighted the weak connections between multiple regions. At the same time, the large-scale use of excitation devices for improving the transient stability of the power grid and various operation modes have led to an increasing number of low-frequency oscillation phenomena. If the oscillation duration becomes longer, it may exacerbate the induction of successive accidents and affect the stability of the system. The actual signals are often mixed with a large amount of noise interference. Therefore, how to quickly and accurately complete the online identification of the modal parameters of low-frequency oscillation signals is of great research significance for dispatchers to take corresponding decision-making control measures.

[0003] The low-frequency oscillation analysis methods can be mainly divided into two categories: The first category is the linearization analysis method based on a mathematical model. The establishment of the mathematical model depends on the accuracy of the model parameters. However, when the dimension of the wide-area model is increasing, it is difficult to obtain relatively accurate model parameter estimates and the eigenvalue calculation time is too long. Obviously, it is not suitable for the online monitoring and analysis of low-frequency oscillations.

[0004] The other is the identification and processing method based on the real-time data of the wide-area measurement system (WAMS). The commonly used identification algorithms mainly include the Fourier transform (FFT), Prony analysis method, intrinsic time-scale decomposition (ITD) method, total least squares-rotational invariance technique (TLS-ESPRIT) and other algorithms. The limitation of FFT is that it cannot reflect the complete oscillation characteristic information; although the Prony method can completely identify the modal parameters of the collected signals, it will produce large errors when processing signals with time-varying characteristics and non-stationary signals; when the ITD algorithm identifies signals with a low signal-to-noise ratio, the anti-interference performance of this method is weak and the parameter extraction is not accurate enough; the TLS-ESPRIT method is a class of signal parameter identification methods based on subspace technology, but the accuracy of the algorithm identification depends to a certain extent on the selection of the modal order. If the order is selected too small, important modal information will be missed; if the order is set too large, false modes will appear, and dispatchers may draw wrong conclusions in practical applications. In terms of denoising preprocessing, the FIRE filter has a certain effect on the accurate identification of parameters under the noise background, but the setting of the filter parameters is relatively complex; due to the use of a fixed threshold in the traditional wavelet transform, it cannot perform adaptive processing according to the characteristics of the noise distribution, and the filtering effect needs to be improved. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a method and a terminal for identifying low-frequency oscillation modes of a power system, so as to achieve more accurate identification of low-frequency oscillation modes.

[0006] In order to solve the above technical problem, the technical solution adopted by the present invention is as follows: A method for identifying low-frequency oscillation modes of a power system, comprising the steps of: S1. Obtain the low-frequency oscillation noisy signal of the power system, and preprocess the noisy signal through an improved adaptive wavelet threshold algorithm; Wherein, the adaptive wavelet threshold algorithm adopts an improved wavelet threshold dynamically adjusted based on the noise standard deviation and the data length; S2. Construct a Hankle matrix and perform singular value decomposition, calculate the cumulative contribution rate of singular values based on PCA, and determine the optimal modal order of the TLS-ESPRIT algorithm; S3. Use TLS-ESPRIT to divide the signal space into subspaces, and determine the oscillation mode parameters by solving the eigenvalues of the rotation factor.

[0007] In order to solve the above technical problem, another technical solution adopted by the present invention is as follows: A terminal for identifying low-frequency oscillation modes of a power system, comprising a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the following steps are implemented: S1. Obtain the low-frequency oscillation noisy signal of the power system, and preprocess the noisy signal through an improved adaptive wavelet threshold algorithm; Wherein, the adaptive wavelet threshold algorithm adopts an improved wavelet threshold dynamically adjusted based on the noise standard deviation and the data length; S2. Construct a Hankle matrix and perform singular value decomposition, calculate the cumulative contribution rate of singular values based on PCA, and determine the optimal modal order of the TLS-ESPRIT algorithm; S3. Use TLS-ESPRIT to divide the signal space into subspaces, and determine the oscillation mode parameters by solving the eigenvalues of the rotation factor.

[0008] The beneficial effect of the present invention is that: a method and a terminal for identifying low-frequency oscillation modes of a power system according to the present invention use a new method combining improved wavelet threshold denoising and PCA-TLS-ESPRIT method for identifying low-frequency oscillation modes of a power system. Through the fusion of improved wavelet threshold denoising and PCA-TLS-ESPRIT, the adaptive preprocessing of noisy signals and data-driven order determination are realized, the noise interference can be effectively suppressed, and the low-frequency oscillation characteristic parameters can be accurately identified. Compared with the traditional method, the anti-interference performance and identification accuracy are improved, and accurate order determination can be achieved, which has a good application prospect in the early warning of low-frequency oscillations in power systems. Description of the Drawings

[0009] Figure 1 This is a schematic flowchart of a method for identifying low-frequency oscillation modes of a power system according to an embodiment of the present invention; Figure 2 This is a flowchart of a method for identifying low-frequency oscillation modes of a power system according to an embodiment of the present invention; Figure 3 This is a waveform diagram of a low-frequency oscillation signal with noise in a method for identifying low-frequency oscillation modes of a power system according to an embodiment of the present invention; Figure 4 This is a comparison diagram of the effects after denoising a low-frequency oscillation signal in a method for identifying low-frequency oscillation modes of a power system according to an embodiment of the present invention; Figure 5 This is a fitting curve diagram of the identification results of each method parameter in an embodiment of the present invention; Figure 6 This is a wiring diagram of the EPRI-8 machine 36-node system in the PSASP software according to an embodiment of the present invention; Figure 7 This is a waveform diagram of a low-frequency oscillation signal with noise in a multi-machine multi-point system of a method for identifying low-frequency oscillation modes of a power system according to an embodiment of the present invention; Figure 8 This is a waveform comparison diagram after denoising a low-frequency oscillation signal in a multi-machine multi-point system of a method for identifying low-frequency oscillation modes of a power system according to an embodiment of the present invention; Figure 9 This is a structural diagram of a terminal for identifying low-frequency oscillation modes of a power system according to an embodiment of the present invention; Label Description: 1. A terminal for identifying low-frequency oscillation modes of a power system; 2. A processor; 3. A memory. Detailed Embodiment

[0010] To describe the technical content, the achieved objectives and the effects of the present invention in detail, the following is described in conjunction with the embodiments and with reference to the drawings.

[0011] Please refer to Figure 1 and Figure 2 , a method for identifying low-frequency oscillation modes of a power system, comprising the steps of: S1. Obtain a low-frequency oscillation noisy signal of the power system, and preprocess the noisy signal by an improved adaptive wavelet threshold algorithm; Wherein, the adaptive wavelet threshold algorithm uses an improved wavelet threshold dynamically adjusted based on the noise standard deviation and the data length; S2. Construct a Hankle matrix and perform singular value decomposition, calculate the cumulative contribution rate of singular values based on PCA, and determine the optimal modal order of the TLS-ESPRIT algorithm; S3. Use TLS - ESPRIT to partition the signal space into sub - spaces, and determine the oscillation mode parameters by solving the eigenvalues of the rotation factor.

[0012] As can be seen from the above description, the beneficial effects of the present invention are as follows: A method for identifying low - frequency oscillation modes in a power system, which combines an improved wavelet threshold denoising method with the PCA - TLS - ESPRIT method, is used for identifying low - frequency oscillation modes in a power system. Through the integration of the improved wavelet threshold denoising method and PCA - TLS - ESPRIT, the adaptive pre - processing of noisy signals and data - driven order determination are realized. It can effectively suppress noise interference and accurately identify the low - frequency oscillation characteristic parameters, improving the anti - interference performance and identification accuracy compared with traditional methods, and can achieve accurate order determination, having good application prospects in the early warning of low - frequency oscillations in power systems.

[0013] Further, step S2 includes the following steps: S21. Construct a Hankel matrix based on the denoised free - oscillation data ; H ; ; Among them, , the number of rows of the matrix H is L, and the number of columns is M; S22. Decompose the matrix using the singular value method: ; Among them, the subscript S and the subscript N correspond to the signal space and the residual noise space respectively. U is an L - order matrix, V is an M - order matrix, represents its conjugate transpose, the matrix , and the diagonal element values in Σ are the th singular value of H, where the numerical values of the singular values in the signal space are much larger than those in the residual noise space; S23. Arrange the singular values in descending order. Taking PCA as the order - determination index, when the influence caused by the current factors is greater than the threshold , at this time the corresponding value is the modal order : ; ; ; Among them, m represents the total number of singular values.

[0014] As can be seen from the above description, the PCA order determination steps based on the Hankle matrix and singular value decomposition are clarified, and the optimal order is automatically determined by calculating the cumulative contribution rate of singular values. Compared with the traditional TLS-ESPRIT manual order determination (such as the order of 10), this method only requires 6 orders to accurately identify the dominant modes in the 8-machine 36-node system, reducing the computational complexity and the interference of false modes, and ensuring the reliability of modal parameter extraction.

[0015] Further, step S3 includes the steps of: S31. According to the order determine the signal space , and divide it into two interleaved subspaces: ; S32. Construct a new matrix , and perform singular value decomposition: ; S33. Divide into 4 square matrices of order P, and find the rotation factor : ; is expressed as follows: ; S34. Solve the oscillation modal parameters of the signal according to the eigenvalues of the rotation factor: ; wherein, is the low-frequency oscillation frequency; is the oscillation attenuation factor; is the damping ratio. [[ID=K]]

[0016] As can be seen from the above description, the TLS-ESPRIT subspace division and rotation factor solution processes are refined, and through the construction of interleaved subspaces and singular value decomposition, high-precision calculations of the oscillation frequency, attenuation factor, and damping ratio are achieved.

[0017] Further, step S1 includes the steps of: S11. Obtain the low-frequency oscillation noisy signal of the power system, perform wavelet transform processing on it, and obtain the wavelet coefficients at each scale; where j represents the decomposition scale, and k represents the kth wavelet coefficient; S12. Use the improved wavelet threshold to perform denoising preprocessing on the noisy signal, and perform signal reconstruction to obtain the denoised signal; wherein, the improved wavelet threshold is dynamically adjusted based on the noise standard deviation and the data length: ; ; where σ represents the standard deviation of the noise, N represents the data length of the signal, means taking the median of the wavelet coefficients.

[0018] As can be seen from the above description, the specific formula for improving the wavelet threshold is defined, and the standard deviation σ of the noise is calculated by the median method, realizing the dynamic adjustment of the threshold. Compared with the fixed-threshold hard / soft denoising, this method can effectively suppress noise interference while retaining the signal details.

[0019] Furthermore, between step S1 and step S2, there is also a step: Judging whether the denoising effect is satisfied through the signal-to-noise ratio SNR index. If the denoising effect is not satisfied, return to step S1. If the denoising effect is satisfied, enter step S2; where the calculation of the signal-to-noise ratio SNR index is specifically: ; where represents the original reference signal, represents the denoised and reconstructed signal.

[0020] As can be seen from the above description, a signal-to-noise ratio SNR iterative verification mechanism is introduced to ensure that the denoising effect meets the standard before modal identification. This closed-loop design avoids parameter deviation caused by incomplete single-time denoising in traditional methods. When the standard is not met, the preprocessing effect can be optimized by adjusting the threshold, ensuring the accuracy of subsequent identification steps.

[0021] Please refer to Figure 9 , a power system low-frequency oscillation mode identification terminal, including a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the following steps are implemented: S1. Obtain the power system low-frequency oscillation noisy signal and preprocess the noisy signal through an improved adaptive wavelet threshold algorithm; where the adaptive wavelet threshold algorithm uses an improved wavelet threshold dynamically adjusted based on the noise standard deviation and data length; S2. Construct a Hankle matrix and perform singular value decomposition, calculate the cumulative contribution rate of singular values based on PCA, and determine the optimal modal order of the TLS-ESPRIT algorithm; S3. Use TLS-ESPRIT to partition the signal space into subspaces and determine the oscillation mode parameters by solving the eigenvalues of the rotation factor.

[0022] As can be seen from the above description, the beneficial effects of the present invention are as follows: A low-frequency oscillation mode identification terminal of a power system according to the present invention uses a new method combining improved wavelet threshold denoising and PCA-TLS-ESPRIT method for low-frequency oscillation mode identification in a power system. Through the integration of improved wavelet threshold denoising and PCA-TLS-ESPRIT, adaptive preprocessing of noisy signals and data-driven order determination are achieved, which can effectively suppress noise interference and accurately identify low-frequency oscillation characteristic parameters. Compared with traditional methods, the anti-interference performance and identification accuracy are improved, and accurate order determination can be achieved, having good application prospects in low-frequency oscillation early warning of power systems.

[0023] Further, step S2 includes the steps: S21. Construct a Hankel matrix according to the denoised free oscillation data ; H ; ; where , the number of rows of matrix H is L, and the number of columns is M; S22. Decompose matrix using the singular value method: ; where the subscript S and the subscript N correspond to the signal space and the residual noise space respectively, U is an L-order matrix, V is an M-order matrix, represents its conjugate transpose, matrix , the diagonal element value of Σ is the th singular value of H, where the numerical value of the singular value in the signal space is much larger than the numerical value of the singular value in the residual noise space; S23. Arrange the singular values in descending order. Taking PCA as the order determination index, when the influence caused by the current factors is greater than the threshold , at this time the corresponding value is the modal order : ; ; ; where m represents the total number of singular values.

[0024] As described above, the PCA order determination steps based on the Hankle matrix and singular value decomposition are clarified, and the optimal order is automatically determined by calculating the cumulative contribution rate of singular values. Compared with the traditional TLS-ESPRIT manual order determination (such as the order of 10), this method only requires 6 orders to accurately identify the dominant modes in an 8-machine 36-node system, reducing the computational complexity and the interference of false modes, and ensuring the reliability of modal parameter extraction.

[0025] Further, step S3 includes the steps: S31. According to the order determine the signal space , and divide it into two interleaved subspaces: ; S32. Construct a new matrix , and perform singular value decomposition: ; S33. Divide into 4 square matrices of order P, and obtain the rotation factor : ; is expressed as follows: ; S34. Solve the oscillation modal parameters of the signal according to the eigenvalues of the rotation factor: ; where is the low-frequency oscillation frequency; is the oscillation attenuation factor; is the damping ratio.

[0026] As described above, the TLS-ESPRIT subspace division and rotation factor solution processes are refined, and through the construction of interleaved subspaces and singular value decomposition, high-precision calculations of the oscillation frequency, attenuation factor, and damping ratio are achieved.

[0027] Further, step S1 includes the steps: S11. Obtain the low-frequency oscillation noisy signal of the power system, perform wavelet transform processing on it, and obtain the wavelet coefficients at each scale; where j represents the decomposition scale and k represents the kth wavelet coefficient; S12. Use the improved wavelet threshold to perform noise reduction preprocessing on the noisy signal, and perform signal reconstruction to obtain the denoised signal; The improved wavelet threshold is dynamically adjusted based on the noise standard deviation and the data length: ; ; wherein, σ represents the standard deviation of the noise, N represents the data length of the signal, denotes that the wavelet coefficient takes the median value.

[0028] As can be seen from the above description, the specific formula of the improved wavelet threshold is defined, and the standard deviation σ of the noise is calculated by the median method, realizing the dynamic adjustment of the threshold. Compared with the fixed-threshold hard / soft denoising, this method can effectively suppress the noise interference while retaining the signal details.

[0029] Furthermore, between step S1 and step S2, there is also a step: Judging whether the denoising effect is satisfied through the signal-to-noise ratio SNR index. If the denoising effect is not satisfied, return to step S1; if the denoising effect is satisfied, enter step S2; wherein, the calculation of the signal-to-noise ratio SNR index is specifically: ; wherein, represents the original reference signal, represents the denoised and reconstructed signal.

[0030] As can be seen from the above description, a signal-to-noise ratio SNR iterative verification mechanism is introduced to ensure that the denoising effect meets the standard before performing modal identification. This closed-loop design avoids parameter deviation caused by incomplete single-time denoising in traditional methods, and can optimize the preprocessing effect by adjusting the threshold when not meeting the standard, ensuring the accuracy of subsequent identification steps.

[0031] A method and terminal for identifying low-frequency oscillation modes of a power system according to the present invention are applicable to the identification of low-frequency oscillation modes of a power system.

[0032] Please refer to Figure 1 and Figure 2 , Example 1 of the present invention is: A method for identifying low-frequency oscillation modes of a power system, comprising the steps of: S1. Obtain a low-frequency oscillation noisy signal of the power system, and preprocess the noisy signal by an improved adaptive wavelet threshold algorithm; wherein, the adaptive wavelet threshold algorithm adopts an improved wavelet threshold dynamically adjusted based on the standard deviation of the noise and the data length; Step S1 includes the steps of: S11. Obtain a low-frequency oscillation noisy signal of the power system, perform wavelet transform processing on it to obtain wavelet coefficients at each scale ; where j represents the decomposition scale, and k represents the k-th wavelet coefficient; S12. Denoise and preprocess the noisy signal using an improved wavelet threshold, and reconstruct the signal to obtain the denoised signal; Among them, the improved wavelet threshold is dynamically adjusted based on the noise standard deviation and data length: ; ; Among them, σ represents the standard deviation of the noise, N represents the data length of the signal, means taking the median of the wavelet coefficients.

[0033] There is also a step between step S1 and step S2: Judge whether the denoising effect is satisfied through the signal-to-noise ratio SNR index. If the denoising effect is not satisfied, return to step S1. If the denoising effect is satisfied, enter step S2; Among them, the calculation of the signal-to-noise ratio SNR index is specifically: ; Among them, represents the original reference signal, represents the denoised and reconstructed signal.

[0034] S2. Construct a Hankle matrix and perform singular value decomposition. Calculate the cumulative contribution rate of singular values based on PCA to determine the optimal modal order of the TLS-ESPRIT algorithm; Step S2 includes the following steps: S21. According to the denoised free oscillation data Construct a Hankle matrix H ; ; Among them, , the matrix H has L rows and M columns; S22. Use the singular value method to decompose the matrix : ; Among them, the subscript S and the subscript N correspond to the signal space and the residual noise space respectively. U is an L-order matrix, V is an M-order matrix, [[ID=, represents its conjugate transpose, the matrix , the diagonal element value in Σ is the th singular value of H, where the numerical value of the singular value in the signal space is much larger than the numerical value of the singular value in the residual noise space; S23. Arrange the singular values in descending order. Using PCA as the order determination index, the current factors have an impact greater than the threshold At this time The corresponding value is the modal order : ; ; ; Among them, m represents the total number of singular values

[0035] S3. Use TLS-ESPRIT to partition the signal space into subspaces, and determine the oscillation mode parameters by solving the eigenvalues of the rotation factor; Step S3 includes the steps: S31. Determine the signal space according to the order , and divide it into two interleaved subspaces: ; S32. Construct a new matrix , and perform singular value decomposition: ; S33. Divide into 4 square matrices of order P, and find the rotation factor : ; It is expressed as follows: ; S34. Solve the oscillation mode parameters of the signal according to the eigenvalues of the rotation factor: ; Among them, is the low-frequency oscillation frequency; is the oscillation attenuation factor; is the damping ratio

[0036] Please refer to Figures 3 - 8 , Example 2 of the present invention is: A method for identifying low-frequency oscillation modes of a power system, which is different from Example 1 in that in this example, a simulation test is provided as evidence for a method for identifying low-frequency oscillation modes of a power system: Construct a low-frequency oscillation noisy signal, which contains a double oscillation mode with frequencies of 0.5 Hz and 1 Hz respectively. The signal is as follows: .

[0037] Since the low-frequency oscillation signal of the actual power system is often mixed with noise interference, 14 dB of Gaussian white noise is added throughout the process. The waveform diagram of the noisy signal is asFigure 3 As shown. The improved wavelet threshold denoising, fixed threshold hard denoising, and fixed threshold soft denoising are respectively used to process the noisy signal, and the comparison diagram of the denoising effect is as Figure 4 shown. The SNR after denoising by the three preprocessing methods is shown in Table 1. By observing the comparison curve of the denoising effect and the data in the table, it can be clearly seen that the improved wavelet threshold denoising method adopted in this paper can better remove noise interference and retain the details of the low-frequency oscillation signal more intact.

[0038] Table 1 Comparison of Denoising Effects

[0039] The improved algorithm of this paper is used to identify the oscillation mode characteristic parameters of the preprocessed signal, and the identification results of the characteristic parameters are compared with those of the Prony method and the traditional TLS-ESPRIT method. The obtained test results are shown in Table 2.

[0040] Table 2 Comparison of Identification Results of Noisy Signals by Multiple Methods

[0041] It can be seen from the data in the table that the errors of the frequency and attenuation factor obtained by the improved method of the present invention are the smallest compared with the actual values, and a high parameter identification accuracy can be achieved in the noisy background; in contrast, the frequency identification errors of the Prony method for the two oscillation modes reach 0.0005 and 0.0311 Hz respectively, and the attenuation factor identification errors reach 0.0241 and 0.0849 respectively, and the identification effect is not ideal; although the TLS-ESPRIT method has a certain anti-interference ability, there is still a certain deviation from the actual value. In the key order determination of the algorithm, the orders of the three methods are 16, 10, and 4 respectively, and the order determination number of the method proposed in this paper is the smallest, which can effectively reduce the influence of false modes on the identification results. It can be seen that the method proposed in this paper has better performance than the other two methods in the oscillation parameter identification.

[0042] To further illustrate the effectiveness of the method of this paper, Figure 5 the fitting curves of the parameter identification results of each method are given, that is, by observing the closeness of each fitting curve to the original signal, to determine whether the identification result is accurate. It can be seen from the figure that the method of this paper can achieve a better fitting effect, and is closer to the original signal than the other two methods, further verifying the effectiveness of the identification method proposed in this paper.

[0043] For the multi-machine and multi-point system: The wiring diagram of the EPRI-8 machine 36-node system in the PSASP software is shown in Figure 6As shown in the figure. The following disturbances are set: A three-phase short-circuit fault with a duration of 0.06 s occurs between node 19 and node 30, and taking G1 as the reference machine, the power angle curve of G5 relative to G1 is collected as the input oscillation signal. To simulate the possible noise interference in the actual signal, Gaussian white noise with a mean of 0 and a variance of 0.6 is added to it, and the noisy signal is as shown in Figure 7 shown. The calculation results of the eigenvalues under this disturbance are shown in Table 3. It can be seen from the table that 0.7775 Hz and 0.9802 Hz are the two dominant oscillation modes corresponding to the input signal.

[0044] Table 3 Calculation Results of PSASP Eigenvalues

[0045] Perform improved wavelet threshold denoising preprocessing on the noisy power angle oscillation input signal. The waveform after denoising is as shown in Figure 8 shown. It can be seen from the figure that the noise interference is significantly suppressed.

[0046] Use the PCA-TLS-ESPRIT method to obtain the oscillation mode characteristic parameters of the denoised signal. Table 4 gives the results of parameter identification of the power angle oscillation curve using the Prony method, the traditional TLS-ESPRIT method, and the method of the present invention respectively.

[0047] Table 4 Comparison of Identification Results of Multiple Methods

[0048] In the power system, a mode with a damping ratio less than 5% is generally regarded as the dominant oscillation mode. From the parameter identification results in Table 4, it can be seen that the frequencies of the two dominant oscillation modes identified by the method of this paper are 0.7759 and 0.9803 Hz respectively, which are consistent with 0.7775 and 0.9802 Hz in the system eigenvalue calculation results in Table 4; while the Prony method has a large error in the identification of the frequency and damping ratio of mode 1, and the order reaches 20, increasing the calculation amount and prone to modal aliasing; the traditional TLS-ESPRIT method is still slightly insufficient in the identification accuracy of frequency and damping ratio compared with the method of this paper. It can be seen that the method proposed in this paper is effective and accurate in a multi-machine system.

[0049] Please refer to Figure 9 , and the third embodiment of the present invention is as follows: A power system low-frequency oscillation mode identification terminal 1 includes a processor 2, a memory 3, and a computer program stored in the memory 3 and executable on the processor 2. When the processor 2 executes the computer program, it implements the steps in a power system low-frequency oscillation mode identification method of the above-mentioned embodiment 1 or 2.

[0050] In summary, a method and a terminal for identifying low-frequency oscillation modes in a power system provided by the present invention use a new method combining improved wavelet threshold denoising and the PCA-TLS-ESPRIT method for identifying low-frequency oscillation modes in a power system. The method proposed by the present invention can effectively suppress noise interference and accurately identify low-frequency oscillation characteristic parameters, improving the anti-interference performance and identification accuracy compared with traditional methods, and can achieve accurate order determination, having a good application prospect in the early warning of low-frequency oscillations in a power system.

[0051] The above are only embodiments of the present invention, and do not limit the patent scope of the present invention accordingly. Any equivalent transformation made using the content of the specification and drawings of the present invention, or directly or indirectly applied in related technical fields, shall equally be included in the patent protection scope of the present invention.

Claims

1. A method for identifying low-frequency oscillation modes of a power system, characterized in that Including the steps: S1. Obtain the noisy signal of low-frequency oscillation in the power system, and preprocess the noisy signal through an improved adaptive wavelet threshold algorithm; wherein, the adaptive wavelet threshold algorithm adopts an improved wavelet threshold dynamically adjusted based on the noise standard deviation and the data length; S2. Construct a Hankle matrix and perform singular value decomposition, calculate the cumulative contribution rate of singular values based on PCA, and determine the optimal modal order of the TLS-ESPRIT algorithm; S3. Use TLS-ESPRIT to partition the signal space into subspaces, and determine the oscillation mode parameters by solving the eigenvalues of the rotation factor.

2. The method for identifying low-frequency oscillation modes of a power system according to claim 1, wherein Step S2 includes the steps: S21. Construct a Hankel matrix based on the free oscillation data after noise reduction H ;​ ； Among them, , the matrix H has L rows and M columns; S22. Decompose the matrix using the singular value method: ； where the subscripts S and N correspond to the signal space and the residual noise space respectively, U is an L-order matrix, V is an M-order matrix, denotes its conjugate transpose, and the matrix , the diagonal element values in Σ are the th singular value of H, where the numerical values of the singular values in the signal space are much larger than those in the residual noise space; S23. Arrange the singular values in descending order. Using PCA as the order determination index, when the influence caused by the current number of factors is greater than the threshold , at this time the corresponding value is the modal order : ； ； ； Among them, m represents the total number of singular values.

3. A method for identifying low-frequency oscillation modes of a power system according to claim 1, characterized in that, Step S3 includes the steps: S31. Determine the signal space according to the order and divide it into two interleaved subspaces: ​ ； S32. Construct a new matrix , and perform singular value decomposition: ； S33. Divide into four square matrices of order P, and find the rotation factor : ； It is shown as follows: ； S34. Solve the oscillation mode parameters of the signal according to the eigenvalues of the rotation factor: ； wherein, is the low-frequency oscillation frequency; is the oscillation attenuation factor; is the damping ratio.

4. A method for identifying low-frequency oscillation modes of a power system according to claim 1, characterized in that, Step S1 includes the steps: S11. Obtain the noisy signal of low-frequency oscillation in the power system, perform wavelet transform processing on it, and obtain the wavelet coefficients at each scale ; where j represents the decomposition scale and k represents the k-th wavelet coefficient; S12. Perform denoising preprocessing on the noisy signal using the improved wavelet threshold, and perform signal reconstruction to obtain the denoised signal; wherein, the improved wavelet threshold is dynamically adjusted based on the noise standard deviation and the data length: ； ； where, σ represents the standard deviation of the noise, and N represents the data length of the signal, indicates that the wavelet coefficients take the median value.

5. A method for identifying low-frequency oscillation modes of a power system according to claim 1, characterized in that, There is also a step between Step S1 and Step S2: Judge whether the denoising effect is satisfied through the signal-to-noise ratio SNR index. If the denoising effect is not satisfied, return to Step S1. If the denoising effect is satisfied, enter Step S2; wherein, the calculation of the signal-to-noise ratio SNR index is specifically: ； Among them, represents the original reference signal, represents the denoised and reconstructed signal.

6. A power system low-frequency oscillation mode identification terminal, comprising a processor, a memory, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, the following steps are implemented: S1. Obtain the noisy signal of low-frequency oscillation in the power system, and preprocess the noisy signal through an improved adaptive wavelet threshold algorithm; wherein, the adaptive wavelet threshold algorithm adopts an improved wavelet threshold dynamically adjusted based on the noise standard deviation and the data length; S2. Construct a Hankle matrix and perform singular value decomposition, calculate the cumulative contribution rate of singular values based on PCA, and determine the optimal modal order of the TLS-ESPRIT algorithm; S3. Use TLS-ESPRIT to partition the signal space into subspaces, and determine the oscillation mode parameters by solving the eigenvalues of the rotation factor.

7. The low-frequency oscillation mode identification terminal of a power system according to claim 6, characterized in that, Step S2 includes the steps: S21. Construct a Hankel matrix based on the free oscillation data after noise reduction ; H ; ； Among them, , the matrix H has L rows and M columns; S22. Use the singular value method to decompose the matrix as follows: ； where the subscripts S and N correspond to the signal space and the residual noise space respectively, U is an L-order matrix, V is an M-order matrix, denotes its conjugate transpose, and the matrix , and the diagonal element values in Σ are the th singular value of H, where the numerical values of the singular values in the signal space are much larger than those in the residual noise space; S23. Arrange the singular values in descending order. Using PCA as the order determination index, when the influence caused by the current number of factors is greater than the threshold , at this time the corresponding value is the modal order : ； ； ； Among them, m represents the total number of singular values.

8. A power system low-frequency oscillation mode identification terminal according to claim 6, characterized in that, Step S3 includes the steps: S31. Determine the signal space according to the order and divide it into two interleaved subspaces: ​ ； S32. Construct a new matrix , and perform singular value decomposition: ； S33. Divide into four square matrices of order P, and find the rotation factor : ； It is represented as follows: ； S34. Solve the oscillation mode parameters of the signal according to the eigenvalues of the rotation factor: ； wherein, is the low-frequency oscillation frequency; is the oscillation decay factor; is the damping ratio.

9. The power system low-frequency oscillation mode identification terminal according to claim 6, characterized in that, Step S1 includes the steps: S11. Obtain the low-frequency oscillation noisy signal of the power system, perform wavelet transform processing on it, and obtain the wavelet coefficients at each scale ; where j represents the decomposition scale and k represents the k-th wavelet coefficient; S12. Perform denoising preprocessing on the noisy signal using the improved wavelet threshold, and perform signal reconstruction to obtain the denoised signal; wherein, the improved wavelet threshold is dynamically adjusted based on the noise standard deviation and the data length: ； ； where, σ represents the standard deviation of the noise, and N represents the data length of the signal, denotes that the wavelet coefficients take the median value.

10. A low-frequency oscillation mode identification terminal for a power system according to claim 6, characterized in that, There is also a step between Step S1 and Step S2: Judge whether the denoising effect is satisfied through the signal-to-noise ratio SNR index. If the denoising effect is not satisfied, return to Step S1. If the denoising effect is satisfied, enter Step S2; wherein, the calculation of the signal-to-noise ratio SNR index is specifically: ； Among them, represents the original reference signal, represents the signal after denoising and reconstruction.

Images