Data processing method for power distribution device

Abstract

The present invention relates to the field of data processing technology, and in particular to a data processing method for an electric power distribution device, comprising: constructing a data point set for reflecting the electric power load in the current area based on the collected electric power consumption data of users in the current area, clustering the data points in the data point set multiple times using the K-means clustering algorithm; obtaining multiple circular rings of each category in each clustering result, and determining the data density of each circular ring; sequentially determining the rationality of the intra-class density change, the rationality of the inter-class distribution, and the degree of preference of each clustering result; and determining the optimal clustering result of the K-means clustering algorithm based on the magnitude of the degree of preference, thereby realizing data processing of the electric power distribution device. The present invention deeply analyzes the distribution characteristics of the data points within the cluster category, taking into account the rationality of the intra-class density change and the rationality of the inter-class distribution, ensuring that the clustering results are more consistent with the spatial distribution and density differences of the actual electric power load, and more truly reflecting the electric power consumption behavior in different areas.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a data processing method for a power distribution device. Background Art

[0002] The development of data processing methods for power distribution devices stems from the needs of smart grid construction and the digital transformation of energy. Its core goal is to improve grid reliability, efficiency, and intelligence. While the widespread use of smart meters, sensors, and IoT technologies has enabled high-frequency data collection, the industry still faces challenges such as data silos, insufficient real-time performance, and security risks.

[0003] Data processing for power distribution devices aims to accurately classify and detect anomalies in electricity usage behavior by cleaning, extracting features, and analyzing patterns in user electricity usage data collected by devices such as smart meters. This not only optimizes grid operation but also assists with grid dispatching and energy distribution by identifying peak loads, balanced, or intermittent electricity usage patterns, ensuring power safety and detecting abnormal behaviors such as electricity theft and equipment failures. It also supports demand-side response, predicts regional electricity usage trends, and promotes renewable energy integration. Ultimately, it provides data-driven decision support for power companies, enabling safe, economical, and sustainable power system operations. The K-means clustering algorithm is an unsupervised learning algorithm commonly used for classifying electricity user behavior patterns and extracting features from electricity data. By clustering data points with similar usage patterns into one category, it helps achieve data processing goals such as identifying peak loads, detecting abnormal electricity usage, and optimizing grid dispatch.

[0004] However, when using the K-means clustering algorithm to cluster user electricity usage data, the number of cluster categories, K, must be manually set in advance. However, for different urban areas, manually setting the number of clusters cannot accurately reflect the various electricity usage situations in the current area. For example, when the number of cluster categories is too small, the clustering results are coarse, and user electricity usage data of different categories may be grouped into the same cluster, which does not accurately reflect electricity usage and reduces the accuracy of electricity behavior classification and anomaly monitoring. When the number of cluster categories is too large, the clustering results may be too detailed, easily grouping similar samples into different clusters. The clustering results are also unstable and sensitive to small changes in user electricity usage data, resulting in the clustering results being unable to effectively guide grid scheduling and energy allocation decisions. Summary of the Invention

[0005] In order to solve the problem that when clustering user electricity consumption data using the K-means clustering algorithm, the number of clusters set manually cannot accurately reflect the various electricity consumption situations in each area; when the number of cluster categories is too small, the electricity consumption situation cannot be accurately reflected, which reduces the accuracy of electricity consumption behavior classification and anomaly monitoring; when the number of cluster categories is too large, the clustering results cannot effectively guide power grid scheduling and energy allocation decisions, the present invention proposes a data processing method for a power layout device, which includes the following steps:

[0006] Based on the collected electricity consumption data of users in the current area, a data point set for reflecting the electricity load in the current area is constructed, and the data points in the data point set are clustered multiple times using the K-means clustering algorithm; any clustering result is recorded as the target result, and any category in the target result is recorded as the current category, and multiple concentric circles of the current category are constructed to obtain multiple rings of the current category. According to the number of data points in each ring of the current category and the area of ​​each ring, the data density of each ring of the current category is determined; according to the data density of each ring of each category in the target result, the rationality of the density change within the target result is determined; the conventional clusters and sparse clusters of the current category, as well as the sparse area of ​​the current category are obtained. Domain, the conventional clusters, sparse clusters and sparse areas all contain a number of rectangles. According to the width and length of the sparse areas of each category in the target result, the number of rectangles in the sparse clusters, the number of rectangles in the sparse areas, the number of data points in the sparse areas, and the number of data points in the conventional clusters, the rationality of the inter-class distribution of the target result is determined; according to the distance from the data points in each category to the intra-class center points of each category in the target result, the distance from the intra-class center points between categories, and the rationality of the intra-class density change and the rationality of the inter-class distribution, the priority of the target result is determined; according to the size of the priority, the optimal clustering result of the K-means clustering algorithm is determined to realize data processing of the power layout device.

[0007] The present invention constructs concentric circles and calculates the data density within each circle, systematically and deeply analyzes the distribution characteristics of data points within the cluster category, taking into account the "rationality of density changes within the category" and the "rationality of distribution between categories", ensuring that the clustering results are more in line with the spatial distribution and density differences of the actual power load, and more truly reflect the power consumption behavior in different regions; by evaluating multiple clustering results, the clustering result with the highest preference is selected, so that the number of cluster categories is no longer manually set, but is automatically determined by the structural characteristics of the data itself, thereby improving the scientificity and effectiveness of classification and anomaly monitoring; after accurate classification, sparse areas and regular areas of power consumption behavior can be more effectively identified, providing a more robust basis for abnormal power consumption detection, thereby enhancing the system's anomaly detection and early warning capabilities; with accurate cluster category division and sparse area positioning, it can provide more targeted regional power load information for the power grid dispatching department, optimize resource allocation, and improve the economy and safety of power grid operation.

[0008] Furthermore, the construction of the data point set for reflecting the electricity load in the current area includes: using the peak period electricity consumption ratio in the user electricity consumption data in the current area as the horizontal coordinate and the average daily electricity consumption as the vertical coordinate to construct a two-dimensional coordinate system to obtain the data point set for reflecting the electricity load in the current area.

[0009] Furthermore, obtaining multiple circular rings of the current category includes: constructing a minimum circumscribed circle containing all data points in the current category with the center point within the current category as the center of the circle, constructing multiple concentric circles of the minimum circumscribed circle according to a preset scaling ratio, dividing the minimum circumscribed circle into multiple circular rings, and sorting the circular rings in sequence from the inside to the outside to obtain multiple circular rings of the current category.

[0010] Furthermore, the data density satisfies:

[0011] Where, The ring of the current category The data density, The ring of the current category The number of data points in is the preset zoom ratio, is the serial number of the ring of the current category, is the radius of the minimum circumscribed circle of the current category, is the maximum and minimum normalization function.

[0012] The present invention eliminates the area of ​​the previous circular ring to accurately calculate the density of data points per unit area in each circular ring, avoiding the confusion caused by simple overall density and reflecting a more realistic spatial distribution of power load. With the help of maximum-minimum normalization, the data density between different categories and different circular rings is scaled to a unified standard range, which facilitates cross-category rationality evaluation and subsequent clustering optimization judgment. Accurate density calculation helps identify the boundaries between conventional clusters and sparse clusters, judges the clustering quality from the data density change trend, and promotes adaptive adjustment of clustering results.

[0013] Furthermore, the rationality of the intra-class density change satisfies:

[0014] Where, The rationality of the intra-class density change of the target result, is the number of categories in the target result, is the number of rings of the category in the target result, Category in the target result Ring The data density, Category in the target result Ring The data density, The length of the preset movement unit.

[0015] The present invention utilizes The deformation of the function smoothes the density differences between adjacent circles, avoiding drastic jumps in simple differences and effectively reflecting the rationality of the decrease or increase in density within the category. Through a comprehensive evaluation of the changes in data density of all categories and all circles, the uniformity and continuity of the data distribution within the category in the clustering results are objectively measured, and abnormal or unreasonable density mutations are identified.

[0016] Furthermore, the obtaining of regular clusters and sparse clusters of the current category includes: obtaining the distance between the intra-class center points of the current category and the remaining categories in the target result, taking the intra-class center points of the current category and the category closest to the current category as the midpoints of any two opposite sides of the square, taking the distance between the intra-class center points of the current category and the category closest to the current category as the side length of the square, starting from any side of the square, dividing the square into multiple rectangles of the same area in sequence to obtain multiple rectangles of the current category; based on the number of data points in the rectangles, clustering the multiple rectangles of the current category using the K-means clustering algorithm to obtain two clusters, recording the cluster with the largest number of data points as the regular cluster, and the other as the sparse cluster, to obtain regular clusters and sparse clusters of the current category.

[0017] Furthermore, the sparse region of the current category is obtained by merging all adjacent rectangles belonging to the same sparse cluster, and recording the region with the largest area after merging as the sparse region in the sparse cluster.

[0018] Furthermore, the rationality of the inter-class distribution satisfies:

[0019] Where, is the rationality of the inter-class distribution of the target results, is the number of categories in the target result, and They are the categories in the target results The width and length of the sparse region, and They are the categories in the target results The number of sparse regions and rectangles within sparse clusters, Category in the target result The number of data points in the sparse region, Category in the target result The mean number of data points within all rectangles of a regular cluster.

[0020] The rationality of inter-class distribution in the present invention combines the geometric dimensions of sparse areas (width-to-length ratio), the ratio of the number of rectangles in sparse areas to sparse clusters, and the density characteristics of data points in sparse areas and conventional clusters to reflect the spatial structure and distribution patterns between categories. The influence of density on region size and shape is expressed in the form of exponential weights, making the distribution of sparse and dense regions in the clustering results more scientific and reasonable, and avoiding excessive ambiguity or overlap of category boundaries.

[0021] Furthermore, the preference satisfies:

[0022] Where, is the preference of the target result, is the number of categories in the target result, Category in the target result The number of data points in Category in the target result No. data points to categories The distance between the center points of the class, Category in the target result With category The distance between the center points of the class, The rationality of the intra-class density change of the target result, The inter-class distribution of the target results is reasonable.

[0023] The optimization degree of the present invention reflects the compactness and discrimination of clustering through the ratio of the average distance from the data points to the center within the class (compactness) to the distance between the class centers (separation), thereby realizing a scientific evaluation of the quality of the clustering structure; combined with the mean of the rationality of the density change within the class and the rationality of the distribution between classes, it ensures that the clustering results are not only compact and clearly differentiated, but also conform to the actual characteristics of the power load in terms of data density distribution.

[0024] Furthermore, the data processing of the power distribution device is realized, including: taking the clustering result corresponding to the maximum value of the preference of all clustering results as the best clustering result of the current area using the K-means clustering algorithm to complete the data processing of the power distribution device.

[0025] The present invention has the following beneficial effects:

[0026] By clustering the same set of data points multiple times and calculating the "rationality of density changes within a class" and "rationality of distribution between classes", the clustering quality under different K values ​​can be objectively evaluated, thereby automatically selecting the optimal K value and avoiding the subjective bias caused by the artificial setting of the K value in traditional methods; introducing concentric circles and ring structures to evaluate the density changes of each cluster data and distinguish between regular clusters and sparse clusters can more finely characterize the distribution characteristics of electricity load in each region and improve the accuracy of the classification of electricity behavior types; by identifying sparse clusters and their sparse areas, abnormal points or abnormal areas in electricity consumption data can be effectively discovered, thereby improving the warning sensitivity for atypical electricity consumption patterns (such as sudden spikes, equipment failures, etc.); finally, the optimal number of cluster categories is obtained, which can avoid the defect of inaccurate clustering caused by the number of cluster categories not conforming to the data distribution characteristics, effectively improving the accuracy of the K-means clustering algorithm and the accuracy of identifying the current area's electricity consumption pattern. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a flowchart of the steps of a data processing method for a power layout device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0028] The following is a clear and complete description of the technical solutions in the embodiments of the present invention. The described embodiments are part of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of the present invention.

[0029] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0030] See also Figure 1 , which shows a flowchart of a data processing method for a power layout device provided by an embodiment of the present invention, the method comprising the following steps:

[0031] In order to solve the problem that when clustering user electricity consumption data using the K-means clustering algorithm, the number of clusters set manually cannot accurately reflect the various electricity consumption situations in each area; when the number of cluster categories is too small, the electricity consumption situation cannot be accurately reflected, which reduces the accuracy of electricity consumption behavior classification and anomaly monitoring; when the number of cluster categories is too large, the clustering results cannot effectively guide power grid scheduling and energy allocation decisions, the present invention proposes a data processing method for a power layout device, which includes the following steps:

[0032] S01: Based on the collected electricity consumption data of users in the current area, a data point set for reflecting the electricity load in the current area is constructed, and the data points in the data point set are clustered multiple times using a K-means clustering algorithm.

[0033] The implementer can set the number of cluster categories for each clustering according to the specific implementation situation. For example, the number of cluster categories can be set to 2, 3, 4, ..., ( is the number of data points).

[0034] Specifically, constructing a data point set for reflecting the electricity load in the current area includes:

[0035] A two-dimensional coordinate system is constructed using the peak period electricity consumption ratio in the user electricity consumption data of the current area as the horizontal axis and the average daily electricity consumption as the vertical axis to obtain a data point set for reflecting the electricity load in the current area.

[0036] S02: Obtain multiple circles of each category in each clustering result, and determine the data density of each circle.

[0037] It's important to note that data density intuitively reflects the compactness of data distribution within the cluster space. Calculating the ring data density effectively assesses the uniformity and rationality of the distribution of data across categories within the clustering results. Therefore, this step, analyzing the ring data density, lays the foundation for subsequent evaluation of clustering effectiveness.

[0038] Any clustering result is recorded as the target result, and any category in the target result is recorded as the current category. Multiple concentric circles of the current category are constructed to obtain multiple rings of the current category. The data density of each ring of the current category is determined based on the number of data points in each ring of the current category and the area of ​​each ring.

[0039] Specifically, obtaining multiple rings of the current category includes:

[0040] With the center point of the current category as the center, construct the minimum circumscribed circle containing all data points in the current category. According to the preset scaling ratio, construct multiple concentric circles of the minimum circumscribed circle, divide the minimum circumscribed circle into multiple rings, and sort the rings from the inside to the outside to obtain multiple rings of the current category.

[0041] Implementers can set the scaling ratio and the number of concentric circles according to the specific implementation situation. For example, the scaling ratio is 0.1 times the length of the radius of the minimum circumscribed circle each time (that is, multiple concentric circles are made with 0.1, 0.2, 0.3, ..., 0.9 times the length of the radius of the minimum circumscribed circle); the number of concentric circles is 9.

[0042] Specifically, the data density satisfies:

[0043] ;

[0044] Where, The ring of the current category The data density, The ring of the current category The number of data points in is the preset zoom ratio, , is the serial number of the ring of the current category, is the radius of the minimum circumscribed circle of the current category, is the maximum and minimum normalization function.

[0045] S03: Determine the rationality of the intra-class density changes of each clustering result.

[0046] It should be noted that the rationality of intra-class density variation is used to quantify whether the trend of data point density changes from the center to the boundary of the cluster results conforms to the ideal distribution. Therefore, by constructing an index of intra-class density variation rationality in this step, we can effectively identify areas of abnormal data distribution in the cluster results.

[0047] According to the data density of each circle in each category of the target result, the rationality of the density change within the target result is determined.

[0048] Specifically, the rationality of the intra-class density change satisfies:

[0049] ;

[0050] Where, The rationality of the intra-class density change of the target result, is the number of categories in the target result, is the number of rings of the category in the target result, Category in the target result Ring The data density, Category in the target result Ring The data density, The length of the preset movement unit.

[0051] Implementers can set the length of the moving unit according to specific implementation conditions, for example, 5.

[0052] in, for function The deformation will As an independent variable, and The function graph moves 5 units to the left. Since the ideal cluster density distribution characteristic is that the closer to the center of the category, the more data points there are and the denser the distribution; the closer to the category boundary, the fewer data points there are and the more discrete the distribution, such a clustering effect not only makes the center point within the category more representative of the entire category, but also can judge whether the current clustering effect is good. When , it indicates the category in the target result Ring The density of data points inside is greater than that of the ring The density of data points within the category Ring Extending outward to the ring When , the density of data points shows a decreasing trend, which is consistent with the ideal intra-class density change trend, and The larger the value, the more reasonable the intra-class density change of the target result is; when When , it indicates the category in the target result Ring The density of data points inside is less than that of the ring The density of data points within the category Ring Extending outward to the ring When , the density of data points shows an increasing trend, which violates the ideal intra-class density change trend, and The smaller it is, the less reasonable the intra-class density change of the target clustering result is. The application and transformation of the function is to make the formula more sensitive to negative values ​​and smooth to positive values. When an area that does not conform to the density change trend appears, its influence on the entire function value becomes greater, making the rationality of the density change within the class smaller. Therefore, the influence of the area with unreasonable density change on the overall rationality can be highlighted. Here, it is also because under a clustering result, as long as there are individual clustering errors and individual unreasonable clustering situations, it can be said that the clustering effect is poor, but the degree still needs to be judged according to the numerical value.

[0053] S04: Determine the rationality of the inter-class distribution of each clustering result.

[0054] It should be noted that the rationality of inter-class distribution is used to measure the distribution characteristics and boundary clarity of data points between different categories, and to judge the scientific nature of the classification of each category in the clustering results. Therefore, this step achieves a quantitative analysis of the rationality of the inter-class distribution of the clustering results by comprehensively evaluating the sparse regions and cluster characteristics of each category.

[0055] Obtain the regular clusters and sparse clusters of the current category, as well as the sparse areas of the current category. The regular clusters, sparse clusters and sparse areas all contain several rectangles. Determine the rationality of the inter-class distribution of the target result based on the width and length of the sparse areas of each category in the target result, the number of rectangles in the sparse clusters, the number of rectangles in the sparse areas, the number of data points in the sparse areas, and the number of data points in the regular clusters.

[0056] Specifically, obtaining the conventional clusters and sparse clusters of the current category includes:

[0057] Get the distance between the center points of the current category and the remaining categories in the target result. Use the center points of the current category and the category closest to the current category as the midpoints of any two opposite sides of a square. Use the distance between the center points of the current category and the category closest to the current category as the side length of the square. Starting from any side of the square, divide the square into multiple rectangles of equal area in sequence to obtain multiple rectangles of the current category.

[0058] Based on the number of data points in the rectangle, the K-means clustering algorithm is used to cluster multiple rectangles of the current category to obtain two clusters. The cluster with the largest number of data points is recorded as a regular cluster, and the other is a sparse cluster, thus obtaining the regular cluster and sparse cluster of the current category.

[0059] Implementers can set the number of rectangles according to specific implementation circumstances, for example, 10.

[0060] Specifically, the sparse area of ​​the current category is obtained in the following manner:

[0061] All adjacent rectangles belonging to the same sparse cluster are merged, and the area with the largest area after the merger is recorded as the sparse area in the sparse cluster (if there are multiple areas with the largest area, one of them is selected arbitrarily).

[0062] Specifically, the inter-class distribution rationality satisfies:

[0063] ;

[0064] Where, is the rationality of the inter-class distribution of the target results, is the number of categories in the target result, and They are the categories in the target results The width and length of the sparse region, and They are the categories in the target results The number of sparse regions and rectangles within sparse clusters, Category in the target result The number of data points in the sparse region, Category in the target result The mean number of data points within all rectangles of a regular cluster.

[0065] in, Representation category The fusion index of the coupling of spatial shape characteristics and data density differences quantifies the rationality of the distribution of sparse areas in the classification results. If there is no rectangular merging in the sparse clusters of a certain category, the fusion index of this category is 0. The larger the value, the more categories The larger the area with sparse data points between the nearest category, the more likely the category is to be The greater the necessity of existence, and the fewer data points bordering another category, the more reasonable the inter-class distribution is; when The smaller the time, the more categories The smaller the area with sparse data points between the nearest category and the nearest class, the more data points there are at the border with the other category. It is very likely that the two categories should be classified into one category, and there may be clustering errors. The less reasonable the inter-class distribution is. When the value is larger, it means that the sparse area occupies most of the sparse cluster, indicating that the sparse area is more likely to be the inherent boundary between the two categories, and the more reasonable the inter-class distribution is; if When the value is larger, it means that the rectangles in the sparse category are discretely distributed, and the sparse area is difficult to represent the boundary. That is, the possibility of a clear boundary between two categories is smaller, and the rationality of the inter-class distribution is smaller. When the value is larger, it means that there are fewer data points in the sparse area, that is, the possibility that the sparse area is the inherent boundary between two categories is greater, and the rationality of the inter-class distribution is greater; when The smaller it is, the more data points the sparse area contains, that is, the smaller the possibility of a clear boundary between the two categories, and the less reasonable the inter-class distribution.

[0066] S05: Determine the priority of each clustering result.

[0067] It should be noted that the optimization degree comprehensively considers the compactness of data within a cluster, the degree of separation between clusters, and the rationality of density and distribution, comprehensively reflecting the quality of the clustering results. Therefore, this step provides a quantitative basis for selecting the best clustering results by calculating the optimization degree.

[0068] The preference of the target result is determined based on the distance between the data points in each category and the center point of each category, the distance between the center points of the categories, and the rationality of the density change within the category and the rationality of the distribution between categories.

[0069] Specifically, the preference satisfies:

[0070] ;

[0071] Where, is the preference of the target result, is the number of categories in the target result, Category in the target result The number of data points in Category in the target result No. data points to categories The distance between the center points of the class, Category in the target result With category The distance between the center points of the class, The rationality of the intra-class density change of the target result, The inter-class distribution of the target results is reasonable.

[0072] in, Reflects the average distance between the data points within the class and the center point. The smaller the value, the more compact the data within the class. It represents the average distance between the center points of the clusters. The larger the value, the higher the separation between the clusters. The ratio of the two, the rationality of the density change within the class, and the rationality of the distribution between classes work together to enable the optimization degree to accurately evaluate the comprehensive quality of the clustering results.

[0073] S06: Determine the optimal clustering result of the K-means clustering algorithm according to the size of the optimization degree, and realize data processing of the power layout device.

[0074] Specifically, the data processing of the power layout device includes:

[0075] The clustering result corresponding to the maximum value of the preference of all clustering results is taken as the optimal clustering result of the current area using the K-means clustering algorithm, thereby completing the classification of user electricity consumption patterns in the current area. Based on this classification, the power company can implement differentiated electricity consumption management, such as optimizing time-of-use electricity prices, recommending energy storage solutions, or troubleshooting abnormal electricity consumption, thereby completing the full-process data processing of power distribution devices from data collection to intelligent applications, providing effective support for power grid optimization and demand-side management, and completing the data processing of power distribution devices.

[0076] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A data processing method for a power distribution device, characterized in that: include: Based on the collected user electricity consumption data in the current area, a data point set for reflecting the electricity load in the current area is constructed, and the data points in the data point set are clustered multiple times using a K-means clustering algorithm; Record any clustering result as the target result, record any category in the target result as the current category, construct multiple concentric circles of the current category, and obtain multiple rings of the current category. According to the number of data points in each ring of the current category and the area of ​​each ring, determine the data density of each ring of the current category; according to the data density of each ring of each category in the target result, determine the rationality of the density change within the target result; obtain the regular clusters and sparse clusters of the current category, as well as the sparse areas of the current category, wherein the regular clusters, sparse clusters and sparse areas all contain several rectangles, and the current category is obtained. The conventional clustering and sparse clustering method includes: obtaining the distance between the center points of the current category and the rest of the categories in the target result, taking the center points of the current category and the category closest to the current category as the midpoints of any two opposite sides of the square, taking the distance between the center points of the current category and the category closest to the current category as the side length of the square, starting from any side of the square, dividing the square into multiple rectangles of the same area in sequence to obtain multiple rectangles of the current category; based on the number of data points in the rectangle, using the K-means clustering algorithm to cluster multiple rectangles of the current category. Clustering is performed to obtain two clusters. The cluster with the largest number of data points is recorded as a regular cluster, and the other is a sparse cluster, thus obtaining the regular cluster and sparse cluster of the current category; the sparse area of ​​the current category is obtained by merging all adjacent rectangles belonging to the sparse cluster, and recording the area with the largest area after merging as the sparse area in the sparse cluster; the target is determined according to the width and length of the sparse area of ​​each category in the target result, the number of rectangles in the sparse cluster, the number of rectangles in the sparse area, the number of data points in the sparse area, and the number of data points in the regular cluster. The rationality of the inter-class distribution of the results; determine the priority of the target result based on the distance between the data points in each category and the intra-class center points of each category, the distance between the intra-class center points of the categories, and the rationality of the intra-class density change and the rationality of the inter-class distribution; determine the optimal clustering result of the K-means clustering algorithm based on the size of the priority, and realize the data processing of the power layout device, including: taking the clustering result corresponding to the maximum value of the priority of all clustering results as the optimal clustering result of clustering in the current area using the K-means clustering algorithm, and completing the data processing of the power layout device.

2. The data processing method of a power distribution device according to claim 1, characterized in that: The step of constructing a data point set for reflecting the electricity load in the current area includes: A two-dimensional coordinate system is constructed using the peak period electricity consumption ratio in the user electricity consumption data of the current area as the horizontal axis and the average daily electricity consumption as the vertical axis to obtain a data point set for reflecting the electricity load in the current area.

3. The data processing method of a power distribution device according to claim 1, characterized in that: The method of obtaining multiple rings of the current category includes: With the center point of the current category as the center, construct the minimum circumscribed circle containing all data points in the current category. According to the preset scaling ratio, construct multiple concentric circles of the minimum circumscribed circle, divide the minimum circumscribed circle into multiple rings, and sort the rings from the inside to the outside to obtain multiple rings of the current category.

4. The data processing method for a power distribution device according to claim 3, characterized in that: The data density satisfies: ； Where, The ring of the current category The data density, The ring of the current category The number of data points in is the preset zoom ratio, is the serial number of the ring of the current category, is the radius of the minimum circumscribed circle of the current category, is the maximum and minimum normalization function.

5. The data processing method for a power distribution device according to claim 1, characterized in that: The rationality of the intra-class density change satisfies: ； Where, The rationality of the intra-class density change of the target result, is the number of categories in the target result, is the number of rings of the category in the target result, Category in the target result Ring The data density, Category in the target result Ring The data density, The length of the preset movement unit.

6. The data processing method for a power distribution device according to claim 1, characterized in that: The rationality of the inter-class distribution satisfies: ； Where, is the rationality of the inter-class distribution of the target results, is the number of categories in the target result, and They are the categories in the target results The width and length of the sparse region, and They are the categories in the target results The number of sparse regions and rectangles within sparse clusters, Category in the target result The number of data points in the sparse region, Category in the target result The mean number of data points within all rectangles of a regular cluster.

7. The data processing method for a power distribution device according to claim 1, characterized in that: The preference satisfies: ； Where, is the preference of the target result, is the number of categories in the target result, Category in the target result The number of data points in Category in the target result No. data points to categories The distance between the center points of the class, Category in the target result With category The distance between the center points of the class, The rationality of the intra-class density change of the target result, The inter-class distribution of the target results is reasonable.

Images