Low-frequency oscillation dominant mode identification method and system, and storage medium

Abstract

The application discloses a low-frequency oscillation dominant mode identification method and system and a storage medium. The method comprises the following steps: S1, acquiring an active power curve of a line, performing EMD decomposition on the active power curve, and removing high-frequency and direct-current components; S2, performing Prony analysis on the active power curve from which the high-frequency and direct-current components are removed, obtaining initial values of parameters of each oscillation mode, and selecting a mode with the largest oscillation energy as a dominant mode; S3, determining an adjustment direction of a certain parameter of the dominant mode; S4, taking the initial value of the parameter obtained by the Prony analysis as a starting point, adjusting the parameter according to the adjustment direction and a set step length, and calculating a correlation coefficient between an adjusted dominant mode curve and an original active power curve; S5, repeating step S4 until the correlation coefficient between the adjusted dominant mode curve and the original active power curve reaches a maximum value; S6, replacing the parameter, and repeating steps S3 to S5 until all parameters that need to be corrected are corrected. The above method can more accurately obtain the low-frequency oscillation dominant mode parameters of a system.

Description

Technical Field

[0001] This invention belongs to the field of power system automation technology, and specifically relates to a method, system and storage medium for identifying low-frequency oscillation dominant modes. Background Technology

[0002] Low-frequency oscillations refer to the relative oscillation of rotors between parallel-operated generator units when an interconnected power system is disturbed, causing oscillations in system power or power angle. The oscillation frequency is generally between 0.1 and 2.5 Hz. With the continuous expansion of interconnected power grids, the widespread application of fast excitation systems, and the increasing line loads, the probability of low-frequency oscillations in the power grid is increasing. Low-frequency oscillations have become one of the important factors restricting the safe and stable operation of the power grid. If relevant information on low-frequency oscillations is not correctly grasped and corresponding measures are not taken in a timely manner, it may cause the interconnected system to disconnect, or even a major blackout.

[0003] The commonly used analysis method for low-frequency oscillations in power systems is the Prony algorithm. Prony is a "global" algorithm that uses a linear combination of a set of exponential terms to fit equally spaced sampled data, thereby obtaining the amplitude, frequency, phase, and damping ratio of the low-frequency oscillation signal. However, the Prony algorithm is extremely sensitive to noise, which can affect the accuracy of parameter identification, especially the accuracy of the damping ratio, thus affecting the dispatchers' judgment of the risk of system oscillations. Summary of the Invention

[0004] Purpose of the invention: The purpose of this invention is to propose a method for identifying low-frequency oscillation dominant modes, which can improve the identification accuracy of low-frequency oscillation dominant mode parameters and reduce the impact of noise.

[0005] Another object of the present invention is to provide a low-frequency oscillation-dominant pattern recognition system capable of implementing the above-described recognition method, and a storage medium storing a computer program containing the recognition method.

[0006] Technical solution: The low-frequency oscillation dominant mode recognition method of the present invention includes the following steps:

[0007] S1: Obtain the active power curve of the line and perform EMD decomposition on the active power curve to remove high-frequency components and DC components.

[0008] S2: Perform Prony analysis on the active power curve after removing high-frequency and DC components to obtain the initial parameter values ​​of each oscillation mode, calculate the oscillation energy of each oscillation mode, and select the mode with the largest energy as the dominant mode.

[0009] S3: Determine the adjustment direction of a certain parameter in the dominant mode;

[0010] S4: Starting with the initial value of this parameter obtained from the Prony analysis, adjust the parameter according to the adjustment direction and set step size, and calculate the correlation coefficient between the adjusted dominant mode curve and the original active power curve.

[0011] S5: Repeat step S4 until the correlation coefficient between the adjusted dominant mode curve and the original active power curve reaches its maximum value.

[0012] S6: Change the parameters and repeat steps S3 to S5 until all parameters that need to be corrected are corrected.

[0013] Furthermore, the parameters in step S3 include phase and damping ratio.

[0014] Furthermore, step S1 includes:

[0015] S1.1: Obtain the active power curve of the line and perform EMD decomposition on the active power curve to obtain m IMF components and one residual component.

[0016] S1.2: Calculate the oscillation frequency of m IMF components, add up the IMF components whose oscillation frequency is within the low-frequency threshold range, and obtain the active power curve after removing the high-frequency component and DC component.

[0017] Furthermore, step S3 includes:

[0018] S3.1: Calculate the correlation coefficient ρ1 between the unadjusted dominant mode curve and the original active power curve;

[0019] S3.2: Increase the initial value of the selected parameter by a set step size, and calculate the correlation coefficient ρ between the adjusted dominant mode curve and the original active power curve. 1+ ;

[0020] S3.3: Calculate ρ 1+ -ρ1, if the difference is greater than 0, the adjustment direction is to increase; if the difference is less than 0, the adjustment direction is to decrease.

[0021] Furthermore, the correlation coefficient ρ between the adjusted dominant mode curve and the original active power curve is calculated using the following formula:

[0022]

[0023] In the formula, N is the number of sampling points for the curve; m i n i These are the values ​​at point i on the original active power curve and the dominant mode curve, respectively; μ m μ n These are the mean values ​​of the original active power curve and the dominant mode curve, respectively; σ m σ nThese are the standard deviations of the original active power curve and the dominant mode curve, respectively.

[0024] Furthermore, the low-frequency threshold range is 0.1-2.5Hz.

[0025] Furthermore, in step S2, the oscillation energy E of the j-th oscillation mode j It is obtained by calculation using the following formula:

[0026]

[0027] In the formula: n is the number of sampling points for the j-th oscillation mode; A i Δt is the fitted value of the i-th sampling point; Δt is the sampling time of the j-th oscillation mode.

[0028] The low-frequency oscillation dominant mode identification method of the present invention includes: a decomposition module for performing EMD decomposition on the active power curve of the line to remove high-frequency components and DC components; a preliminary analysis module for using Prony analysis to obtain the initial parameter values ​​of the dominant mode of the active power curve after removing high-frequency components and DC components; and a correction module for adjusting the initial parameter values ​​of the dominant mode and calculating the correlation coefficient between the adjusted dominant mode curve and the original active power curve, with the parameter value at which the correlation coefficient reaches its maximum value being the determining parameter value of the low-frequency oscillation dominant mode.

[0029] The storage medium of the present invention stores a computer program, characterized in that the computer program is configured to implement the above-mentioned low-frequency oscillation dominant mode recognition method when executed.

[0030] Beneficial effects: Compared with the prior art, the present invention has the following advantages: by adjusting the dominant mode parameters identified by the Prony method according to a certain step size, and by calculating the correlation coefficient between the adjusted dominant mode curve and the original active power curve, and using the parameter value when the correlation coefficient is the largest as the final dominant mode determination parameter, the accuracy of the low-frequency oscillation information obtained by identification is improved, and the influence of noise on the identification accuracy is removed. Attached Figure Description

[0031] Figure 1 This is a flowchart of the low-frequency oscillation dominant mode recognition method according to the first embodiment of the present invention;

[0032] Figure 2 This is a flowchart of the low-frequency oscillation dominant mode recognition method according to the second embodiment of the present invention. Detailed Implementation

[0033] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0034] Reference Figure 1The low-frequency oscillation dominant pattern recognition method according to an embodiment of the present invention includes the following steps:

[0035] S1: Obtain the active power curve of the line and perform EMD decomposition on the active power curve to remove high-frequency components and DC components.

[0036] S2: Perform Prony analysis on the active power curve after removing high-frequency and DC components to obtain the initial parameter values ​​of each oscillation mode, calculate the oscillation energy of each oscillation mode, and select the mode with the largest energy as the dominant mode.

[0037] S3: Determine the adjustment direction of a certain parameter in the dominant mode;

[0038] S4: Starting with the initial value of this parameter obtained from the Prony analysis, adjust the parameter according to the adjustment direction and set step size, and calculate the correlation coefficient between the adjusted dominant mode curve and the original active power curve.

[0039] S5: Repeat step S4 until the correlation coefficient between the adjusted dominant mode curve and the original active power curve reaches its maximum value.

[0040] S6: Change the parameters and repeat steps S3 to S5 until all parameters that need to be corrected are corrected.

[0041] The above technical solution removes high-frequency and DC components from the original active power curve through EMD decomposition, eliminating a large number of stray components and improving the accuracy of Prony analysis. Simultaneously, based on the initial parameter values ​​from the Prony analysis results, the initial parameter values ​​are adjusted according to set step values, and the correlation coefficient between the adjusted dominant mode curve and the original active power curve is calculated to determine if the adjustment direction is correct. When the correlation coefficient reaches its maximum value, the adjusted parameters are closest to the actual parameter values. Through this method, the Prony analysis results are further corrected, improving the accuracy of identifying the dominant mode parameters of low-frequency oscillation signals, overcoming the Prony algorithm's sensitivity to noise, and improving the accuracy of dispatchers' judgment of system oscillation risks.

[0042] Reference Figure 2 In practice, the amplitude and frequency of the dominant mode analyzed by Prony are less affected by noise and can be considered to be no different from the actual amplitude and frequency values. Therefore, generally only the phase and damping ratio analyzed by Prony are corrected.

[0043] The correlation coefficient ρ between the dominant mode curve and the original active power curve is calculated by the following formula:

[0044]

[0045] In the formula, N is the number of sampling points for the curve; mi n i These are the values ​​at point i on the original active power curve and the dominant mode curve, respectively; μ m μ n These are the mean values ​​of the original active power curve and the dominant mode curve, respectively; σ m σ n These represent the standard deviations of the original active power curve and the dominant mode curve, respectively.

[0046] When the correlation coefficient between the dominant mode curve and the original active power curve is at its maximum, it can be considered that the dominant mode curve can reflect the oscillation characteristics of the original active power curve to the greatest extent. Therefore, during correction, the parameters analyzed by Prony need to be corrected in the direction of increasing correlation coefficient. The correction process involves first increasing the initial value of the parameter to be corrected by a set step size, and then calculating the correlation coefficient ρ1 between the unadjusted dominant mode curve and the original active power curve, and the correlation coefficient ρ2 between the adjusted dominant mode curve and the original active power curve. 1+ And calculate ρ 1+ -ρ1, if the difference is greater than 0, the adjustment direction is to increase; if the difference is less than 0, the adjustment direction is to decrease. In this embodiment, the phase adjustment step size is preferably 0.1 rad, and the damping ratio adjustment step size is preferably 0.1%.

[0047] In practice, step S1 includes:

[0048] S1.1: Obtain the active power curve of the line and perform EMD decomposition on the active power curve to obtain m IMF components and one residual component.

[0049] S1.2: Calculate the oscillation frequency of m IMF components, add up the IMF components whose oscillation frequency is within the low-frequency threshold range, and obtain the active power curve after removing the high-frequency component and DC component.

[0050] The preferred low-frequency threshold range is 0.1-2.5Hz.

[0051] In step S2, the oscillation energy E of the j-th oscillation mode j It is obtained by calculation using the following formula:

[0052]

[0053] In the formula: n is the number of sampling points for the j-th oscillation mode; A i Δt is the fitted value of the i-th sampling point; Δt is the sampling time of the j-th oscillation mode.

[0054] Since the correct correction direction for the parameters that need to be corrected has been determined in step S3, the difference ρ between the correlation coefficient of the dominant mode curve after the i-th correction and the original curve and the correlation coefficient calculated after the (i-1)-th correction is... i -ρ i-1 When the value is ≤0, it means that the parameter is corrected in the determined adjustment direction, and the correlation coefficient begins to decrease. The parameter after the (i-1)th correction is the parameter that is closest to the true value, which is a parameter that more accurately reflects the low-frequency oscillation of the system.

[0055] The low-frequency oscillation dominant mode identification system of this invention includes a decomposition module, a preliminary analysis module, and a correction module. The decomposition module performs EMD decomposition on the active power curve of the line to remove high-frequency and DC components. The preliminary analysis module uses Prony analysis to obtain initial parameter values ​​for the dominant mode of the active power curve after removing high-frequency and DC components. The correction module adjusts the initial parameter values ​​of the dominant mode and calculates the correlation coefficient between the adjusted dominant mode curve and the original active power curve. The parameter value at which the correlation coefficient reaches its maximum value is used as the determined parameter value for the low-frequency oscillation dominant mode. The storage medium of this invention stores a computer program instantiated from the above-described low-frequency oscillation dominant mode identification method.

Claims

1. A method for recognizing low-frequency oscillation dominant patterns, characterized in that, Includes the following steps: S1: Obtain the active power curve of the line and perform EMD decomposition on the active power curve to remove high-frequency components and DC components. S2: Perform Prony analysis on the active power curve after removing high-frequency and DC components to obtain the initial parameter values ​​of each oscillation mode, calculate the oscillation energy of each oscillation mode, and select the mode with the largest energy as the dominant mode. S4: Determine the adjustment direction of a specific parameter in the dominant mode; including: S4.1: Calculate the correlation coefficient ρ1 between the unadjusted dominant mode curve and the original active power curve; S4.2: Increase the initial value of the selected parameter by a set step size, and calculate the correlation coefficient ρ between the adjusted dominant mode curve and the original active power curve. 1+ ; S4.3: Calculate ρ 1+ -ρ1, if the difference is greater than 0, the adjustment direction is to increase; if the difference is less than 0, the adjustment direction is to decrease; the parameters include phase and damping ratio; S5: Starting with the initial value of this parameter obtained from Prony analysis, adjust the parameter according to the adjustment direction and set step size, and calculate the correlation coefficient between the adjusted dominant mode curve and the original active power curve. S6: Repeat step S5 until the correlation coefficient between the adjusted dominant mode curve and the original active power curve reaches its maximum value. S7: Change the parameters and repeat steps S4 to S6 until all parameters that need to be corrected are corrected.

2. The low-frequency oscillation dominant mode recognition method according to claim 1, characterized in that, Step S1 includes: S1.1: Obtain the active power curve of the line and perform EMD decomposition on the active power curve to obtain m IMF components and one residual component. S1.2: Calculate the oscillation frequency of m IMF components, add up the IMF components whose oscillation frequency is within the low-frequency threshold range, and obtain the active power curve after removing the high-frequency component and DC component.

3. The low-frequency oscillation dominant mode recognition method according to claim 1, characterized in that, The correlation coefficient ρ between the adjusted dominant mode curve and the original active power curve is calculated by the following formula: In the formula, N is the number of sampling points for the curve; m i n i These are the values ​​at point i on the original active power curve and the dominant mode curve, respectively; μ m μ n These are the mean values ​​of the original active power curve and the dominant mode curve, respectively; σ m σ n These are the standard deviations of the original active power curve and the dominant mode curve, respectively.

4. The low-frequency oscillation dominant mode recognition method according to claim 1, characterized in that, The low-frequency threshold range is 0.1-2.5Hz.

5. The low-frequency oscillation dominant mode recognition method according to claim 1, characterized in that, In step S2, the oscillation energy E of the j-th oscillation mode j It is obtained by calculation using the following formula: In the formula: n is the number of sampling points for the j-th oscillation mode; A i Δt is the fitted value of the i-th sampling point; Δt is the sampling time of the j-th oscillation mode.

6. A low-frequency oscillation-dominant pattern recognition system, characterized in that, include: Decomposition module: Used to perform EMD decomposition on the active power curve of the line, removing high-frequency components and DC components; Preliminary Analysis Module: Used to obtain initial parameter values ​​for the dominant mode of the active power curve after removing high-frequency and DC components using Prony analysis; The correction module adjusts the initial parameter values ​​of the dominant mode and calculates the correlation coefficient between the adjusted dominant mode curve and the original active power curve. The parameter value at which the correlation coefficient reaches its maximum value is used to determine the parameter values ​​for the low-frequency oscillation dominant mode. This includes: Calculate the correlation coefficient ρ1 between the unadjusted dominant mode curve and the original active power curve; increase the initial value of the selected parameter by a set step size, and calculate the correlation coefficient ρ1 between the adjusted dominant mode curve and the original active power curve. 1+ ; Calculate ρ 1+ -ρ1, if the difference is greater than 0, the adjustment direction is to increase; if the difference is less than 0, the adjustment direction is to decrease; the parameters include phase and damping ratio; 7. A storage medium storing a computer program, characterized in that, The computer program is designed to implement the low-frequency oscillation dominant pattern recognition method according to any one of claims 1 to 5 at runtime.

Images