CN106548301B - Power consumer clustering method and device - Google Patents

Power consumer clustering method and device Download PDF

Info

Publication number
CN106548301B
CN106548301B CN201611081166.6A CN201611081166A CN106548301B CN 106548301 B CN106548301 B CN 106548301B CN 201611081166 A CN201611081166 A CN 201611081166A CN 106548301 B CN106548301 B CN 106548301B
Authority
CN
China
Prior art keywords
clustering
power
preset
time point
time period
Prior art date
Legal status (The legal status 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 status listed.)
Active
Application number
CN201611081166.6A
Other languages
Chinese (zh)
Other versions
CN106548301A (en
Inventor
李秋硕
肖勇
钱斌
李鹏
孙宇军
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
China South Power Grid International Co ltd
Original Assignee
China South Power Grid International Co ltd
Power Grid Technology Research Center of China Southern Power Grid Co Ltd
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 China South Power Grid International Co ltd, Power Grid Technology Research Center of China Southern Power Grid Co Ltd filed Critical China South Power Grid International Co ltd
Priority to CN201611081166.6A priority Critical patent/CN106548301B/en
Publication of CN106548301A publication Critical patent/CN106548301A/en
Application granted granted Critical
Publication of CN106548301B publication Critical patent/CN106548301B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0637—Strategic management or analysis, e.g. setting a goal or target of an organisation; Planning actions based on goals; Analysis or evaluation of effectiveness of goals
    • 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
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Energy or water supply

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Human Resources & Organizations (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Economics (AREA)
  • Strategic Management (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Marketing (AREA)
  • Health & Medical Sciences (AREA)
  • Tourism & Hospitality (AREA)
  • Educational Administration (AREA)
  • General Business, Economics & Management (AREA)
  • Water Supply & Treatment (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Quality & Reliability (AREA)
  • Public Health (AREA)
  • Game Theory and Decision Science (AREA)
  • General Health & Medical Sciences (AREA)
  • Primary Health Care (AREA)
  • Probability & Statistics with Applications (AREA)
  • Operations Research (AREA)
  • Development Economics (AREA)
  • Artificial Intelligence (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Supply And Distribution Of Alternating Current (AREA)

Abstract

The invention discloses a power consumer clustering method and device, wherein the starting time points of preset clustering numbers are selected from a starting time point set obtained after sequencing according to a preset selection rule to initialize the starting time points of the clustering centers of the preset numbers, the starting time points of the preset clustering numbers are selected from an ending time point set obtained after sequencing according to the preset selection rule to initialize the ending time points of the clustering centers of the preset numbers, and then clustering is carried out through a K-means clustering algorithm to obtain a clustering result. When the clustering centers with the preset clustering numbers are initialized by the power consumer clustering method and the device, the time period with the preset clustering numbers is not selected randomly from the peak time period of the power load of the power consumers to serve as the initialized clustering centers, and the new time period formed by the starting time point and the ending time point corresponding to each ranking number is initialized to the clustering centers with the preset clustering numbers, so that more accurate clustering results can be obtained, and the clustering accuracy is improved.

Description

Power consumer clustering method and device
Technical Field
The invention relates to the technical field of electric power, in particular to a method and a device for clustering electric power users.
Background
In the process that the power system provides electric energy for users, the electricity utilization conditions of different power users may be different, and the classification of the power users has important influence on the economic analysis, operation and planning of the power system. At present, a commonly adopted power consumer clustering method is a K-means (K mean value) clustering algorithm, which mainly selects K data from a data set S as centers of initial clustering and clusters each data in the data set with the center closest to the center. Firstly, randomly selecting K data as initial centers, calculating the distance from each data to each selected center, assigning the data to the nearest center to form a class, calculating the mean value of each class, and circularly and repeatedly executing until the function convergence meeting the clustering criterion is met. Wherein, inputting: the initial data set S and the number of classes K. And (3) outputting: k clustering categories meet the convergence of a square error criterion function, and the specific working steps are as follows: 1) randomly selecting K data from the data set as an initial clustering center; 2) calculating the distance from each data to each selected center, and assigning the data to the nearest center to form a class; 3) taking the mean value of the formed class data as a corresponding new clustering center, and updating each data to the most similar class; 4) a clustering criterion function E (typically the mean square error) is calculated until the criterion function E starts to converge.
The K-means clustering algorithm firstly needs to initialize a clustering center, that is, the initial clustering center has a large influence on the subsequent clustering effect, however, when the K-means clustering algorithm is adopted for clustering the power consumers, K data are randomly selected from a data set as the initial centers, and then the subsequent clustering step is carried out, so that the accuracy of the power consumer clustering cannot be ensured.
Disclosure of Invention
Therefore, it is necessary to provide a power consumer clustering method and device for improving the power consumer clustering accuracy, aiming at the problem that the power consumer clustering is inaccurate.
A power consumer clustering method comprises the following steps:
acquiring a peak time period of the power load of each power consumer, wherein the peak time period comprises a starting time point and an ending time point;
sequencing the starting time points in the peak time periods of the power consumers according to a preset sequencing rule to obtain a starting time point set, and sequencing the ending time points in the peak time periods of the power consumers to obtain an ending time point set;
obtaining ranking numbers of preset clustering numbers according to a preset selection rule, selecting a starting time point corresponding to each ranking number from the starting time point set, and selecting an ending time point corresponding to each ranking number from the ending time point set;
constructing a new time period of the preset clustering number according to the starting time point corresponding to each ranking number and the ending time point corresponding to each ranking number;
and respectively initializing the new time periods of the preset clustering number to be clustering centers of the preset clustering number, and clustering the power consumers according to the peak time period of the power loads of the power consumers, the clustering centers and a K-means clustering algorithm to obtain clustering results of the power consumers.
The invention also provides a power consumer clustering device, which comprises:
the load peak time period acquisition module is used for acquiring the peak time period of the power load of each power consumer, wherein the peak time period comprises a starting time point and an ending time point;
the sorting module is used for sorting the starting time points in the peak time period of each power consumer according to a preset sorting rule to obtain a starting time point set, and sorting the ending time points in the peak time period of each power consumer to obtain an ending time point set;
the selection module is used for obtaining ranking numbers of preset clustering numbers according to a preset selection rule, selecting a starting time point corresponding to each ranking number from the starting time point set, and selecting an ending time point corresponding to each ranking number from the ending time point set;
the construction module is used for constructing a new time period of the preset clustering number according to the starting time point corresponding to each ranking number and the ending time point corresponding to each ranking number;
and the clustering module is used for respectively initializing the new time periods of the preset clustering number to clustering centers of the preset clustering number, clustering the power consumers according to the peak time period of the power load of each power consumer, the clustering centers and a K-means clustering algorithm, and obtaining the clustering results of the power consumers.
According to the power consumer clustering method and device, the starting time points of the peak time periods of the power consumers are arranged according to the time sequence, the starting time points of the preset clustering number are selected from the starting time point set obtained after sequencing according to the preset selection rule to initialize the starting time points of the preset clustering centers, similarly, the starting time points of the preset clustering number are selected from the ending time point set obtained after sequencing according to the preset selection rule to initialize the ending time points of the preset clustering centers, namely, the clustering centers of the preset clustering number are initialized, and then the power consumers are clustered according to the peak time periods of the power loads of the power consumers, the clustering centers and the K-means clustering algorithm to obtain the clustering results of the power consumers. When the clustering centers with the preset clustering numbers are initialized by the power consumer clustering method and the device, the time period with the preset clustering numbers is not selected randomly from the peak time period of the power load of the power consumers as the initialized clustering centers, but the clustering centers with the preset clustering numbers are initialized according to the new time period formed by the starting time point corresponding to each ranking number and the ending time point corresponding to each ranking number, so that the clustering centers with the preset clustering numbers can be initialized reasonably, and the power consumers are clustered according to a K-means clustering algorithm after the clustering centers are initialized, so that more accurate clustering results can be obtained, and the clustering accuracy is improved.
Drawings
FIG. 1 is a flow chart of a power consumer clustering method according to an embodiment;
FIG. 2 is a sub-flowchart of a power consumer clustering method according to another embodiment;
FIG. 3 is a block diagram of an electric power consumer clustering device according to an embodiment;
fig. 4 is a sub-block diagram of a power consumer clustering device according to another embodiment.
Detailed Description
Referring to fig. 1, an embodiment of a power consumer clustering method is provided, which includes the following steps:
s110: and acquiring the peak time period of the power load of each power consumer.
Wherein the peak time period includes a start time point and an end time point. The power system supplies power to the power consumers, each power consumer corresponds to a respective power load, and the power consumers have peak time periods of the respective power loads, namely time periods corresponding to power peak hours. In this embodiment, specifically, the peak time period of the electrical load of each electrical consumer within the preset time period is obtained, for example, within one day of the preset time period, in the time period in which the electrical load of the electrical consumer is higher than 70% of the maximum value of the electrical load, if the data interval is 1 hour, the time at which the first electrical load exceeds 70% of the maximum value of the electrical load is pushed forward by 30 minutes to be the peak start time, that is, the start time point, and if the time is 0 point, the peak start time is 0 point. And the time point of the end of the peak, namely the end time point, is deduced backwards by 30 minutes when the last continuous electric load value is higher than 70% of the maximum electric load value. If the data interval is 15 minutes, the time when the first electrical load exceeds 70% of the maximum value of the electrical load is advanced by 15 minutes to be the peak start time, and if the time is 0 point, the peak start time is 0 point. And if the time is 23 points and 45 points, the peak ending time is 23 points and 45 points.
S120: and sequencing the starting time points in the peak time periods of the power consumers according to a preset sequencing rule to obtain a starting time point set, and sequencing the ending time points in the peak time periods of the power consumers to obtain an ending time point set.
Due to the fact that the electricity utilization conditions of the power consumers are different, the corresponding electricity loads are different, namely the peak time periods of the corresponding electricity loads may be different, and therefore the starting time point and the ending time period in the peak time periods of the power consumers may be different. In this embodiment, according to a preset sorting rule, the start time points in the peak time period of each power consumer are sorted to obtain a start time point set, and the end time points in the peak time period of each power consumer are sorted to obtain an end time point set. The preset ordering rule may be an ordering rule with time from small to large, or an ordering rule with time from large to small.
S130: obtaining the ranking numbers of the preset clustering numbers according to a preset selection rule, selecting the starting time point corresponding to each ranking number from the starting time point set, and selecting the ending time point corresponding to each ranking number from the ending time point set.
When power consumers are clustered, a preset clustering number is preset, after a start time point and an end time point in a peak time period of each power consumer are respectively sequenced to obtain a start time point set and an end time point set, the start time point of the preset clustering number needs to be selected from the start time point set, the end time point of the preset clustering number needs to be selected from the end time set, in this embodiment, ranking numbers of the preset clustering number are obtained through a preset selection rule, the start time point corresponding to each ranking number is selected from the start time point set, and the end time point corresponding to each ranking number is selected from the end time point set.
S140: and constructing a new time period of the preset clustering number according to the starting time point corresponding to each row position and the ending time point corresponding to each row position.
S150: and respectively initializing the new time periods with the preset clustering number into clustering centers with the preset clustering number, and clustering the power consumers according to the peak time period of the power load of each power consumer, the clustering centers and a K-means clustering algorithm to obtain the clustering results of each power consumer.
And obtaining the ranking of the preset clustering number, thereby obtaining the starting time point and the ending time point corresponding to the ranking, and constructing a new time period of the preset clustering number according to the starting time point corresponding to each ranking and the ending time point corresponding to each ranking. When clustering is performed by using the K-means clustering algorithm, firstly, the clustering centers of the preset clustering numbers need to be initialized, in this embodiment, new time periods of the preset clustering numbers are respectively initialized to the clustering centers of the preset clustering numbers, and then, clustering is performed on the power consumers by using the K-means clustering algorithm to obtain clustering results, wherein the clustering characteristics adopted during clustering are peak time periods of the power consumers.
The power consumer clustering method includes the steps that starting time points of peak time periods of power consumers are arranged according to a time sequence, starting time points of preset clustering numbers are selected from a starting time point set obtained after sequencing according to a preset selection rule to initialize starting time points of the preset clustering centers, similarly, starting time points of the preset clustering numbers are selected from an ending time point set obtained after sequencing according to the preset selection rule to initialize ending time points of the preset clustering centers, namely, the clustering centers of the preset clustering numbers are initialized, then the power consumers are clustered according to the peak time periods of power loads of the power consumers, the clustering centers and a K-means clustering algorithm, and clustering results of the power consumers are obtained. When the clustering centers with the preset clustering numbers are initialized by the power consumer clustering method, the time period with the preset clustering numbers is not selected from the peak time period of the power load of the power consumers randomly as the initialized clustering centers, but the new time period formed by the starting time point corresponding to each ranking number and the ending time point corresponding to each ranking number is initialized to the clustering centers with the preset clustering numbers, so that the clustering centers with the preset clustering numbers can be initialized reasonably, and the power consumers are clustered according to a K-means clustering algorithm after the clustering centers are initialized, so that more accurate clustering results can be obtained, and the clustering accuracy is improved.
In the process that the power system provides electric energy for users, the electricity utilization conditions of different power users may be different, and the clustering result of the power users has an important influence on economic analysis, operation and planning of the power system, so that after the power users are clustered to obtain the clustering result, the power system can be planned according to the clustering result of the power users, for example, electric energy with higher power can be provided for the power users with larger electricity loads. By improving the accuracy of the clustering result of the power consumer, the power system can be effectively and accurately planned, so that the power system can stably operate.
In one embodiment, the step of initializing the new time periods of the preset clustering number to the clustering centers of the preset clustering number respectively, and clustering the power consumers according to the peak time period of the power loads of the power consumers, the clustering centers and the K-means clustering algorithm to obtain the clustering results of the power consumers comprises:
s251: and respectively initializing the new time periods with the preset clustering number into the clustering centers with the preset clustering number.
And respectively initializing the new time periods with the preset clustering number into the clustering centers with the preset clustering number, so as to obtain the clustering centers with the preset clustering number.
S252: and respectively calculating the coincidence rate of the peak time period of the power load of each power consumer and each clustering center.
S253: and respectively dividing the power consumers to the clustering centers corresponding to the peak time periods of the power consumers and the maximum coincidence rate of each clustering center to obtain the clustering categories of the preset clustering numbers.
S254: and calculating the mean value of the starting time point and the mean value of the ending time point of the peak time period corresponding to the power consumer in each cluster category.
S255: and updating the clustering center of the clustering category where the power consumer is located into a time period from the mean value of the starting time point to the mean value of the ending time point corresponding to the clustering category.
S256: and judging whether the updated clustering centers have clustering centers different from the corresponding clustering centers before updating.
When the updated cluster center is the same as the corresponding cluster center before updating, S257 is executed: and determining the clustering result of each power user as the clustering category of the preset clustering number.
And when the updated clustering centers have clustering centers different from the corresponding clustering centers before updating, returning to the step of respectively calculating the coincidence rate of the peak time period of the power load of each power consumer and each clustering center.
After the clustering centers are initialized, the coincidence rate of the peak time periods of the power loads of the power consumers and the clustering centers is calculated, then the power consumers are classified into the clustering centers corresponding to the peak time periods of the power consumers and the maximum coincidence rate of the clustering centers, the clustering categories with preset clustering numbers are obtained, and the initial clustering of the power consumers is realized. After the cluster types with the preset cluster number are obtained, calculating the starting time point mean value and the ending time point mean value of the peak time period corresponding to the power users in each cluster type, updating the cluster centers of the cluster types where the power users are located into the time periods from the starting time point mean value to the ending time point mean value corresponding to the cluster types, and repeating the steps until the cluster centers with the preset cluster number do not change any more.
In one embodiment, the coincidence rate is the coincidence period between the peak time period of the electric load of the electric power consumer and the clustering center divided by the longer time period between the peak time period of the electric load of the electric power consumer and the clustering center.
The larger the coincidence rate is, the larger the time period in which the peak time period of the power load of the power consumer and the peak time period of the power load of the power consumer are summed with the longer time period in the clustering center is, the closer the peak time period of the power load of the power consumer and the clustering center is, and the higher the probability that the power consumer corresponding to the peak time period of the power load is classified into the clustering center is.
In one embodiment, obtaining the ranking number of the preset cluster number according to the preset selection rule includes:
obtaining the ranking number of the preset cluster number according to the following formula:
Ph＝round(m/(k+1))*h-1。
wherein, PhH is more than or equal to 1 and less than or equal to k, k is the preset clustering number, m is the number of power users, and the round function is a rounded function.
The above-mentioned power consumer clustering method is specifically described in an embodiment.
When the power consumers need to be clustered, the peak time periods in the power loads can be analyzed, and the clustering of the power consumers can be realized by performing clustering analysis on the peak time periods. The invention relates to a power consumer clustering method based on a K-means clustering algorithm (a time-period K-means algorithm), wherein the preset clustering number K is 10, and the power consumers are assumed to be 100, namely the peak time periods needing clustering are 100 in total and need to be clustered into 10 classes. The specific clustering steps are as follows:
(1) initializing 10 cluster centers
The start times (start time points) of all peak periods to be clustered are ranked in time order, and a set of start time points (set as set 1) is obtained as follows:
set 1 ═ start time point 1, start time point 2, start time point 3, …, start time point 100.
Where start time point 1 is earlier than start time point 2, start time point 2 is earlier than start time point 3, and so on, start time point 100 is the largest. Similarly, the end times (end time points) of all peak periods that need to be clustered are chronologically ranked, and an end time set (set 2) is obtained as follows:
set 2 ═ end time point 1, end time point 2, end time point 3, …, end time point 100}
Where end time point 1 is earlier than end time point 2, end time point 2 is earlier than end time point 3, and so on, end time point 100 is the largest.
According to the formula: round (m/(k + 1)). h-1 (note: round is a rounded function, h is 1, 2, 3,. k, k is a preset number of clusters, and m is the number of the clustered samples, namely the number of power consumers) selects the ranking numbers which are respectively 8, 17, 26, 35, 44, 53, 62, 71, 80 and 89.
Selecting a starting time point corresponding to each row position number in the set 1, selecting an ending time point corresponding to each row position number in the set 2, and constructing a new time period with preset clustering numbers according to the starting time point corresponding to each row position number and the ending time point corresponding to each row position number, namely selecting the 8 th elements of the set 1 and the set 2 to form a new time period, such as: { starting time point 8, ending time point 8}, similarly selecting 17 th, 26 th, 35 th, 44 th, 53 th, 62 th, 71 th, 80 th, 89 th elements in the set 1 and the set 2 respectively, thereby forming 10 new time periods respectively, and taking the 10 new time periods as initial center points of the clusters, namely initializing 10 cluster centers.
(2) Performing first clustering
When clustering is carried out, the coincidence rate from the peak time period of the electric load of each power consumer to the time period of the 10 initialized clustering centers is calculated (the coincidence rate is the maximum of the coincidence time period/the peak time period of the electric load of the power consumer and the time period of the clustering centers), the clustering center with the maximum coincidence rate is selected as the clustering center, and thus the corresponding clustering center is found by the peak time periods of the electric loads of all the power consumers.
(3) Updating cluster centers
And respectively averaging the initial time and the end time of all time periods of the same class, taking the time point average value and the end time point average value as the initial time point and the end time point of the new clustering center of the class, namely updating the clustering centers, and performing the same operation on all other clustering classes to obtain 10 updated clustering centers.
(4) Clustering continues until convergence
And recalculating the coincidence rate of the peak time periods of the power loads of all the power consumers and the updated clustering centers by using the generated 10 updated clustering centers, selecting the clustering center with the highest coincidence rate, and continuously repeating the steps until the 10 clustering centers are not changed any more.
Referring to fig. 3, a power consumer clustering device according to an embodiment includes:
the load peak time period obtaining module 310 is configured to obtain a peak time period of the electrical load of each power consumer, where the peak time period includes a start time point and an end time point.
The sorting module 320 is configured to sort the start time points in the peak time period of each power consumer according to a preset sorting rule to obtain a start time point set, and sort the end time points in the peak time period of each power consumer to obtain an end time point set.
The selecting module 330 is configured to obtain the ranking numbers of the preset cluster numbers according to a preset selection rule, select a starting time point corresponding to each ranking number from the starting time point set, and select an ending time point corresponding to each ranking number from the ending time point set.
The constructing module 340 is configured to construct a new time period with a preset number of clusters according to the starting time point corresponding to each row number and the ending time point corresponding to each row number.
And the clustering module 350 is configured to initialize the new time periods with the preset clustering numbers to the clustering centers with the preset clustering numbers, and cluster the power consumers according to the peak time period of the power load of each power consumer, the clustering centers, and the K-means clustering algorithm to obtain the clustering results of each power consumer.
The electric power user clustering device arranges the starting time points of the peak time periods of all the electric power users according to the time sequence, initializes the starting time points of the clustering centers of the preset number according to the starting time points of the preset clustering number selected from the starting time point set obtained after sequencing according to the preset selection rule, also initializes the ending time points of the clustering centers of the preset number according to the starting time points of the preset clustering number selected from the ending time point set obtained after sequencing according to the preset selection rule, namely initializes the clustering centers of the preset clustering number, and then clusters all the electric power users according to the peak time periods of the electric loads of all the electric power users, the clustering centers and the K-means clustering algorithm to obtain the clustering results of all the electric power users. When the clustering centers with the preset clustering numbers are initialized by the power consumer clustering device, the time period with the preset clustering numbers is not selected from the peak time period of the power consumer's power load at random to serve as the initialized clustering centers, but the new time period formed by the starting time point corresponding to each ranking number and the ending time point corresponding to each ranking number is initialized to be the clustering centers with the preset clustering numbers, so that the clustering centers with the preset clustering numbers can be initialized reasonably, and the power consumers are clustered according to a K-means clustering algorithm after the clustering centers are initialized, so that more accurate clustering results can be obtained, and the clustering accuracy is improved.
Referring to fig. 4, in one embodiment, the clustering module includes:
the initializing module 451 is configured to initialize the new time periods with the preset number of clusters to the cluster centers with the preset number of clusters.
And the coincidence rate calculating module 452 is configured to calculate the coincidence rate between the peak time period of the power load of each power consumer and each cluster center.
The cluster category obtaining module 453 is configured to divide the power consumers into cluster centers corresponding to peak time periods of the power consumers and the maximum coincidence rates of the cluster centers, and obtain cluster categories of a preset number of clusters.
The mean value calculating module 454 is configured to calculate a starting time point mean value and an ending time point mean value of the peak time period corresponding to the power consumer in each cluster category.
The updating module 455 is configured to update the cluster center of the cluster category where the power consumer is located into a time period from the starting time point mean value to the ending time point mean value corresponding to the cluster category.
The determining module 456 is configured to determine whether there is a cluster center in the updated cluster centers that is different from the corresponding cluster center before updating.
The determining module 457 is configured to determine that the clustering result of each power user is a clustering category of a preset clustering number when the updated clustering center is the same as the corresponding pre-updated clustering center.
In one embodiment, the power consumer clustering device further includes:
and a returning module 458, configured to, when there is a cluster center in the updated cluster centers that is different from the corresponding cluster center before updating, calculate, by the returning coincidence rate calculation module, a coincidence rate between the peak time period of the power load of each power consumer and each cluster center, respectively.
In one embodiment, the coincidence rate is the coincidence period between the peak time period of the electric load of the electric power consumer and the clustering center divided by the longer time period between the peak time period of the electric load of the electric power consumer and the clustering center.
In one embodiment, the selection module is further specifically configured to obtain the ranking number of the preset cluster number according to the following formula:
Ph＝round(m/(k+1))*h-1。
wherein, PhH is more than or equal to 1 and less than or equal to k, k is the preset clustering number, m is the number of power users, and the round function is a rounded function.
The above power consumer clustering device is a device for implementing the above power consumer clustering method, and the technical features are in one-to-one correspondence, which is not described herein again.
The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
The above examples only show some embodiments of the present invention, and the description thereof is more specific and detailed, but not construed as limiting the scope of the present invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the inventive concept, which falls within the scope of the present invention. Therefore, the protection scope of the present patent shall be subject to the appended claims.

Claims (8)

1. A power consumer clustering method is characterized by comprising the following steps:
acquiring a peak time period of the power load of each power consumer, wherein the peak time period comprises a starting time point and an ending time point;
sequencing the starting time points in the peak time periods of the power consumers according to a preset sequencing rule to obtain a starting time point set, and sequencing the ending time points in the peak time periods of the power consumers to obtain an ending time point set;
obtaining ranking numbers of preset clustering numbers according to a preset selection rule, selecting a starting time point corresponding to each ranking number from the starting time point set, and selecting an ending time point corresponding to each ranking number from the ending time point set;
constructing a new time period of the preset clustering number according to the starting time point corresponding to each ranking number and the ending time point corresponding to each ranking number;
respectively initializing the new time periods of the preset clustering number to clustering centers of the preset clustering number, clustering the power users according to the peak time period of the power load of each power user, the clustering centers and a K-means clustering algorithm to obtain clustering results of each power user, and providing electric energy with higher power for one type of power users with larger power loads according to the clustering results of the power users;
the step of initializing the new time periods of the preset clustering number to the clustering centers of the preset clustering number respectively, clustering the power consumers according to the peak time period of the power loads of the power consumers, the clustering centers and the K-means clustering algorithm, and obtaining the clustering results of the power consumers comprises the following steps:
initializing the new time periods of the preset clustering number to be clustering centers of the preset clustering number respectively;
respectively calculating the coincidence rate of the peak time period of the power load of each power consumer and each cluster center;
dividing the power consumers into the clustering centers corresponding to the peak time periods of the power consumers and the maximum coincidence rate of each clustering center respectively to obtain the clustering categories of the preset clustering number;
calculating a starting time point mean value and an ending time point mean value of the peak time period corresponding to the power consumer in each clustering category;
updating the clustering center of the clustering category where the power consumer is located into a time period from the starting time point mean value to the ending time point mean value corresponding to the clustering category;
and when the updated clustering center is the same as the corresponding clustering center before updating, determining the clustering result of each power consumer as the clustering category of the preset clustering number.
2. The power consumer clustering method according to claim 1, further comprising the steps of:
and when the updated cluster center has a cluster center different from the corresponding cluster center before updating, returning to the step of respectively calculating the coincidence rate of the peak time period of the power load of each power consumer and each cluster center.
3. The power consumer clustering method according to claim 1, wherein the coincidence ratio is the coincidence period between the peak time period of the power consumer's electrical load and the clustering center divided by the longer time period between the peak time period of the power consumer's electrical load and the clustering center.
4. The power consumer clustering method according to claim 1, wherein the obtaining the ranking number of the preset clustering number according to the preset selection rule comprises:
obtaining the ranking number of the preset cluster number according to the following formula:
Ph＝round(m/(k+1))*h-1；
wherein, the PhH is more than or equal to 1 and less than or equal to k, k is the preset clustering number, m is the number of the power users, and the round function is a rounded function.
5. An electric power consumer clustering device, comprising:
the load peak time period acquisition module is used for acquiring the peak time period of the power load of each power consumer, wherein the peak time period comprises a starting time point and an ending time point;
the sorting module is used for sorting the starting time points in the peak time period of each power consumer according to a preset sorting rule to obtain a starting time point set, and sorting the ending time points in the peak time period of each power consumer to obtain an ending time point set;
the selection module is used for obtaining ranking numbers of preset clustering numbers according to a preset selection rule, selecting a starting time point corresponding to each ranking number from the starting time point set, and selecting an ending time point corresponding to each ranking number from the ending time point set;
the construction module is used for constructing a new time period of the preset clustering number according to the starting time point corresponding to each ranking number and the ending time point corresponding to each ranking number;
the clustering module is used for respectively initializing the new time periods of the preset clustering number to clustering centers of the preset clustering number, clustering each power consumer according to the peak time period of the power load of each power consumer, the clustering centers and a K-means clustering algorithm, and obtaining a clustering result of each power consumer;
the clustering module comprises:
the initialization module is used for respectively initializing the new time periods of the preset clustering number into clustering centers of the preset clustering number;
the coincidence rate calculation module is used for calculating the coincidence rate of the peak time period of the power load of each power consumer and each cluster center respectively;
the cluster category acquisition module is used for dividing the power consumers into the cluster centers corresponding to the peak time periods of the power consumers and the maximum coincidence rate of each cluster center respectively to obtain the cluster categories of the preset cluster number;
the mean value calculating module is used for calculating a starting time point mean value and an ending time point mean value of the peak time period corresponding to the power consumer in each clustering category;
the updating module is used for updating the clustering center of the clustering category where the power consumer is located into a time period from the starting time point mean value to the ending time point mean value corresponding to the clustering category;
and the determining module is used for determining the clustering result of each power consumer as the clustering category of the preset clustering number when the updated clustering center is the same as the corresponding clustering center before updating.
6. The power consumer clustering device according to claim 5, further comprising:
and the returning module is used for returning to the coincidence rate calculating module to respectively calculate the coincidence rate of the peak time period of the power load of each power consumer and each clustering center when the updated clustering center has a clustering center different from the corresponding clustering center before updating.
7. The electric power consumer clustering device according to claim 5, wherein the coincidence ratio is a coincidence period between a peak time period of the electric power consumer's electric load and the clustering center divided by a longer time period between the peak time period of the electric power consumer's electric load and the clustering center.
8. The electric power consumer clustering device according to claim 5, wherein the selection module is further specifically configured to obtain the ranking number of the preset clustering number according to the following formula:
Ph＝round(m/(k+1))*h-1；
wherein, the PhH is more than or equal to 1 and less than or equal to k, k is the preset clustering number, m is the number of the power users, and the round function is a rounded function.
CN201611081166.6A 2016-11-30 2016-11-30 Power consumer clustering method and device Active CN106548301B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201611081166.6A CN106548301B (en) 2016-11-30 2016-11-30 Power consumer clustering method and device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201611081166.6A CN106548301B (en) 2016-11-30 2016-11-30 Power consumer clustering method and device

Publications (2)

Publication Number Publication Date
CN106548301A CN106548301A (en) 2017-03-29
CN106548301B true CN106548301B (en) 2020-11-03

Family

ID=58396771

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201611081166.6A Active CN106548301B (en) 2016-11-30 2016-11-30 Power consumer clustering method and device

Country Status (1)

Country Link
CN (1) CN106548301B (en)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110826641B (en) * 2019-11-13 2023-07-07 上海积成能源科技有限公司 System and method for classifying residential electricity consumption conditions based on cluster analysis
CN110990387B (en) * 2019-11-29 2024-02-27 广东电网有限责任公司 Power failure record error processing method and device for electric energy metering equipment
CN112464059B (en) * 2020-12-08 2024-03-22 深圳供电局有限公司 Distribution network user classification method, device, computer equipment and storage medium

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106022578A (en) * 2016-05-12 2016-10-12 国网天津市电力公司 Residential electricity peak-valley-flat dividing method based on data dimension increasing and K-means clustering

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140320388A1 (en) * 2013-04-25 2014-10-30 Microsoft Corporation Streaming k-means computations
CN105630801A (en) * 2014-10-30 2016-06-01 国际商业机器公司 Method and apparatus for detecting deviated user
CN105303470A (en) * 2015-11-26 2016-02-03 国网辽宁省电力有限公司大连供电公司 Electric power project planning and construction method based on big data

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106022578A (en) * 2016-05-12 2016-10-12 国网天津市电力公司 Residential electricity peak-valley-flat dividing method based on data dimension increasing and K-means clustering

Also Published As

Publication number Publication date
CN106548301A (en) 2017-03-29

Similar Documents

Publication Publication Date Title
CN106548301B (en) Power consumer clustering method and device
CN111160617B (en) Power daily load prediction method and device
CN106600119B (en) K-means-based power consumer clustering method and device
Mikhailiuk et al. Active sampling for pairwise comparisons via approximate message passing and information gain maximization
CN103714135B (en) MapReduce recommendation method and system of second-degree interpersonal relationships of massive users
KR101852527B1 (en) Method for Dynamic Simulation Parameter Calibration by Machine Learning
Seridi et al. Multi-objective evolutionary algorithm for biclustering in microarrays data
CN114818915A (en) Data clustering method and device, storage medium and electronic device
CN110796159A (en) Power data classification method and system based on k-means algorithm
JP2019053491A (en) Neural network evaluation apparatus, neural network evaluation method, and program
CN108288111B (en) Thermal power plant exhaust smoke temperature reference value determining method and device based on association rules
CN109801073B (en) Risk user identification method, apparatus, computer device and storage medium
KR20160058891A (en) Method and system for rating measured values taken from a system
Hernaez et al. A cluster-based approach to compression of quality scores
JP7440395B2 (en) Optimal solution search device and optimal solution search program
CN111949530B (en) Test result prediction method and device, computer equipment and storage medium
Gu et al. An evolutionary algorithm based on decomposition for multimodal optimization problems
CN112561138A (en) Power load prediction method, power load prediction device, computer equipment and storage medium
Purshouse et al. Elitism, Sharing and Ranking Choices in Evolutionary Multi-Criterion Optimisation
Denysiuk et al. A new hybrid evolutionary multiobjective algorithm guided by descent directions
RU80604U1 (en) AUTOMATED RESOURCE DISTRIBUTION SYSTEM FOR OPTIMUM SOLUTION OF TARGET TASKS
CN109949070B (en) User viscosity evaluation method, device, computer equipment and storage medium
Tran et al. A sequential decision approach to ordinal preferences in recommender systems
da Silva et al. Clustering distributed short time series with dense patterns
CN109325976A (en) It is divided and simplified figure diameter algorithm and system based on K scattergram of dynamic

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant
TR01 Transfer of patent right
TR01 Transfer of patent right

Effective date of registration: 20210603

Address after: 510700 3rd, 4th and 5th floors of building J1 and 3rd floor of building J3, No.11 Kexiang Road, Science City, Luogang District, Guangzhou City, Guangdong Province

Patentee after: China South Power Grid International Co.,Ltd.

Address before: 510080 water Donggang 8, Dongfeng East Road, Yuexiu District, Guangzhou, Guangdong.

Patentee before: China South Power Grid International Co.,Ltd.

Patentee before: POWER GRID TECHNOLOGY RESEARCH CENTER. CHINA SOUTHERN POWER GRID