A data window optimization method for minimizing the amplitude weighted fitting error of wide-area measurement points

Abstract

The data window optimization method with the goal of minimizing the amplitude weighted fitting error of wide-area measuring points includes: detecting the fluctuation of the injected power of each plant and calculating the fluctuation amplitude, and the maximum amplitude corresponds to the initial frequency; at the current moment, the initial frequency is used to obtain the current data window and perform Prony analysis on the injected power to obtain the oscillation mode combination and oscillation mode decomposition error as the multi-measurement point amplitude weighted fitting error; the comparison result with the preset error limit is used for the current moment t cur The error of weighted fitting of amplitude of multiple measuring points is corrected, and the data window set of this period is determined according to the moment obtained after correction; the maximum moving sampling step is used as the constraint condition, and the minimum correction value of weighted fitting error of amplitude of multiple measuring points is used as the optimization goal, and the optimal data window is obtained in the data window set of this period. The present invention realizes stable online oscillation grouping, supports the power grid control center to conduct panoramic observation of the power grid oscillation process, and provides a basis for decision-making and post-analysis of power grid oscillation events.

Description

Technical Field

[0001] The present invention relates to the technical field of power system safety and stability analysis, and more specifically, to a data window optimization method with the goal of minimizing the amplitude weighted fitting error of wide-area measuring points. Background Art

[0002] Low-frequency oscillation is one of the classic problems of power systems. Many studies have proposed algorithms for low-frequency oscillation analysis. However, there are still deficiencies in the engineering of online monitoring and analysis of low-frequency oscillations. Existing oscillation monitoring applications use spectrum analysis algorithms such as FFT, Prony, and EMD to calculate the amplitude, frequency, and other information of the dominant oscillation mode online. When the calculated mode meets the preset amplitude threshold, a low-frequency oscillation alarm is issued. However, due to the presence of various interference components in the operation of the power grid, there are multiple oscillation modes in an oscillation process, and the dominant oscillation mode changes over time, making it difficult to accurately perform real-time spectrum analysis. The modal parameters obtained during the continuous analysis process are unstable, making it difficult to use for online display and analysis of the occurrence, development process, and inter-regional impact of low-frequency oscillations in the wide-area power grid, which is not conducive to the dispatching and operation personnel to accurately understand the power grid oscillation situation and make rapid disposal decisions.

[0003] Prior art 1 (CN101447676B) "A method for analyzing low-frequency oscillations in a power system" uses a high-performance digital filter to filter raw data; uses a short-time Fourier algorithm to identify the spectrum and dominant oscillation of the filtered data; and then uses the Prony algorithm to analyze the oscillation mode and identify the dominant oscillation.

[0004] Prior art 2 (CN110311392A) "A method for identifying power system oscillation modes and modes based on SPDMD" uses the time amplitude coefficient to characterize the energy weight corresponding to each oscillation mode, and screens out oscillation modes and oscillation modes with larger energy weights; transforms the matrix composed of the first N-1 columns of the measured data sequence to calculate the initial amplitude coefficient, introduces the penalty function term on the Frobenius norm term and combines ADMM to determine the position of the non-zero elements inside the initial amplitude coefficient, encodes the initial amplitude coefficient in the matrix, and uses the La-type multiplier method to adjust the numerical value of its internal non-zero elements to obtain the optimal amplitude coefficient; determines the dominant oscillation mode through the corresponding relationship between the non-zero elements in the optimal amplitude coefficient and each oscillation mode, and completes the coordinated identification and evaluation of the dominant oscillation mode, the dominant participation factor and the coherent group. The present invention realizes the identification of power system oscillation modes and modes, and makes up for the defect that the DMD algorithm cannot determine the number of dominant oscillation modes.

[0005] Therefore, it is necessary to study a data window optimization method that aims to minimize the error of wide-area multi-point amplitude weighted fitting. Summary of the invention

[0006] In order to solve the deficiencies in the prior art, the purpose of the present invention is to provide a data window optimization method with the goal of minimizing the amplitude weighted fitting error of wide-area measuring points. The method can obtain the analysis result with the minimum overall error of the oscillation mode of the injected power of the nodes in the whole network within each analysis step under the condition of controllable calculation amount, and apply the oscillation mode available in the engineering to analyze the oscillation situation of the whole network, thereby realizing stable online oscillation grouping during the oscillation process, supporting the operation analysts of the power grid control center to conduct panoramic observation of the power grid oscillation process, assisting in determining the location of the oscillation source, and providing a basis for the decision-making and post-analysis of power grid oscillation event handling.

[0007] The present invention adopts the following technical solution.

[0008] A data window optimization method for minimizing the weighted fitting error of wide-area measurement point amplitudes, the method comprising:

[0009] Step 1, obtaining the injected power of each plant and station into the power grid;

[0010] Step 2, detecting the fluctuation of the power injected by each plant and station, and when it is detected that the fluctuation of the power injected by at least one plant and station exceeds a preset limit, calculating the fluctuation amplitude of the power injected by each plant and station;

[0011] Step 3, sort the fluctuation amplitudes of the injected power of each plant from large to small, select the injected power and fluctuation amplitudes of the first n plants, and record the frequency of the plant with the largest amplitude as the initial frequency f0;

[0012] Step 4: According to the current time t cur , use the initial frequency f0 to obtain the current data window;

[0013] Step 5, using the current data window to perform Prony analysis on the injected power to obtain an oscillation mode combination of the injected power; based on the oscillation mode combination, the error of the oscillation mode decomposition is calculated;

[0014] Step 6, using the product of the error of oscillation mode decomposition and the normalized amplitude as the multi-measurement point amplitude weighted fitting error;

[0015] Step 7, based on the current time t cur and the current data window, and compare the multi-point amplitude weighted fitting error with the preset error limit to calculate the current time t cur The error of weighted fitting of amplitudes of multiple measuring points is corrected, and the data window set of this period is determined according to the corrected time.

[0016] Step 8, taking the maximum moving sampling step length as the constraint condition and the minimum multi-point amplitude weighted fitting error correction value as the optimization goal, obtain the optimal data window in the data window set of this period.

[0017] Preferably, step 1 comprises:

[0018] Step 1.1, forming a monitoring equipment set with the primary equipment of the power grid equipped with a phasor measurement device; the monitoring equipment set includes: a plant station, an AC line, a transformer, and a generator;

[0019] Step 1.2, with a fixed sampling period T chk , from the real-time database of the monitoring platform, with an initial window length T def Get the current active power of each device in the monitoring device set; where the fixed sampling period T chk Select an integer second from 1 to 10 seconds; configure and modify the fixed sampling period T through the relational database of the monitoring platform chk ; Among them, the initial window length T def Not less than 10 seconds;

[0020] Step 1.3, use the active power of each device to calculate the injected power of each plant input into the grid.

[0021] Preferably, step 2 comprises:

[0022] Step 2.1, with the initial window length T def The power injection of each plant is tested for fluctuations, and the window length is an integer of 1 to 60 seconds. For low-frequency oscillations, the initial window length T def Not less than 10 seconds; configure and modify the initial window length through the relational database of the monitoring platform;

[0023] Step 2.2, when it is detected that at least one injected power fluctuation exceeds a preset limit, the fluctuation amplitude of the injected power of each plant is calculated by using an FFT algorithm; wherein the preset limit value range is 1MW to 10MW.

[0024] Preferably, step 3 comprises:

[0025] Step 3.1, sort the fluctuation amplitude of the injected power of each plant from large to small, select the injected power and fluctuation amplitude of the first n plants, and form the injected power set P = {P1, P2, ... P j …,P n} and amplitude set A={A1,A2,…A j …,A n}, where P j and A j denote the injected power and amplitude of the jth plant, respectively, j = 1, 2, …, n;

[0026] Step 3.2, calculate the spectrum of the plant with the maximum amplitude A1, and record the frequency corresponding to the maximum amplitude A1 as the initial frequency f0.

[0027] Preferably, step 4 comprises:

[0028] Step 4.1, calculate the length T0 of the period window according to the initial frequency f0, that is

[0029] Step 4.2, take the current time t cur Push forward the length of a period window T0 to get the current data window (t cur -T0,t cur ].

[0030] Preferably, step 5 comprises:

[0031] Step 5.1, using the current data window (t cur -T0,t cur ] for the injection power set P = {P1, P2, ... P j …,P n}Perform Prony analysis to obtain the oscillation mode combination of injected power;

[0032] Step 5.2, use the percentage error formula to calculate the error of the oscillation mode decomposition, and get the error set PE = {PE1, PE2, ... PE j …,PE n}, PE j Represents the error of the oscillation mode decomposition of the j-th plant, j = 1, 2, …, n.

[0033] Preferably, step 6 comprises:

[0034] Step 6.1, normalize the amplitude set A to obtain the normalized amplitude set A nom ={A nom_1 ,A nom_2 ,…A nom_ j…,A nom_n}, A nom_j represents the normalized amplitude of the jth plant, j = 1, 2, ..., n;

[0035] Step 6.2, using the error set PE and the normalized amplitude set A nom The product of the amplitude of multiple measuring points is used as the weighted fitting error PE cur , satisfying the following relationship:

[0036]

[0037] Among them, the multi-point amplitude weighted fitting error PE cur is the current data window (t cur -T0,t cur ] is the error value under .

[0038] Preferably, step 7 comprises:

[0039] Step 7.1: weight the multi-point amplitude fitting error PE cur and the first preset error limit E set1 For comparison,

[0040] Among them, the first preset error limit E set1 1%; configure and modify the first preset error limit E through the monitoring platform relational database set1 ;

[0041] Step 7.2, when PE cur ≤E set1 When the current data window (t cur -T0,t cur ] as a periodic data window; when PE cur >E set1 When the current time t cur Move back i sampling steps respectively to get the i-th correction time t cur-ik , and enter step 7.3, where each sampling step is k, i = 1, 2, ..., m, m ...; and ik is not greater than the fixed sampling period T chk ;

[0042] Step 7.3, using the i-th correction time t cur-ik Repeat steps 4 to 6 to obtain the i-th corrected error PE of the multi-point amplitude weighted fitting cur-ik ;

[0043] Step 7.4, find the minimum value of the correction error from the adjacent correction errors, that is, the minimum value of the correction error PE cur-ik Meet PE cur-(i-1)k ≥PE cur-ik ≤PE cur-(i+1)k ; and the minimum value PE of the corrected error cur-ik and the second preset error limit E set2 For comparison,

[0044] Among them, the second preset error limit E set2 10%; configure and modify the second preset error limit E through the monitoring platform relational database set2 ;

[0045] Step 7.5, when PE cur-ik ≤E set2 , then the i-th correction time t cur-ik The corresponding data window (t cur-ik -T0,t cur-ik ] as the periodic data window; otherwise, return to step 4 and use the i+1th correction time t cur-(i+1)kand initial frequency f0 to obtain the current data window, and repeat steps 5 to 7;

[0046] Step 7.6, after the above steps, the periodic data window set W = {W i}, where W i is the i-th modified time t cur-ik The corresponding periodic data window (t cur-ik -T0,t cur-ik ], periodic data window W i Corresponding correction error PE cur-ik .

[0047] Preferably, step 8 comprises:

[0048] Step 8.1: Configure and modify the maximum sampling step size k through the relational database of the monitoring platform max , where the maximum sampling step length k max The length T0 of the data window shall not exceed 1 cycle;

[0049] Step 8.2, extract the element W of the periodic data window set i The corresponding correction error PE cur-ik , forming the corrected error set

[0050] Step 8.3, when the moving sampling step k reaches the maximum sampling step k max x, when any element in the corrected error set does not exceed the second preset error limit E set2 When , the periodic data window corresponding to the minimum element in the modified error set is taken as the optimal data window; otherwise, there is no optimal data window, and return to step 4.

[0051] Preferably, step 8 further includes: selecting an optimal data window to perform spectrum analysis on the injected power of each plant station to obtain oscillation mode results of the injected power of all plants stations.

[0052] The beneficial effect of the present invention lies in that, compared with the prior art, the present invention optimizes the data window for low-frequency oscillation modal decomposition by taking the minimum error of wide-area multi-measurement-point amplitude weighted fitting as the goal, and obtains the analysis data window with the minimum error of wide-area power grid plant power injection oscillation modal decomposition within the analysis step range. The analysis window is used to analyze the oscillation mode of the entire network, and the analysis result with the minimum overall error of the oscillation mode of the power injection of the entire network nodes within each analysis step can be obtained under the condition of controllable calculation amount. The oscillation mode available in the engineering is used to track the oscillation situation of the entire network, thereby realizing stable online oscillation clustering during the oscillation process, and supporting and assisting in the identification of oscillation sources.

[0053] The power injected into wide-area power stations is initially screened according to amplitude and continuous oscillation modal decomposition is performed. The optimal data window for wide-area power grid oscillation modal analysis is determined with the goal of minimizing the amplitude-weighted fitting error. This achieves wide-area optimal online oscillation modal decomposition and accurately identifies the impact of multiple types of oscillation events on the power grid, such as ultra-low frequency, low frequency, medium and high frequency. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 It is a flowchart of the steps of a data window optimization method with the goal of minimizing the weighted fitting error of wide-area measurement points amplitude. DETAILED DESCRIPTION

[0055] The present application is further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present application.

[0056] like Figure 1 A data window optimization method is provided with the goal of minimizing the weighted fitting error of wide-area measurement point amplitudes. The method comprises steps 1 to 8.

[0057] Step 1: Obtain the injected power of each plant and station into the power grid.

[0058] Preferably, step 1 comprises:

[0059] Step 1.1, forming a monitoring equipment set with primary equipment of the power grid equipped with a phasor measurement device; the monitoring equipment set includes but is not limited to: power plants, AC lines, transformers and generators;

[0060] Step 1.2, with a fixed sampling period T chk , from the real-time database of the monitoring platform, with an initial window length T def Get the current active power of each device in the monitoring device set; where the fixed sampling period T chk Select an integer number of seconds from 1 to 10 seconds and fix the sampling period T chk Preferably 1 second; configure and modify the fixed sampling period T through the relational database of the monitoring platform chk ; Among them, the initial window length T def Not less than 10 seconds;

[0061] Step 1.3, use the active power of each device to calculate the injected power of each plant input into the grid.

[0062] Step 2: Detect the fluctuation of the power injected by each plant and station. When it is detected that the fluctuation of the power injected by at least one plant and station exceeds the preset limit, calculate the fluctuation amplitude of the power injected by each plant and station.

[0063] Preferably, step 2 comprises:

[0064] Step 2.1, with the initial window length T def The power injection of each plant is tested for fluctuations, and the window length is an integer of 1 to 60 seconds. For low-frequency oscillations, the initial window length T def Not less than 10 seconds; configure and modify the initial window length through the relational database of the monitoring platform. For low-frequency oscillation, the initial window length T def Preferably 10 seconds;

[0065] Step 2.2, when it is detected that at least one injected power fluctuation exceeds a preset limit, the fluctuation amplitude of the injected power of each plant is calculated by using an FFT algorithm; wherein the preset limit value range is 1MW to 10MW.

[0066] Step 3: sort the fluctuation amplitude of the injected power of each plant from large to small, select the injected power and fluctuation amplitude of the first n plants, and record the frequency of the plant with the largest amplitude as the initial frequency f0.

[0067] Preferably, step 3 comprises:

[0068] Step 3.1, sort the fluctuation amplitude of the injected power of each plant from large to small, select the injected power and fluctuation amplitude of the first n plants, and form the injected power set P = {P1, P2, ... P j …,P n} and amplitude set A={A1,A2,…A j …,A n}, where P j and A j denote the injected power and amplitude of the jth plant, respectively, j = 1, 2, …, n;

[0069] Step 3.2, calculate the spectrum of the plant with the maximum amplitude A1, and record the frequency corresponding to the maximum amplitude A1 as the initial frequency f0.

[0070] Step 4: According to the current time t cur , use the initial frequency f0 to obtain the current data window.

[0071] Preferably, step 4 comprises:

[0072] Step 4.1, calculate the length T0 of the period window according to the initial frequency f0, that is

[0073] Step 4.2, take the current time t cur Push forward the length of a period window T0 to get the current data window (t cur -T0,t cur ].

[0074] Step 5, using the current data window to perform Prony analysis on the injected power, to obtain the oscillation mode combination of the injected power; based on the oscillation mode combination, the error of the oscillation mode decomposition is calculated.

[0075] Preferably, step 5 comprises:

[0076] Step 5.1, using the current data window (t cur -T0,t cur ] for the injection power set P = {P1, P2, ... P j …,P n}Perform Prony analysis to obtain the oscillation mode combination of injected power;

[0077] Step 5.2, use the percentage error formula to calculate the error of the oscillation mode decomposition, and get the error set PE = {PE1, PE2, ... PE j …,PE n}, PE j Represents the error of the oscillation mode decomposition of the j-th plant, j = 1, 2, …, n.

[0078] Step 6: Use the product of the error of oscillation mode decomposition and the normalized amplitude as the multi-measurement point amplitude weighted fitting error.

[0079] Preferably, step 6 comprises:

[0080] Step 6.1, normalize the amplitude set A to obtain the normalized amplitude set A nom ={A nom_1 ,A nom_2 ,…A nom_ j…,A nom_n}, A nom_j represents the normalized amplitude of the jth plant, j = 1, 2, ..., n;

[0081] Step 6.2, using the error set PE and the normalized amplitude set A nom The product of the amplitude of multiple measuring points is used as the weighted fitting error PE cur , satisfying the following relationship:

[0082]

[0083] Among them, the multi-point amplitude weighted fitting error PE cur is the current data window (t cur -T0,t cur ] is the error value under .

[0084] Step 7, based on the current time t cur and the current data window, and compare the multi-point amplitude weighted fitting error with the preset error limit to calculate the current time tcur The error of weighted fitting of amplitudes of multiple measuring points is corrected, and the data window set of this period is determined according to the corrected time.

[0085] Preferably, step 7 comprises:

[0086] Step 7.1: weight the multi-point amplitude fitting error PE cur and the first preset error limit E set1 For comparison,

[0087] Among them, the first preset error limit E set1 Preferably 1%; configure and modify the first preset error limit E through the monitoring platform relational database set1 ;

[0088] Step 7.2, when PE cur ≤E set1 When the current data window (t cur -T0,t cur ] as a periodic data window; when PE cur >E set1 When the current time t cur Move back i sampling steps respectively to get the i-th correction time t cur-ik , and enter step 7.3, where each sampling step is k, i = 1, 2, ..., m, m ...; and ik is not greater than the fixed sampling period T chk ;

[0089] Step 7.3, using the i-th correction time t cur-ik Repeat steps 4 to 6 to obtain the i-th corrected error PE of the multi-point amplitude weighted fitting cur-ik ;

[0090] Step 7.4, find the minimum value of the correction error from the adjacent correction errors, that is, the minimum value of the correction error PE cur-ik Meet PE cur-(i-1)k ≥PE cur-ik ≤PE cur-(i+1)k ; and the minimum value PE of the corrected error cur-ik and the second preset error limit E set2 For comparison,

[0091] Among them, the second preset error limit E set2 Preferably 10%; configure and modify the second preset error limit E through the monitoring platform relational database set2 ;

[0092] Step 7.5, when PE cur-ik ≤E set2 , then the i-th correction time t cur-ik The corresponding data window (tcur-ik -T0,t cur-ik ] as the periodic data window; otherwise, return to step 4 and use the i+1th correction time t cur-(i+1)k and initial frequency f0 to obtain the current data window, and repeat steps 5 to 7;

[0093] Step 7.6, after the above steps, the periodic data window set W = {W i}, where W i is the i-th modified time t cur-ik The corresponding periodic data window (t cur-ik -T0,t cur-ik ], periodic data window W i Corresponding correction error PE cur-ik .

[0094] Step 8, taking the maximum moving sampling step length as the constraint condition and the minimum multi-point amplitude weighted fitting error correction value as the optimization goal, obtain the optimal data window in the data window set of this period.

[0095] Preferably, step 8 comprises:

[0096] Step 8.1: Configure and modify the maximum sampling step size k through the relational database of the monitoring platform max , where the maximum sampling step length k max The length of the data window is no more than 1 cycle T0; where the maximum sampling step length k max Preferably, it is 5, and the maximum step size that can be set is the step size corresponding to the length T0 of one cycle data window.

[0097] Step 8.2, extract the element W of the periodic data window set i The corresponding correction error PE cur-ik , forming the corrected error set

[0098] Step 8.3, when the moving sampling step k reaches the maximum sampling step k ma x, when any element in the corrected error set does not exceed the second preset error limit E set2 When , the periodic data window corresponding to the minimum element in the modified error set is taken as the optimal data window; otherwise, there is no optimal data window, and return to step 4.

[0099] Preferably, step 8 further includes: selecting an optimal data window to perform spectrum analysis on the injected power of each plant station to obtain oscillation mode results of the injected power of all plants stations.

[0100] The data window optimization method module proposed by the present invention with the goal of minimizing the wide-area multi-measurement point amplitude weighted fitting error is run on the dispatching automation system platform, which receives the voltage phasor, current phasor and power collected and sent by the broadband measurement device located in the plant station through the data network at a rate of 50 frames per second, and stores them in the platform's time series real-time library and time series history library. Taking the monitoring of 0.1Hz to 1.0Hz low-frequency oscillation signal as an example, it is explained to call the data window optimization method of the present invention with the goal of minimizing the wide-area multi-measurement point amplitude weighted fitting error to perform real-time online oscillation mode analysis.

[0101] The beneficial effect of the present invention lies in that, compared with the prior art, the present invention optimizes the data window for low-frequency oscillation modal decomposition by taking the minimum error of wide-area multi-measurement-point amplitude weighted fitting as the goal, and obtains the analysis data window with the minimum error of wide-area power grid plant power injection oscillation modal decomposition within the analysis step range. The analysis window is used to analyze the oscillation mode of the entire network, and the analysis result with the minimum overall error of the oscillation mode of the power injection of the entire network nodes within each analysis step can be obtained under the condition of controllable calculation amount. The oscillation mode available in the engineering is used to track the oscillation situation of the entire network, thereby realizing stable online oscillation clustering during the oscillation process, and supporting and assisting in the identification of oscillation sources.

[0102] The power injected into wide-area power stations is initially screened according to amplitude and continuous oscillation modal decomposition is performed. The optimal data window for wide-area power grid oscillation modal analysis is determined with the goal of minimizing the amplitude-weighted fitting error. This achieves wide-area optimal online oscillation modal decomposition and accurately identifies the impact of multiple types of oscillation events on the power grid, such as ultra-low frequency, low frequency, medium and high frequency.

[0103] The applicant of the present invention has made a detailed explanation and description of the implementation examples of the present invention in conjunction with the drawings in the specification. However, those skilled in the art should understand that the above implementation examples are only preferred implementation schemes of the present invention, and the detailed description is only to help readers better understand the spirit of the present invention, but not to limit the scope of protection of the present invention. On the contrary, any improvements or modifications based on the inventive spirit of the present invention should fall within the scope of protection of the present invention.

Claims

1. A data window optimization method with the goal of minimizing the weighted fitting error of wide-area measurement point amplitudes, characterized in that: The method comprises: Step 1, obtaining the injected power of each plant and station into the power grid; Step 2, detecting the fluctuation of the power injected by each plant and station, and when it is detected that the fluctuation of the power injected by at least one plant and station exceeds the preset limit, calculating the fluctuation amplitude of the power injected by each plant and station; Step 3, sort the fluctuation amplitudes of the injected power of each plant from large to small, select the injected power and fluctuation amplitudes of the first n plants, and record the frequency of the plant with the largest amplitude as the initial frequency f0; Step 4: According to the current time t cur , use the initial frequency f0 to obtain the current data window; Step 5, using the current data window to perform Prony analysis on the injected power to obtain an oscillation mode combination of the injected power; based on the oscillation mode combination, the error of the oscillation mode decomposition is calculated; Step 6, using the product of the error of oscillation mode decomposition and the normalized amplitude as the multi-measurement point amplitude weighted fitting error; Step 7, based on the current time t cur and the current data window, and compare the multi-point amplitude weighted fitting error with the preset error limit to calculate the current time t cur The error of weighted fitting of amplitudes of multiple measuring points is corrected, and the data window set of this period is determined according to the corrected time. Step 8, taking the maximum moving sampling step length as the constraint condition and the minimum multi-point amplitude weighted fitting error correction value as the optimization goal, obtain the optimal data window in the data window set of this period.

2. The data window optimization method according to claim 1, wherein the method aims at minimizing the wide-area measurement point amplitude weighted fitting error, Step 1 includes: Step 1.1, forming a monitoring equipment set with the primary equipment of the power grid equipped with a phasor measurement device; the monitoring equipment set includes: a plant station, an AC line, a transformer, and a generator; Step 1.2, with a fixed sampling period T chk , from the real-time database of the monitoring platform, with an initial window length T def Get the current active power of each device in the monitoring device set; where the fixed sampling period T chk Select an integer second from 1 to 10 seconds; configure and modify the fixed sampling period T through the relational database of the monitoring platform chk ; Among them, the initial window length T def Not less than 10 seconds; Step 1.3, use the active power of each device to calculate the injected power of each plant input into the grid.

3. The data window optimization method with the goal of minimizing the wide-area measurement point amplitude weighted fitting error according to claim 2 is characterized in that: Step 2 includes: Step 2.1, with the initial window length T def The power injection of each plant is tested for fluctuations, and the window length is an integer of 1 to 60 seconds. For low-frequency oscillations, the initial window length T def Not less than 10 seconds; configure and modify the initial window length through the relational database of the monitoring platform; Step 2.2, when it is detected that at least one injected power fluctuation exceeds a preset limit, the fluctuation amplitude of the injected power of each plant is calculated by using an FFT algorithm; wherein the preset limit ranges from 1MW to 10MW.

4. The data window optimization method with the goal of minimizing the wide-area measurement point amplitude weighted fitting error according to claim 3 is characterized in that: Step 3 includes: Step 3.1, sort the fluctuation amplitude of the injected power of each plant from large to small, select the injected power and fluctuation amplitude of the first n plants, and form the injected power set P = {P1, P2, ... P j …, P n } and amplitude set A = {A1, A2, ... A j …, A n }, where P j and A j denote the injected power and amplitude of the jth plant, respectively, j = 1, 2, ..., n; Step 3.2, calculate the spectrum of the plant with the maximum amplitude A1, and record the frequency corresponding to the maximum amplitude A1 as the initial frequency f0.

5. The data window optimization method with the goal of minimizing the wide-area measurement point amplitude weighted fitting error according to claim 4 is characterized in that: Step 4 includes: Step 4.1, calculate the length T0 of the period window according to the initial frequency f0, that is Step 4.2, take the current time t cur Push forward the length of a period window T0 to get the current data window (t cur -T0,t cur ].

6. The data window optimization method with the goal of minimizing the wide-area measurement point amplitude weighted fitting error according to claim 5, characterized in that: Step 5 includes: Step 5.1, using the current data window (t cur -T0,t cur ] for the injection power set P = {P1, P2, ... P j …, P n }Perform Prony analysis to obtain the oscillation mode combination of injected power; Step 5.2, use the percentage error formula to calculate the error of the oscillation mode decomposition, and get the error set PE = {PE1, PE2, ...PE j …, PE n }, PE j Represents the error of the oscillation modal decomposition of the j-th plant, j = 1, 2, …, n.

7. The data window optimization method according to claim 6, wherein the method is characterized in that: Step 6 includes: Step 6.1, normalize the amplitude set A to obtain the normalized amplitude set A nom ={A nom_1 , A nom_2 , …A nom_j …, A nom_n }, A nom_j represents the normalized amplitude of the jth plant, j = 1, 2, ..., n; Step 6.2, using the error set PE and the normalized amplitude set A nom The product of the amplitude of multiple measuring points is used as the weighted fitting error PE cur , satisfying the following relationship: Among them, the multi-point amplitude weighted fitting error PE cur is the current data window (t cur -T0,t cur ] under the error value.

8. The data window optimization method according to claim 7, wherein the method aims at minimizing the wide-area measurement point amplitude weighted fitting error, Step 7 includes: Step 7.1: weight the multi-point amplitude fitting error PE cur and the first preset error limit E set1 For comparison, Among them, the first preset error limit E set1 1%; configure and modify the first preset error limit E through the monitoring platform relational database set1 ; Step 7.2, when PE cur ≤E set1 When the current data window (t cur -T0,t cur ] as a periodic data window; when PE cur >E set1 When the current time t cur Move back i sampling steps respectively to get the i-th correction time t cur-ik , and enter step 7.3, where each sampling step is k, i = 1, 2, ..., m, m ...; and ik is not greater than the fixed sampling period T chk ; Step 7.3, using the i-th correction time t cur-ik Repeat steps 4 to 6 to obtain the i-th corrected error PE of the multi-point amplitude weighted fitting cur-ik ; Step 7.4, find the minimum value of the correction error from the adjacent correction errors, that is, the minimum value of the correction error PE cur-ik Meet PE cur-(i-1)k ≥PE cur-ik ≤PE cur-(i+1)k ; and the minimum value PE of the corrected error cur-ik and the second preset error limit E set2 For comparison, Among them, the second preset error limit E set2 10%; configure and modify the second preset error limit E through the monitoring platform relational database set2 ; Step 7.5, when PE cur-ik ≤E set2 , then the i-th correction time t cur-ik The corresponding data window (t cur-ik -T0,t cur-ik ] as the periodic data window; otherwise, return to step 4 and use the i+1th correction time t cur-(i+1)k and initial frequency f0 to obtain the current data window, and repeat steps 5 to 7; Step 7.6, after the above steps, the periodic data window set W = {W i }, where W i is the i-th modified time t cur-ik The corresponding periodic data window (t cur-ik -T0,t cur-ik ], periodic data window W i Corresponding correction error PE cur-ik .

9. The data window optimization method with the goal of minimizing the wide-area measurement point amplitude weighted fitting error according to claim 8, characterized in that: Step 8 includes: Step 8.1: Configure and modify the maximum sampling step size k through the relational database of the monitoring platform max , where the maximum sampling step length k max The length T0 of the data window shall not exceed 1 cycle; Step 8.2, extract the element W of the periodic data window set i The corresponding correction error PE cur-ik , forming the corrected error set Step 8.3, when the moving sampling step k reaches the maximum sampling step k max Under the constraint condition of set2 When , the periodic data window corresponding to the minimum element in the modified error set is taken as the optimal data window; otherwise, there is no optimal data window, and return to step 4.

10. The data window optimization method with the goal of minimizing the wide-area measurement point amplitude weighted fitting error according to claim 9, characterized in that: Step 8 also includes: selecting an optimal data window to perform spectrum analysis on the injected power of each plant station to obtain the oscillation mode results of the injected power of all plants stations.

Images