Power quality zoning treatment method based on time sequence and data driving

Abstract

The invention discloses an electric energy quality zoning treatment method based on a time sequence and data driving, belongs to the technical field of electric energy quality zoning treatment, and solves the problems of poor dynamism and adaptability and low calculation efficiency in the existing electric energy quality treatment process. The method comprises the following steps: respectively collecting and processing voltage data of a plurality of continuous sampling points of each node of a power grid to obtain an electric energy quality time sequence of each node; according to the power quality time sequence of each node, obtaining a behavior similarity distance between different nodes; calculating the comprehensive coupling degree between the corresponding nodes according to the behavior similarity distance between the different nodes; carrying out regional division on the power grid according to the comprehensive coupling degree between different nodes; and based on the dominant node of each power grid subarea, performing cooperative treatment of the corresponding power grid subarea.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power quality zoned governance, and particularly relates to a power quality zoned governance method based on time series and data driving. BACKGROUND

[0002] With the rapid development of power systems, power quality problems have increasingly become an important factor affecting the stability of power systems, the safe operation of equipment, and the efficiency of industrial production. Whether the power quality is good or not directly relates to the stability of the power grid, the safe operation of equipment, and the efficiency of production. Power quality disturbances are usually manifested as voltage deviation, frequency fluctuation, harmonic pollution, and flicker, which can cause equipment failure, electrical equipment damage, and production efficiency decline, and even can cause large-scale power system failure.

[0003] Currently, the main governance methods for power quality are local governance, global governance, and zoned governance. Local governance handles local power quality problems by installing compensation devices, filters, and other equipment at key nodes of the power system. However, this method is limited to a specific area and cannot cope with the propagation effect of power quality pollution in the power grid. Global governance optimizes the power quality of each node in the power grid through overall modeling, but due to the large scale of the power grid and the complexity of power quality pollution factors, the governance process has a large amount of calculation, and there is a problem of poor timeliness in dealing with dynamic power quality problems.

[0004] Zoned governance is a governance method based on the coupling relationship of power quality at nodes of the power grid. The idea is to divide the power grid into several governance areas, and the nodes within each area have a strong coupling degree, while the coupling degree between different areas is low. Zoned governance can use targeted governance strategies according to the specific power quality situation of different areas. However, existing power quality zoned governance methods rely mainly on electrical distance or static power grid models and do not effectively consider the dynamic changes and propagation characteristics of power quality disturbances. Traditional methods rely on manually designed features and rules for spatiotemporal feature analysis of power quality, lack dynamic optimization strategies based on data driving, and cannot adapt to real-time changes in power quality problems.

[0005] In recent years, with the continuous development of big data and artificial intelligence technology, data-driven power quality governance methods have received widespread attention. By collecting and analyzing real-time power quality data at grid nodes, the time-domain characteristics, frequency-domain characteristics, and spatial distribution patterns of power quality disturbances can be accurately identified, enabling dynamic monitoring and optimization of power quality. Based on machine learning, deep learning, and other data mining techniques, many data-driven power quality zoning governance methods have been proposed, which can divide regions and make intelligent decisions based on real-time data, improving the efficiency and accuracy of power quality governance. However, existing technologies still face some challenges, such as the need for large amounts of labeled data for training, difficulty in feature selection, and complex model training, which limit their application in power grid. SUMMARY

[0006] In view of the above analysis, the embodiments of the present application aim to provide a power quality zoning governance method based on time series and data driving, to solve the problems of poor dynamicity and adaptability, and low computational efficiency in the existing power quality governance process.

[0007] The present application discloses a power quality zoning governance method based on time series and data driving, which comprises: Collecting and processing voltage data of continuous multiple sampling points of each node in the power grid to obtain power quality time series of each node; According to the power quality time series of each node, the behavior similarity distance between different nodes is obtained; According to the behavior similarity distance between different nodes, the comprehensive coupling degree between corresponding nodes is calculated, and the power grid is divided into regions according to the comprehensive coupling degree between different nodes; Based on the dominant node of each power grid partition, the collaborative governance of the corresponding power grid partition is carried out.

[0008] Based on the above scheme, the present application further improves as follows: Further, the comprehensive coupling degree between nodes and is represented as: (1) Wherein, represents the behavior similarity distance between nodes and , represents the physical electrical distance between nodes and ; , respectively represent nodes , ​Importance weight of 、 is a scale parameter used to adjust the contribution of different factors.

[0009] Furthermore, the power grid is divided into regions based on the comprehensive coupling degree between different nodes, and the following operations are performed: According to the comprehensive coupling degree between every two nodes in the power grid, a comprehensive coupling degree matrix of all nodes in the power grid is constructed; the comprehensive coupling degree matrix is ​​used as the adjacency matrix of the graph; According to the adjacency matrix, calculate the Laplace matrix; Perform eigendecomposition on the Laplace matrix to construct a new feature space; The K-means clustering algorithm is applied in the new feature space to cluster the nodes in the power grid and realize the regional division of the power grid.

[0010] Furthermore, the power quality time series of each node is obtained by performing the following operations: The power grid monitoring system is used to collect voltage data of multiple consecutive sampling points at each node of the power grid to form the original voltage waveform data of the corresponding node; Based on the original voltage waveform data of each node, multiple power quality indicators of the corresponding node are calculated, including voltage deviation, harmonic voltage content rate and three-phase imbalance; Each power quality indicator of each node is standardized and arranged in sequence according to the sampling point order to construct a power quality subsequence of the corresponding power quality indicator.

[0011] Furthermore, the behavior similarity distance between different nodes is obtained by performing the following operations: Perform piecewise linear representation and linear regression fitting on each power quality subsequence of each node to obtain multiple segmented time subsequences of the corresponding power quality subsequence; For each power quality subsequence, time series pattern matching is performed on the segmented time subsequences between different nodes to obtain the behavioral similarity distance between different nodes.

[0012] Furthermore, the node and The behavioral similarity distance between Expressed as: (2) in, 、 、 Represents nodes respectively and The DTW distance between the power quality time subseries of voltage deviation index, harmonic content index and three-phase imbalance index.

[0013] Further, the node and the sub-sequence of time segments in the same item of power quality time sub-sequence at index , at index , the matching distance is represented as: (3) wherein, the cumulative matching distance between the first data points of the sub-sequence of time segments of the node and the first data points of the sub-sequence of time segments of the node is represented as, is the weighted distance between the data at index and the data at index ; the time weight factor at index ; the time weight factor at index ; assuming the lengths of the sub-sequence of time segments , are , the matching distance between the sub-sequence of time segments , is represented as, , , the matching distance between the sub-sequence of time segments , ; the DTW distance between the nodes in the same item of power quality time sub-sequence is obtained by summing the matching distances between all the sub-sequence of time segments of different nodes in the same item of power quality time sub-sequence.

[0014] Further, the weighted distance is represented as: (4) wherein, the data of the sub-sequence of time segments of the node at index , the data of the sub-sequence of time segments of the node at index .

[0015] Further, the collaborative governance of the corresponding power grid partition is performed by: For each power grid partition, the power quality indicators of the dominant node are monitored in real time, and if the trigger condition of the corresponding power quality indicator is met, the coordinated action instruction of the corresponding power quality indicator dimension is triggered in the corresponding power grid partition. The power quality management equipment of the corresponding power quality indicator dimension in the corresponding power grid partition executes power quality management according to the corresponding coordinated action instruction, and realizes the coordinated management of the corresponding power grid partition.

[0016] Further, the node with the maximum average value of the comprehensive coupling degree with all other nodes in each power grid partition is taken as the dominant node.

[0017] Compared with the prior art, the present application can at least achieve one of the following beneficial effects: In view of the existing research problems, the present application proposes a power quality partition management method based on time series and data driving. This method combines time series analysis, similarity analysis and clustering algorithm, accurately extracts the time-frequency features and spatial features of power quality by monitoring the power quality data of each node in the power grid in real time, and realizes accurate partitioning of the power grid based on the coupling degree between nodes. Each region selects the node with the maximum coupling degree as the dominant control node, and finally forms an intelligent system that can dynamically adjust the management area and optimize the strategy according to real-time monitoring data. Through the data-driven modeling method, the present application can dynamically adjust the management area and optimize the management strategy, effectively cope with the complexity and variability of power quality disturbance in the power grid, thereby improving the stability and power supply quality of the power grid, and solving the problems of poor dynamicity and adaptability, low calculation efficiency in the existing power quality management process.

[0018] The present application relates to power quality management technology, especially a power quality partition management strategy combining data driving and time series pattern recognition. This technology realizes regional management of power quality pollution through data-driven modeling method, combined with time series analysis and self-supervised learning, especially for power quality problems caused by high-penetration distributed photovoltaic access and power electronic loads. First, based on the power quality monitoring data of each node in the distribution network, time series feature extraction and pattern recognition technology are used to analyze the power quality time series and extract local trend features and time features. Then, through similarity analysis and clustering algorithm, the coupling relationship of power quality pollution between nodes is identified, and the power quality management area is divided according to the coupling degree. Then, according to the regional division, the dominant control node of each region is determined, and targeted management measures such as passive filter, capacitor and other devices are implemented. Through these methods, the present application can effectively realize the coordinated management of power quality pollution and ensure the stability and high-quality power supply of the power system.

[0019] The technical solutions in the present application can be combined with each other to achieve more preferred combination solutions. Other features and advantages of the present application will be described in the following description, and some advantages will become apparent from the description, or will be learned by practice of the present application. The objects and other advantages of the present application can be achieved and obtained from the content particularly pointed out in the description and the drawings. BRIEF DESCRIPTION OF DRAWINGS

[0020] The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiments of the application and together with the description serve to explain the principles of the application. Figure 1 A flowchart of the time series and data driven power quality zoning governance method provided by the embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0021] The preferred embodiments of the present application will be specifically described below in conjunction with the drawings, wherein the drawings constitute a part of the present application and are used to illustrate the principles of the embodiments of the present application, and are not considered as limiting the scope of the present application.

[0022] One specific embodiment of the present application discloses a time series and data driven power quality zoning governance method, and a flowchart is shown in Figure 1 .

[0023] Step S1: Collecting voltage data of continuous multiple sampling points of each node of the power grid respectively and processing to obtain power quality time series of each node.

[0024] The power quality time series constructed in the embodiment includes multiple power quality time sub-sequences, and the specific implementation process is described as follows.

[0025] Step S11: Collecting voltage data of continuous multiple sampling points of each node of the power grid respectively by using the power grid monitoring system to form original voltage waveform data of the corresponding node.

[0026] In the implementation process, the obtained original voltage waveform data can be preprocessed to ensure data integrity and accuracy. For example, Kalman filtering algorithm can be used to remove noise in voltage data. Kalman filtering algorithm can dynamically adjust filtering parameters according to noise characteristics, accurately remove measurement errors or abnormal fluctuations caused by environmental factors, adapt to different noise environments, and improve the filtering accuracy of subsequent calculations. If there is missing data in the collected voltage data, a generative adversarial network (GANs) can be used to generate high-quality filling data, better preserve the rules and complex relationships of voltage data, and ensure data integrity, laying a foundation for accurate calculation of subsequent power quality indicators. In addition, for periodic fluctuations, wavelet packet transform (WPT) is used to provide higher time-frequency resolution than Fourier transform, effectively separate high-frequency noise and low-frequency trend, and improve data quality and accuracy of subsequent analysis.

[0027] Step S12: According to the original voltage waveform data of each node, the multiple power quality indicators of the corresponding node are calculated, including: voltage deviation, harmonic voltage content rate and three-phase imbalance.

[0028] Voltage deviation defined as the deviation of the measured voltage from the system nominal voltage, and the calculation formula is: (1) wherein, is the measured voltage, is the system nominal voltage.

[0029] Harmonic voltage content rate defined as the ratio of the effective value of harmonic voltage to the effective value of fundamental voltage, reflecting the distortion degree of voltage waveform, and the calculation formula is: (2) wherein, is the effective value of fundamental voltage, is the effective value of the th harmonic voltage, the highest harmonic number calculated; in the implementation process, Fourier transform is performed on the original voltage waveform data, and , and are obtained.

[0030] Three-phase voltage imbalance defined as the degree of asymmetry of three-phase voltage amplitude or phase, usually represented by the ratio of negative sequence component to positive sequence component, and the calculation formula is: (3) wherein, is the positive sequence voltage component, is the negative sequence voltage component. In the implementation process, Fourier transform and sequence component extraction are performed on the original voltage waveform data, so that and are obtained. Among them, the positive sequence voltage component refers to the component in the three-phase voltage that is symmetrical and the phase difference is 120°. It is usually used to describe a balanced three-phase system. The negative sequence voltage component refers to the component in the three-phase voltage that is symmetrical and the phase difference is -120°. It is usually used to describe an unbalanced three-phase system.

[0031] Step S13: Each item of power quality index of each node is standardized and arranged in order according to the sampling points, and the power quality subsequence of the corresponding power quality index is constructed.

[0032] In the implementation process, the following standardization formula can be used: (4) Among them, is an item of power quality index of node , , respectively represent the minimum value and the maximum value of all nodes in the corresponding power quality index. Through this process, the power quality data can be normalized to the interval [-1, 1], ensuring the stability of the subsequent data analysis process.

[0033] Based on the above description, it can be known that the power quality time sequence constructed in the embodiment specifically includes the voltage deviation subsequence, the harmonic voltage content rate subsequence and the three-phase imbalance subsequence.

[0034] Step S2: According to the power quality time sequence of each node, the behavior similarity distance between different nodes is obtained.

[0035] Step S21: Each item of power quality subsequence of each node is represented by piecewise linear representation and linear regression fitting, and a plurality of segmented time subsequences of the corresponding power quality subsequence are obtained.

[0036] Specifically, the local trend feature of each item of power quality time subsequence is extracted by piecewise linear representation (PLR). Each item of power quality subsequence is divided into a plurality of time periods, and each time period is fitted by linear regression method, so as to extract the local trend feature of the time sequence. This process helps to reduce the redundant data in the original time sequence and can highlight the key trend of power quality change.

[0037] Step S22: For each item of power quality subsequence, the segmented time subsequences between different nodes are matched in time sequence mode, and the behavior similarity distance between different nodes is obtained.

[0038] In traditional DTW algorithm, the matching distance of time series is usually equal, but in power quality time series, the time points of dramatic changes such as voltage sag and flicker carry more information than the smooth period. By introducing a weight factor combining the amplitude and trend strength of time series, the matching accuracy of key time points can be improved.

[0039] For any segmented time sub-sequence of node , the time weight factor at index can be defined as: (5) where , are adjustment coefficients; represents the data of segmented time sub-sequence of node at index ; , represent the maximum and minimum value in the segmented time sub-sequence, respectively; represents the second-order difference of , which represents the trend change strength; represents the maximum value of the second-order difference of all data in the segmented time sub-sequence.

[0040] In specific real-time process, formula (5) is used for the importance of any segmented time sub-sequence at a specific index, exemplarily, , take 0.7 and 0.3, respectively.

[0041] (6) Based on DTW algorithm, the matching distance of the segmented time sub-sequence of node and in the same power quality time sub-sequence (segmented time corresponds to the same observation period) at index , at index can be represented as: (7) where represents the cumulative matching distance between the first data points of and the first data points of , is the data of at index and the data of at index​ weighted distance between the data at the index denotes the time weight factor at the index , denotes the time weight factor at the index ,

[0042] DTW algorithm allows the index and not equal, to deal with the possible small time delay between different nodes. The weighted distance is expressed as: (8) The algorithm realizes the minimum total distance of the matching path by dynamic programming, and gets the optimal time series alignment. Assuming that the length of the segmented time subsequence , is , , the length of the subsequence refers to the cumulative distance from the beginning of the sequence to the current index point, then denotes the matching distance between the segmented time subsequence , , which reflects the similarity of the power quality behavior of nodes and in a certain dimension.

[0043] The sum of the matching distances between all segmented time subsequences of different nodes in the same item of power quality time subsequence is obtained, and the DTW distance between the corresponding nodes in the same item of power quality time subsequence is obtained, to reflect the similarity of the power quality behavior of nodes and in a certain dimension.

[0044] The DTW distances of each item of power quality subsequence between nodes are weighted and fused to obtain the behavior similarity distance between the corresponding nodes.

[0045] The behavior similarity distance between nodes is expressed as: (9) wherein, , , denote the DTW distance of the power quality time subsequence between nodes and in the voltage deviation index, the harmonic content index, and the three-phase imbalance index, respectively.​ 、 、 respectively represent the global importance weight of the voltage deviation index, the harmonic content index, and the three-phase imbalance index, 、 、 and the sum is 1.

[0046] Step S3: Calculate the comprehensive coupling degree between corresponding nodes according to the behavior similarity distance between different nodes; and divide the power grid into regions according to the comprehensive coupling degree between different nodes.

[0047] The embodiment is based on node coupling degree to divide the power grid. To this end, it is necessary to first calculate a coupling degree index that can comprehensively reflect the strength of the association between nodes.

[0048] First, the importance weight of each node is calculated respectively. The importance weight of node is denoted as : (10) wherein, denotes the comprehensive fluctuation rate of power quality of node , , 、 、 respectively represent the standard deviation of voltage deviation, harmonic voltage content rate, and three-phase imbalance; is a smoothing coefficient to avoid zero denominator, and the value is 0.01 here; denotes the total number of all nodes in the power grid.

[0049] The comprehensive coupling degree between node and is denoted as: (11) wherein, denotes the physical electrical distance (the shortest electrical distance calculated by Dijkstra algorithm) between node and ; 、 respectively represent the importance weight of node 、 ; 、 are scale parameters for adjusting the contribution degree of different factors. The higher the value of the comprehensive coupling degree, the closer the association between the two nodes, and the more they should be divided into the same region.

[0050] According to the comprehensive coupling degrees between two nodes in the power grid, a comprehensive coupling degree matrix of all nodes in the power grid is constructed; based on the comprehensive coupling degree matrix, a clustering algorithm is used to divide the power grid into regions. Preferably, the spectral clustering algorithm is used to divide the power grid into regions, which can effectively solve the graph partitioning problem based on the similarity (coupling degree) between nodes. The algorithm can abstract all nodes and their coupling relationships in the power grid into a weighted undirected graph, and the optimal partitioning is realized through the following specific steps.

[0051] Step S31: Taking the comprehensive coupling degree matrix as the adjacency matrix of the graph , the matrix element represents the comprehensive coupling degree between the power grid node and the node , that is, the weight of the edge. The higher the weight, the closer the association between the two nodes in the power quality behavior.

[0052] Step S32: According to the adjacency matrix , the Laplacian matrix is calculated.

[0053] Specifically, according to the adjacency matrix , a diagonal matrix is constructed, and the element on the diagonal is the sum of all elements in the first row of the adjacency matrix , that is, , which physically represents the total coupling strength of the node with all other nodes in the network.

[0054] Subsequently, the Laplacian matrix is calculated, which is defined as . The Laplacian matrix contains the structure and connectivity information of the graph, which is the core of spectral analysis.

[0055] Step S33: Perform eigenvalue decomposition on the Laplacian matrix to construct a new feature space.

[0056] Eigenvalue decomposition is performed on the Laplacian matrix to calculate the first smallest eigenvalues and their corresponding eigenvectors of the Laplacian matrix , where is the number of regions that the power grid is expected to be divided into.

[0057] The eigenvectors are combined by columns to construct a new dimensional feature matrix .

[0058] Feature Matrix No. A row is a dimensional vector, considered as a grid node The coordinate representation in the new, low-dimensional feature space. This process achieves dimensionality reduction of node relationships by retaining the most important structural information of the graph. of indivual dimensional vector as the new feature space.

[0059] Step S34: Applying the K-means clustering algorithm in the new feature space to cluster the nodes in the power grid to achieve regional division of the power grid.

[0060] Specifically, the feature matrix of Row (i.e. indivual dimensional vector) as input, and apply the standard K-means clustering algorithm to divide it into Each cluster is a collection of several nodes, forming a power grid partition, thereby realizing the regional division of the power grid, that is, forming a power quality management partition.

[0061] Through the above steps, the present invention transforms the complex graph segmentation problem based on coupling degree into a clustering problem that is easier to handle in low-dimensional Euclidean space, thereby automatically dividing the power grid into The power quality management zones are designed with strong coupling between internal nodes and weak coupling between regions.

[0062] Step S4: Based on the dominant nodes of each grid partition, collaborative governance of the corresponding grid partition is performed.

[0063] After completing the grid area division, in order to achieve efficient and coordinated governance of each grid partition, it is necessary to determine a dominant node for each grid partition and use it as the core to solve the power quality problems of the corresponding grid partition.

[0064] Specifically, the node with the largest average value of comprehensive coupling degree with all other nodes in each grid partition is regarded as the dominant node, and the determination formula is described as follows. The dominant node is determined by the following formula : (12) in, Indicates the A collection of grid partitions The formula calculates the number of nodes in Each node in the grid partition The comprehensive coupling degree to all other nodes in the area is calculated, and the node with the maximum average value of the comprehensive coupling degree is selected as the leading node of the corresponding power grid partition.

[0065] The leading node is the "core" or "propagation source" of power quality problems in the area, and it is most closely associated with other nodes in the area. Therefore, governing the leading node can effectively transmit and radiate the effect to other nodes in the entire area.

[0066] Preferably, in this embodiment, a set of collaborative governance strategies is established and implemented for each power grid partition. The strategies convert static analysis results into dynamic control practices, and achieve precise response to power quality problems by defining clear control logic for each area in advance. The specific implementation process is described as follows.

[0067] Step S41: For each power grid partition, real-time monitoring of various power quality indicators of the leading node is performed, and if the trigger condition of the corresponding power quality indicator is met, the collaborative action instruction of the corresponding power quality indicator dimension is triggered in the corresponding power grid partition.

[0068] In the specific implementation process, first, a regional state monitoring and triggering mechanism is established, and the leading node of each power grid partition is taken as the core monitoring point and state representative of the power quality status of the corresponding power grid partition. Through the monitoring system, the various power quality indicators of the leading node are continuously collected and analyzed at a preset sampling period of not more than 200 ms, including voltage deviation , harmonic voltage content , and three-phase voltage imbalance .

[0069] A collaborative governance rule base containing multiple levels of response rules is constructed for each power grid partition. The rule base takes the real-time monitoring data of the leading node as input, and automatically determines whether regional collaborative governance needs to be started according to preset threshold values that comply with the national grid safety operation guide. The rule base contains specific control procedures for different power quality disturbances, ensuring the accuracy and reliability of control. This embodiment also gives an example of a collaborative rule base, which is described as follows.

[0070] (1) For voltage out-of-limit problems, the rule base includes voltage collaborative compensation rules, the trigger condition of which is set as: if the amplitude of the voltage deviation of the leading node is greater than the voltage control upper threshold value set as 7% or lower than the voltage control lower threshold value set as -7%, and the duration of this state exceeds the preset time set as 2 seconds , the collaborative compensation action is triggered.

[0071] ​​(2) For the problem of harmonic pollution, the rule base includes a harmonic cooperative suppression rule, and the trigger condition is set as: if the harmonic voltage content rate of the dominant node is greater than the harmonic control threshold set as 5% then the cooperative filtering action is triggered.

[0072] (3) For the problem of three-phase imbalance, the rule base also includes an imbalance cooperative governance rule, and the trigger condition is set as: if the three-phase voltage imbalance of the dominant node is greater than the imbalance control threshold set as 2% then the imbalance compensation action is triggered.

[0073] When any rule in the rule base is triggered, the system will generate a clear control target and issue a unified cooperative action instruction to all related controllable governance devices in the area corresponding to the rule. The control target is to restore the corresponding power quality index of the dominant node that triggered the problem to within the safety threshold, specifically to restore the within 5%, or to restore the within 4%, or to restore the within 1.5%.

[0074] Step S42: The power quality governance devices in the corresponding power quality index dimension of the corresponding power grid partition execute power quality governance according to the corresponding cooperative action instruction, realizing cooperative governance of the corresponding power grid partition.

[0075] Specifically, the cooperative action instruction can be distinguished according to the type of the triggered rule. If it is voltage overrun, cooperative compensation instructions are issued to all reactive power compensation devices in the area, such as static reactive power compensators STATCOM, to compensate for the reactive power in the power system, improve the power factor, and reduce grid loss. If it is harmonic over-standard, cooperative filtering instructions are issued to all active power filters APF in the area to dynamically compensate for harmonic currents, improve power quality, and reduce harmonic pollution. If it is three-phase imbalance over-standard, cooperative instructions to inject negative sequence compensation currents are issued to governance devices with independent phase control capability in the area to adjust the phase of voltage and current through phase control for solving voltage imbalance, flicker and other problems.

[0076] Through the above process, the complex analysis results in the early stage are converted into a set of clear, automated, and capable of coping with multiple disturbance scenarios control logic, providing a clear basis for subsequent device deployment and adjustment.

[0077] ​​In the implementation process, the embodiment can also select and deploy equipment according to the management task. When any of the above management rules is triggered, the system determines the type of equipment that needs to be started or deployed in the corresponding power grid partition according to the type of the triggered rule. The deployment logic follows the following hierarchical decision principle to ensure the targeted deployment of management resources: (13) wherein, represents the type of equipment deployed or started in the power grid partition , , , , respectively, the maximum harmonic content ratio, the maximum voltage deviation and the maximum three-phase imbalance degree monitored in the region, , , corresponding management thresholds, consistent with the triggering threshold in step S4. The above formula defines which type of management resource should be mobilized first when a specific power quality problem occurs.

[0078] In the process of executing power quality management according to the corresponding cooperative action instruction, the power quality management equipment can perform a dynamic closed-loop feedback adjustment process. The compensation rate or output power of the deployed or started power quality management equipment is not fixed, but is adjusted in real time according to the management effect. The closed-loop feedback adjustment process includes the following operations: the power quality management equipment starts to execute the cooperative action instruction; at the same time, the system continuously monitors the changes of the corresponding power quality indicators of the dominant node at a higher frequency not higher than 50ms. The system dynamically adjusts the compensation rate of all cooperative action devices in the region through a proportional-integral (PI) controller according to the error between the real-time monitoring value of the dominant node indicator and the set control target . The calculation formula of the compensation rate can be expressed as: (14) wherein, is the compensation rate instruction issued to the power quality management equipment, is the real-time error between the monitoring value and the target value, and are the proportional and integral coefficients of the PI controller. The closed-loop adjustment process continues until the power quality indicators of the dominant node are stabilized in the preset safe range, the cooperative management process is completed, and the system returns to the normal monitoring state. Based on the above operation, the power quality management equipment in each region can be dynamically deployed and closed-loop adjusted to ensure the rapidity and stability of the management effect.

[0079] In summary, the present application has the following effects: (1) The present application can effectively identify the time and space characteristics of power quality disturbance without a large amount of prior knowledge by introducing a data-driven modeling method, combining time series analysis and pattern recognition technology, and providing accurate basis for the development of power quality management strategy.

[0080] (2) The clustering algorithm and node coupling degree analysis method used in the present application can accurately divide the management area in the power grid, ensure that the node power quality characteristics in each area are highly consistent, and effectively reduce the redundant adjustment in cross-area management, thereby improving the management efficiency.

[0081] (3) The present application introduces the leading node of each power grid partition and performs corresponding power grid partition collaborative management based on the leading node, thereby ensuring that the management measures can quickly adapt and automatically adjust in the dynamic environment of power grid power quality change, and improving the flexibility and real-time performance of the management.

[0082] Those skilled in the art can understand that all or part of the processes of the above-mentioned embodiments can be completed by a computer program instructing relevant hardware, and the program can be stored in a computer readable storage medium. The computer readable storage medium includes a magnetic disk, an optical disk, a read-only memory, a random access memory, etc.

[0083] The above is only a preferred specific embodiment of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. A time series and data-driven based power quality zoned governance method, characterized in that, The method comprises: Respectively collect the voltage data of each node of the power grid at a plurality of continuous sampling points and process to obtain the power quality time sequence of each node; According to the power quality time sequence of each node, the behavior similarity distance between different nodes is obtained; According to the behavior similarity distance between different nodes, the comprehensive coupling degree between the corresponding nodes is calculated; and according to the comprehensive coupling degree between different nodes, the power grid is divided into regions; Based on the dominant node of each power grid partition, the corresponding power grid partition is cooperatively managed.

2. The time series and data-driven based power quality zonal governance method according to claim 1, wherein, Node And The integrated coupling degree between Is represented as: (1) wherein, represents the behavioral similarity distance between nodes and , represents the physical electrical distance between nodes and ; , respectively represent the importance weight of nodes , ; , are scale parameters for adjusting the contribution degree of different factors.

3. The time series and data-driven based power quality zonal governance method of claim 2, wherein, According to the comprehensive coupling degree between the nodes in the power grid, the comprehensive coupling degree matrix of all nodes in the power grid is constructed; The comprehensive coupling degree matrix is used as the adjacency matrix of the graph; According to the adjacency matrix, the Laplacian matrix is calculated; The Laplacian matrix is decomposed to construct a new feature space; In the new feature space, the K-means clustering algorithm is applied to cluster the nodes in the power grid, and the regional division of the power grid is realized. The power quality time sequence of each node is obtained by performing the following operations:

4. The time series and data-driven based power quality zonal governance method according to any one of claims 1-3, characterized in that, Using the power grid monitoring system, the voltage data of each node of the power grid at a plurality of continuous sampling points is collected to form the original voltage waveform data of the corresponding node; According to the original voltage waveform data of each node, the multiple power quality indicators of the corresponding node are calculated, including voltage deviation, harmonic voltage content and three-phase imbalance; Each item of power quality indicators of each node is standardized and arranged in order according to the sampling point sequence to construct the power quality subsequence of the corresponding power quality indicator. The behavior similarity distance between different nodes is obtained by performing the following operations:

5. The time series and data-driven based power quality zonal governance method of claim 4, wherein, Each item of power quality subsequence of each node is linearly represented and linearly regressed to obtain a plurality of segmented time subsequences of the corresponding power quality subsequence; For each item of power quality subsequence, the time sequence pattern matching of the segmented time subsequences between different nodes is performed to obtain the behavior similarity distance between different nodes. The matching distance between all segmented time subsequences of the same item of power quality time subsequence of different nodes is summed to obtain the DTW distance of the corresponding nodes in the same item of power quality time subsequence.

6. The time series and data-driven based power quality zonal governance method of claim 5, wherein, Node and between the behaviors of the nodes is represented as: (2) wherein, , , respectively represent the DTW distance of the power quality time sub-sequences of the voltage deviation index, the harmonic content index, the three-phase imbalance index between the nodes and .

7. The time series and data-driven based power quality zonal governance method of claim 6, wherein, Node And Segmented time sub-sequences in the same power quality time sub-sequence At index , Match distance at index , Is represented as: (3) wherein denotes the cumulative matching distance between the first data points of the first data set and the first data points of the second data set, is the weighted distance between the data at index and the data at index ; denotes the time weight factor at index ; denotes the time weight factor at index ; denotes the time weight factor at index ; Assuming segmented time sub-sequences , of length , , denote the matching distance between segmented time sub-sequences , , The cooperative management of the corresponding power grid partition is performed by the following operations:

8. The time series and data-driven based power quality zonal governance method of claim 7, wherein, weighted distance is represented as: (4) in, Representation node The segmented time subsequences are indexed The data at Representation node The segmented time subsequences are indexed The data at.

9. The time series and data-driven based power quality zonal governance method of claim 4, wherein, For each power grid partition, the power quality indicators of the dominant node are monitored in real time, and if the trigger condition of the corresponding power quality indicator is met, the cooperative action instruction of the corresponding power quality indicator dimension is triggered in the corresponding power grid partition; The power quality management equipment of the corresponding power quality indicator dimension in the corresponding power grid partition executes the power quality management according to the corresponding cooperative action instruction to realize the cooperative management of the corresponding power grid partition. The node with the maximum average value of the comprehensive coupling degree with all other nodes in each power grid partition is taken as the dominant node.

10. The time series and data-driven based power quality zonal governance method of claim 9, wherein, ​

Images