Online clustering method of broadband measurement signals in power grid based on principal component and K-means algorithm

Abstract

The present application discloses an online clustering method for broadband measurement signals of power grids based on principal components and K-means algorithms, including: acquiring broadband measurement signals of power grid equipment in real time and online at a fixed period, and constructing a broadband measurement signal sample matrix X; performing dimensionality reduction processing on the broadband measurement signal sample matrix X through principal component analysis to obtain an input sample matrix Xˊ for cluster analysis; performing cluster optimization and division of the input sample matrix Xˊ according to the topological relationship of power grid plants and equipment to form a cluster set of the input sample matrix Xˊ; and performing online clustering of all elements contained in the cluster set using the K-means algorithm to obtain the distribution of the dominant broadband oscillation mode. The present invention identifies the spatiotemporal distribution of broadband oscillation modes by clustering analysis of broadband measurement signals after dimensionality reduction processing, ensures accurate identification of broadband oscillation events in wide-area power grids, and improves the online perception capability of the harmonic distribution characteristics of power grids.

Description

Technical Field

[0001] The invention belongs to the technical field of power system safety and stability analysis, and relates to an online clustering method for broadband measurement signals of a power grid based on principal components and a K-means algorithm. Background Art

[0002] With the changes in the grid structure and the introduction of a large number of power electronic equipment, the characteristics of the grid are becoming increasingly complex, and the impact of interharmonics and harmonics on the safe operation of the grid is gradually emerging. The resulting broadband oscillation problem seriously threatens the stability of power system operation.

[0003] Power grid dispatching and control centers at all levels have installed broadband measurement devices in key power plants and stations and new energy gathering areas to achieve real-time synchronous collection of 1-300Hz interharmonic phasors and 2nd-50th harmonic phasor broadband signals, which are then sent to the power grid dispatching technical support system of the dispatching and control center for monitoring and analysis.

[0004] Due to the complex mechanism of broadband oscillation, statistical analysis of broadband measurement signals in the power grid dispatching and control center can obtain the key broadband oscillation modes and broadband distribution characteristics of the power grid, and improve the level of broadband oscillation monitoring of the power grid. Traditional statistical methods usually analyze broadband measurement signals through dimensions such as plants, equipment, and oscillation frequency, which requires certain preconditions, and there is little interaction between different dimensions, which is not conducive to operation analysts to grasp the spatiotemporal distribution characteristics of broadband oscillations in the power grid.

[0005] The clustering algorithm method can better mine the rich oscillation information contained in the broadband measurement signal and present the development trend of the power grid oscillation mode changes at multiple levels. However, it does not pre-process the broadband measurement signal according to the oscillation characteristics of the power system, nor does it optimize the clustering calculation process in combination with the power grid topology relationship. It is difficult to adapt to the real-time online analysis requirements and the growing number of broadband monitoring devices. Summary of the invention

[0006] In order to address the deficiencies in the prior art, the present application provides an online clustering method for power grid broadband measurement signals based on principal components and K-means algorithms, so that operation analysts of the power grid control center can make full use of broadband measurement signals to online identify the distribution trend of broadband oscillation modes in the wide-area power grid, obtain key broadband oscillation modes of the power grid in real time, and timely adjust the frequency band range of the currently tracked power grid broadband oscillation events, support dynamic updating and incremental loading of online monitoring frequency bands for power grid oscillation events, ensure panoramic observation of important power grid oscillation events, and facilitate dispatching and operation analysts to fully grasp the harmonic distribution characteristics of the power grid, and provide a basis for power grid oscillation event handling decisions and post-analysis.

[0007] In order to achieve the above object, the present invention adopts the following technical solution:

[0008] The online clustering method of power grid broadband measurement signals based on principal component and K-means algorithm includes the following steps:

[0009] Step 1: Obtain broadband measurement signals of power grid equipment online in real time at a fixed periodicity, and construct a broadband measurement signal sample matrix X;

[0010] Step 2: Perform dimensionality reduction processing on the broadband measurement signal sample matrix X in step 1 through principal component analysis to obtain the input sample matrix Xˊ for cluster analysis;

[0011] Step 3: According to the topological relationship of power grid plants and equipment, the input sample matrix Xˊ is clustered and optimized to form a cluster set of the input sample matrix Xˊ;

[0012] Step 4: Use the K-means algorithm to perform online clustering on all elements contained in the cluster set in step 3 to obtain the distribution of the dominant broadband oscillation mode.

[0013] The present invention further includes the following preferred embodiments:

[0014] Preferably, in step 1, the broadband measurement signal refers to a broadband measurement signal of 1-300 Hz interharmonic phasor and 2-50th harmonic phasor;

[0015] The power grid equipment includes various plants and stations, AC lines, transformers and generators.

[0016] Preferably, step 1 specifically includes:

[0017] Step 1.1: Screen the power grid equipment with broadband measurement signal configuration information to form a set of participating monitoring equipment;

[0018] Step 1.2: Periodically and real-timely acquire broadband measurement signals of the monitoring equipment set;

[0019] Step 1.3: Construct a broadband measurement signal sample matrix X with participating monitoring equipment as rows and interharmonic frequency bands or harmonic orders as columns;

[0020] The number of rows of X is the number of monitoring devices, and the number of columns is the number of interharmonic frequency bands or harmonic orders.

[0021] Preferably, step 1.1 is specifically as follows: selecting the power grid primary equipment models of power plants, AC lines, transformers and generators with 1-300Hz interharmonic phasors and 2-50th harmonic phasor broadband measurement signal configuration information from the real-time library of the dispatching technical support system platform to form a set of participating monitoring equipment.

[0022] Preferably, in step 1.2, the fixed period is 1 second.

[0023] Preferably, step 2 specifically includes:

[0024] Step 2.1: Calculate the mean value of the broadband measurement signal corresponding to each column in the broadband measurement signal sample matrix X;

[0025] Step 2.2: Use the mean calculated in step 2.1 to center the sample matrix X columns to obtain the matrix Z;

[0026] Step 2.3: Use singular value decomposition to find Z T The eigenvalue matrix λ and eigenvector matrix Y of Z;

[0027] Step 2.4: Arrange the eigenvalues ​​in the eigenvalue matrix λ in descending order and select m eigenvalues;

[0028] Step 2.5: Obtain the eigenvectors corresponding to the m eigenvalues ​​in the eigenvector matrix Y, and calculate the reduced-dimensional matrix Xˊ as the input sample matrix Xˊ for cluster analysis.

[0029] Preferably, in step 2.4, the set threshold is configured or modified through a database.

[0030] Preferably, in step 2.4, the m eigenvalues ​​are screened in such a manner that if the ratio of the sum of the first m eigenvalues ​​to the sum of all eigenvalues ​​is greater than a set threshold, the m eigenvalues ​​are selected.

[0031] Preferably, the set threshold is 80%.

[0032] Preferably, step 3 is specifically:

[0033] The sets of monitoring devices corresponding to the rows of each element in each column of the input sample matrix Xˊ are topologically analyzed respectively, and the devices belonging to the same electrical island are classified into a cluster, thereby forming a cluster set of the input sample matrix Xˊ.

[0034] Beneficial effects achieved by this application:

[0035] The present invention can identify the spatiotemporal distribution of broadband oscillation modes by clustering analysis of broadband measurement signals after dimensionality reduction processing, ensure accurate identification of broadband oscillation events in wide-area power grids, and improve the online perception capability of power grid harmonic distribution characteristics.

[0036] The present invention adopts principal component analysis to perform dimensionality reduction preprocessing on the broadband measurement signals of the monitoring equipment, preliminarily screens the monitoring equipment, frequency bands or frequency points involved in cluster analysis, and reduces the number of broadband measurement signal samples for cluster analysis; combined with the power grid topology relationship of the monitoring equipment, the sample clusters with additional physical meanings are divided to carry out K-means algorithm cluster analysis, the initial cluster center selection method is optimized, the cluster analysis efficiency is improved, and the online clustering of broadband measurement signals is realized, which is convenient for operation analysts to timely grasp the changes in the broadband oscillation mode of the power grid, provide information support for the online monitoring of power grid oscillation events, effectively improve the level of broadband oscillation monitoring of the power grid, and effectively ensure the safe and stable operation of the power grid. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 It is a flow chart of the online clustering method of power grid broadband measurement signal based on principal component and K-means algorithm of the present invention;

[0038] Figure 2 It is a flow chart for implementing the online clustering method of broadband measurement signals of power grid based on principal components and K-means algorithm of the present invention. DETAILED DESCRIPTION

[0039] 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.

[0040] When the online clustering method of power grid broadband measurement signals based on principal components and K-means algorithm proposed in the present invention is implemented, it runs on the dispatching technology support system platform of the power grid dispatching control center. The platform receives 1-300 Hz interharmonic phasors and 2-50th harmonic phasor broadband measurement signals collected and uploaded 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.

[0041] Taking the 2nd to 50th harmonic phasor broadband measurement signal as an example, the online clustering method of the broadband measurement signal of the power grid based on the principal component and K-means algorithm of the present invention is described. Figure 1 and 2 As shown, the method of the present invention specifically comprises the following steps:

[0042] Step 1: Obtain broadband measurement signals of power grid equipment online in real time at a fixed periodicity, and construct a broadband measurement signal sample matrix X;

[0043] Step 1.1: Obtain the power grid primary equipment models such as power plants, AC lines, transformers, and generators with 2nd to 50th harmonic signal configuration information from the real-time library of the dispatching technical support system platform to form a set of participating monitoring equipment;

[0044] Step 1.2: Obtain the 2nd to 50th harmonic signal of each second of the participating monitoring equipment set from the real-time library of the scheduling technical support system platform in real time online with a period of 1 second;

[0045] During specific implementation, the period can be configured or modified through the database, and the default period is 1 second.

[0046] This embodiment uses 2-50 harmonics for analysis. Each column of the sample matrix X represents 100 Hz, 150 Hz, 200 Hz, ..., 2500 Hz, and each row represents a device having harmonic signals of each order participating in monitoring. All elements of the sample matrix X are harmonic signals of the devices participating in monitoring.

[0047] Step 1.3: Construct a sample matrix X with participating monitoring devices as rows and 2nd to 50th harmonic signals as columns. The number of rows in X is 120, which is the number of participating monitoring devices, and the number of columns is 49.

[0048] Step 2: Perform dimensionality reduction processing on the broadband measurement signal sample matrix X in step 1 through principal component analysis to obtain the input sample matrix Xˊ for cluster analysis;

[0049] Step 2.1: Calculate the mean value of the broadband measurement signal corresponding to each column in the broadband measurement signal sample matrix X;

[0050] Step 2.2: Use the mean calculated in step 2.1 to center the sample matrix X columns to obtain the matrix Z;

[0051] Step 2.3: Use singular value decomposition (SVD) to find Z T The eigenvalue matrix λ and eigenvector matrix Y of Z;

[0052] Step 2.4: Arrange the eigenvalues ​​in the eigenvalue matrix λ in descending order and select 5 eigenvalues;

[0053] The screening method is: if the ratio of the sum of the first five eigenvalues ​​to the sum of all eigenvalues ​​is greater than the set threshold, then the five eigenvalues ​​are selected;

[0054] During specific implementation, the threshold value may be configured or modified through the database, and the default value is 80%.

[0055] Step 2.5: Obtain the eigenvectors corresponding to the five eigenvalues ​​in the eigenvector matrix Y, and calculate the reduced-dimensional matrix Xˊ as the input sample matrix Xˊ for cluster analysis.

[0056] Step 2 initially screened the monitoring equipment, frequency bands or frequency points involved in cluster analysis.

[0057] Step 3: According to the topological relationship of power grid plants and equipment, the input sample matrix Xˊ is clustered and optimized to form a cluster set of the input sample matrix Xˊ;

[0058] Topological analysis is performed on the sets of monitoring devices corresponding to the rows of each element in each column of the input sample matrix Xˊ, and the devices belonging to the same electrical island are classified into one cluster, thereby forming a cluster set of the input sample Xˊ;

[0059] XˊEach column represents a frequency band or frequency point, and each row represents the equipment involved in monitoring;

[0060] Step 4: Use the K-means algorithm to perform online clustering on all elements contained in the cluster set in step 3 (i.e., all elements of the Xˊ matrix after cluster division) to obtain the distribution of the dominant broadband oscillation mode.

[0061] The present invention can identify the spatiotemporal distribution of broadband oscillation modes by clustering analysis of broadband measurement signals after dimensionality reduction processing, ensure accurate identification of broadband oscillation events in wide-area power grids, and improve the online perception capability of power grid harmonic distribution characteristics.

[0062] The present invention adopts principal component analysis to perform dimensionality reduction preprocessing on the broadband measurement signals of the monitoring equipment, preliminarily screens the monitoring equipment, frequency bands or frequency points involved in cluster analysis, and reduces the number of broadband measurement signal samples for cluster analysis; combined with the power grid topology relationship of the monitoring equipment, the sample clusters with additional physical meanings are divided to carry out K-means algorithm cluster analysis, the initial cluster center selection method is optimized, the cluster analysis efficiency is improved, and the online clustering of broadband measurement signals is realized, which is convenient for operation analysts to timely grasp the changes in the broadband oscillation mode of the power grid, provide information support for the online monitoring of power grid oscillation events, effectively improve the level of broadband oscillation monitoring of the power grid, and effectively ensure the safe and stable operation of the power grid.

[0063] 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. An online clustering method for broadband measurement signals of power grid based on principal component and K-means algorithm, characterized by: The method comprises the following steps: Step 1: Obtain broadband measurement signals of power grid equipment online in real time at a fixed periodicity, and construct a broadband measurement signal sample matrix X; Step 2: Perform dimensionality reduction processing on the broadband measurement signal sample matrix X in step 1 through principal component analysis to obtain the input sample matrix Xˊ for cluster analysis, which specifically includes: Step 2.1: Calculate the mean value of the broadband measurement signal corresponding to each column in the broadband measurement signal sample matrix X; Step 2.2: Use the mean calculated in step 2.1 to center the sample matrix X columns to obtain the matrix Z; Step 2.3: Use singular value decomposition to find Z T The eigenvalue matrix λ and eigenvector matrix Y of Z; Step 2.4: Arrange the eigenvalues ​​in the eigenvalue matrix λ in descending order and select m eigenvalues; The screening method of m eigenvalues ​​is: if the ratio of the sum of the first m eigenvalues ​​to the sum of all eigenvalues ​​is greater than the set threshold, the m eigenvalues ​​are selected; Step 2.5: Obtain the eigenvectors corresponding to the m eigenvalues ​​in the eigenvector matrix Y, and calculate the reduced-dimensional matrix Xˊ as the input sample matrix Xˊ for cluster analysis; Step 3: According to the topological relationship of power grid stations and equipment, the input sample matrix Xˊ is clustered and optimized to form a cluster set of the input sample matrix Xˊ, specifically: Perform topological analysis on the sets of monitoring devices corresponding to the rows of each element in each column of the input sample matrix Xˊ, and classify the devices belonging to the same electrical island into a cluster, thereby forming a cluster set of the input sample matrix Xˊ; Step 4: Use the K-means algorithm to perform online clustering on all elements contained in the cluster set in step 3 to obtain the distribution of the dominant broadband oscillation mode.

2. The online clustering method for power grid broadband measurement signals based on principal component and K-means algorithm according to claim 1, characterized in that: In step 1, the broadband measurement signal refers to the broadband measurement signal of 1-300 Hz interharmonic phasor and 2-50th harmonic phasor; The power grid equipment includes various plants and stations, AC lines, transformers and generators.

3. The online clustering method for power grid broadband measurement signals based on principal component and K-means algorithm according to claim 1, characterized in that: Step 1 specifically includes: Step 1.1: Screen the power grid equipment with broadband measurement signal configuration information to form a set of participating monitoring equipment; Step 1.2: Periodically and real-timely acquire broadband measurement signals of the monitoring equipment set; Step 1.3: Construct a broadband measurement signal sample matrix X with participating monitoring equipment as rows and interharmonic frequency bands or harmonic orders as columns; The number of rows of X is the number of monitoring devices, and the number of columns is the number of interharmonic frequency bands or harmonic orders.

4. The online clustering method for power grid broadband measurement signals based on principal component and K-means algorithm according to claim 3 is characterized by: Step 1.1 is specifically as follows: select the power grid primary equipment models of power plants, AC lines, transformers and generators with 1-300Hz interharmonic phasors and 2-50th harmonic phasor broadband measurement signal configuration information from the real-time library of the dispatching technical support system platform to form a set of participating monitoring equipment.

5. The online clustering method for power grid broadband measurement signals based on principal component and K-means algorithm according to claim 4 is characterized by: In step 1.2, the fixed period is 1 second.

6. The online clustering method for power grid broadband measurement signals based on principal component and K-means algorithm according to claim 1, characterized in that: In step 2.4, the threshold value is configured or modified through the database.

7. The online clustering method for power grid broadband measurement signals based on principal component and K-means algorithm according to claim 1, characterized in that: The set threshold is 80%.

Images