Data collection system and method based on power distribution
By iterative clustering and feature extraction of power distribution data, dynamically adjusting the merger and splitting of cluster clusters, the problem of iterative self-organizing clustering algorithm relying on fixed thresholds is solved, and the data denoising effect and acquisition accuracy are improved, so as to achieve Accurate analysis of power distribution data.
Patent Information
- Application Number
- CN202411080974.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-08
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2044-08-08
AI Technical Summary
The iterative self-organized clustering algorithm relies on fixed thresholds when denoising power distribution data, resulting in inaccurate clustering results, affecting the data acquisition accuracy and inability to conduct accurate data analysis.
By iteratively clustering the power distribution data, the characteristic values and correlations of the cluster cluster are extracted using data features, the merger and splitting of cluster clusters are dynamically adjusted, and the necessity of merger and splitting is compared with the preset threshold value, and the combination and splitting are optimized. Clustering results ensure the accuracy and adaptability of clustering results.
The accuracy of clustering results is improved, so that clustering results can adapt to the actual distribution of data, improve the data denoising effect, and ensure the acquisition accuracy and analysis of power distribution data.
Smart Images

Figure CN118626893B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of data processing, and in particular to a data acquisition system and method based on power distribution. Background Art
[0002] With the rapid development of social economy, the sharp increase in electricity demand has put forward higher requirements for the stability, security and intelligence level of the power grid. The rise of smart grids has prompted the power system to transform towards deep automation and informatization. By collecting key parameters such as voltage, current intensity, power factor and power quality and conducting subsequent data analysis, it can ensure the efficient operation of the power grid and achieve early fault warning and rapid emergency response.
[0003] During the data collection process of the power distribution system, the collected data is often contaminated by noise due to environmental factors, equipment aging, and electromagnetic interference, which will affect the accuracy and reliability of data collection. In order to ensure the accuracy of subsequent data analysis, clustering algorithms are often used for data denoising. For example, the patent document CN106650228B introduces a noise data processing method based on an improved k-means algorithm. This method selects the cluster center through the farthest priority strategy, and updates the cluster according to the center until the cluster is stable. Then, the environmental threshold is used to filter and remove noise data, which effectively improves the collection accuracy of air temperature data. However, the clustering algorithm is constrained by the artificially set number of clusters, which in turn affects the clustering quality and data denoising effect, resulting in the accuracy of subsequent data analysis being affected. In order to overcome the limitation of the artificially set number of clusters, the prior art often uses an iterative self-organizing clustering algorithm. The algorithm can automatically adjust the number of clusters according to the data distribution during the clustering process, optimize the clustering results by dynamically merging and splitting cluster clusters, and does not require continuous human intervention. In theory, it can achieve a better clustering effect, thereby more accurately removing noise and ensuring high accuracy and high quality of data collection.
[0004] However, the iterative self-organizing clustering algorithm relies on a fixed threshold to guide the merging and splitting of clusters. When it is actually applied to denoising power distribution data, due to the regional and temporal differences in power distribution data, that is, when the data is unevenly distributed, the cluster division may be inaccurate, affecting the clustering effect and the precise removal of noise, affecting the accuracy of data collection, and resulting in the inability to perform accurate data analysis later. Summary of the invention
[0005] In order to solve the problem that when the iterative self-organizing clustering algorithm clusters power distribution data, the strategy of merging and splitting clusters is guided by a fixed threshold, which affects the accuracy of the clustering results, resulting in the inability to accurately denoise the collected distribution data, and affecting the accuracy of the collected data, the present invention proposes a data collection system and collection method based on power distribution, which specifically adopts the following scheme:
[0006] In one aspect, the present invention provides a data collection method based on electric power distribution, comprising:
[0007] Preliminarily collecting power distribution data as initial collected data, performing iterative clustering on the initial collected data, and generating multiple clusters in each iteration;
[0008] Select a cluster and obtain the characteristic value of the cluster by extracting data characteristics;
[0009] Selecting a cluster to be merged corresponding to the cluster, and calculating the necessity of merging the cluster with the corresponding cluster to be merged according to the product of the difference in characteristic values and the difference in the number of data between the cluster and the corresponding cluster to be merged, and the correlation between the cluster and the corresponding cluster to be merged, wherein the necessity of merging is negatively correlated with the product and positively correlated with the correlation;
[0010] Taking the accumulated value of the difference between the characteristic values of the cluster and the remaining clusters as the difference degree of the cluster, and calculating the necessity of splitting the cluster according to the product of the difference degree of the cluster and the variance of the data of the cluster, the necessity of splitting the cluster is positively correlated with the product;
[0011] Based on the comparison result of the necessity of splitting and merging with the preset threshold value, the clustering clusters are merged and split to obtain multiple final clustering clusters, and data denoising is performed according to the final clustering clusters to obtain final collected data.
[0012] The above scheme iteratively clusters the initially collected power distribution data, and performs feature extraction and differential analysis on multiple clusters generated in each iteration. Specifically, for each cluster, the corresponding cluster to be merged is selected, and the difference in characteristic values and the difference in the number of data between the cluster and the corresponding cluster to be merged are analyzed, which can accurately reflect the difference between the two. The size of the difference can reflect the necessity of merging the two. The smaller the difference, the more similar the two are, and the more they need to be merged. The difference between each cluster and the remaining clusters, as well as the variance of the data of each cluster, are further analyzed. The necessity of splitting the cluster clusters is calculated. The variance reflects the stability of the data of the cluster clusters. The more unstable the data in a cluster cluster is and the greater the difference between the data and the other cluster clusters is, the more the cluster cluster needs to be split. By dynamically adjusting the merging and splitting of cluster clusters, the accuracy of the clustering results is improved, so that the clustering results can adapt to the actual distribution of the data and better fit the complex diversity of the power distribution data, thereby improving the denoising effect of the data denoising based on the clustering results, ensuring the collection accuracy of the power distribution data, and enabling subsequent accurate data analysis based on the collected power distribution data.
[0013] Preferably, the necessity of merging the cluster with the corresponding cluster to be merged satisfies the following relationship:
[0014]
[0015] In the formula, For the Clusters and corresponding clusters to be merged The necessity of merger For the The eigenvalues of the clusters, for The characteristic value of For the The number of clusters. for The number of data, For the Clusters and corresponding clusters to be merged The relevance of is the normalization function.
[0016] The above scheme realizes the accurate quantitative evaluation of the necessity of merging clusters with the corresponding clusters to be merged. Based on the evaluation results, it can not only effectively identify similar clusters to be merged and avoid unnecessary cluster subdivision, but also ensure that the merging decision takes into account both the statistical characteristics of the data (eigenvalues and quantity) and the intrinsic connection (correlation) between the data, thereby improving the accuracy and practicality of cluster analysis.
[0017] Preferably, the necessity of splitting the clusters satisfies the following relationship:
[0018]
[0019] In the formula, For the The necessity of splitting clusters, For the The variance of the data for each cluster, is a natural constant, For the The degree of difference between the clusters, and , For the The eigenvalues of the clusters, is the cluster number, the value is taken from All integers in the range, For the The eigenvalues of the clusters, is the number of clusters, are preset parameters.
[0020] The necessity of splitting calculated by the above scheme reflects the rationality of the splitting decision, which not only improves the adaptability of the model to complex data structures and optimizes the classification accuracy, but also The adjustment balances the split sensitivity, making the split strategy more flexible and accurate.
[0021] Preferably, the method of selecting the cluster to be merged corresponding to the cluster is: calculating the Euclidean distance between the center of the cluster and the centers of the remaining clusters, and taking the cluster with the closest Euclidean distance as the corresponding cluster to be merged.
[0022] Preferably, the method of obtaining the characteristic value of the cluster cluster by extracting data features includes: calculating the data aggregation degree of the cluster cluster based on the accumulated distance value between all the data of the cluster cluster and the center of the cluster cluster, and the data aggregation degree is negatively correlated with the accumulated distance value; obtaining the data change trend line of the cluster cluster by data fitting, and calculating the data trend characteristic value of the cluster cluster based on the mean distance between all the data and the change trend line, and the data trend characteristic value is negatively correlated with the mean distance; using the number of data of the cluster cluster, the data aggregation degree, and the data trend characteristic value as the characteristic vector of the cluster cluster, and using the norm of the characteristic vector as the characteristic value of the cluster cluster.
[0023] The above scheme comprehensively evaluates the characteristics of clusters through detailed data feature extraction strategies, significantly enhancing the depth and accuracy of analysis. The negative correlation between data aggregation degree and trend characteristic values ensures the sensitivity of feature extraction and effectively reflects the compactness and dynamic changes of clusters. The construction of cluster feature vectors provides a unified and refined description, which facilitates subsequent comparison and decision-making between clusters.
[0024] Preferably, merging and splitting the cluster clusters based on the comparison results of the splitting necessity and merging necessity with preset threshold sizes includes: presetting a splitting necessity threshold and a merging necessity threshold; if the merging necessity of the cluster cluster and the corresponding cluster cluster to be merged is greater than the merging necessity threshold, merging the two; if the splitting necessity of the cluster cluster is greater than the splitting necessity threshold, splitting the cluster cluster.
[0025] The strategy of the above scheme improves the efficiency and accuracy of cluster processing by setting clear splitting and merging thresholds and automating the decision-making process. It avoids unnecessary splitting operations and reduces the computational burden, while ensuring the timeliness of merging and effectively optimizing resource utilization.
[0026] Preferably, the data aggregation degree of the clustering clusters , For the The degree of data aggregation of clusters. For the The first The data and The Euclidean distance between the centers of the clusters, For the The number of data in the clusters, is a natural exponential function.
[0027] Preferably, the data trend characteristic value , For the The data trend characteristic value of the clusters, For the The average vertical distance between all data in a cluster and the data change trend line.
[0028] Preferably, performing data denoising according to the final clustering to obtain final collected data includes:
[0029] For any final cluster, calculate the mean of the Euclidean distance between any data point in the cluster and the center of the cluster;
[0030] If the Euclidean distance between a data point and the center of the cluster is greater than the mean of the Euclidean distances, the data point is a noise data point;
[0031] Smooth the value of the noise data point to the value of the center of the cluster to complete data denoising;
[0032] All final clusters are denoised and used as the final collected data.
[0033] The above scheme realizes accurate identification and effective elimination of noise data points by accurately calculating the mean Euclidean distance between the data points in each cluster and the cluster center, significantly enhancing the purity of the data and the reliability of analysis. By smoothly replacing the identified noise points with the cluster center values, not only the interference is removed, but also the continuity and trend of the data sequence are maintained, providing a high-quality data foundation for the accurate monitoring and efficient operation of the power system.
[0034] On the other hand, the present invention also provides a data acquisition system based on electric power distribution, the acquisition system includes a memory and a processor, the memory stores a computer program, and the processor executes the computer program to implement the steps of the acquisition method.
[0035] The present invention has the following effects:
[0036] The present invention obtains more accurate clustering results by performing a refined analysis of the clustering process and dynamically adjusting the merging and splitting of cluster clusters. The clustering results can adapt to the actual distribution of data and are more in line with the complexity and diversity of power distribution data. Accurate clustering results ensure the effect of data denoising and improve the accuracy of the collected power distribution data. Accurate data analysis can be performed based on the power distribution data. BRIEF DESCRIPTION OF THE DRAWINGS
[0037] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:
[0038] Figure 1 It is a schematic flow chart of the method of the present invention. DETAILED DESCRIPTION
[0039] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.
[0040] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.
[0041] Reference Figure 1 , a data collection method based on electric power distribution includes steps S1-S5, which are specifically as follows:
[0042] S1: Preliminary collection of power distribution data, and iterative clustering, with each iteration generating multiple clusters.
[0043] In modern power distribution systems, in order to ensure efficient operation and maintenance of the system, sensors are deployed at key nodes such as substations, transmission and distribution lines, and various types of power equipment. These sensors monitor key parameters such as current, voltage, power factor, and power quality in real time at a frequency of once per second (1 Hz), providing valuable data support for health status assessment and fault prevention of the power system. This embodiment takes the current data monitored by the core equipment of the transformer as an example, and preliminarily collects 1 hour of current data as the initial collected data. Taking the initial collected data as an example, it aims to demonstrate a set of efficient data denoising processes to ensure the denoising effect and the collection accuracy of power distribution data. The high-precision distribution data can be used to accurately analyze the operation status of the power grid. The data denoising process is also applicable to the analysis and processing of other parameter data in the distribution system.
[0044] In the data denoising process, the initial collected data is first clustered, and then data denoising is performed based on the clustering results to obtain the final collected data, ensuring the accuracy of the collected data. This step first explains the clustering:
[0045] This step uses an iterative self-organizing clustering algorithm to iteratively cluster the initial collected data. The difference between the iterative self-organizing clustering algorithm and the K-MEANS algorithm is that the iterative self-organizing clustering algorithm will split or merge the clusters generated in each iteration during the clustering process to automatically optimize the number of clusters. When the cluster center distance between two clusters is less than or equal to the set distance threshold, the two clusters are merged. When the number of data in a cluster exceeds the set number threshold, the cluster is split.
[0046] Specifically, we first set the expected number of cluster centers to 5, perform iterative clustering on the collected data, set the number of iterations to 10, and the iterative self-organizing clustering algorithm will iterate 10 times. After each iteration, the multiple clusters generated by this iteration will be split or merged until 10 iterations are completed. The final clustering result is often obtained in which the number of clusters contained in the final clustering result is different from the expected number of cluster centers.
[0047] S2: Select any cluster and obtain the feature value of the cluster by extracting data features.
[0048] In view of the changes of power load in different geographical locations and different time periods, the current data presents a significant uneven distribution. This unbalanced data distribution pattern may lead to two types of problems in cluster analysis: first, some clusters are difficult to accurately represent the data characteristics of the region due to sparse data; second, the data-dense clusters may mask the subtle changes inside and mistakenly regard the data points of normal fluctuations as abnormal noise. In the clustering process, the traditional iterative self-organizing clustering algorithm simply relies on the distance between cluster centers or the number of data points in the cluster to merge and split clusters. It may ignore the specific characteristics of data distribution and mistakenly mark some normal data as outliers, reducing the accuracy and practicality of the analysis. Therefore, the method of in-depth analysis of the internal data characteristics of clusters is used to guide the adjustment of cluster structure, which can more accurately determine when to merge too sparse clusters to integrate similar patterns, or split dense clusters at the right time, so as to ensure that the clustering results can not only accurately reflect the real distribution of data, but also provide strong data support for the evaluation of the operation status of the power system, and improve the accuracy of data analysis and the effectiveness of decision-making.
[0049] The method of obtaining the characteristic value of the cluster by extracting data characteristics includes:
[0050] S21: Calculate the data aggregation degree of the cluster according to the accumulated value of the distance between all the data of the cluster and the center of the cluster. The data aggregation degree of the cluster is calculated according to the following formula:
[0051]
[0052] In the formula, For the The degree of data aggregation of clusters. For the The first The data and The Euclidean distance between the centers of the clusters, For the The number of data in the clusters, is a natural exponential function.
[0053] This formula uses the cumulative distance between all the data in the cluster and the center of the cluster as the negative exponent of the natural exponential function to calculate the degree of data aggregation. The degree of data aggregation is negatively correlated with the cumulative distance value. The smaller the cumulative distance value, the smaller the distance between the data in the cluster and the center of the cluster. The higher the degree of aggregation of the data in the cluster, that is, the more concentrated the data in the cluster, the higher the consistency of the data, which intuitively reflects the compactness of the data within the cluster.
[0054] Furthermore, the degree of data aggregation provides a quantitative basis for the decision of cluster merging or splitting. Clusters with small distance and high degree of aggregation may not need to be split. Otherwise, splitting should be considered, which is conducive to optimizing the clustering structure and improving the adaptability and explanatory power of the model.
[0055] S22: Obtain the data change trend line of the cluster cluster by data fitting. Specifically, the data fitting can adopt linear regression analysis, and the data trend characteristic value of the cluster cluster is calculated according to the mean distance between all data and the change trend line. The data trend characteristic value satisfies the following relationship:
[0056]
[0057] In the formula, For the The data trend characteristic value of the clusters, For the The average value of the vertical distance between all data in a cluster and the data change trend line. The smaller the average value of the vertical distance, the more obvious the change trend of the data in the cluster. The larger the average value of the vertical distance, the less obvious the change trend of the data in the cluster. The data trend characteristic value is negatively correlated with the distance mean.
[0058] When the change trend of the data within the cluster is closer to the data change trend line, the data trend characteristic value of the cluster is higher, reflecting a significant data change trend. Otherwise, it means that the data change trend within the cluster is not obvious. This method provides a quantitative basis for analyzing the dynamic change trend of data, which is conducive to understanding the inherent laws of data changes over time or conditions.
[0059] S23: Taking the number of data, data aggregation degree, and data trend characteristic value of the cluster as the characteristic vector of the cluster , For the The feature vectors of the clusters, For the The number of clusters. For the The degree of data aggregation of clusters. For the The data trend characteristic value of the cluster cluster is obtained, and the norm of the characteristic vector is used as the characteristic value of the cluster cluster, such as The one-norm of ,use As the The eigenvalue of a cluster can reflect the data distribution characteristics of the cluster. The norm of the eigenvector refers to the value obtained by adding the absolute values of all elements in the eigenvector, which is used to measure the size of the eigenvector.
[0060] This method of constructing feature vectors not only takes into account the degree of data concentration and data change trends within the cluster, but also the amount of data, reflecting the comprehensive characteristics of the cluster more comprehensively. The norm of the feature vector, as the eigenvalue of the cluster, is the overall strength measure of the feature vector, measuring the comprehensive influence of the concentration, trend and quantity of data distribution, making the evaluation of the cluster more systematic and quantitative, facilitating further decision support and optimization, such as cluster merging and splitting strategies, and improving the efficiency and accuracy of the overall analysis.
[0061] In summary, through this step, a comprehensive and detailed cluster feature analysis system is formed for each cluster, and each cluster can obtain a feature value according to the method of this step.
[0062] S3: Calculate the necessity of merging each cluster with the corresponding cluster to be merged, and the necessity of splitting each cluster.
[0063] The necessity of merging the cluster with the corresponding cluster to be merged satisfies the following relationship:
[0064]
[0065] In the formula, For the Clusters and corresponding clusters to be merged The necessity of merger For the The eigenvalues of the clusters, for The characteristic value of For the The number of clusters. for The number of data, For the Clusters and corresponding clusters to be merged The relevance of is the normalization function.
[0066] The method for obtaining the cluster to be merged corresponding to the cluster is as follows:
[0067] Calculate the Euclidean distance between the center of the cluster and the centers of the remaining clusters, and use the cluster with the closest Euclidean distance as the corresponding cluster to be merged. For example, there are 3 clusters in total. Take the first cluster as the analysis target, calculate the distance between the center of the first cluster and the centers of the second and third clusters, and select the cluster closest to the first cluster from the second and third clusters as the cluster to be merged corresponding to the first cluster.
[0068] The method for obtaining the correlation is: arranging the data of the cluster and the corresponding cluster to be merged in time sequence, and using (Dynamic Time Warping) algorithm obtains the correlation between the cluster and the corresponding cluster to be merged, The algorithm is a dynamic programming algorithm for calculating the similarity (that is, correlation) of two time series. It is particularly suitable for processing time series data of different lengths or non-aligned times.
[0069] This formula realizes the accurate quantitative evaluation of the necessity of merging clusters with the corresponding clusters to be merged. Based on the evaluation results, it can not only effectively identify similar clusters to be merged and avoid unnecessary cluster subdivision, but also ensure that the merging decision takes into account both the statistical characteristics of the data (eigenvalues and quantity) and the intrinsic connection (correlation) between the data, thereby improving the accuracy and practicality of cluster analysis.
[0070] The necessity of splitting the clusters satisfies the following relationship:
[0071]
[0072] In the formula, For the The necessity of splitting clusters, For the The variance of the data of the clusters. The variance reflects the stability of the data within the clusters. The larger the variance, the lower the stability of the data within the clusters. is a natural constant, For the The degree of difference between the clusters, and , For the The eigenvalues of the clusters, is the cluster number, the value is taken from All integers in the range, For the The eigenvalues of the clusters, is the number of clusters, is the preset parameter, , To constrain the result between 0 and 1, The degree of difference between a cluster reflects the degree of difference between the cluster and the other clusters. The greater the difference between the cluster and the other clusters, the greater the variance of the data within the cluster, which means that it is more necessary to split the cluster.
[0073] The necessity of splitting calculated by this formula reflects the rationality of the splitting decision. It not only improves the model's adaptability to complex data structures and optimizes classification accuracy, but also balances the splitting sensitivity by adjusting parameter A, making the splitting strategy more flexible and accurate, and effectively responding to changes in various data distributions.
[0074] S4: Based on the comparison result of the split necessity and the merge necessity with a preset threshold value, the clustering clusters are merged and split to obtain a plurality of final clustering clusters.
[0075] The split necessity threshold and the merge necessity threshold are preset, both of which are 0.5. If the merging necessity of the cluster cluster and the corresponding cluster cluster to be merged is greater than the merging necessity threshold, the two are merged. If the split necessity of the cluster cluster is greater than the split necessity threshold, the cluster cluster is split.
[0076] According to this merging and splitting rule, multiple clusters in the clustering result generated in each iteration are dynamically adjusted until the iteration ends to obtain a final clustering result, which includes several clusters.
[0077] This strategy sets a threshold to guide the dynamic optimization of cluster structure, achieving the adaptability and flexibility of cluster analysis, which is specifically reflected in:
[0078] Dynamically adjust and optimize the clustering structure: Automatically decide whether to merge or split clusters based on the comparison between the necessity of merging and splitting and the preset threshold. This process dynamically adjusts the clustering structure so that the clustering division is more in line with the actual data distribution, thereby improving the accuracy of the clustering results.
[0079] Prevent over-clustering: The set merging necessity threshold prevents over-splitting, while the splitting necessity threshold avoids unnecessary merging, ensuring that the clustering clusters can capture the subtle structure in the data while avoiding over-segmentation noise interference, thus maintaining the effectiveness and simplicity of clustering.
[0080] Iteration converges to stable results: Each iteration adjusts the clustering according to the rules until convergence is reached. The final clustering result obtained at the end of the iteration ensures the stability and accuracy of the clustering, reflects the true clustering structure of the data, and is conducive to subsequent analysis and decision-making.
[0081] Improve efficiency and effectiveness: This dynamic adjustment strategy avoids the subjectivity of manually presetting the number of clusters. Through iterative automatic optimization, it improves processing efficiency while maintaining the adaptability and accuracy of the model. It can quickly respond to and accurately reflect changes in power distribution data, and guide timely maintenance.
[0082] S5: performing data denoising according to the final clustering clusters to obtain final collected data.
[0083] Since the final clustering cluster is an accurate clustering result, the accuracy of the clustering operation has been improved through the above steps. Then, data denoising is performed on the accurate clustering result to obtain the final high-quality data.
[0084] The process of data denoising, specifically:
[0085] Taking any final cluster as an example, first obtain the Euclidean distance of the data in the cluster from the cluster center, and obtain the mean Euclidean distance of all data points in the cluster and the cluster center, and then analyze each data point in the cluster. If the Euclidean distance between a data point and the cluster center is greater than the mean Euclidean distance, it is considered that the data point deviates from the cluster where it is located, and it is marked as a noise data point, and the value of the data point is smoothed to the value of the cluster center, that is, the value of the noise data point is adjusted to the value of the cluster center where it is located, thereby realizing the data denoising process. For all final clusters, the noise data points are marked and data denoising is performed according to the above method. All final clusters after data denoising are used as the final collected data. The final collected data has high accuracy, which ensures the quality of the collected data. In the subsequent data analysis process, the operating status of the distribution network can be accurately analyzed according to the final collected data for timely maintenance.
[0086] The present invention also provides a data acquisition system based on power distribution, the acquisition system includes a memory and a processor, the memory stores a computer program, the processor executes the computer program to implement the steps of the acquisition method, clusters the initially collected power distribution data, obtains accurate clustering results, performs data denoising based on the accurate clustering results, and obtains the final collected data. It is equivalent to having a built-in data denoising program in the data acquisition system, the data denoising program integrates the steps of the acquisition method, clusters the collected data to identify noise data and smoothes it, obtains the final collected data after denoising, and the final collected data after denoising can be input into the monitoring terminal to monitor the operating status of the distribution network.
[0087] In the description of this specification, "plurality" or "several" means at least two, such as two, three or more, etc., unless otherwise clearly and specifically defined.
[0088] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.
Claims
1. A data collection method based on electric power distribution, characterized in that: include: Preliminarily collecting power distribution data as initial collected data, performing iterative clustering on the initial collected data, and generating multiple clusters in each iteration; The initial collected data is current; Select a cluster and obtain the characteristic value of the cluster by extracting data characteristics; Selecting a cluster to be merged corresponding to the cluster, and calculating the necessity of merging the cluster with the corresponding cluster according to the product of the difference in characteristic values and the difference in the number of data between the cluster and the corresponding cluster, and the correlation between the cluster and the corresponding cluster; Taking the accumulated value of the difference between the characteristic values of the cluster and the remaining clusters as the difference degree of the cluster, and calculating the necessity of splitting the cluster according to the product of the difference degree of the cluster and the variance of the data of the cluster; Based on the comparison result of the necessity of splitting and merging with the preset threshold value, the clustering clusters are merged and split to obtain multiple final clustering clusters, and data denoising is performed according to the final clustering clusters to obtain final collected data; The necessity of the merger satisfies the following relationship: ; In the formula, For the Clusters and corresponding clusters to be merged The necessity of merger For the The eigenvalues of the clusters, for The characteristic value of For the The number of clusters. for The number of data, For the Clusters and corresponding clusters to be merged The relevance of is the normalization function; The necessity of splitting satisfies the following relationship: ; In the formula, For the The necessity of splitting clusters, For the The variance of the data for each cluster, is a natural constant, For the The degree of difference between the clusters, and , For the The eigenvalues of the clusters, is the cluster number, the value is taken from All integers in the range, For the The eigenvalues of the clusters, is the number of clusters, is the preset parameter; The step of obtaining the characteristic value of the cluster by extracting the data characteristics comprises: Calculating the data aggregation degree of the cluster according to the accumulated distance between all the data of the cluster and the center of the cluster, wherein the data aggregation degree is negatively correlated with the accumulated distance value; Obtaining a data change trend line of the cluster cluster by data fitting, and calculating a data trend characteristic value of the cluster cluster according to a distance mean of all data and the change trend line, wherein the data trend characteristic value is negatively correlated with the distance mean; The number of data, the degree of data aggregation, and the data trend characteristic value of the cluster are used as the characteristic vector of the cluster, and the norm of the characteristic vector is used as the characteristic value of the cluster.
2. The data collection method based on electric power distribution according to claim 1 is characterized in that: The method for selecting the cluster to be merged corresponding to the cluster is: The Euclidean distance between the center of the cluster and the centers of the remaining clusters is calculated, and the cluster with the closest Euclidean distance is used as the corresponding cluster to be merged.
3. The data collection method based on electric power distribution according to claim 1 is characterized in that: Merging and splitting the clusters based on the comparison result of the necessity of splitting and the necessity of merging with the preset threshold size includes: Preset split necessity threshold and merge necessity threshold; If the necessity of merging the cluster and the corresponding cluster to be merged is greater than the merging necessity threshold, the two are merged; if the necessity of splitting the cluster is greater than the splitting necessity threshold, the cluster is split.
4. The data collection method based on electric power distribution according to claim 1 is characterized in that: The degree of data aggregation of the cluster , For the The degree of data aggregation of clusters. For the The first Data and The Euclidean distance between the centers of the clusters, For the The number of data in the clusters, is a natural exponential function.
5. The data collection method based on electric power distribution according to claim 1 is characterized in that: The data trend characteristic value , For the The data trend characteristic value of the clusters, For the The average vertical distance between all data in a cluster and the data change trend line.
6. The data collection method based on electric power distribution according to claim 3 is characterized in that: Data denoising is performed according to the final clustering cluster to obtain final collected data, including: For any final cluster, calculate the mean of the Euclidean distance between any data point in the cluster and the center of the cluster; If the Euclidean distance between a data point and the center of the cluster is greater than the mean of the Euclidean distances, the data point is a noise data point; Smooth the value of the noise data point to the value of the center of the cluster to complete data denoising; All final clusters are denoised and used as the final collected data.
7. The data acquisition system based on power distribution is characterized by: The acquisition system includes a memory and a processor, the memory stores a computer program, and the processor executes the computer program to implement the steps of the acquisition method according to any one of claims 1 to 6.
Citation Information
Patent Citations
Improved k-means algorithm for noise removal and its implementation system
CN106650228B
Building energy consumption monitoring method and system
CN117421618A
Intelligent management system for short message data
CN117880765A
Energy storage method based on open source gap system
CN118017564A
Real-time state monitoring method for power distribution circuit
CN118214168A
