A harmonic responsibility division method considering non-synchronization of monitoring data and harmonic impedance change

Abstract

This invention discloses a harmonic responsibility allocation method considering the asynchronicity of monitoring data and harmonic impedance changes. The method includes: preprocessing the harmonic monitoring data to obtain preprocessed harmonic monitoring data; aligning the asynchronous feeder harmonic current data and bus harmonic voltage data in the preprocessed harmonic monitoring data to obtain aligned harmonic monitoring data; dividing the aligned harmonic monitoring data into data clusters for different scenarios; and then using correlation analysis to allocate harmonic responsibility based on the data in each cluster. This invention avoids the inaccuracy of calculation results caused by directly using raw data for harmonic responsibility allocation; based on the DTW algorithm matching, the ShapeDTW algorithm is used to incorporate the local shape information around the sequence points into the dynamic programming matching process to achieve data matching and alignment; and considering the impact of system harmonic impedance changes on harmonic responsibility allocation, thus avoiding doubts about the rationality and applicability of the harmonic responsibility allocation results.

Description

Technical Field

[0001] This invention relates to a harmonic responsibility allocation method that takes into account the asynchrony of monitoring data and the changes in harmonic impedance, belonging to the field of power harmonic analysis technology. Background Technology

[0002] Under the "dual carbon" goal, the energy structure has undergone significant changes, with a large number of distributed energy sources and nonlinear loads being connected to the power system, increasing the number of harmonic sources and exacerbating the problem of harmonic pollution in the power system. To avoid disputes arising from harmonic pollution and to effectively manage harmonics, an international "reward and punishment scheme" for harmonic control has been proposed. Accurate allocation of harmonic responsibility is a crucial prerequisite for implementing this scheme and achieving precise management. To suppress harmonics at their source, it is necessary to accurately define the harmonic responsibility that each power user should bear for harmonic distortion at the point of common coupling (PCC).

[0003] Traditional harmonic liability allocation relies on the equivalent circuit model of a specific harmonic in the power system, using harmonic voltage measured at the common node and harmonic current measured at each feeder. However, due to the time-varying nature of power systems, the physical model-based harmonic liability allocation method faces difficulties in accurately obtaining system parameters. Harmonic monitoring data, on the other hand, can reflect the harmonic source situation at the monitoring points, making the direct use of historical harmonic monitoring data for multi-harmonic source harmonic liability assessment a promising approach.

[0004] However, previous methods based on correlation analysis of monitoring data did not take into account the asynchronous nature of the monitoring data and the impact of changes in the corresponding harmonic impedance on the division of responsibilities. Summary of the Invention

[0005] This invention provides a harmonic responsibility allocation method that takes into account the asynchrony of monitoring data and the changes in harmonic impedance, so as to allocate harmonic responsibility using correlation analysis after taking into account the asynchrony of monitoring data and the changes in harmonic impedance.

[0006] The technical solution of this invention is:

[0007] According to one aspect of the present invention, a harmonic responsibility allocation method considering the asynchronous nature of monitoring data and harmonic impedance changes is provided, comprising: preprocessing the harmonic monitoring data to obtain preprocessed harmonic monitoring data; aligning the asynchronous feeder harmonic current data and bus harmonic voltage data in the preprocessed harmonic monitoring data to obtain aligned harmonic monitoring data; dividing the aligned harmonic monitoring data into data clusters for different scenarios, and then allocating harmonic responsibility based on the data in each cluster using a correlation analysis method.

[0008] The preprocessing of the harmonic monitoring data specifically involves using a segmented aggregation approximation algorithm to preprocess the harmonic monitoring data.

[0009] The process of aligning the asynchronous feeder harmonic current data and bus harmonic voltage data in the preprocessed harmonic monitoring data specifically involves:

[0010] S2.1. Select sequence x as the base sequence and sequence y as the sequence to be aligned; take each element of the two sequences as the center and cut off lengths L respectively. x and L y From the subsequences of sequence x, we can obtain the subsequence matrix X′ of sequence x and the subsequence matrix Y′ of sequence y;

[0011] S2.2, For each row of matrix X′, the data sequence x′ i Each row of the matrix Y′ is compared with the data sequence y′. j Perform optimal matching when y′ j The minimum value is obtained at point x, which represents the element x in sequence x. i With the element y of sequence y j match;

[0012] S2.3, Keep the baseline sequence x unchanged, and compare it with x i Matching y j The i-th element assigned to the new sequence y′, i.e., y′ i =y j Reconstruct sequence y according to the above rules to form a new sequence y′.

[0013] The process involves dividing the aligned harmonic monitoring data into data clusters for different scenarios, and then using correlation analysis to assign harmonic responsibility based on the data from each cluster. Specifically:

[0014] S3.1. Traverse the elements in the aligned harmonic monitoring data sample set E, determine whether the element is a core object, and if so, add it to the set Ω; otherwise, continue to determine the next element until all elements are traversed.

[0015] S3.2. In the set Ω, select any unprocessed object point p, mark the point as processed, find all directly density-reachable points of the point, and sort all directly density-reachable points in order of reachability distance and store them in the set S.

[0016] S3.3 If S is an empty set, return to step S3.2; if S is not empty, select the sample point q with the smallest reach distance in set S, mark it as processed, store the point in the ordered list M, and determine whether point q is a core object point. If yes, continue to step S3.4; otherwise, return to step S3.3.

[0017] S3.4, Find all directly density-reachable points a from point q. q (j), if a q (j) If it already exists in M, then no action is taken; otherwise, check if a already exists in S. q (j), if it exists, continue to step S3.5; if it does not exist, skip to step S3.6.

[0018] S3.5, If the new reachable distance d for the current object is... r '(i) is less than the old reachable distance d' r If (i), then replace its corresponding reachable distance with rd'(i), reorder S according to reachable distance, and return to step S3.3;

[0019] S3.6, Insertion point a q (j), reorder S according to reachability distance, and return to step S3.3;

[0020] S3.7. Process all elements in sample set E according to steps S3.2-S3.6.

[0021] With the processing order as the x-axis, the reachable distance d r (i) is the ordinate, generating an ordered queue graph; based on the preset neighborhood radius ε, if d r If (i) < ε, then the reachable distance is valid. Cluster them into one class, output the trough data, and obtain the final clustering result. After the clustering is completed, the monitoring dataset is divided into data clusters of different scenarios. Then, based on the data of each cluster, the harmonic responsibility is divided using the correlation analysis method.

[0022] According to another aspect of the present invention, a harmonic responsibility allocation system considering the asynchronicity of monitoring data and harmonic impedance changes is provided, comprising:

[0023] The first acquisition module is used to preprocess the harmonic monitoring data to obtain preprocessed harmonic monitoring data.

[0024] The first acquisition module is used to align the asynchronous feeder harmonic current data and bus harmonic voltage data in the preprocessed harmonic monitoring data to obtain aligned harmonic monitoring data.

[0025] The segmentation module is used to divide the aligned harmonic monitoring data into data clusters for different scenarios, and then use correlation analysis to assign harmonic responsibility based on the data of each cluster.

[0026] The beneficial effects of this invention are as follows: This invention preprocesses harmonic monitoring data, avoiding the inaccuracy of calculation results caused by directly using raw data for harmonic responsibility allocation; furthermore, based on DTW algorithm matching, the ShapeDTW algorithm is used to merge the local shape information around the sequence points into the dynamic programming matching process, realizing data matching and alignment, and achieving synchronization between data; further still, the influence of system harmonic impedance changes on harmonic responsibility allocation is considered, thereby avoiding doubts about the rationality and applicability of the harmonic responsibility allocation results. Attached Figure Description

[0027] Figure 1 This is a flowchart of the present invention;

[0028] Figure 2 This is a schematic diagram of harmonic monitoring data acquisition according to the present invention.

[0029] Figure 3 This is a scatter plot of the linear correlation between harmonic sources;

[0030] Figure 4 This is a scatter plot of harmonic monitoring data under harmonic impedance variation;

[0031] Figure 5 This is a scatter plot of harmonic monitoring data under harmonic impedance changes. Detailed Implementation

[0032] The invention will be further described below with reference to the accompanying drawings and embodiments, but the scope of the invention is not limited to the description.

[0033] Example 1: As Figure 1-5 As shown, according to one aspect of the present invention, a harmonic responsibility allocation method considering the asynchronous nature of monitoring data and harmonic impedance changes is provided, comprising: preprocessing the harmonic monitoring data to obtain preprocessed harmonic monitoring data; aligning the asynchronous feeder harmonic current data and bus harmonic voltage data in the preprocessed harmonic monitoring data to obtain aligned harmonic monitoring data; dividing the aligned harmonic monitoring data into data clusters for different scenarios, and then allocating harmonic responsibility based on the data in each cluster using a correlation analysis method.

[0034] Furthermore, the principle of this invention is analyzed as follows:

[0035] Power quality monitoring device monitoring points such as Figure 2 As shown, the monitoring device is generally deployed at the common connection point of the 10kV bus, with a data sampling interval of 3 minutes, and can obtain bus harmonic voltage data and harmonic current data of m feeders.

[0036] On the one hand, because bus harmonic voltage acquisition and feeder harmonic current acquisition are handled by different monitoring devices, and power quality monitoring devices often use a local clock as a reference for data acquisition, this results in asynchrony in data acquisition from different monitoring points. On the other hand, since the measurement data output by the harmonic monitoring devices currently used by power grid companies are generally statistical values ​​within a monitoring period, such as maximum, minimum, average, and 95% probability high values, misalignment and offset often occur between different sets of data in terms of data time, further exacerbating the asynchrony of harmonic monitoring data. Even with clock correction methods such as network time synchronization, it is difficult to achieve perfect synchronization between data. Therefore, this paper uses the ShapeDTW algorithm to align the asynchronic data.

[0037] On the other hand, in actual operating power systems, when the data acquisition cycle of harmonic monitoring devices is long, the system harmonic impedance may change due to changes in system operation mode, load switching, and reactive power compensation device switching. This will significantly affect how to accurately assign the harmonic responsibility to harmonic source s, and the harmonic responsibility to harmonic source s will vary considerably at different times. Taking the strictly aligned monitoring data of a substation in the Yunnan power grid with no significant harmonic impedance changes as an example, the linear correlation scatter plot between the amplitude of the 7th harmonic voltage at the busbar and the amplitude of the 7th harmonic current of the connected feeder harmonic source is shown in the figure below. Figure 3 As shown, extend Figure 3 The data collection period includes the time periods of substation operation mode changes and load switching. The scatter plot of the monitoring data is shown below. Figure 4 As shown. From Figure 4 It can be seen that the correlation between the data has changed significantly, and the harmonic responsibility of the harmonic source within this period can no longer be measured by a fixed correlation coefficient range. To address this, this invention uses the OPTICS clustering algorithm to separate the datasets within different correlation periods, grouping the datasets with relatively stable correlations into clusters, and then calculating the correlation coefficients of different clusters to allocate harmonic responsibility.

[0038] Specifically, the present invention includes the following:

[0039] I. Data Preprocessing

[0040] Power quality monitoring data is characterized by high noise levels, and directly using the raw data for harmonic responsibility allocation leads to low accuracy in the calculation results. Noise reduction of the harmonic monitoring data can effectively improve this problem. This paper employs the Piecewise Aggregation Approximation (PAA) algorithm to preprocess the harmonic monitoring data. The harmonic monitoring data is represented as a time series v = {v1, v2, ..., v...} i ,…,v m}, vi Let m represent the i-th monitoring data point, and m represent the sequence length.

[0041] This invention employs the classic PAA algorithm to perform noise reduction processing on harmonic monitoring data as shown in equation (1):

[0042]

[0043] In the formula, ω is the length of the time window, and v′ j For the j-th harmonic monitoring data after preprocessing, the data after PAA denoising is v′={v′1,v′2,…,v′ j ,…,v′ n The data is of length n. After PAA processing, the original information is retained and the data noise is significantly reduced.

[0044] II. Data Alignment Based on ShapeDTW Algorithm

[0045] Power quality monitoring data is a typical type of time-series data, and the data sequences collected from different monitoring points exhibit local shifts. Dynamic time warping (DTW) can map different time-series data points at different times, effectively handling local shifts in the sequence and measuring the similarity between two non-time series. When performing curve matching with local shifts, the DTW distance correspondence and the traditional Euclidean distance correspondence are as follows: Figure 5 As shown, Euclidean distance is not suitable for measuring curves that are time-shifted but locally similar, while DTW distance can accurately measure their correspondence and perform matching.

[0046] By solving for the optimal matching path and alignment method, the distance matrix between the bus monitoring harmonic voltage sequence and the feeder monitoring harmonic current sequence can be obtained, thus enabling a quantitative analysis of the correlation between the two. Let the feeder harmonic current sequence and the bus harmonic voltage sequence be x={x1,x2,…,x m} and y = {y1, y2, ..., y n Let the sequence lengths be m and n respectively. First, construct an m x n distance matrix M, where M[i,j] represents the i-th number x in sequence x. i The j-th number y in sequence y j The Euclidean distance. Secondly, let the cumulative distance matrix be M. c The initial values ​​of its first row and first column are calculated as shown in equation (2):

[0047]

[0048] The calculation method for the remaining part of the cumulative distance matrix is ​​shown in equation (3), where 2≤i≤m, 2≤j≤n, and i, j∈N.

[0049]

[0050] Finally, the DTW distance between sequences x and y is determined. As shown in equation (3), the cumulative distance calculation process is equivalent to calculating the optimal alignment of sequences x and y, and then accumulating the distance under the optimal matching method at the end of the matrix. This yields the overall minimum cumulative distance DTW(x,y) = M for the two sets of sequence data. c [i,j]. Its value represents the degree of similarity between two sets of sequences in terms of trend and time characteristics; the larger the value, the higher the similarity.

[0051] In practical engineering applications, when using the DTW algorithm for point-to-point data matching of bus harmonic voltage and feeder harmonic current, ill-conditioned matching may occur, where the distance between two points is minimized but local shape similarity is ignored. To effectively avoid this ill-conditioned matching and ensure that monitoring data sequence points with similar local shapes tend to match, this paper, based on the DTW algorithm matching, employs the ShapeDTW algorithm to incorporate the local shape information around the sequence points into the dynamic programming matching process, thus achieving data matching alignment. The steps of the ShapeDTW algorithm for data matching alignment are as follows:

[0052] (1) The harmonic responsibility division in this paper studies the responsibility relationship between two sequences x and y. Here, sequence x is chosen as the reference sequence, and sequence y is the sequence to be aligned. Taking each element of the two sequences as the center, the lengths of the cuts are L respectively. x (L x <<m) and L y (L y From the subsequences of << n), we can obtain the subsequence matrix X′(m×L) of sequence x. x The subsequence matrix Y′(n×L) of sequence y and sequence y. y dimension);

[0053] (2) For each row data sequence x′(i) (i=1,2,…,m) of matrix X′, perform optimal matching with each row data sequence y′(j) (j=1,2,…,n) of matrix Y′ according to the principle shown in equation (4). Let equation (4) be applied to y′ j The minimum value is obtained at point x, which represents the element x in sequence x. i With the element y of sequence y j match;

[0054] min{DTW(x′ i ,y′1),DTW(x′ i ,y′2),…,DTW(x′ i ,y′ j (4)

[0055] (3) Keep the baseline sequence x unchanged, and compare it with x i Matching y j The i-th element assigned to the new sequence y′, i.e., y′ i =y j Reconstruct sequence y according to the above rules to form a new sequence y′.

[0056] III. Scenarios Based on OPTICS Clustering

[0057] By aligning asynchronous feeder harmonic current data and bus harmonic voltage data using the method presented in this paper, data synchronization can be achieved. However, for harmonic monitoring data containing harmonic impedance variations, the rationality and applicability of the harmonic responsibility allocation results will be questionable if the impact of system harmonic impedance variations on harmonic responsibility allocation is not considered. Therefore, this paper divides the operating scenarios corresponding to different system harmonic impedances into different clusters, and then allocates harmonic responsibility for the harmonic sources in each scenario.

[0058] The OPTICS clustering algorithm is a density-based clustering algorithm, an improvement upon the DBSCAN clustering algorithm. Unlike DBSCAN, which is susceptible to the influence of the input neighborhood parameters (ε, MinPts), resulting in different clustering results for different parameter values, the OPTICS algorithm achieves flexible clustering by generating an ordered queue reflecting the density-based clustering structure of each sample point. Theoretically, the OPTICS algorithm can cluster data of different densities, obtaining clusters of arbitrarily separable shapes.

[0059] Suppose the dataset is X = {x1, x2, ..., x...} i ,…,x n}, with a neighborhood radius of ε and a minimum number of elements of ε, the core object is defined as: if card(x i If x ≥ MinPts, then x i Let card(x) be the core object point of X. i ) represents point x i The number of elements contained in the neighborhood; direct density reachability is defined as: if x i Belongs to x j The neighborhood of x and x j If x is the core object point, then x is called... i It is x j The direct density reachable point; the core distance is defined as: x i The core distance is such that x i The minimum radius of the neighborhood of a core object point; the reachable distance is defined as: x j Regarding x i The reachable distance is x i Core distance and x j With xi The maximum Euclidean distance between them. The scene partitioning steps based on the OPTICS clustering algorithm using harmonic impedance as the clustering basis are as follows:

[0060] (1) Traverse the elements in the sample set E and determine whether the element is a core object. If it is, add it to the set Ω. Otherwise, continue to determine the next element until all elements are traversed.

[0061] (2) Select any unprocessed object point p in the set Ω, mark the point as processed, find all directly density reachable points of the point, and sort all directly density reachable points in order according to the reachable distance and store them in the set S;

[0062] (3) If S is an empty set, return to step (2); if S is not empty, select the sample point q with the smallest reachable distance in set S, mark it as processed, store the point in the ordered list M, and determine whether point q is a core object point. If yes, continue to step (4); otherwise, return to step (3).

[0063] (4) Find all directly density-reachable points a from point q. q (j), if a q (j) If it already exists in M, then no action is taken; otherwise, check if a already exists in S. q If (j) exists, continue to step (5); otherwise, skip to step (6).

[0064] (5) If at this time the new reachable distance d of the current object r '(i) is less than the old reachable distance d' r If (i), then replace its corresponding reachability distance with rd'(i), reorder S according to reachability distance, and return to step (3);

[0065] (6) Insertion point a q (j), reorder S according to reachability distance, and return to step (3);

[0066] (7) Process all elements in sample set E according to steps (2)-(6).

[0067] With the processing order as the x-axis, the reachable distance d r (i) is the ordinate, generating an ordered queue graph. Based on the ordered queue graph, a suitable neighborhood radius ε is selected, if d... r If (i) < ε, then the reachable distance is valid. Cluster them into one class, output the trough data, and obtain the final clustering result. After OPTICS clustering is completed, the monitoring dataset is divided into data clusters of different scenarios. Then, based on the data of each cluster, the harmonic responsibility can be divided using the correlation analysis method.

[0068] IV. Harmonic Responsibility Indicators

[0069] To more effectively characterize the linear relationship between feeder harmonic current and bus harmonic voltage, this paper uses correlation analysis to assign harmonic responsibility. The trends of bus harmonic voltage and feeder harmonic current are correlated to a certain extent. The correlation varies significantly across different feeders and operating scenarios, requiring quantitative analysis of the correlation coefficient to determine the harmonic responsibility of each harmonic source. The linear relationship between variables is described by the degree of correlation between them, and the correlation can be represented by the coefficient r(x,y), as shown in equation (5):

[0070]

[0071] Where, x i and y i These are the i-th element of the feeder harmonic current sequence x and the i-th element of the bus harmonic voltage sequence y, respectively. and Let r(x,y) be the mean of sequence x and the mean of sequence y, respectively. r(x,y)∈[-1,1]. When the value of r(x,y) approaches 1, it indicates that there is a good positive correlation between the two sequences; when the value of r(x,y) approaches -1, it indicates that there is a good negative correlation between the two sequences; when the value of r(x,y) approaches 0, it indicates that there is no correlation between the two sequences.

[0072] Within the same scene time period, the harmonic liability remains relatively stable. If the monitoring period contains k time periods of the same scene, the harmonic liability r for the a-th time period of the same scene (a=1,2,…,k) is... a The calculation is shown in equation (6):

[0073] r a =r(x(a),y(a))(6)

[0074] In the formula, x(a) and y(a) are the feeder harmonic current subsequence and the bus harmonic voltage subsequence under the a-th same scene time period, respectively.

[0075] To more intuitively compare the relative magnitudes of harmonic liability for each feeder and to facilitate the implementation of harmonic liability reward and punishment schemes, this paper focuses on r a Normalization is performed. During normalization, the combined harmonic load from all feeder sources on the bus is used as the baseline value of 1. The harmonic responsibilities of all feeders are then summed to 1. A negative value indicates that the feeder did not generate harmonics but was forced to absorb them, making it a victim in the harmonic responsibility allocation. Its value represents the harmonic load from all harmonic sources acting on that feeder. The normalization process is as follows:

[0076] For all r a The positive value r in a+ Perform the processing as shown in equation (7), where b is a positive value r. a + The number of B a (i) is the coefficient of responsibility that the i-th feeder should bear in the a-th scenario (assuming that the bus is connected to N harmonic source feeders, i = 1, 2, ..., N).

[0077]

[0078] Let E be the responsibility of the feeder that generates harmonics after normalization. a + (i) The feeder that absorbs harmonics is forced to assume the responsibility of E a - (i), the calculation process is shown in equation (8). Where r a - For r a A negative value in r indicates that the feeder is forced to bear the harmonic responsibility; c is a negative value. a - The number of.

[0079]

[0080] Taking into account the time range corresponding to harmonic liability in each scenario, the harmonic liability of the i-th feeder in the first to k-th time periods of the same scenario is calculated cumulatively. Then, the total harmonic liability F of feeder i in the monitoring period is calculated as shown in Equation (9). i .

[0081]

[0082] In the formula, E a (i) represents the harmonic responsibility value of the i-th feeder after normalization in the a-th scenario, and T(a) represents the duration of the a-th time period in the same scenario within the monitoring period. Subsequently, the normalization method shown in equations (7)-(8) is used to normalize the total harmonic responsibility, resulting in a more intuitive and easier-to-compare total harmonic responsibility for each feeder that takes into account the time range.

[0083] According to another aspect of the present invention, a harmonic responsibility allocation system considering the asynchronicity of monitoring data and harmonic impedance changes is provided, comprising: a first acquisition module for preprocessing harmonic monitoring data to obtain preprocessed harmonic monitoring data; the first acquisition module for aligning asynchronous feeder harmonic current data and bus harmonic voltage data in the preprocessed harmonic monitoring data to obtain aligned harmonic monitoring data; and a allocation module for dividing the aligned harmonic monitoring data into data clusters for different scenarios, and then allocating harmonic responsibility based on each cluster using a correlation analysis method. For parts of the modules not described in detail above, please refer to the relevant descriptions in the embodiments.

[0084] The specific embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention.

Claims

1. A method for harmonic liability allocation considering the asynchronicity of monitoring data and harmonic impedance changes, characterized in that, include: The harmonic monitoring data is preprocessed to obtain preprocessed harmonic monitoring data; Align the asynchronous feeder harmonic current data and bus harmonic voltage data in the preprocessed harmonic monitoring data to obtain aligned harmonic monitoring data; The aligned harmonic monitoring data is divided into data clusters for different scenarios, and then the harmonic responsibility is assigned based on the correlation analysis method on the basis of each cluster. The process involves dividing the aligned harmonic monitoring data into data clusters for different scenarios, and then using correlation analysis to assign harmonic responsibility based on the data from each cluster. Specifically: S3.1 Traversing the aligned harmonic monitoring data sample set E For each element in the set Ω, determine whether it is a core object. If it is, add it to the set Ω. Otherwise, continue to the next element until all elements have been traversed. S3.2, Randomly select an unprocessed object point in set Ω. p Mark the point as processed, find all directly density-reachable points of the point, and sort all directly density-reachable points in order of reachability distance and store them in a set. S middle; S3.3, if S If it is an empty set, then return to step S3.2; if S If it is not empty, then select the set. S The sample point with the smallest reachable distance q Mark it as processed and store it in an ordered list. M In the middle, and determine the point q Is it a core object point? If yes, continue to step S3.4; otherwise, return to step S3.

3. S3.4 Finding the point q All directly achievable density points a q ( j ),like a q ( j (Already exists) M If the result is neutral, no action is taken; otherwise, a judgment is made. S Does it already exist in China? a q ( j If the condition exists, continue to step S3.5; otherwise, skip to step S3.

6. S3.5 If the new reachable distance of the current object is at this time d r ' ( i (less than the old reachable distance) d r ( i If ), then its corresponding reachable distance is replaced with rd' ( i ),right S Reorder by reachability distance and return to step S3.3; S3.6, Insertion Point a q ( j ),right S Reorder by reachability distance and return to step S3.3; S3.7, Sample set E All elements in the process are processed according to steps S3.2-S3.6; Distance reached by using the processing order as the x-axis d r ( i Using ) as the ordinate, an ordered queue graph is generated; based on the preset neighborhood radius ε, if d r ( i If ε < ε, then the reachable distance is valid. Cluster them into one class, output the trough data, and obtain the final clustering result. After clustering, the monitoring dataset is divided into data clusters of different scenarios. Then, based on the data of each cluster, the harmonic responsibility is divided using the correlation analysis method.

2. The harmonic responsibility allocation method considering the asynchronicity of monitoring data and harmonic impedance changes according to claim 1, characterized in that, The preprocessing of the harmonic monitoring data specifically involves using a segmented aggregation approximation algorithm to preprocess the harmonic monitoring data.

3. The harmonic responsibility allocation method considering the asynchronicity of monitoring data and harmonic impedance changes according to claim 1, characterized in that, The process of aligning the asynchronous feeder harmonic current data and bus harmonic voltage data in the preprocessed harmonic monitoring data specifically involves: S2.1, Select sequence x As the baseline sequence, the sequence y Given the sequences to be aligned; using each element of the two sequences as the center, cut off lengths of respectively... L x and L y The subsequence can be used to obtain the sequence. x subsequence matrix X 'and sequence y subsequence matrix Y '; S2.2, For matrices X Each row of data sequence x ' i respectively with matrix Y Each row of data sequence y ' j Perform optimal matching when y ' j The sequence reaches its minimum value at a certain point, which indicates that the sequence... x elements x i with sequence y elements y j match; S2.3, Save the reference sequence x Unchanged, will be with x i Matching y j Assigned to new sequence y The first in ' i Each element, namely y ' i = y j The sequence is processed according to the above rules. y Reconstruction, forming a new sequence y '.

4. A harmonic responsibility allocation system that considers the asynchronicity of monitoring data and harmonic impedance changes, characterized in that, Implementing the harmonic responsibility allocation method of claim 1, which considers the asynchronicity of monitoring data and harmonic impedance changes, includes: The first acquisition module is used to preprocess the harmonic monitoring data to obtain preprocessed harmonic monitoring data. The first acquisition module is used to align the asynchronous feeder harmonic current data and bus harmonic voltage data in the preprocessed harmonic monitoring data to obtain aligned harmonic monitoring data. The segmentation module is used to divide the aligned harmonic monitoring data into data clusters for different scenarios, and then use correlation analysis to assign harmonic responsibility based on the data of each cluster.

Images