A non-synchronous monitoring data screening method for interval harmonic state estimation

Abstract

The application provides a non-synchronous monitoring data screening method for interval harmonic state estimation, selects a certain time sequence length, constructs a measurement time sequence of maximum and minimum values of harmonic monitoring data, screens suspicious bad data in the maximum and minimum values of harmonic monitoring data by analyzing the autocovariance variation law of the maximum and minimum values of harmonic monitoring data, calculates a correlation coefficient matrix of the maximum and minimum values of harmonic monitoring data in a current sampling period and a last sampling period, finds out a strong correlation data set of suspicious bad data in the two periods through the correlation coefficient matrix and compares the change conditions of the two data sets, thereby screening bad data in the suspicious bad data, removes the bad data in the harmonic monitoring data and constructs an interval harmonic state estimation model, converts the interval harmonic state estimation model into a linear optimization model based on a Taylor expansion formula, solves the linear optimization model by using a linear programming algorithm and obtains a solution interval of a harmonic state variable.

Description

Technical Field

[0001] This invention relates to the field of power system interval harmonic state estimation technology, and in particular to a method for screening asynchronous monitoring data for interval harmonic state estimation. Background Technology

[0002] With the integration of numerous distributed power sources and power electronic devices into the power grid, harmonic pollution in the power system is becoming increasingly severe. To ensure the safe and stable operation of the power grid, harmonic mitigation is necessary, and harmonic state estimation can provide strong technical support for harmonic mitigation.

[0003] The measurement data for interval harmonic state estimation (IHSE) of power systems consists of harmonic monitoring data. Therefore, the quality of harmonic monitoring data is crucial to the accuracy of IHSE. Harmonic monitoring data originates from power quality monitoring devices (PQMDs). PQMDs record the maximum, minimum, 95% probability maximum, and average values ​​of power quality data within a statistical period (typically 3 minutes), and the detection periods of PQMDs at different monitoring points are asynchronous. IHSE selects the maximum and minimum values ​​of harmonic monitoring data within the statistical period to construct the interval harmonic measurement dataset, and also requires selecting a sampling period from a specific PQMD as the estimation period. Therefore, when the system state changes significantly, the asynchronous nature of PQMDs may cause some measurement data in IHSE to deviate from actual values, becoming defective data. With the large-scale integration of renewable energy into the grid, the grid state exhibits strong randomness and volatility, thus limiting the application of harmonic monitoring data in IHSE. Further research is needed to identify defective data in harmonic monitoring data to provide a high-quality measurement dataset for IHSE. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to provide a method for screening asynchronous monitoring data for interval harmonic state estimation, which can accurately screen out bad data in harmonic monitoring data and provide high-quality and sufficiently redundant measurement data for interval harmonic state estimation.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a method for screening asynchronous monitoring data for interval harmonic state estimation, comprising the following steps:

[0006] S1: Construct the measurement time series of the maximum and minimum values ​​of harmonic monitoring data, and calculate the covariance matrix of the maximum and minimum values ​​of harmonic monitoring data respectively;

[0007] S2: Calculate the standardized error of the maximum and minimum values ​​of the harmonic monitoring data respectively;

[0008] S3: Set a threshold for the standardization error, mark the data whose standardization error is greater than the threshold among the maximum and minimum values ​​of the harmonic monitoring data as suspicious bad data, and record their measurement number;

[0009] S4: Calculate the correlation coefficient matrix of the maximum and minimum values ​​of harmonic monitoring data in the current sampling period and the previous sampling period;

[0010] S5: Identify the strongly correlated datasets of the suspected bad data in the current sampling period and the previous sampling period; compare the changes of the strongly correlated datasets of the suspected bad data in the two sampling periods. If the two datasets remain unchanged or change very little, the suspected bad data is considered normal data; if the two datasets change significantly, the suspected bad data is considered bad data.

[0011] S6: Remove bad data from the harmonic monitoring data, convert the maximum and minimum values ​​of the harmonic monitoring data into the form of interval numbers, and thus construct the interval harmonic measurement dataset;

[0012] S7: Take the harmonic voltage amplitude and harmonic voltage phase angle as state variables to construct an interval harmonic state estimation model;

[0013] S8: Convert the interval harmonic state estimation model into a linear optimization model, and solve it using a linear programming algorithm to obtain a new iterative solution interval;

[0014] S9: Set convergence conditions and determine whether the convergence conditions are met. If the convergence conditions are met, end the calculation and output the interval harmonic state estimation results. If the convergence conditions are not met, set k = k + 1 and replace the initial solution interval with the latest iterative solution interval, and re-execute steps S8-S9.

[0015] In a preferred embodiment, step 1 takes the nth sampling time period as an example:

[0016]

[0017] In the formula: q ii (n) represents the autovariance of the i-th measurement data during the sampling period n; i (k) represents the time series of the i-th measurement data; k represents the time scale; q represents the average value of the i-th measurement over the time series length; ij (n) represents the covariance between the i-th and j-th measurement data in the sampling period n; E represents the expected value; M represents the total number of measurement data; N represents the length of the time series, and N = 100.

[0018] In a preferred embodiment, in step 2, the standardized errors of the maximum and minimum values ​​of the harmonic monitoring data are calculated respectively:

[0019]

[0020] In the formula: z i (n) represents the value of the i-th measurement data in the sampling period n; || represents the absolute value;

[0021] Set the standardization error threshold to 2.5, mark the data with a standardization error greater than 2.5 among the maximum and minimum values ​​of harmonic monitoring data as suspicious bad data, and record their measurement number.

[0022] In a preferred embodiment, in step 4, the correlation coefficient matrix of the maximum and minimum values ​​of the harmonic monitoring data in the current sampling period and the previous sampling period is calculated. Taking sampling period n as an example:

[0023]

[0024] In the formula: p ii (n) represents the diagonal elements of the correlation coefficient matrix during the sampling period n; p ij (n) represents the off-diagonal elements of the correlation coefficient matrix during the sampling period n;

[0025] Identify datasets with strong correlation between suspicious bad data in the current sampling period and the previous sampling period, when |p ij When |>0.7, the i-th measurement data and the j-th measurement data are considered to have a strong correlation, and thus the strongly correlated dataset of suspicious bad data in the two sampling periods is recorded;

[0026] By comparing the changes of the suspected bad data with the strongly correlated datasets in the two sampling periods, if the two datasets remain unchanged or change very little, the suspected bad data is considered normal data; if the two datasets change significantly, the suspected bad data is considered bad data.

[0027] In a preferred embodiment, in step 7, the harmonic voltage amplitude and harmonic voltage phase angle are taken as state variables to construct an interval harmonic state estimation model:

[0028]

[0029] In the formula: x i and denoted by the lower and upper bounds of the i-th state variable; x represents the interval vector of the state variable; h represents the measurement function; z represents the interval harmonic measurement dataset; and st represents the constraint condition.

[0030] In a preferred embodiment, in step 8, the interval harmonic state estimation model is converted into a linear optimization model, and the initial solution interval of the state variables is set to x. (0) At the midpoint of the initial solution interval After Taylor expansion of the measurement function and ignoring higher-order terms, we obtain the following expression:

[0031]

[0032] In the formula: and x (0) x represents (0) The upper and lower bounds of the measurement function, J represents the measurement function at... The Jacobian matrix after the first-order Taylor expansion at ; let Then we have:

[0033]

[0034] In the formula: z and Let z be the lower and upper bounds; let The solution to the model is then transformed into solving two linear optimization problems as follows:

[0035]

[0036]

[0037] The correction amount Δx can be obtained by solving equations (7)-(8) using a linear programming algorithm. The first approximate solution for the upper and lower bounds of the system state variables is then obtained as follows:

[0038]

[0039] Will replace Substituting the corrected equation and repeating the iterative calculation yields an estimated solution range that more closely approximates the true solution. The convergence condition for ending the iteration is set as follows:

[0040]

[0041] In the formula: k represents the number of iterations; ε represents the convergence threshold, which is taken as ε = 10. -3 ;

[0042] If the convergence condition is met, the calculation ends and the interval harmonic state estimation result is output; if the convergence condition is not met, k = k + 1 is set and the latest iterative solution interval is replaced with the initial solution interval to re-iterate and solve.

[0043] Compared with existing technologies, this invention has the following advantages: First, this invention selects a certain time series length to construct measurement time series of the maximum and minimum values ​​of harmonic monitoring data. By analyzing the autovariate variation law of the maximum and minimum values ​​of harmonic monitoring data, suspicious abnormal data in the maximum and minimum values ​​of harmonic monitoring data are screened out. This method can narrow down the screening range of abnormal data in harmonic monitoring data. Second, the correlation coefficient matrix of the maximum and minimum values ​​of harmonic monitoring data in the current sampling period and the previous sampling period is calculated. The strongly correlated datasets of suspicious abnormal data in the two periods are found through the correlation coefficient matrix, and the changes in the two datasets are compared, thereby screening out abnormal data among the suspicious abnormal data. This method can accurately screen out abnormal data in harmonic monitoring data, providing high-quality and sufficiently redundant measurement data for interval harmonic state estimation. Finally, abnormal data in the harmonic monitoring data are removed, and an interval harmonic state estimation model is constructed. Based on the Taylor expansion formula, the interval harmonic state estimation model is transformed into a linear optimization model, which is solved using a linear programming algorithm. The estimation results can provide network state information more accurately. Attached Figure Description

[0044] Figure 1 This is a flowchart of a preferred embodiment of the present invention. Detailed Implementation

[0045] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0046] It should be noted that the following detailed descriptions are illustrative and intended to provide further explanation of this application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0047] It should be noted that the terminology used herein is for the purpose of describing particular implementations only and is not intended to limit the exemplary implementations according to this application; as used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise; furthermore, it should be understood that when the terms “comprising” and / or “including” are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or combinations thereof.

[0048] A method for filtering asynchronous monitoring data for interval harmonic state estimation, referenced Figure 1 This includes the following steps:

[0049] S1: Construct the measurement time series of the maximum and minimum values ​​of harmonic monitoring data, and calculate the covariance matrix of the maximum and minimum values ​​of harmonic monitoring data respectively;

[0050] S2: Calculate the standardized error of the maximum and minimum values ​​of the harmonic monitoring data respectively;

[0051] S3: Set a threshold for the standardization error, mark the data whose standardization error is greater than the threshold among the maximum and minimum values ​​of the harmonic monitoring data as suspicious bad data, and record their measurement number;

[0052] S4: Calculate the correlation coefficient matrix of the maximum and minimum values ​​of harmonic monitoring data in the current sampling period and the previous sampling period;

[0053] S5: Identify the strongly correlated datasets of the suspected bad data in the current sampling period and the previous sampling period; compare the changes of the strongly correlated datasets of the suspected bad data in the two sampling periods. If the two datasets remain unchanged or change very little, the suspected bad data is considered normal data; if the two datasets change significantly, the suspected bad data is considered bad data.

[0054] S6: Remove bad data from the harmonic monitoring data, convert the maximum and minimum values ​​of the harmonic monitoring data into the form of interval numbers, and thus construct the interval harmonic measurement dataset;

[0055] S7: Take the harmonic voltage amplitude and harmonic voltage phase angle as state variables to construct an interval harmonic state estimation model;

[0056] S8: Convert the interval harmonic state estimation model into a linear optimization model, and solve it using a linear programming algorithm to obtain a new iterative solution interval;

[0057] S9: Set convergence conditions and determine whether the convergence conditions are met. If the convergence conditions are met, end the calculation and output the interval harmonic state estimation results. If the convergence conditions are not met, set k = k + 1 and replace the initial solution interval with the latest iterative solution interval, and re-execute steps S8-S9.

[0058] The specific implementation method is as follows:

[0059] Construct time series of the maximum and minimum values ​​of harmonic monitoring data, and calculate the covariance matrices of the maximum and minimum values ​​of the harmonic monitoring data respectively (since the calculation process is consistent, the maximum or minimum value is not emphasized in the formula). Take the nth sampling period as an example:

[0060]

[0061] In the formula: q ii (n) represents the autovariance of the i-th measurement data during the sampling period n; i (k) represents the time series of the i-th measurement data; k represents the time scale; q represents the average value of the i-th measurement over the time series length; ij(n) represents the covariance between the i-th and j-th measurement data in the sampling period n; E represents the expected value; M represents the total number of measurement data; N represents the length of the time series, and N = 100.

[0062] Calculate the standardized errors of the maximum and minimum values ​​of the harmonic monitoring data separately (since the calculation process is consistent, the maximum or minimum value is not emphasized in the formula):

[0063]

[0064] In the formula: z i (n) represents the value of the i-th measurement data in the sampling period n; || represents the absolute value.

[0065] Set the standardization error threshold to 2.5, mark the data with a standardization error greater than 2.5 among the maximum and minimum values ​​of harmonic monitoring data as suspicious bad data, and record their measurement number.

[0066] Calculate the correlation coefficient matrix of the maximum and minimum values ​​of harmonic monitoring data in the current sampling period and the previous sampling period, taking sampling period n as an example (since the calculation process is the same, the maximum or minimum value is not emphasized in the formula):

[0067]

[0068] In the formula: p ii (n) represents the diagonal elements of the correlation coefficient matrix during the sampling period n; p ij (n) represents the off-diagonal elements of the correlation coefficient matrix during the sampling period n.

[0069] Identify datasets with strong correlation between suspicious bad data in the current sampling period and the previous sampling period, when |p ij When |>0.7, the i-th measurement data and the j-th measurement data are considered to have a strong correlation, and thus the strongly correlated dataset of suspicious bad data in the two sampling periods is recorded.

[0070] By comparing the changes of the suspected bad data with the strongly correlated datasets in the two sampling periods, if the two datasets remain unchanged or change very little, the suspected bad data is considered normal data; if the two datasets change significantly, the suspected bad data is considered bad data.

[0071] After removing undesirable data from the harmonic monitoring data, the maximum and minimum values ​​of the harmonic monitoring data are converted into interval numbers, thereby constructing an interval harmonic measurement dataset. Using harmonic voltage amplitude and harmonic voltage phase angle as state variables, an interval harmonic state estimation model is constructed:

[0072]

[0073] In the formula: x i and denoted by the lower and upper bounds of the i-th state variable; x represents the interval vector of the state variable; h represents the measurement function; z represents the interval harmonic measurement dataset; and st represents the constraint condition.

[0074] The interval harmonic state estimation model is transformed into a linear optimization model, and the initial solution interval of the state variables is x. (0) At the midpoint of the initial solution interval After Taylor expansion of the measurement function and ignoring higher-order terms, we obtain the following expression:

[0075]

[0076] In the formula: and x (0) x represents (0) The upper and lower bounds of the measurement function, J represents the measurement function at... Let be the Jacobian matrix obtained by performing a first-order Taylor expansion at . Then we have:

[0077]

[0078] In the formula: z and Let z denote the lower and upper bounds of z. The solution to the model is then transformed into solving two linear optimization problems as follows:

[0079]

[0080]

[0081] The correction amount Δx can be obtained by solving equations (7)-(8) using a linear programming algorithm. The first approximate solution for the upper and lower bounds of the system state variables is then obtained as follows:

[0082]

[0083] Will replace Substituting the corrected equation and repeating the iterative calculation yields an estimated solution range that more closely approximates the true solution. The convergence condition for ending the iteration is set as follows:

[0084]

[0085] In the formula: k represents the number of iterations; ε represents the convergence threshold, which is taken as ε = 10. -3 .

[0086] If the convergence condition is met, the calculation ends and the interval harmonic state estimation result is output; if the convergence condition is not met, k = k + 1 is set and the latest iterative solution interval is replaced with the initial solution interval to re-iterate and solve.

Claims

1. A method for screening asynchronous monitoring data for interval harmonic state estimation, characterized in that... Includes the following steps: S1: Construct the measurement time series of the maximum and minimum values ​​of harmonic monitoring data, and calculate the covariance matrix of the maximum and minimum values ​​of harmonic monitoring data respectively; S2: Calculate the standardized error of the maximum and minimum values ​​of the harmonic monitoring data respectively; S3: Set a threshold for the standardization error. Mark data whose standardization error for the maximum and minimum values ​​of harmonic monitoring data is greater than the threshold as suspicious or defective data and record their measurement number. S4: Calculate the correlation coefficient matrix p of the maximum and minimum values ​​of harmonic monitoring data in the current sampling period and the previous sampling period. ij ; S5: Identify datasets with strong correlation between suspicious bad data in the current sampling period and the previous sampling period, when |p ij When |>0.7, the i-th measurement data and the j-th measurement data are considered to have a strong correlation. Therefore, the strongly correlated dataset of the suspected defective data in the two sampling periods is recorded. By comparing the changes of the strongly correlated dataset of the suspected defective data in the two sampling periods, it is determined whether the suspected defective data is normal data or defective data. S6: Remove bad data from the harmonic monitoring data, convert the maximum and minimum values ​​of the harmonic monitoring data into the form of interval numbers, and thus construct the interval harmonic measurement dataset; S7: Take the harmonic voltage amplitude and harmonic voltage phase angle as state variables to construct an interval harmonic state estimation model; S8: Convert the interval harmonic state estimation model into a linear optimization model, and solve it using a linear programming algorithm to obtain a new iterative solution interval; S9: Set convergence conditions, determine whether the convergence conditions are met, and if the convergence conditions are met, end the calculation and output the interval harmonic state estimation results. If the convergence condition is not met, set k=k+1 and replace the initial solution interval with the latest iterative solution interval, then re-execute steps S8-S9; In step 7, the harmonic voltage amplitude and harmonic voltage phase angle are taken as state variables to construct an interval harmonic state estimation model: （4） In the formula: and represents the lower and upper bounds of the i-th state variable; x represents the interval vector of the state variable; h represents the measurement function; z represents the interval harmonic measurement dataset; st represents the constraint condition; In step 8, the interval harmonic state estimation model is converted into a linear optimization model, and the initial solution interval of the state variables is set to x. (0) At the midpoint of the initial solution interval After Taylor expansion of the measurement function and ignoring higher-order terms, we obtain the following expression: （5） In the formula: ; and x represents (0) The upper and lower bounds of the measurement function, J represents the measurement function at... The Jacobian matrix after the first-order Taylor expansion at ; let Then we have: （6） In the formula: and Let z be the lower and upper bounds; let , Therefore, solving the model is transformed into solving two linear optimization problems as follows: （7） （8） The correction amount is obtained by solving equations (7)-(8) using a linear programming algorithm. , Therefore, the first approximate solution for the upper and lower bounds of the system state variables is obtained as follows: （9） Will replace Substituting the corrected equation and iterating repeatedly, we obtain an estimated solution interval that more closely approximates the true solution. The convergence condition for ending the iteration is set as follows: （10） In the formula: k represents the number of iterations; This represents the convergence threshold, taken as... =10 -3 ; If the convergence condition is met, the calculation ends and the interval harmonic state estimation result is output; if the convergence condition is not met, k=k+1 is set and the latest iterative solution interval is replaced with the initial solution interval to solve the problem again.

2. The asynchronous monitoring data screening method for interval harmonic state estimation according to claim 1, characterized in that, The construction of the measurement time series of the maximum and minimum values ​​of harmonic monitoring data, and the calculation of the covariance matrices of the maximum and minimum values ​​of the harmonic monitoring data respectively, include: （1） In the formula: q ii (n) represents the autovariance of the i-th measurement data during the sampling period n; i (k) represents the time series of the i-th measurement data; k represents the time scale; q represents the average value of the i-th measurement over the time series length; ij (n) represents the covariance between the i-th and j-th measurement data in the sampling period n; E represents the expected value; M represents the total number of measurement data; N represents the length of the time series, and N=100.

3. The asynchronous monitoring data screening method for interval harmonic state estimation according to claim 1, characterized in that, The standardized errors for calculating the maximum and minimum values ​​of harmonic monitoring data respectively include: （2） In the formula: z i (n) represents the value of the i-th measurement data in the sampling period n; | represents the absolute value; The standardization error threshold is set to 2.

5. Data with a standardization error greater than 2.5 for the maximum and minimum values ​​of harmonic monitoring data are marked as suspicious or defective data, and their measurement numbers are recorded.

4. The asynchronous monitoring data screening method for interval harmonic state estimation according to claim 1, characterized in that, The correlation coefficient matrix p of the maximum and minimum values ​​of the harmonic monitoring data in the current sampling period and the previous sampling period is calculated. ij include: （3） In the formula: n is the sampling time period, p ii (n) represents the diagonal elements of the correlation coefficient matrix during the sampling period n; p ij (n) represents the off-diagonal elements of the correlation coefficient matrix during the sampling period n.

Images