CN108734603A - Big data phase recognition methods based on cluster - Google Patents

Abstract

The present invention relates to one kind to be handled based on K-means clustering algorithms for user's phase information in radio area, to complete the classification for user node in radio area.Key step of the present invention includes：Acquire the voltage data of the intelligent electric meter of Operation of Electric Systems, voltage data is handled by PCA, scatterplot is projected according to Data Dimensionality Reduction and sets initial cluster center, and Clustering processing is carried out to the affiliated phase information of voltage using K-means algorithms, is realized to the Classification and Identification for user's phase in radio area.Advantage is that the big data phase recognition methods based on cluster are directly clustered the information of voltage for belonging to phase using K-means methods, it can be with accurate judgement user's phase information, power distribution station network topology and three-phase imbalance are administered, support is provided, the quantity of rationally adjustment and planning access per phase user, to maintain the stabilization of Operation of Electric Systems；The drawbacks of extra hardware expense need not be increased, need not be arranged and compare threshold value, compensate for current phase identification technology.

Description

Clustering-based Big Data Phase Recognition Method

technical field

The invention relates to the field of classification of node phases in a user station area, in particular to a K-means clustering big data phase identification method, which belongs to the application field of distribution network electrical parameter measurement.

Background technique

In the distribution network operation and management of the power company, many issues related to the phase of the user's electric energy meter are involved, such as: the adjustment of the three-phase power imbalance in the transformer station area. When the management personnel find that there is a serious three-phase power imbalance in the power system When adjusting the user's phase, because you don't know the specific phase of the user, you don't know which users to choose for adjustment, and you can't evaluate the new customer's access to the network so that they can be installed on the most reasonable phase line. For another example, when conducting line loss analysis, it is found that when a certain phase line loss is large, there are hidden dangers of line aging and power theft. However, because the phase of the user is not known, the inspection scope cannot be narrowed, and the workload is very heavy. User phase recognition is very important. At present, there are three methods for user phase identification: the installation phase identification of user nodes can be carried out through the power line carrier communication technology; the secondary circuit identification can be constructed by using the phase meter; the staff can identify on the spot according to the color of the access user line. The current phase identification technology has many deficiencies: although the installation phase identification of user nodes can be carried out through power line carrier communication technology, phase identification cannot be performed for station areas without power carrier; , the degree of automation needs to be improved, and it is necessary to build a secondary circuit, which requires a large investment and workload; in the process of user access processing, there is a lack of unified requirements for the color of the power line, and it is impossible to accurately identify the phase according to the color of the user line.

The user phase identification method based on big data analysis is particularly important. At present, my country has basically realized the full coverage of smart energy meters. The user phase identification method based on big data analysis does not need to increase additional hardware overhead. The collected user phase voltage information is processed and classified by writing algorithms, and the user phase voltage information can be accurately obtained. The network topology of the distribution station area can greatly reduce the work intensity of the front-line inspection personnel and improve the operation and management level of the power company.

Although in the field of big data phase identification methods, the classification of user phases can be completed by linear correlation method, but this method will affect the classification results because of the selection of threshold; the clustering-based user phase identification method is an effective big data phase identification The method can be used as a test standard for methods such as phase identification of power carriers.

Contents of the invention

A clustering-based big data phase identification method, using K-means clustering algorithm to process user phase information in the power supply station area, characterized in that: the voltage data of the smart meter running in the power system is collected, and the voltage data is processed by PCA, according to The initial clustering center is set by data dimensionality reduction and projection scatter points, and the K-means algorithm is used to cluster and group the phase information of the voltage, so as to realize the classification and identification of user phases in the power supply station area. The clustering-based big data phase identification method directly uses the K-means method to cluster the voltage information of the same phase, which can accurately determine the user phase information, provide support for the network topology of the distribution station area and the three-phase imbalance control, and make reasonable adjustments And plan the number of users connected to each phase to maintain the stability of the power system; there is no need to add additional hardware overhead, and there is no need to set a comparison threshold, which makes up for the shortcomings of the current phase identification technology.

The effective effects of the present invention are: 1. The clustering-based big data phase identification method directly uses the K-means method to cluster the voltage information of the phase, which can accurately determine the phase information of the user, and the network topology of the distribution station area and the three-phase Unbalance governance provides support to maintain the stability of power system operation; 2. There is no need for special tool identification such as carrier phase, which reduces workload, does not require additional hardware overhead, and does not need to set comparison thresholds. Measure voltage function phases through smart energy meters The precise identification makes up for the disadvantages of the current phase identification technology; 3 realizes cross-region identification, which can solve the problem of file entry errors in the marketing system of power companies, solve the problem of line loss statistics errors caused by it, and rationally adjust and plan access The number of users provides support for the network topology and three-phase unbalance governance of the distribution station area, and evaluates the access of new customers to the network in advance.

Description of drawings

Fig. 1 is the flow chart of the main program of the system of the present invention.

Figure 2 is a flowchart of the K-mean K-means subroutine.

Fig. 3 Simulation experiment verification of clustering-based big data phase recognition.

Fig. 4 The experimental verification of clustering-based big data phase recognition.

Detailed ways

The present invention will be further described in detail below in combination with specific embodiments.

The transformer station area in the distribution network is equipped with a master meter, and the three-phase phase of the transformer station master meter is determined and can be used as a reference;

The user's phase information is measured by the smart energy meter, and the system collects the user's phase voltage data information;

System data import algorithm, calling PCA to process voltage data for the first time, data dimensionality reduction, and projecting data to a two-dimensional plane (scatter diagram representation);

Select the initial cluster center according to the shape of the scatter diagram, and record the coordinate value of the initial cluster center;

The user phases in the same station area are A, B and C, set the number of clustering groups K=3, and declare the intermediate variables of the K-means clustering algorithm;

Input the initial cluster center coordinates, the number of cluster centers, and the number of iterations;

Carry out K-means program processing;

Return the cluster classification results: user phase classification results, cluster center point coordinates, square error function value, number of iterations and the number of midpoints in each cluster;

According to the classification results of the Kmeans algorithm, the data is projected onto a 2-dimensional plane, and different colors are used to indicate the user phase classification in the station area to complete the phase identification;

Call the K-means program to cluster and group the phase information of the voltage. The specific steps are:

(1) PCA dimensionality reduction to process the voltage data of the user station area;

(2) Input the selected initial clustering center of the PCA scatter diagram, the number of clustering groups K=3 and set the maximum number of iterations and square error threshold;

(3) Declare the Kmeans array and intermediate variables;

(4) Use the Euclidean distance formula to calculate the distance from the cluster center to each voltage point;

(5) Update the classification results according to the point to the nearest cluster center, and redistribute the voltage data to the relevant phase group;

(6) Update the cluster center coordinates according to the average value of the coordinates of each point in each cluster;

(7) Update the square error function value;

(8) Perform steps 3, 4, 5, 6, and 7 reciprocally until the number of iterations reaches the maximum number of iterations, or the square error is less than the threshold, and returns the classification result, cluster center coordinates, square error function value, iteration number, and each cluster The number of data in .

The system main program flow chart of the present invention is as shown in accompanying drawing 1.

The block diagram of K-means programming is shown in Figure 2.

The scheme proposed by the present invention is verified by following simulation experiments and measured data:

The simulation experiment data contains 300 users (including 300 user data collected in three phases A, B, and C), and each user collects 1000 data, and the data is brought into the program to obtain the following results:

Among them, the three colors shown in Figure 3 represent different clusters, and the red dots represent the final cluster centers. The output results are automatically saved under the data folder, displaying the number of clustering groups and the final clustering center in each cluster, and adding a label to each data, and saving the data in each cluster in CSV form , the conclusion about user phase clustering is consistent with the simulation design.

The measured data includes 25 users (including 25 user data collected in the three phases of A, B, and C), and each user collects 500 data, and the result of bringing the data into the program is shown in Figure 4, and the user phases are clustered The conclusion is consistent with the experimental conclusion.

The contents not described in detail in the description of the present invention belong to the prior art known to those skilled in the art.

Claims (4)

1. A large data phase identification method based on clustering, which uses the K-means clustering algorithm to process the user phase information in the power supply station area, and is characterized in that: the voltage data of the smart electric meter operating in the power system is collected, and the voltage data is processed by PCA , set the initial clustering center according to the data dimensionality reduction projection scatter points, and use the K-means algorithm to cluster and group the phase information of the voltage, so as to realize the classification and identification of the user phase in the power supply station area. 2. According to the described clustering-based big data phase recognition method according to claim 1, its technical feature is that: the phase information of users in the power supply station area is specifically processed by the clustering algorithm as follows: (1) Perform PCA dimension reduction processing on the voltage data; (2) According to the phase data of the user station area collected by the smart meter, the user phases in the same station area are A, B and C three phases, set the number of clustering groups K=3, and obtain the scatter diagram according to the PCA dimension reduction process , select three voltage data as the initial clustering center of the phase; (3) Call the K-means program; (4) Return the cluster classification results: user phase classification results, cluster center point coordinates, square error function value, number of iterations and the number of midpoints in each cluster; (5) According to the classification results of the Kmeans algorithm, the data is projected onto a 2-dimensional plane, and different colors are used to represent the user phase classification in the station area to complete the phase identification. 3. according to the described big data phase recognition method based on clustering of right 2, its technical feature is: call K-means program to carry out clustering grouping process to voltage belonging phase information, concrete steps are: (1) PCA dimensionality reduction to process the voltage data of the user station area; (2) Input the selected initial clustering center of the PCA scatter diagram, the number of clustering groups K=3 and set the maximum number of iterations and square error threshold; (3) Declare the Kmeans array and intermediate variables; (4) Use the Euclidean distance formula to calculate the distance from the cluster center to each voltage point; (5) Update the classification results according to the point to the nearest cluster center, and redistribute the voltage data to the relevant phase group; (6) Update the cluster center coordinates according to the average value of the coordinates of each point in each cluster; (7) Update the square error function value; (8) Perform steps 3, 4, 5, 6, and 7 reciprocally until the number of iterations reaches the maximum number of iterations, or the square error is less than the threshold, and returns the classification result, cluster center coordinates, square error function value, iteration number, and each cluster The number of data in . 4. According to the clustering-based big data phase identification method described in right 2 or right 3, its technical features are: process voltage data through two PCA; first PCA process voltage data selects 3 phase initial clustering centers, concrete steps In order to process the voltage data through PCA dimension reduction and form a two-dimensional plane user voltage data scatter diagram, select the initial cluster center according to the shape of the scatter diagram, which can effectively reduce the error of the clustering result; the second PCA process user voltage data Down to two dimensions in order to bring the data into the K-means clustering algorithm.