WO2022110557A1 - 一种台区户变关系异常诊断方法及装置 - Google Patents

一种台区户变关系异常诊断方法及装置 Download PDF

Info

Publication number
WO2022110557A1
WO2022110557A1 PCT/CN2021/077141 CN2021077141W WO2022110557A1 WO 2022110557 A1 WO2022110557 A1 WO 2022110557A1 CN 2021077141 W CN2021077141 W CN 2021077141W WO 2022110557 A1 WO2022110557 A1 WO 2022110557A1
Authority
WO
WIPO (PCT)
Prior art keywords
abnormal
station area
user
voltage data
clustering
Prior art date
Application number
PCT/CN2021/077141
Other languages
English (en)
French (fr)
Inventor
黄瑞
刘谋海
肖宇
刘朝阳
叶志
刘小平
杨茂涛
陈浩
卿曦
王智
叶浏青
曾文伟
Original Assignee
国网湖南省电力有限公司
国网湖南省电力有限公司供电服务中心（计量中心）
国家电网有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 国网湖南省电力有限公司, 国网湖南省电力有限公司供电服务中心（计量中心）, 国家电网有限公司 filed Critical 国网湖南省电力有限公司
Publication of WO2022110557A1 publication Critical patent/WO2022110557A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G06F18/232—Non-hierarchical techniques
    • G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R35/00—Testing or calibrating of apparatus covered by the other groups of this subclass
    • G01R35/04—Testing or calibrating of apparatus covered by the other groups of this subclass of instruments for measuring time integral of power or current
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
    • G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/08—Probabilistic or stochastic CAD
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2113/00—Details relating to the application field
    • G06F2113/04—Power grid distribution networks
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
    • Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
    • Y04S10/00—Systems supporting electrical power generation, transmission or distribution
    • Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications

Definitions

  • the invention relates to the technical field of electricity consumption information collection systems, and in particular, to a method and device for diagnosing abnormal relationship between households in a station area.
  • the power consumption information collection system containing massive data can not only directly reflect the operation status of the distribution network, but also indirectly reflect the topology relationship of the distribution network.
  • Existing distribution network topology check mainly includes: line-to-line relationship check, feeder topology check, household change relationship and phase check, line-to-house relationship check.
  • the correct low-voltage distribution network topology, especially the correct household-change relationship is the basis for the refinement of the current distribution network management and the reduction of losses.
  • the electricity consumption information collection system is prone to the problem of incorrect relationship between households in the station area.
  • the household change relationship is inconsistent with the actual situation, resulting in abnormal situation in the calculation of line loss in the station area.
  • the main form of household change relationship errors in low-voltage station areas is that the user files recorded in the main station system do not match the actual information, and usually a user in a station area is mistakenly connected to an adjacent station area, or a feeder transformer is mistakenly connected Connected to the adjacent feeder, this situation will affect the correct household change relationship and hinder the normal line loss calculation.
  • the prior art there are mainly two ways to verify the relationship between household changes in the station area: manual special equipment on-site identification and online automatic identification.
  • the manual method needs to arrange staff to identify on-site, and the identification efficiency is low and the cost is high.
  • Online automatic identification can solve the problem of manual identification.
  • the current online automatic identification method is mainly based on data mining technology, by extracting the characteristics of abnormal users in the station area to identify the station area with the same characteristics, for example, by extracting the similarity feature of the user voltage curve to realize the abnormal diagnosis of the household change relationship in the station area.
  • the voltage In the low-voltage distribution network, due to the uncertainty of the load everywhere, the voltage usually fluctuates accordingly.
  • the load with electrical distance is closer, and the voltage fluctuation curve is similar, while the similarity of the voltage fluctuation curve of the load with far electrical distance is lower.
  • the similarity of the user voltage curve can be selected as the basis for the abnormal diagnosis of the household change relationship in the station area.
  • the above-mentioned online automatic identification method needs to rely on the feature extraction of abnormal users, and can only be used for a small number of abnormal users in the station area.
  • the present invention provides a method and device for diagnosing abnormal relationship between households in a station area with a simple implementation method, high diagnostic efficiency and high precision, which can realize the same station area. Diagnosis of one or more abnormal users and multiple abnormal users in different stations.
  • the technical scheme proposed by the present invention is:
  • a method for diagnosing abnormality of household change relationship in Taiwan district comprising the steps of:
  • the K-means clustering method is used to cluster the dimension-reduced voltage data, and in the clustering process, the initial clustering center is selected according to the maximum and minimum values of the data dimension, and the designated area is found out. Abnormal users with abnormal household change relationship;
  • Station area diagnosis Calculate the Pearson correlation coefficient between the abnormal user and each station area master table, sort each of the calculated Pearson correlation coefficients, and diagnose the abnormal user according to the sorting result. the correct station area.
  • step of selecting the initial cluster center in the step S02 includes:
  • zij is the dimensionality reduction voltage data
  • ⁇ r is the updated cluster center of each cluster
  • is the maximum distance between each sample and the current nearest cluster center
  • k initial cluster centers are selected according to the following formula
  • ⁇ r is the initial cluster center
  • ma is the maximum number of each dimension
  • mi is the minimum number of each dimension.
  • the station area corresponding to the largest Pearson correlation coefficient in the sorting result is specifically selected as the correct station area to which the abnormal user belongs.
  • the step S03 also includes first calculating the correlation values between the voltage data of all the abnormal electricity meters found, and judging the abnormal electricity meters whose correlation value is less than the preset threshold as belonging to the same area, from the judgment as One of the abnormal meters in the same station is selected as the meter to be detected, and then the Pearson correlation coefficient between the meter to be detected and the total meter of each station is calculated to diagnose the correct station to which it belongs.
  • clustering is performed on all the found abnormal electricity meters to form multiple clusters, and the abnormal electricity meters that are the cluster centers of each cluster are obtained as the electricity meters to be detected, and then the electricity meters to be detected and each station are calculated separately.
  • the Pearson correlation coefficient between the district total meters is used to diagnose the correct station area to which the target abnormal meter belongs, and the correct station area to which each abnormal meter belongs in the entire cluster is obtained from the correct station area to which the target abnormal electricity meter belongs.
  • the step S03 further includes taking the diagnosed correct station area as the destination station area, acquiring the user voltage data of the destination station area and returning to step S02 to re-run. Perform clustering processing, and finally confirm whether the abnormal user belongs to the destination station area according to the clustering result.
  • a principal component analysis method is specifically used to perform dimensionality reduction processing.
  • a device for diagnosing abnormal relationship between households in Taiwan area comprising:
  • the dimension reduction processing module is used to obtain the voltage data of each user in the designated station area and perform dimension reduction processing to obtain the dimension reduction voltage data;
  • the clustering processing module is used for clustering the dimension-reduced voltage data using K-means clustering method, and in the clustering process, selects the initial clustering center according to the maximum and minimum values of the data dimension, and finds out the designated station area Abnormal users with abnormal intra-internal change relationship are regarded as users to be detected;
  • the station area diagnosis module is used to calculate the Pearson correlation coefficient between the to-be-detected user and each station area master table, sort each of the calculated Pearson correlation coefficients, and diagnose the abnormality according to the sorting result The correct zone to which the user belongs.
  • a device for diagnosing abnormal relationship between households in a station area comprising a processor and a memory, the memory is used to store a computer program, the processor is used to execute the computer program, and the processor is used to execute the computer program to execute method as above.
  • the present invention uses improved K-means clustering to extract voltage data features after obtaining the voltage data of the general meter of the station area and the user's electricity meter to reduce the dimension, and selects the initial cluster center according to the maximum and minimum values of the data dimension during clustering, It makes it possible to quickly and accurately find all abnormal users in the station area, and use the improved Pearson correlation coefficient method for the abnormal users found.
  • the size of the Pearson correlation coefficient between the abnormal users and the general table of each station area is calculated It can accurately extract the characteristics of abnormal users in multiple adjacent station areas, and realize the diagnosis of the correct station area to which multiple abnormal users belong, so that one or more abnormal users in the same station area and multiple abnormal users in different station areas can be diagnosed. It can effectively realize the accurate detection and analysis of abnormal users in all user situations.
  • the present invention can efficiently realize feature information mining among voltage data.
  • a better clustering center is selected according to the maximum and minimum data dimensions, so that different feature anomalies can be obtained.
  • the Pearson correlation coefficient to characterize the degree of correlation between abnormal users and the station area, by taking the maximum correlation coefficient between the user to be detected and the general table of each station area as the correct station area, it can be effectively avoided.
  • the selection of the reference value of the Pearson correlation coefficient improves the detection efficiency and accuracy, and is especially suitable for the automatic diagnosis of the household change relationship when there are multiple abnormal users in the same station area and multiple abnormal users in the adjacent station area.
  • the present invention further divides the abnormal station area according to the correlation, determines the abnormal electricity meter whose correlation value is less than the preset threshold value as belonging to the same station area, and selects one of the abnormal electricity meters determined as the same station area as the The meter to be tested, and then calculate the Pearson correlation coefficient between the meter to be tested and the total meter of each station area to diagnose the correct station area, so that only a small amount of calculation is needed to realize the station area diagnosis of all abnormal electricity meters, no need Diagnosing all abnormal electricity meters one by one can further improve the efficiency of diagnosis and avoid a lot of unnecessary calculations, especially when there is a large-scale abnormality in the division of electricity meters, it can quickly diagnose the correct station area to which all electricity meters belong.
  • FIG. 1 is a schematic flowchart of the implementation of the method for diagnosing the abnormality of the household change relationship in the station area according to the present embodiment.
  • FIG. 2 is a schematic diagram of the principle of selecting a voltage data clustering center in this embodiment.
  • FIG. 3 is a schematic diagram of a simulation result obtained in a specific application example.
  • FIG. 4 is a schematic diagram of the data length influence analysis result obtained in the specific application embodiment.
  • the steps of the method for diagnosing abnormality in the relationship between household changes in Taiwan districts of the present embodiment include:
  • K-means clustering method is used to cluster the dimensionality-reduced voltage data, and in the clustering process, the initial clustering center is selected according to the maximum and minimum values of the data dimension, and the household change in the designated station area is found.
  • Abnormal users with abnormal relationships
  • Station area diagnosis Calculate the Pearson correlation coefficient between the abnormal user and the general table of each station area, sort the calculated Pearson correlation coefficients, and diagnose the correct station area to which the abnormal user belongs according to the sorting result.
  • the improved K-means clustering is used to extract the characteristics of the voltage data, and the initial cluster center is selected according to the maximum and minimum values of the data dimension during clustering, and Instead of randomly selecting the initial cluster center as in the traditional method, all abnormal users in the station area can be quickly and accurately found.
  • the improved Pearson correlation coefficient method is used for the found abnormal users. The relationship between the Pearson correlation coefficients between the two can diagnose the correct station area, and can accurately extract the characteristics of abnormal users in multiple adjacent station areas, and realize the diagnosis of the correct station area to which multiple abnormal users belong, so that one or more users in the same station area can be diagnosed. It can effectively realize the accurate detection and analysis of abnormal users in the case of one abnormal user and multiple abnormal users in different stations.
  • the PCA method of principal component analysis is used to perform dimension reduction processing on the original voltage data.
  • the voltage data of the user's electricity meter in the station area is collected every 1 hour for 24 hours a day.
  • the traditional clustering algorithm will face the problem that the high-dimensional data contains massive redundant and irrelevant information. Clustering high-dimensional data greatly reduces performance, and it is difficult for the clustering algorithm to achieve high stability.
  • the PCA method is used to perform dimensionality reduction processing on the original data, which can facilitate subsequent clustering processing on the voltage data to find abnormal users. .
  • Step S101 Embody p-dimensional voltage data features in m-dimension based on PCA.
  • the m-dimensional data information is also called principal components.
  • n users in the station area are The p-dimension voltage data of the electric meter is expressed as:
  • x i and x j are the voltage data of the station area, and x is the average value of the voltage data.
  • the eigenvalue ⁇ 1 ⁇ 2 ⁇ ... ⁇ p ⁇ 0 can be obtained.
  • Step S102 Determine the number m of principal components according to the principal component contribution rate method (CPV), and calculate the contribution rate CPV i and the cumulative contribution rate CPV a of each principal component as:
  • CPV i is the contribution rate of the i-th principal component
  • ⁇ m is the eigenvalue corresponding to the m-th (m ⁇ p) principal component
  • CPV a is the control limit
  • K-means clustering is based on the iterative theory to find the maximum number of iteration steps or make the clustering error function converge to obtain the cluster center.
  • the initial clustering center is randomly selected. If the initial clustering center is not selected properly, it will greatly affect the final clustering result.
  • This embodiment improves the traditional K-means clustering. According to the maximum and minimum data dimensions, the initial cluster centers are selected based on the principle of obtaining the initial cluster centers that are as far away from each other as possible, rather than random selection as in the traditional method. The initial cluster center can avoid the problem of wrong or inappropriate selection of the initial cluster center, thereby greatly improving the detection accuracy and efficiency of abnormal users.
  • step S02 of this embodiment The specific steps of selecting the initial cluster center in step S02 of this embodiment include:
  • ⁇ r is the initial cluster center
  • ma is the maximum number of each dimension
  • mi is the minimum number of each dimension
  • zij is the dimensionality reduction voltage data
  • ⁇ r is the updated cluster center of each cluster
  • is the maximum distance between each sample and the current nearest cluster center
  • k sample points are selected as k initial clustering centers for the dimension-reduced voltage data, and for each sample data in the voltage, the Euclidean distance to the existing nearest clustering center is calculated separately, which is calculated by Equation (9). The larger the distance value, the greater the probability of the sample being selected as the next clustering center.
  • the sample data is classified into the nearest clustering center category to obtain k clusters, and the probability of the sample being selected as the clustering center can be calculated as follows:
  • D(z ij ) is the distance from the sample to the cluster center.
  • the obtained 2-dimensional data are A(0.1, 0.1), B(0.2, 0.2), C(0.2, 0), D(0.4, 0.6 ), E(0.5, 0.6), F(0.5, 0.5), G(0.6, 0.5), and then perform cluster analysis on the voltage data after dimension reduction, and select C 1 (0.2, 0.3) in the first clustering, C 2 (0.5, 0.3) two points as the first and second cluster centers are shown in Figure 2.
  • cluster analysis is performed on all the voltage data of the user's electricity meters in the adjacent stations, so as to find out the abnormality of the voltage data of all the users' electricity meters in the adjacent stations.
  • the Pearson correlation coefficient can represent the linear correlation between the two variables, in this embodiment, after the abnormal user is found, the Pearson correlation coefficient is used to determine the correlation between the abnormal user and the voltage data of the user's electricity meter between the master meters of each station area. level to diagnose the correct zone to which the abnormal user to be detected belongs.
  • the Pearson correlation coefficient is used to measure the relationship between the two voltage data X and Y, and it can measure the strength of the linear correlation between the two variables.
  • the overall correlation coefficient is:
  • ⁇ X and ⁇ Y are expected values
  • ⁇ X and ⁇ Y are the population standard deviation
  • cov(X, Y) is the covariance
  • Equation (13) can also be used for the Pearson correlation coefficient of the sample, and the calculated voltage sample correlation coefficient is:
  • the Pearson correlation coefficient does not change due to changes in the position and scale of the two variables, that is, moving X to a+bX and moving Y to c+dY, where a, b, c, and d are constants , has no effect on the correlation coefficient between the two variables, and the Pearson correlation coefficient formula holds for the population and the sample at the same time, so it can be concluded that a more general linear transformation will change the correlation coefficient.
  • the expected ⁇ Y and the variance ⁇ 2 Y of the voltage data Y of the meter in the station area are calculated as:
  • the overall correlation coefficient can be calculated as
  • ⁇ X, Y is the overall Pearson correlation coefficient
  • E(X) are the expected values of the voltage data X, Y.
  • r x,y is the sample Pearson correlation coefficient
  • Z x , Z y are standardized variables
  • S x , S y are the sample standard deviations of X and Y, represents the sample mean
  • n is the number of samples.
  • the Pearson correlation coefficient is
  • the criteria for evaluating the Pearson correlation coefficient are shown in Table 3.
  • the Pearson correlation coefficient between the user to be detected and the general table of each station area is calculated, and then the calculated Pearson correlation coefficients are sorted according to Sort the results to diagnose the correct station area of the abnormal user, without calculating the Pearson correlation coefficient between the abnormal user and all voltage data in the station area, which can greatly reduce the amount of calculation and improve the diagnosis efficiency, and there is no need to set the Pearson correlation coefficient reference for diagnosis. It is only necessary to sort the Pearson correlation coefficient between the user to be detected and the total table of each station area, which can effectively avoid the selection of the reference value of the Pearson correlation coefficient, and can make full use of the Pearson correlation coefficient between adjacent stations. Improve the final diagnosis accuracy and reduce the complexity of diagnosis.
  • step S03 of this embodiment the station area corresponding to the largest Pearson correlation coefficient in the sorting result is taken as the correct station area to which the abnormal user belongs, that is, the station area corresponding to the largest correlation coefficient is the correct station area to which the user to be detected belongs.
  • the above method in this embodiment can efficiently realize feature information mining among voltage data.
  • a better clustering center is selected according to the maximum and minimum data dimensions, so that different
  • you can It can effectively avoid the selection of the reference value of the Pearson correlation coefficient, thereby improving the detection efficiency and accuracy.
  • Step S03 in this embodiment also includes first calculating the correlation values between the voltage data of all the abnormal electricity meters found, and judging the abnormal electricity meters whose correlation value is less than the preset threshold as belonging to the same area, and from the judgment as the same One of the abnormal meters in the district is selected as the meter to be detected, and then the Pearson correlation coefficient between the meter to be detected and the total meter of each station is calculated to diagnose the correct station to which it belongs.
  • the abnormal station area is further divided according to the correlation, and the abnormal electricity meter whose correlation value is less than the preset threshold is determined as belonging to the same station area, and one is selected from the abnormal electricity meters determined as the same station area as the waiting area. Detect the electricity meters, and then calculate the Pearson correlation coefficient between the meter to be tested and the total meter of each station to diagnose the correct station area, so that only a small amount of calculation can be used to realize the station area diagnosis of all abnormal electricity meters. Diagnosing all abnormal electricity meters can further improve the efficiency of diagnosis and avoid a lot of unnecessary calculations, especially when there is a large-scale abnormality in the division of electricity meters, it can quickly diagnose the correct station area to which all electricity meters belong.
  • all the found abnormal electricity meters are firstly clustered to form multiple clusters, the target abnormal electricity meters as the cluster center of each cluster are obtained, and then the target abnormal electricity meters and the total meters of each station area are calculated respectively.
  • the Pearson correlation coefficient between them is used to diagnose the correct station area to which the target abnormal electricity meter belongs, and the correct station area to which each abnormal electricity meter belongs in the entire cluster is obtained from the correct station area to which the target abnormal electricity meter belongs. That is, the abnormal electricity meters are firstly classified by clustering, and then only the Pearson correlation coefficient between each cluster center and the total meter of the station area needs to be calculated, and all abnormal electricity meters can be diagnosed quickly and accurately by the clustering method.
  • the method further includes using the diagnosed correct station area as the destination station area, acquiring the user voltage data of the destination station area, and returning to step S02 to perform the clustering process again.
  • the clustering result finally confirms whether the abnormal user belongs to the destination station area.
  • the stations diagnosed according to the above steps may also have diagnostic errors. If the voltage characteristics of the two stations are very close, the obtained Pearson correlation coefficient is also relatively close, and the correct station is determined directly based on the Pearson correlation coefficient. Errors may exist.
  • the user voltage data of the station area is further obtained and clustered again. If the diagnosed station area is correct, the clustering can be performed correctly again. Therefore, the clustering process can be performed according to the re-clustering process. The result verifies whether the diagnosed station area is correct, and further improves the accuracy of diagnosis.
  • the data used are 137 of a certain station area.
  • the user's voltage data collected every 1 hour every 24 hours is simulated.
  • the 7 users in the adjacent station area of this station area mistakenly access the station area.
  • the situation is simulated and analyzed, 7 users from 3 stations are added to 1 station, and diagnostic analysis is performed to calculate the Pearson correlation coefficient between the 7 users to be detected and the total table of 5 stations, as shown in Figure 3(b).
  • the stations corresponding to the maximum Pearson correlation coefficient between the seven users to be detected and the total table of station areas are all 3 stations, and the actual corresponding stations also belong to 3 stations.
  • the inventive method has high accuracy in the case that n users in an adjacent station area access the station area by mistake.
  • the station area corresponding to the one with the largest coefficient is the correct station area to which the user to be detected belongs, indicating that the two users to be detected belong to the 2 station area and the 3 station area respectively, which are the same as the real results. Accuracy and validity in the case that one user in the adjacent n stations accesses the station by mistake.
  • this embodiment proposes an abnormal user identification accuracy rate index to reflect the identification accuracy of the household change relationship in the station area.
  • the index is defined as the number of abnormal users diagnosed and the total users to be diagnosed. Number ratio, the larger the value is, the more accurate the recognition result is.
  • the voltage data length is defined as the user voltage time dimension, and when the voltage values from 2 moments in a day to the voltage values at 24 moments in a day are selected, the recognition accuracy of the above method of the present invention is as follows: shown in Figure 4. It can be seen from Fig.
  • the recognition accuracy rate of the method of the present invention shows a trend of sharp rise at first and then a gentle rise with the increase of the voltage data length, and gradually converges in 11 dimensions, and is stable at 90%.
  • the above correct rates indicate that the longer the voltage data length is, the more accurate the method of the present invention is to diagnose the abnormality of the household change relationship.
  • the commonly used improved grey correlation analysis method, BP neural network method and the above method of the present invention are selected for simulation comparison. Taking multiple stations and multiple users mistakenly accessing other stations as an example, 7 users in two adjacent station areas in the station area are placed in the station area for diagnosis and identification. The simulation comparison is shown in Table 5. .
  • the abnormal diagnosis results of the three detection methods are different.
  • the number of abnormal users diagnosed by the improved grey correlation analysis method is 7, and the correct recognition rate is 50%.
  • the number of abnormal users identified by the BP neural network method is 9, and the correct recognition rate is 9. 64.29%, the number of abnormal users identified by the algorithm in this paper is 13, and the correct recognition rate is 92.86%.
  • Comprehensive judgment can obtain the correct station area of another abnormal user. That is, the above-mentioned method of the present invention for diagnosing abnormal household-variable relationships by improving K-means clustering and Pearson correlation coefficient has a higher diagnostic accuracy rate than common detection methods.
  • This embodiment also provides a device for diagnosing abnormal relationship between households in the station area, and the device includes:
  • the dimension reduction processing module is used to obtain the voltage data of each user in the designated station area and perform dimension reduction processing to obtain the dimension reduction voltage data;
  • the clustering processing module is used to cluster the dimensionality-reduced voltage data using the K-means clustering method, and during the clustering process, select the initial clustering center according to the maximum and minimum data dimensions, and find out the households in the designated station area.
  • the abnormal user with abnormal relationship is regarded as the user to be detected;
  • the station area diagnosis module is used to calculate the Pearson correlation coefficient between the user to be detected and the general table of each station area, sort the calculated Pearson correlation coefficients, and diagnose the correct station area to which the abnormal user belongs according to the sorting result. .
  • the station area diagnosis module further includes first calculating the correlation values between the voltage data of all the abnormal electricity meters found, and judging the abnormal electricity meters whose correlation value is less than the preset threshold value as belonging to the same station area.
  • One of the abnormal meters in the same station is selected as the meter to be detected, and then the Pearson correlation coefficient between the meter to be detected and the total meter of each station is calculated to diagnose the correct station to which it belongs.
  • the station diagnostic module specifically performs clustering processing on all the found abnormal electricity meters to form multiple clusters, obtains the abnormal electricity meters that are the cluster centers of each cluster as the electricity meters to be detected, and then calculates the The Pearson correlation coefficient between the meter to be detected and the total meter of each station area is used to diagnose the correct station area to which the target abnormal electricity meter belongs. belong to the correct station area.
  • the station area diagnosis module further includes taking the diagnosed correct station area as the destination station area, acquiring the user voltage data of the destination station area and returning it to the clustering processing module to re-run Perform clustering processing, and finally confirm whether the abnormal user belongs to the destination station area according to the clustering result.
  • the apparatus for diagnosing the abnormality of the household change relationship in the station area in this embodiment is in one-to-one correspondence with the above-mentioned method for diagnosing the abnormality of the household change relationship in the station area, which will not be repeated here.
  • the apparatus for diagnosing abnormality of household change relationship in a station area of the present invention may further include a processor and a memory, the memory is used for storing a computer program, and the processor is used for executing the computer program, and it is characterized in that, the processor is used for executing the computer program , in order to perform the above-mentioned abnormal diagnosis method of household change relationship in Taiwan area.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Probability & Statistics with Applications (AREA)
  • Medical Informatics (AREA)
  • Software Systems (AREA)
  • Computer Hardware Design (AREA)
  • Geometry (AREA)
  • Remote Monitoring And Control Of Power-Distribution Networks (AREA)
  • Supply And Distribution Of Alternating Current (AREA)

Abstract

一种台区户变关系异常诊断方法及装置，该方法步骤包括：S01.获取指定台区的总表和用户电表电压数据并进行降维处理，得到降维电压数据；S02.对降维电压数据采用K-means聚类方法进行聚类，并在聚类过程中根据数据维度的最大和最小值选取初始聚类中心，查找出指定台区内户变关系存在异常的异常用户；S03.分别计算异常用户与各个台区总表之间的皮尔逊相关系数，将计算出的各所述皮尔逊相关系数进行排序，根据排序结果诊断出异常用户所属的正确台区。该方法能够实现同一台区一个及多个异常用户、不同台区多个异常用户的诊断，且具有实现方法简单、诊断效率以及精度高等优点。

Description

一种台区户变关系异常诊断方法及装置 【技术领域】
本发明涉及用电信息采集系统技术领域，尤其涉及一种台区户变关系异常诊断方法及装置。
【背景技术】
随着电网建设的快速发展，包含海量数据的用电信息采集系统不仅可直接反映配电网的运行状况，且能间接反映配电网的拓扑关系。现有配电网拓扑结构校验主要包括：线变关系校验、馈线拓扑校验、户变关系及相位校验、线户关系校验。正确的低压配电网拓扑，尤其正确的户变关系是当前配电网管理精细化和降耗减损的基础。然而用电信息采集系统易出现台区户变关系错误问题，存在部分台区线路临时改变使户变关系档案更新不及时或记录错误等原因，导致用户进线端和集中器归属关系记录不准确，户变关系与实际不符，造成台区线损计算时出现异常情况。目前低压台区户变关系错误主要形式为主站系统中记录的用户档案与实际信息不相符，且通常把某个台区用户错误挂接至相邻台区，或把某个馈线变压器错误挂接至相邻馈线，该情况会影响正确的户变关系，阻碍正常线损计算。若将实际属于台区A的用户1档案信息，错归为台区B所有，则线损计算时将造成台区A线损偏小，而台区B线损偏大结果发生。因此，台区户变关系的准确诊断与分析对当前配电网的精益化管理具有重要意义。
现有技术中校验台区户变关系主要有人工专用设备现场识别和在线自动识别两种方式，其中人工方式需安排工作人员至现场识别，识别效率低、成本高，在线自动识别可以解决人工方式存在的上述问题，目前在线自动识别方法主要是基于数据挖掘技术，通过提取台区异常用户特征来识别特征相符的台区，如通过提取用户电压曲线相似性特征实现台区户变关系异常诊，在低压配电网中由于各处负荷的不确定性，电压通常也随之波动，电气距离较近负荷，其电压波动曲线较相似，而电气距离较远负荷其电压波动曲线相似度较低，因此可选取用户电压曲线相似性作为台区户变关系异常诊断的依据。但是上述在线自动识别方法需要依赖于异常用户的特征提取，仅能针对台区中出现少量异常用户的情况，对于多达数百用户台区，若存在多个相邻台区存在数量较多的异常用户时，大量的异常用户特征提取困难，且提取的特征所能够表征的信息有限，就难以快速、准确的实现多用户台区和多个异常用户情况的诊断与分析。
【发明内容】
本发明要解决的技术问题就在于：针对现有技术存在的技术问题，本发明提供一种实现方法简单、诊断效率以及精度高的台区户变关系异常诊断方法及装置，能够实现同一台区一个及多个异常用户、不同台区多个异常用户的诊断。
为解决上述技术问题，本发明提出的技术方案为：
一种台区户变关系异常诊断方法，步骤包括：
S01.降维处理：获取指定台区的总表和用户电表电压数据并进行降维处理，得到降维电压数据；
S02.聚类处理：对所述降维电压数据采用K-means聚类方法进行聚类，并在聚类过程中根据数据维度的最大和最小值选取初始聚类中心，查找出指定台区内户变关系存在异常的异常用户；
S03.台区诊断：分别计算所述异常用户与各个台区总表之间的皮尔逊相关系数，将计算出的各所述皮尔逊相关系数进行排序，根据排序结果诊断出所述异常用户所属的正确台区。
进一步的，所述步骤S02中选取初始聚类中心的步骤包括：
S201.使用所述降维电压数据中每一维最大和最小值选取k个样本点作为k个初始聚类中心；
S202.分别计算所述降维电压数据中每个样本数据距离最近聚类中心的距离D(z ij)，即：
Figure PCTCN2021077141-appb-000001
其中，z ij为所述降维电压数据；
S203.取所述降维电压数据中每个样本数据距离聚类中心的最大距离作为下一个聚类中心，以更新各簇的聚类中心，即：
μ r＝arg max|D(z ij)|
其中，μ r为各簇更新的聚类中心，max|D(z ij)|为每个样本与当前最近一个聚类中心的最大距离；
S204.重复执行步骤S202、步骤S203，直至聚类误差函数E收敛或达到最大步数，得到最终的k个聚类中心[μ 1,μ 2…,μ k]。
进一步的，所述步骤S201中具体按照下式选取k个初始聚类中心；
μ r＝ma(i)+(mi(i)-ma(i))+rand(),r＝1,2,...,k
其中，μ r为初始聚类中心，ma为每一维最大数，mi为每一维最小数。
进一步的，所述步骤S03中具体取排序结果中最大皮尔逊相关系数所对应的台区作为所述异常用户所属的正确台区。
进一步的，所述步骤S03中还包括先分别计算查找出的所有异常电表的电压数据之间的相关性值，将相关性值小于预设阈值的异常电表判定为属于同一台区，从判定为同一台区的各异常电表中选取出一个作为待检测电表，然后分别计算待检测电表与各个台区总表之间的皮尔逊相关系数以诊断出所属的正确台区。
进一步的，具体对查找出的所有异常电表先进行聚类处理，形成多个聚类，获取作为每个聚类的聚类中心的异常电表作为待检测电表，然后分别计算待检测电表与各个台区总表之间的皮尔逊相关系数以诊断出目标异常电表所属的正确台区，由诊断出的目标异常电表所属的正确台区得到对应的整个聚类中各异常电表所属的正确台区。
进一步的，所述步骤S03诊断出所述异常用户所属的正确台区后，还包括以诊断出的所述正确台区作为目的台区，获取目的台区的用户电压数据并返回步骤S02以重新进行聚类处理，根据聚类结果最终确认异常用户是否属于目的台区。
进一步的：所述步骤S01中具体采用主成分分析方法进行降维处理。
一种台区户变关系异常诊断装置，包括：
降维处理模块，用于获取指定台区的各用户电压数据并进行降维处理，得到降维电压数据；
聚类处理模块，用于对所述降维电压数据采用K-means聚类方法进行聚类，并在聚类过程中根据数据维度的最大和最小值选取初始聚类中心，查找出指定台区内户变关系存在异常的异常用户作为待检测用户；
台区诊断模块，用于分别计算所述待检测用户与各个台区总表之间的皮尔逊相关系数，将计算出的各所述皮尔逊相关系数进行排序，根据排序结果诊断出所述异常用户所属的正确台区。
一种台区户变关系异常诊断装置，包括处理器以及存储器，所述存储器用于存储计算机程序，所述处理器用于执行所述计算机程序，所述处理器用于执行所述计算机程序，以执行如上述方法。
与现有技术相比，本发明的优点在于：
1、本发明通过获取台区总表和用户电表电压数据降维后，使用改进的K-means聚类提取电压数据特征，在聚类时根据数据维度的最大和最小值选取初始聚类中心，使得可以快速准确的查找出台区内所有的异常用户，同时对查找出的异常用户，使用改进皮尔逊相 关系数方法，由异常用户与各个台区总表之间的皮尔逊相关系数之间的大小关系诊断出正确的台区，能够准确提取多个相邻台区异常用户间特征，实现多异常用户所属正确台区诊断，从而在同一台区一个及多个异常用户、不同台区多个异常用户情况下均能有效实现异常用户的准确检测与分析。
2、本发明能够高效实现电压数据间的特征信息挖掘，通过在使用K-means聚类方法提取电压数据特征时，根据数据维度的最大最小值选取更优聚类中心，使得可以得到不同特征异常用户，同时通过使用皮尔逊相关系数表征异常用户与台区之间相关程度的基础上，通过取待检测用户与各个台区总表间的最大相关系数作为所求的正确台区，可以有效避免皮尔逊相关系数参考值的选取，从而提高检测效率以及精度，尤其适用于同一个台区存在多台异常用户、相邻台区存在多台异常用户时的户变关系自动诊断。
3、本发明进一步先将异常台区按照相关性进行划分，将相关性值小于预设阈值的异常电表判定为属于同一台区，从判定为同一台区的各异常电表中再选取出一个作为待检测电表，然后分别计算待检测电表与各个台区总表之间的皮尔逊相关系数以诊断出所属的正确台区，使得仅需要少量的计算即可实现所有异常电表的台区诊断，无需一一对所有异常电表进行诊断，可以进一步提高诊断的效率，避免大量不必要的计算，尤其是在存在大规模的电表划分异常时，能够快速的诊断出所有电表所属的正确台区。
【附图说明】
图1是本实施例台区户变关系异常诊断方法的实现流程示意图。
图2是本实施例中电压数据聚类中心选取的原理示意图。
图3是在具体应用实施例中得到的仿真结果示意图。
图4是在具体应用实施例中得到的数据长度影响分析结果示意图。
【具体实施方式】
以下结合说明书附图和具体优选的实施例对本发明作进一步描述，但并不因此而限制本发明的保护范围。
如图1所示，本实施例台区户变关系异常诊断方法的步骤包括：
S01.降维处理：获取指定台区的总表和用户电表电压数据并进行降维处理，得到降维电压数据；
S02.聚类处理：对降维电压数据采用K-means聚类方法进行聚类，并在聚类过程中根据数据维度的最大和最小值选取初始聚类中心，查找出指定台区内户变关系存在异常的异常用户；
S03.台区诊断：分别计算异常用户与各个台区总表之间的皮尔逊相关系数，将计算出 的各皮尔逊相关系数进行排序，根据排序结果诊断出异常用户所属的正确台区。
本实施例通过获取台区总表和用户电表电压数据降维后，使用改进的K-means聚类提取电压数据特征，在聚类时根据数据维度的最大和最小值选取初始聚类中心，而不是如传统的随机选取初始聚类中心，可以快速准确的查找出台区内所有的异常用户，同时对查找出的异常用户，使用改进皮尔逊相关系数方法，由异常用户与各个台区总表之间的皮尔逊相关系数之间的大小关系诊断出正确的台区，能够准确提取多个相邻台区异常用户间特征，实现多异常用户所属正确台区诊断，从而在同一台区一个及多个异常用户、不同台区多个异常用户情况下均能有效实现异常用户的准确检测与分析。
本实施例具体使用主成分分析PCA方法对原始电压数据进行降维处理。通常台区用户电表电压数据为一天24小时，每隔1小时采集一次，对于数据维度为24维的高维数据，传统聚类算法将面临高维数据包含海量冗余、不相干信息问题，直接对高维数据聚类极大降低性能，聚类算法难以实现高稳定性，本实施例通过采用PCA方法对原始数据进行降维处理，可以便于后续对电压数据进行聚类处理而查找出异常用户。
本实施例使用主成分分析PCA对原始电压数据进行降维处理的详细步骤为：
步骤S101.基于PCA将p维电压数据特征在m维上体现，该m维数据信息也被称为主成分，为在原有p维电压数据特征上构造出m维新特征，将台区n个用户电表p维电压数据表示为：
Figure PCTCN2021077141-appb-000002
通过PCA得出的主成分之间互不相关，计算电压数据相关系数具体为：
Figure PCTCN2021077141-appb-000003
式中，x i、x j为台区电压数据，x为电压数据均值。
由式(2)可得电压数据相关系数矩阵为：
Figure PCTCN2021077141-appb-000004
式中，r ij(i,j＝1,2,…,p)为台区电压数据x i、x j的相关系数，r ij＝r ji。
根据式(3)解特征方程得到：
|λI-R|＝0  (4)
根据数值大小得特征值λ 1≥λ 2≥…≥λ p≥0，同理可得对应于特征值λ i的特征向量e i(i＝1,2,…,p)。
步骤S102.根据主成分贡献率法(CPV)确定主成分个数m，计算各主成分的贡献率CPV i与累计贡献率CPV a分别为：
Figure PCTCN2021077141-appb-000005
Figure PCTCN2021077141-appb-000006
式中，CPV i为第i个主成分的贡献率，λ m为第m(m≤p)个主成分所对应的特征值，CPV a为控制限。
通过式(5)和(6)计算包含原始电压数据绝大部分信息的主成分个数m，使用式(1)中原始电压数据矩阵X的m个特征向量作线性组合得到主成分，则有：
Figure PCTCN2021077141-appb-000007
式中，z ij(i＝1,2,…,n,j＝1,2,…,m)为台区用户电表电压数据第i个样本第j个主成分。
相同台区用户电压数据波动相似性使其具有相同特征，本实施例基于该特性使用K-means聚类检测出异常用户。K-means聚类是根据迭代理论，求出最大迭代步数或使得聚类误差函数收敛得到聚类中心，该聚类方法为无监督学习算法，可以适用于检测台区中异常用户，但是传统的K-means聚类方法中是采用随机选取初始聚类中心的方式，若初始聚类中心选取不合适，会极大影响最终聚类结果。本实施例对传统的K-means聚类进行改进，通过根据数据维度的最大和最小值，基于尽可能得到相互距离远的初始聚类中心原则选取初始聚类中心，而不是如传统方法随机选取初始聚类中心，可以避免出现初始聚类中心选择错误或不合适的问题，从而大大提高异常用户的检测精度以及效率。
本实施例步骤S02中选取初始聚类中心的具体步骤包括：
S201.使用降维电压数据中每一维最大和最小值选取k个样本点作为k个初始聚类中心，具体按照下式(8)选取k个初始聚类中心；
μ r＝ma(i)+(mi(i)-ma(i))+rand(),r＝1,2,...,k  (9)
其中，μ r为初始聚类中心，ma为每一维最大数，mi为每一维最小数；
S202.分别计算降维电压数据中每个样本数据距离最近聚类中心的距离D(z ij)，即：
Figure PCTCN2021077141-appb-000008
其中，z ij为降维电压数据；
S203.取降维电压数据中每个样本数据距离聚类中心的最大距离作为下一个聚类中心，以更新各簇的聚类中心，即：
μ r＝arg max|D(z ij)|   (11)
其中，μ r为各簇更新的聚类中心，max|D(z ij)|为每个样本与当前最近一个聚类中心的最大距离；
S204.重复执行步骤S202、步骤S203，直至聚类误差函数E收敛或达到最大步数，得到最终的k个聚类中心[μ 1,μ 2…,μ k]。
本实施例先对于降维电压数据选取k个样本点作为k个初始聚类中心，针对电压中每个样本数据，分别计算其到已有最近聚类中心欧式距离，由式(9)所求距离值越大表示该样本被选取作为下一次聚类中心的概率越大，样本数据被分别归为最近聚类中心类别后得到k个簇，计算样本被选取作为聚类中心的概率可得：
Figure PCTCN2021077141-appb-000009
式中，D(z ij)为样本到聚类中心的距离。
在具体应用实施例中对台区用户电表电压数据通过PCA降维后，得到2维数据为A(0.1,0.1)，B(0.2,0.2)，C(0.2,0)，D(0.4,0.6)，E(0.5,0.6)，F(0.5,0.5)，G(0.6,0.5)，再对降维后电压数据进行聚类分析，并在首次聚类中选择C 1(0.2,0.3)，C 2(0.5,0.3)两点作为第1个和第2个聚类中心如图2所示。由图2可知，计算每一簇中其他电压数据样本与该簇当前已有最近聚类中心距离D(z ij)，以及各簇中每个电压数据样本被选取作为下一个聚类中心的概率P(z ij)分别如表1和2所示。
表1中心点1聚类距离与概率
Figure PCTCN2021077141-appb-000010
Figure PCTCN2021077141-appb-000011
表2中心点2聚类距离与概率
Figure PCTCN2021077141-appb-000012
由表1可见，对于第一簇数据A点被选为下一个聚类中心的概率最大，对于第二簇数据D点被选为下一个聚类中心的概率最大。而由图2可知，A，D分别为距离初始聚类中心C 1和C 2点最远的两个点。
由式(9)求得每个样本与当前最近一个聚类中心的距离并选取最大值，基于相互距离尽可能远原则选取该样本为该簇新聚类中心点，则按照式(10)即可计算出最佳的新聚类中心点作为各簇下一个聚类中心；重复上述式(9)和(11)不断移动聚类中心直至聚类误差函数收敛或达到最大迭代步数，平方误差SSE函数为：
Figure PCTCN2021077141-appb-000013
根据式(12)直至平方误差SSE收敛或达到最大步数，选出k个聚类中心[μ 1,μ 2…,μ k]，并分别实现以μ r为聚类中心的台区用户电表电压数据聚类求得台区内异常用户。
本实施例通过按照上述改进的K-means聚类分析步骤对相邻各台区内所有的用户电表电压数据进行聚类分析，即可查找出相邻各台区内所有的用户电表电压数据异常的用户。由于皮尔逊相关系数可以表征两个变量间的线性相关性，本实施例在查找出异常用户后，通过使用皮尔逊相关系数来判断异常用户与各台区总表之间用户电表电压数据的相关程度，以诊断出待检测异常用户所属的正确台区。
皮尔逊相关系数用来度量两个电压数据X与Y之间的相互关系，可度量两个变量线性相关的强弱，其总体相关系数为：
Figure PCTCN2021077141-appb-000014
式中，μ X，μ Y为期望值，σ X、σ Y为总体标准差，cov(X,Y)为协方差。
式(13)对于样本的皮尔逊相关系数同样可用，计算电压样本相关系数为：
Figure PCTCN2021077141-appb-000015
式中，
Figure PCTCN2021077141-appb-000016
为标准化变量；
Figure PCTCN2021077141-appb-000017
为样本均值；S x、S y为样本标准差。
皮尔逊相关系数因两个变量的位置和尺度的变化并不会引起该系数的改变，即把X移动到a+bX和把Y移动到c+dY，其中a、b、c和d是常数，对两个变量间相关系数毫无影响，且对于总体以及样本皮尔逊相关系数公式同时成立，由此可得出更一般的线性变换则会改变相关系数。
由于台区电表电压数据X的期望μ X与方差σ 2 X分别为：
Figure PCTCN2021077141-appb-000018
Figure PCTCN2021077141-appb-000019
同理，计算台区电表电压数据Y的期望μ Y与方差σ 2 Y分别为：
Figure PCTCN2021077141-appb-000020
Figure PCTCN2021077141-appb-000021
根据两个电压数据X与Y间期望变换公式为：
E[(X-E(X))(Y-E(Y))]＝E(XY)-E(X)E(Y)  (19)
可计算总体相关系数为
Figure PCTCN2021077141-appb-000022
式中，ρ X,Y为总体皮尔逊相关系数，E(X)、E(Y)为电压数据X、Y的期望值。
同理，由式(19)计算电压数据样本皮尔逊相关系数可得：
Figure PCTCN2021077141-appb-000023
式中，r x,y为样本皮尔逊相关系数，Z x、Z y为标准化变量，S x、S y为X、Y样本标准差，
Figure PCTCN2021077141-appb-000024
表示样本均值，n为样本数。
皮尔逊相关系数为|r x,y|≤1，r x,y大于零表示两者为正相关方向，小于零表示为负相关方向，评价皮尔逊相关系数的标准如表3所示。
表3皮尔逊相关系数关联度标准
Figure PCTCN2021077141-appb-000025
由表3可见，若直接使用皮尔逊相关系数表征两个电压数据X与Y之间相关程度，皮尔逊相关系数的阈值取值会决定相关程度的判定结果，但实际应用中就难以准确的设定参 考值以判断某用户是否属于某台区，且当通过聚类查找到台区内的异常用户后，由于该台区及邻近几个台区用户较多，若计算待检测用户与各个台区间之间的皮尔逊相关系数，即计算异常用户与台区中所有电表数据之间的皮尔逊相关系数，还会存在工作量巨大的问题。
本实施例在通过聚类查找到台区内的异常用户后，通过计算待检测用户与各台区总表之间的皮尔逊相关系数，再对计算出的各皮尔逊相关系数进行排序，依据排序结果来诊断异常用户所属的正确台区，无需计算异常用户与台区所有电压数据之间的皮尔逊相关系数，可以大大减少计算量，提高诊断效率，同时无需设置诊断的皮尔逊相关系数参考值，只需对待检测用户与各个台区总表间的皮尔逊相关系数进行大小排序，有效避免皮尔逊相关系数参考值的选取，可以充分利用相邻台区之间的皮尔逊相关系数，大大提高最终的诊断精度、降低诊断的复杂度。
本实施例步骤S03中具体取排序结果中最大皮尔逊相关系数所对应的台区作为异常用户所属的正确台区，即相关系数最大者对应台区为待检测用户所属正确台区。
本实施例上述方法，能够高效实现电压数据间的特征信息挖掘，通过在使用K-means聚类方法提取电压数据特征时，根据数据维度的最大最小值选取更优聚类中心，使得可以得到不同特征异常用户，同时通过使用皮尔逊相关系数表征异常用户与台区之间相关程度的基础上，通过取待检测用户与各个台区总表间的最大相关系数作为所求的正确台区，可以有效避免皮尔逊相关系数参考值的选取，从而提高检测效率以及精度，尤其适用于同一个台区存在多台异常用户、相邻台区存在多台异常用户时的户变关系自动诊断。
本实施例步骤S03中还包括先分别计算查找出的所有异常电表的电压数据之间的相关性值，将相关性值小于预设阈值的异常电表判定为属于同一台区，从判定为同一台区的各异常电表中选取出一个作为待检测电表，然后分别计算待检测电表与各个台区总表之间的皮尔逊相关系数以诊断出所属的正确台区。
当相邻台区中存在大量的异常用户时，若一一分别计算异常用户与各台区总表的相关系数进行台区诊断，依然会存在需要进行大量计算的问题。本实施例进一步先将异常台区按照相关性进行划分，将相关性值小于预设阈值的异常电表判定为属于同一台区，从判定为同一台区的各异常电表中再选取出一个作为待检测电表，然后分别计算待检测电表与各个台区总表之间的皮尔逊相关系数以诊断出所属的正确台区，使得仅需要少量的计算即可实现所有异常电表的台区诊断，无需一一对所有异常电表进行诊断，可以进一步提高诊断的效率，避免大量不必要的计算，尤其是在存在大规模的电表划分异常时，能够快速的诊断出所有电表所属的正确台区。
本实施例具体对查找出的所有异常电表先进行聚类处理，形成多个聚类，获取作为每个聚类的聚类中心的目标异常电表，然后分别计算目标异常电表与各个台区总表之间的皮尔逊相关系数以诊断出目标异常电表所属的正确台区，由诊断出的目标异常电表所属的正确台区得到对应的整个聚类中各异常电表所属的正确台区。即采用聚类的方式先将异常电表进行分类，后续仅需要计算各聚类中心与台区总表之间的皮尔逊相关系数，利用聚类的方式快速、精准的对所有异常电表进行诊断。
本实施例步骤S03诊断出异常用户所属的正确台区后，还包括以诊断出的正确台区作为目的台区，获取目的台区的用户电压数据并返回步骤S02以重新进行聚类处理，根据聚类结果最终确认异常用户是否属于目的台区。按照上述步骤诊断出的台区，也可能会存在诊断误差，如若两个台区的电压特性非常接近，所得到的皮尔逊相关系数也较为接近，直接依据皮尔逊相关系数来确定正确台区即可能会存在误差。本实施例在诊断出异常用户的台区后，进一步获取该台区的用户电压数据重新进行聚类处理，若诊断的台区正确则再次聚类时可以正确分类，因而可以依据再次聚类处理的结果验证所诊断的台区是否正确，进一步提高诊断的精度。
为验证本发明上述按照改进的K-means聚类分析查找出异常用户、以及使用最大皮尔逊相关系数诊断出正确台区的有效性，在具体实施例中使用数据为某个台区其中137个用户的每24个小时每隔1小时采集的电压数据进行仿真。先判断出异常用户，对于需校验用户，获取相邻台区总表电压数据，诊断待校验用户所属正确台区，分别采用以下各种情况对本发明上述方法进行验证，结果如图3所示。
(1)1台区1用户分析
首先针对该台区相邻1个台区中的1个用户错接入该台区的情况进行仿真和分析，将2台区1用户加入1台区中，并对该用户进行异常诊断，找出该用户所属正确台区，计算待检测用户与5个台区总表间的皮尔逊相关系数如图3(a)所示。由图3(a)可见，该用户与2台区总表间的皮尔逊相关系数为0.9910，在5个台区中为最大值，本实施例取相关系数最大者所对应台区为待检测用户所属的正确台区，即该用户属于2台区，与真实结果相同，由此验证本发明上述方法在相邻1个台区中1个用户错接入该台区情况下的准确性和可行性。
为验证本发明方法在1个台区多个用户错接入其他台区情况下的准确性和可行性，针对该台区相邻1个台区中的7个用户错接入该台区的情况进行仿真和分析，将3台区7个用户加入1台区中，并进行诊断分析，计算待检测7个用户与5个台区总表间的皮尔逊相关系数如图3(b)所示。计算待检测7个用户与5个台区总表间的皮尔逊相关系数，并将 皮尔逊相关系数所对应台区与实际台区作比较结果如表4所示。
表4皮尔逊相关系数测试结果
Figure PCTCN2021077141-appb-000026
由表4可见，可直观得出7个待检测用户与台区总表间最大皮尔逊相关系数对应的台区均为3台区，实际对应台区也均属于3台区，由此可见本发明方法在相邻1个台区中n个用户错接入该台区情况下的具有较高的准确度性。
为验证本发明上述方法在多个台区用户错接入其他台区情况下的准确性和可行性，针对该台区相邻2个台区中的各1个用户错接入该台区的情况进行仿真和分析，分别将2台区1个用户和3台区1个用户加入1台区，并进行诊断分析，计算待检测2个用户与5个台区总表间的皮尔逊相关系数如图3(c)所示。由图3(c)可知，待检测用户1仅与2台区总表间的皮尔逊相关系数超过0.9900，待检测用户2仅与3台区总表间的皮尔逊相关系数超过0.9800，根据相关系数最大者所对应的台区即为待检测用户所属的正确台区，表明2个待检测用户分别属于2台区、3台区，与真实结果相同，由此可验证本发明上述方法在相邻n个台区中1个用户错接入该台区情况下的准确性和有效性。
为验证本发明方法在多个台区多个用户错接入其他台区情况下的准确性和可行性，针对该台区相邻2个台区中各7个用户错接入该台区的情况进行仿真和分析，分别将2台区7个用户和3台区7个用户加入1台区，并进行诊断分析，计算待检测14个用户与5个台区总表之间的皮尔逊相关系数如图3(d)所示。由图3(d)可见，用户1到用户7与台区3皮尔逊相关系数最大，用户8、用户10到用户14与台区2皮尔逊相关系数最大，用户9与台区1皮尔逊相关系数最大。因此除用户9外，1到7个待检测用户均属于3台区，8到14个待检测用户均属于2台区。表明除用户9外，其余待检测用户诊断结果与真实结果相同，又因聚类结果中用户9已被识别为1台区异常用户，故从1台区排除，且用户9与台区2总表皮尔逊相关系数大于其他台区，故综合判断可得出用户9所属正确台区。由此可见本发明在相邻n个台区中n个用户错接入该台区情况下仍有较高的准确性。
为分析用户电能表电压数据长度对识别结果的影响，本实施例提出异常用户识别正确率指标，以反映台区户变关系识别准确性，该指标定义为诊断出异常用户数与待诊断总用户数比值，数值越大表明识别结果越准确，电压数据长度定义为用户电压时间维度，选取一天中2个时刻的电压值到24个时刻的电压值情况下，本发明上述方法的识别正确率如图4所示。由图4可见，当电压数据长度低于10维时，本发明方法识别正确率随着电压数据长度的增加，呈现先急剧上升后平缓上升的趋势，并在11维逐渐收敛，稳定在90％以上的正确率，表明电压数据长度越大，本发明方法对户变关系异常诊断结果越准确。
为比较不同户变关系异常诊断的方法，选取常用的改进灰色关联度分析法、BP神经网络法以及本发明上述方法进行仿真比较。以多个台区多个用户错接入其他台区为例，将该台区相邻2个台区中的各7个用户放入该台区进行诊断识别，其仿真对比如表5所示。
表5与常用算法诊断结果对比
Figure PCTCN2021077141-appb-000027
由表5可见，3种检测方法异常诊断结果不同，其中改进灰色关联度分析法异常用户诊断数为7，正确识别率为50％，BP神经网络法识别异常用户数为9，正确识别率为64.29％，本文算法识别异常用户数为13，正确识别率为92.86％，综合判断可得到另外一个异常用户所属正确台区。即本发明上述通过改进K-means聚类和皮尔逊相关系数进行户变关系异常诊断的方法，相比常用检测方法具有更高的诊断准确率。
本实施例还提供台区户变关系异常诊断装置，该装置包括：
降维处理模块，用于获取指定台区的各用户电压数据并进行降维处理，得到降维电压数据；
聚类处理模块，用于对降维电压数据采用K-means聚类方法进行聚类，并在聚类过程中根据数据维度的最大和最小值选取初始聚类中心，查找出指定台区内户变关系存在异常的异常用户作为待检测用户；
台区诊断模块，用于分别计算待检测用户与各个台区总表之间的皮尔逊相关系数，将计算出的各皮尔逊相关系数进行排序，根据排序结果诊断出异常用户所属的正确台区。
本实施例中，台区诊断模块还包括先分别计算查找出的所有异常电表的电压数据之间的相关性值，将相关性值小于预设阈值的异常电表判定为属于同一台区，从判定为同一台 区的各异常电表中选取出一个作为待检测电表，然后分别计算待检测电表与各个台区总表之间的皮尔逊相关系数以诊断出所属的正确台区。
本实施例中，台区诊断模块具体对查找出的所有异常电表先进行聚类处理，形成多个聚类，获取作为每个聚类的聚类中心的异常电表作为待检测电表，然后分别计算待检测电表与各个台区总表之间的皮尔逊相关系数以诊断出目标异常电表所属的正确台区，由诊断出的目标异常电表所属的正确台区得到对应的整个聚类中各异常电表所属的正确台区。
本实施例中，台区诊断模块诊断出异常用户所属的正确台区后，还包括以诊断出的正确台区作为目的台区，获取目的台区的用户电压数据并返回聚类处理模块以重新进行聚类处理，根据聚类结果最终确认异常用户是否属于目的台区。
本实施例台区户变关系异常诊断装置与上述台区户变关系异常诊断方法为一一对应，在此不再一一赘述。
在另一实施例中，本发明台区户变关系异常诊断装置还可以为包括处理器以及存储器，存储器用于存储计算机程序，处理器用于执行计算机程序，其特征在于，处理器用于执行计算机程序，以执行如上述台区户变关系异常诊断方法。
上述只是本发明的较佳实施例，并非对本发明作任何形式上的限制。虽然本发明已以较佳实施例揭露如上，然而并非用以限定本发明。因此，凡是未脱离本发明技术方案的内容，依据本发明技术实质对以上实施例所做的任何简单修改、等同变化及修饰，均应落在本发明技术方案保护的范围内。

Claims (10)

  1. 一种台区户变关系异常诊断方法，其特征在于，步骤包括：
    S01.降维处理：获取指定台区的总表和用户电表电压数据并进行降维处理，得到降维电压数据；
    S02.聚类处理：对所述降维电压数据采用K-means聚类方法进行聚类，并在聚类过程中根据数据维度的最大和最小值选取初始聚类中心，查找出指定台区内户变关系存在异常的异常用户；
    S03.台区诊断：分别计算所述异常用户与各个台区总表之间的皮尔逊相关系数，将计算出的各所述皮尔逊相关系数进行排序，根据排序结果诊断出所述异常用户所属的正确台区。
  2. 根据权利要求1所述的台区户变关系异常诊断方法，其特征在于，所述步骤S02中选取初始聚类中心的步骤包括：
    S201.使用所述降维电压数据中每一维最大和最小值选取k个样本点作为k个初始聚类中心；
    S202.分别计算所述降维电压数据中每个样本数据距离最近聚类中心的距离D(z ij)，即：
    Figure PCTCN2021077141-appb-100001
    其中，z ij为所述降维电压数据；
    S203.取所述降维电压数据中每个样本数据距离聚类中心的最大距离作为下一个聚类中心，以更新各簇的聚类中心，即：
    μ r＝arg max|D(z ij)|
    其中，μ r为各簇更新的聚类中心，max|D(z ij)|为每个样本与当前最近一个聚类中心的最大距离；
    S204.重复执行步骤S202、步骤S203，直至聚类误差函数E收敛或达到最大步数，得到最终的k个聚类中心[μ 1,μ 2…,μ k]。
  3. 根据权利要求2所述的台区户变关系异常诊断方法，其特征在于，所述步骤S201中具体按照下式选取k个初始聚类中心；
    μ r＝ma(i)+(mi(i)-ma(i))+rand(),r＝1,2,...,k
    其中，μ r为初始聚类中心，ma为每一维最大数，mi为每一维最小数。
  4. 根据权利要求1或2或3所述的台区户变关系异常诊断方法，其特征在于，所述步骤S03中具体取排序结果中最大皮尔逊相关系数所对应的台区作为所述异常用户所属的正确台区。
  5. 根据权利要求1或2或3所述的台区户变关系异常诊断方法，其特征在于，所述步骤S03中还包括先分别计算查找出的所有异常电表的电压数据之间的相关性值，将相关性值小于预设阈值的异常电表判定为属于同一台区，从判定为同一台区的各异常电表中选取出一个作为待检测电表，然后分别计算待检测电表与各个台区总表之间的皮尔逊相关系数以诊断出所属的正确台区。
  6. 根据权利要求5所述的台区户变关系异常诊断方法，其特征在于，具体对查找出的所有异常电表先进行聚类处理，形成多个聚类，获取作为每个聚类的聚类中心的异常电表作为待检测电表，然后分别计算待检测电表与各个台区总表之间的皮尔逊相关系数以诊断出目标异常电表所属的正确台区，由诊断出的目标异常电表所属的正确台区得到对应的整个聚类中各异常电表所属的正确台区。
  7. 根据权利要求1或2或3所述的台区户变关系异常诊断方法，其特征在于，所述步骤S03诊断出所述异常用户所属的正确台区后，还包括以诊断出的所述正确台区作为目的台区，获取目的台区的用户电压数据并返回步骤S02以重新进行聚类处理，根据聚类结果最终确认异常用户是否属于目的台区。
  8. 根据权利要求1或2或3所述的台区户变关系异常诊断方法，其特征在于，所述步骤S01中具体采用主成分分析方法进行降维处理。
  9. 一种台区户变关系异常诊断装置，其特征在于，包括：
    降维处理模块，用于获取指定台区的各用户电压数据并进行降维处理，得到降维电压数据；
    聚类处理模块，用于对所述降维电压数据采用K-means聚类方法进行聚类，并在聚类过程中根据数据维度的最大和最小值选取初始聚类中心，查找出指定台区内户变关系存在异常的异常用户作为待检测用户；
    台区诊断模块，用于分别计算所述待检测用户与各个台区总表之间的皮尔逊相关系数，将计算出的各所述皮尔逊相关系数进行排序，根据排序结果诊断出所述异常用户所属的正确台区。
  10. 一种台区户变关系异常诊断装置，其特征在于，包括处理器以及存储器，所述存储器用于存储计算机程序，所述处理器用于执行所述计算机程序，其特征在于，所述处理器用于执行所述计算机程序，以执行如权利要求1～7中任意一项所述方法。
PCT/CN2021/077141 2020-11-25 2021-02-22 一种台区户变关系异常诊断方法及装置 WO2022110557A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202011343787.3A CN112699913B (zh) 2020-11-25 2020-11-25 一种台区户变关系异常诊断方法及装置
CN202011343787.3 2020-11-25

Publications (1)

Publication Number Publication Date
WO2022110557A1 true WO2022110557A1 (zh) 2022-06-02

Family

ID=75506037

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/077141 WO2022110557A1 (zh) 2020-11-25 2021-02-22 一种台区户变关系异常诊断方法及装置

Country Status (2)

Country Link
CN (1) CN112699913B (zh)
WO (1) WO2022110557A1 (zh)

Cited By (19)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114825647A (zh) * 2022-06-30 2022-07-29 国网浙江省电力有限公司宁波供电公司 一种基于拓扑关系的台区线损监测方法及相关设备
CN115235676A (zh) * 2022-09-26 2022-10-25 山东济矿鲁能煤电股份有限公司阳城煤矿 一种矿井液压支架压力异常自动报警方法
CN115423250A (zh) * 2022-07-28 2022-12-02 国网浙江省电力有限公司营销服务中心 一种台区户变关系分析方法
CN115438035A (zh) * 2022-10-27 2022-12-06 江西师范大学 一种基于kpca和混合相似度的数据异常处理方法
CN115470858A (zh) * 2022-09-13 2022-12-13 国网天津市电力公司 基于ddtw距离和最小最大距离算法的配电台区拓扑辨识方法
CN116482471A (zh) * 2023-06-21 2023-07-25 四川中电启明星信息技术有限公司 一种基于电压时空聚合曲线的户变关系识别方法
CN116499419A (zh) * 2023-06-27 2023-07-28 西安高商智能科技有限责任公司 一种舵机旋转角度异常检测方法及系统
CN116522181A (zh) * 2023-06-28 2023-08-01 广东电网有限责任公司 一种配电台区日线损特征提取方法及系统
CN116303480B (zh) * 2023-05-22 2023-08-04 新乡市万新电气有限公司 基于云计算的电能表误差校验方法
CN116599055A (zh) * 2023-05-26 2023-08-15 联桥科技有限公司 一种低压配网台区的拓扑网络识别方法及系统
CN116595426A (zh) * 2023-07-17 2023-08-15 济南大陆机电股份有限公司 一种工业物联网数据智能采集管理系统
CN116881745A (zh) * 2023-09-07 2023-10-13 深圳市特安工业科技有限公司 基于大数据的压力变送器异常监测方法
CN116910596A (zh) * 2023-07-26 2023-10-20 江苏方天电力技术有限公司 一种基于改进dbscan聚类的用户窃电分析方法、设备及存储介质
CN116954342A (zh) * 2023-09-18 2023-10-27 惠州市鑫晖源科技有限公司 基于物联网的全方位高效散热主机电源运行监测方法
CN117129790A (zh) * 2023-10-26 2023-11-28 山西思极科技有限公司 一种电力系统故障诊断系统
CN117148023A (zh) * 2023-10-31 2023-12-01 威海海泰电子有限公司 一种智能电源适配器生产检测方法及系统
CN117648657A (zh) * 2023-12-13 2024-03-05 青岛市建筑设计研究院集团股份有限公司 一种城市规划多源数据优化处理方法
CN118051862A (zh) * 2024-04-16 2024-05-17 洛阳禾安工程技术服务有限公司 一种建筑涂料的检测方法及系统
CN118070080A (zh) * 2024-04-17 2024-05-24 山东中电仪表有限公司 一种多功能电能表的用户电耗数据智能分析方法及系统

Families Citing this family (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113297539A (zh) * 2021-05-07 2021-08-24 天云融创数据科技（北京）有限公司 户变关系异常识别方法、装置、电子设备和介质
CN113657171A (zh) * 2021-07-20 2021-11-16 国网上海市电力公司 基于图小波神经网络的低压配电网台区拓扑识别方法
CN113804965B (zh) * 2021-09-28 2024-05-31 北京来也网络科技有限公司 基于rpa和ai的异常计量点识别方法和装置
CN113744089B (zh) * 2021-11-08 2022-02-15 广东电网有限责任公司佛山供电局 一种台区户变关系辨识方法及设备
CN114152909A (zh) * 2021-11-29 2022-03-08 国网江苏省电力有限公司营销服务中心 基于大数据的中高压失准分析系统
CN114240690A (zh) * 2021-12-17 2022-03-25 国网辽宁省电力有限公司沈阳供电公司 基于多聚合用户电表数据的用户用电异常监测方法
CN115203286A (zh) * 2022-05-25 2022-10-18 广州番禺职业技术学院 一种基于知识驱动的台区户变关系识别方法
CN115344567A (zh) * 2022-10-18 2022-11-15 国网天津市电力公司营销服务中心 一种适用于边缘计算的低压台区数据清洗治理方法及装置
CN115542062B (zh) * 2022-11-07 2024-01-09 北京志翔科技股份有限公司 户变关系异常的识别方法、装置、设备和存储介质
CN118033249A (zh) * 2024-04-11 2024-05-14 国网江苏省电力有限公司常州供电分公司 台区相位识别方法和装置

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109325545A (zh) * 2018-10-16 2019-02-12 南方电网科学研究院有限责任公司 低压配电网拓扑结构校验方法、装置、设备及存储介质
CN109461096A (zh) * 2018-11-05 2019-03-12 杭州致成电子科技有限公司 一种基于电量关联性的户变关系异常判别方法
EP3489508A1 (en) * 2016-07-21 2019-05-29 Zhejiang Windey Co., Ltd. Method and device for automatically calibrating wind alignment error of wind power generation unit
CN110120668A (zh) * 2019-05-08 2019-08-13 许昌许继软件技术有限公司 一种台区拓扑自动识别方法及系统
CN111126429A (zh) * 2019-11-10 2020-05-08 国网浙江省电力有限公司 一种基于PCA降维和K-Means聚类的低压台区用户接入点识别方法

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8095521B2 (en) * 2007-03-30 2012-01-10 Amazon Technologies, Inc. Recommendation system with cluster-based filtering of recommendations
CN104376057A (zh) * 2014-11-06 2015-02-25 南京邮电大学 一种基于最大最小距离和K-means的自适应聚类方法
CN109685128A (zh) * 2018-12-18 2019-04-26 电子科技大学 一种MB-kmeans++聚类方法及基于其的用户会话聚类方法
CN110084718B (zh) * 2019-04-25 2023-08-22 国网湖南省电力有限公司 一种台区户变关系精准校验方法及装置
CN110727662A (zh) * 2019-09-10 2020-01-24 国网浙江省电力有限公司电力科学研究院 基于相关性分析的低压台区用户相位识别方法和系统
CN111061821B (zh) * 2019-11-12 2024-02-02 国网山东省电力公司聊城供电公司 基于改进k值聚类算法的低压配电网拓扑校验方法及系统
CN111444241B (zh) * 2020-03-26 2020-12-22 南京工程学院 一种基于数据挖掘的台区线损异常关联用户精准定位方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP3489508A1 (en) * 2016-07-21 2019-05-29 Zhejiang Windey Co., Ltd. Method and device for automatically calibrating wind alignment error of wind power generation unit
CN109325545A (zh) * 2018-10-16 2019-02-12 南方电网科学研究院有限责任公司 低压配电网拓扑结构校验方法、装置、设备及存储介质
CN109461096A (zh) * 2018-11-05 2019-03-12 杭州致成电子科技有限公司 一种基于电量关联性的户变关系异常判别方法
CN110120668A (zh) * 2019-05-08 2019-08-13 许昌许继软件技术有限公司 一种台区拓扑自动识别方法及系统
CN111126429A (zh) * 2019-11-10 2020-05-08 国网浙江省电力有限公司 一种基于PCA降维和K-Means聚类的低压台区用户接入点识别方法

Cited By (30)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114825647A (zh) * 2022-06-30 2022-07-29 国网浙江省电力有限公司宁波供电公司 一种基于拓扑关系的台区线损监测方法及相关设备
CN115423250A (zh) * 2022-07-28 2022-12-02 国网浙江省电力有限公司营销服务中心 一种台区户变关系分析方法
CN115470858A (zh) * 2022-09-13 2022-12-13 国网天津市电力公司 基于ddtw距离和最小最大距离算法的配电台区拓扑辨识方法
CN115235676A (zh) * 2022-09-26 2022-10-25 山东济矿鲁能煤电股份有限公司阳城煤矿 一种矿井液压支架压力异常自动报警方法
CN115235676B (zh) * 2022-09-26 2022-12-09 山东济矿鲁能煤电股份有限公司阳城煤矿 一种矿井液压支架压力异常自动报警方法
CN115438035A (zh) * 2022-10-27 2022-12-06 江西师范大学 一种基于kpca和混合相似度的数据异常处理方法
CN116303480B (zh) * 2023-05-22 2023-08-04 新乡市万新电气有限公司 基于云计算的电能表误差校验方法
CN116599055B (zh) * 2023-05-26 2023-12-01 联桥科技有限公司 一种低压配网台区的拓扑网络识别方法及系统
CN116599055A (zh) * 2023-05-26 2023-08-15 联桥科技有限公司 一种低压配网台区的拓扑网络识别方法及系统
CN116482471B (zh) * 2023-06-21 2023-09-08 四川中电启明星信息技术有限公司 一种基于电压时空聚合曲线的户变关系识别方法
CN116482471A (zh) * 2023-06-21 2023-07-25 四川中电启明星信息技术有限公司 一种基于电压时空聚合曲线的户变关系识别方法
CN116499419A (zh) * 2023-06-27 2023-07-28 西安高商智能科技有限责任公司 一种舵机旋转角度异常检测方法及系统
CN116499419B (zh) * 2023-06-27 2023-10-20 西安高商智能科技有限责任公司 一种舵机旋转角度异常检测方法及系统
CN116522181A (zh) * 2023-06-28 2023-08-01 广东电网有限责任公司 一种配电台区日线损特征提取方法及系统
CN116522181B (zh) * 2023-06-28 2024-03-29 广东电网有限责任公司 一种配电台区日线损特征提取方法及系统
CN116595426A (zh) * 2023-07-17 2023-08-15 济南大陆机电股份有限公司 一种工业物联网数据智能采集管理系统
CN116595426B (zh) * 2023-07-17 2023-09-26 济南大陆机电股份有限公司 一种工业物联网数据智能采集管理系统
CN116910596A (zh) * 2023-07-26 2023-10-20 江苏方天电力技术有限公司 一种基于改进dbscan聚类的用户窃电分析方法、设备及存储介质
CN116881745A (zh) * 2023-09-07 2023-10-13 深圳市特安工业科技有限公司 基于大数据的压力变送器异常监测方法
CN116881745B (zh) * 2023-09-07 2023-12-19 深圳市特安工业科技有限公司 基于大数据的压力变送器异常监测方法
CN116954342B (zh) * 2023-09-18 2023-12-15 惠州市鑫晖源科技有限公司 基于物联网的全方位高效散热主机电源运行监测方法
CN116954342A (zh) * 2023-09-18 2023-10-27 惠州市鑫晖源科技有限公司 基于物联网的全方位高效散热主机电源运行监测方法
CN117129790A (zh) * 2023-10-26 2023-11-28 山西思极科技有限公司 一种电力系统故障诊断系统
CN117129790B (zh) * 2023-10-26 2024-01-23 山西思极科技有限公司 一种电力系统故障诊断系统
CN117148023A (zh) * 2023-10-31 2023-12-01 威海海泰电子有限公司 一种智能电源适配器生产检测方法及系统
CN117148023B (zh) * 2023-10-31 2024-01-26 威海海泰电子有限公司 一种智能电源适配器生产检测方法及系统
CN117648657A (zh) * 2023-12-13 2024-03-05 青岛市建筑设计研究院集团股份有限公司 一种城市规划多源数据优化处理方法
CN117648657B (zh) * 2023-12-13 2024-05-14 青岛市建筑设计研究院集团股份有限公司 一种城市规划多源数据优化处理方法
CN118051862A (zh) * 2024-04-16 2024-05-17 洛阳禾安工程技术服务有限公司 一种建筑涂料的检测方法及系统
CN118070080A (zh) * 2024-04-17 2024-05-24 山东中电仪表有限公司 一种多功能电能表的用户电耗数据智能分析方法及系统

Also Published As

Publication number Publication date
CN112699913A (zh) 2021-04-23
CN112699913B (zh) 2023-08-29

Similar Documents

Publication Publication Date Title
WO2022110557A1 (zh) 一种台区户变关系异常诊断方法及装置
CN111505433B (zh) 一种低压台区户变关系纠错及相位识别方法
CN106709662B (zh) 一种电力设备运行工况划分方法
CN110008254B (zh) 一种变电设备台账核查处理方法
CN109829497B (zh) 一种基于监督学习的台区用户识别及判别方法
CN112800231B (zh) 电力数据校验方法、装置、计算机设备和存储介质
CN111476435B (zh) 基于密度峰值的充电桩负荷预测方法
CN108334894B (zh) 基于无监督机器学习的变压器油温异常识别方法
CN108133225A (zh) 一种基于支持向量机的覆冰闪络故障预警方法
CN110687361A (zh) 一种用于识别低压配网台区相序的方法及系统
CN111401785A (zh) 一种基于模糊关联规则的电力系统设备故障预警方法
CN110544047A (zh) 一种不良数据辨识方法
CN111709668A (zh) 基于数据挖掘技术的电网设备参数风险识别方法及装置
CN117214637B (zh) 充电枪耐压测试方法、装置、设备及存储介质
CN113408548A (zh) 变压器异常数据检测方法、装置、计算机设备和存储介质
CN113125903A (zh) 线损异常检测方法、装置、设备及计算机可读存储介质
CN116011158A (zh) 一种低压台区的拓扑识别方法、系统及装置
CN112001644A (zh) 一种配电网运行可靠性检测方法、装置、终端及存储介质
CN113484817A (zh) 基于tsvm模型的智能电能表自动化检定系统异常检测方法
CN115033591A (zh) 一种电费数据异常智能检测方法、系统、存储介质及计算机设备
CN117154716B (zh) 一种分布式电源接入配电网的规划方法及系统
CN109389517B (zh) 一种量化线路损耗影响因素的分析方法及装置
CN117236022A (zh) 互感器剩余寿命预测模型训练方法、应用方法及电子设备
Li et al. Wind pressure coefficients zoning method based on an unsupervised learning algorithm
CN109784777B (zh) 基于时序信息片段云相似度度量的电网设备状态评估方法

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21896077

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 21896077

Country of ref document: EP

Kind code of ref document: A1