CN116187808A - Electric power package recommendation method based on virtual power plant user-package label portrait - Google Patents

Electric power package recommendation method based on virtual power plant user-package label portrait Download PDF

Info

Publication number
CN116187808A
CN116187808A CN202211595823.4A CN202211595823A CN116187808A CN 116187808 A CN116187808 A CN 116187808A CN 202211595823 A CN202211595823 A CN 202211595823A CN 116187808 A CN116187808 A CN 116187808A
Authority
CN
China
Prior art keywords
user
package
users
label
tag
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.)
Pending
Application number
CN202211595823.4A
Other languages
Chinese (zh)
Inventor
李忆
乐鹰
周保中
吕若佳
张继广
毕圣
赵琦
吴思翰
郭超
孙诗洁
谢康
许皓文
孙宇航
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.)
Shaoxing Research Institute Of Zhejiang University
Huadian Electric Power Research Institute Co Ltd
Zhejiang University City College ZUCC
Original Assignee
Shaoxing Research Institute Of Zhejiang University
Huadian Electric Power Research Institute Co Ltd
Zhejiang University City College ZUCC
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 Shaoxing Research Institute Of Zhejiang University, Huadian Electric Power Research Institute Co Ltd, Zhejiang University City College ZUCC filed Critical Shaoxing Research Institute Of Zhejiang University
Priority to CN202211595823.4A priority Critical patent/CN116187808A/en
Publication of CN116187808A publication Critical patent/CN116187808A/en
Pending legal-status Critical Current

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
    • G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
    • G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
    • 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
    • G06Q30/00—Commerce
    • G06Q30/02—Marketing; Price estimation or determination; Fundraising
    • G06Q30/0201—Market modelling; Market analysis; Collecting market data
    • 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
    • G06Q30/00—Commerce
    • G06Q30/02—Marketing; Price estimation or determination; Fundraising
    • G06Q30/0201—Market modelling; Market analysis; Collecting market data
    • G06Q30/0202—Market predictions or forecasting for commercial activities
    • 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
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J13/00—Circuit arrangements for providing remote indication of network conditions, e.g. an instantaneous record of the open or closed condition of each circuitbreaker in the network; Circuit arrangements for providing remote control of switching means in a power distribution network, e.g. switching in and out of current consumers by using a pulse code signal carried by the network
    • H02J13/00001—Circuit arrangements for providing remote indication of network conditions, e.g. an instantaneous record of the open or closed condition of each circuitbreaker in the network; Circuit arrangements for providing remote control of switching means in a power distribution network, e.g. switching in and out of current consumers by using a pulse code signal carried by the network characterised by the display of information or by user interaction, e.g. supervisory control and data acquisition systems [SCADA] or graphical user interfaces [GUI]
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J3/00—Circuit arrangements for ac mains or ac distribution networks
    • H02J3/008—Circuit arrangements for ac mains or ac distribution networks involving trading of energy or energy transmission rights
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
    • H02J2203/10—Power transmission or distribution systems management focussing at grid-level, e.g. load flow analysis, node profile computation, meshed network optimisation, active network management or spinning reserve management
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
    • H02J2203/20—Simulating, e g planning, reliability check, modelling or computer assisted design [CAD]
    • 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

Landscapes

  • Business, Economics & Management (AREA)
  • Engineering & Computer Science (AREA)
  • Strategic Management (AREA)
  • Development Economics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Economics (AREA)
  • Finance (AREA)
  • Human Resources & Organizations (AREA)
  • Accounting & Taxation (AREA)
  • General Business, Economics & Management (AREA)
  • Marketing (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Physics & Mathematics (AREA)
  • Game Theory and Decision Science (AREA)
  • Health & Medical Sciences (AREA)
  • Power Engineering (AREA)
  • Tourism & Hospitality (AREA)
  • Data Mining & Analysis (AREA)
  • Educational Administration (AREA)
  • Public Health (AREA)
  • Operations Research (AREA)
  • Water Supply & Treatment (AREA)
  • General Health & Medical Sciences (AREA)
  • Primary Health Care (AREA)
  • Quality & Reliability (AREA)
  • Human Computer Interaction (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses an electric power package recommendation method based on virtual power plant user-package label portrait, which comprises the following steps: selecting common tag attributes in the electric power package to establish a characteristic tag set; acquiring implicit scores of the users on package labels through historical package purchase records, and constructing a user-package label portrait model considering time weights; according to the load data of the training user set, comprehensively considering the electricity utilization level and the electricity utilization habit to calculate the similarity between training users; clustering training users according to a spectral clustering algorithm; determining a weight coefficient of each label according to a cluster contour coefficient of a training user clustering result, and constructing a user similarity evaluation model based on the differentiated label weights; and predicting potential scores of the target user on the packages according to the economic scores of the neighbor users on the packages, and recommending the most economical electric packages for the user. The invention can discover the load characteristics of the user under the condition that only the historical package purchase record of the user is obtained, and realizes the accurate recommendation of the electric package.

Description

Electric power package recommendation method based on virtual power plant user-package label portrait
Technical Field
The invention relates to the technical field of intelligent electricity selling, in particular to an electric power package recommendation method based on virtual power plant user-package label portrait.
Background
With further deepening of the innovation of the electricity selling side, the electric energy gradually recovers the commodity attribute in the electric power market, and the electric power package is an electric power product designed for different types of users. However, in the face of the huge and diverse number of retail packages that are promoted in the power market, the complex package charging mechanisms and the high cost of information collection necessarily present a significant degree of difficulty to users in finding economical and demand-satisfying packages.
The virtual power plant comprises different types of users on the load side, reasonable electric package recommendation is provided for the virtual power plant users, on one hand, the electricity consumption cost can be reduced for the virtual power plant users, the time cost in package selection can be saved, and on the other hand, the viscosity of the users can be increased for the virtual power plant, and new users can be attracted.
However, many of the existing researches are performed on the premise of acquiring the data of the smart meter of the power consumer, and the required consumer characteristics are obtained by directly using the total electricity record of the power consumer or extracting key energy information from the total electricity record. But in the absence of load data, these methods cannot be further implemented in the face of users newly joining a virtual power plant. In this regard, the invention provides a power retail package recommendation method based on virtual power plant user-package label portrait, and the user can obtain personalized package suggestions by providing historical package purchase records.
Disclosure of Invention
In order to solve the problems and the demands in the background technology, the invention provides an electric package recommendation method based on virtual power plant user-package label portrait. By the method, the load characteristics of the user can be discovered according to the historical electric package purchase information of the user under the condition of lacking of the load data of the user, and economic package purchase suggestions are provided for the user.
The invention solves the problems by adopting the following technical scheme: a power package recommendation method based on virtual power plant user-package label portrait is characterized by comprising the following steps:
step 1: selecting common tag attributes in the electric package to establish a characteristic tag set, and preprocessing package tag data by adopting a normalization scoring mode;
step 2: acquiring implicit scores of the users on package labels through historical package purchase records, and constructing a user-package label portrait model considering time weights;
step 3: according to load data of a sample user, comprehensively considering the power consumption level and the power consumption habit, and calculating the similarity of a training user load curve;
step 4: clustering training users according to a spectral clustering algorithm;
step 5: determining a weight coefficient of each label according to a cluster contour coefficient of a sample user clustering result, and constructing a user similarity evaluation model based on the differentiated label weights;
step 6: and predicting potential scores of the target user on the packages according to the economic scores of the neighbor users on the packages, and recommending the most economical electric packages for the user.
Compared with the prior art, the invention has the following advantages and effects: according to the invention, under the condition of lacking of user load data, the load characteristics of the user can be discovered according to the historical electric package purchase information of the user. When the user-package label portrait is constructed, experience accumulation and preference transfer of purchasing power packages by the user in a long time scale are considered, and time weight is introduced so as to accurately reflect the actual demands and preferences of the user; and matching the neighbor users with similar behavior with the target user by adopting the differentiated label weights. Thus, the most economical power packages can be recommended to the user based on the concept of collaborative filtering.
Drawings
Fig. 1 is a diagram of a power package recommendation implementation framework of the present invention.
Detailed Description
The present invention will be described in further detail by way of examples with reference to the accompanying drawings, which are illustrative of the present invention and not limited to the following examples.
Examples.
Referring to fig. 1, in this embodiment, a power package recommendation method based on a virtual power plant user-package label portrait includes the following steps:
step 1: referring to a typical foreign electric package, an electric package feature tag set is constructed from the dimensions of a pricing mechanism, electric energy cost, clean energy proportion, star grade score of an electric company, package exit cost and the like. And then, preprocessing package label data by adopting normalized scores according to different data types of package characteristic labels.
The star rating and clean energy ratio of electricity companies belong to the numeric-benefit type label, and the normalization can be expressed as:
Figure BDA0003997258440000021
in the middle of
Figure BDA0003997258440000022
Normalized score, x, for benefit tag j representing package i i,j The actual value of tag j representing package i.
The electric energy cost and the package exit cost belong to a numerical value-cost type label, and the normalization can be expressed as follows:
Figure BDA0003997258440000031
in the method, in the process of the invention,
Figure BDA0003997258440000032
representing packagesNormalized score for cost type label j for i.
The pricing mode label and the time-of-use electricity price label are nominal labels, the label normalization scores of the flat rate and the unified electricity price attribute are set to 0, and the label normalization scores of the variable rate and the time-of-use electricity price attribute are set to 1.
Step 2: and obtaining the electric power package set purchased by the user according to the historical electric power package purchase record of the user. Considering that recently purchased packages more accurately reflect the needs and preferences of users than earlier purchased packages, the impact of time on the user's actual needs is corrected by introducing a time decay function when generating user tag portraits using historical package tag information. The user-package tab representation taking into account the time weights can be expressed as:
Figure BDA0003997258440000033
γ(t)＝e -λ(T-t)
wherein L is u,j Representing the portrait score of user u with respect to tag j. T represents the historical package purchase number for user u. T represents the time sequence of the user purchasing the package, and the closer T is to T, the closer the user purchasing the package is to the current time.
Figure BDA0003997258440000034
Package i representing user u t purchase u(t) Scoring of label j. λ is a time decay coefficient, λ is greater than or equal to 0, and the greater the value of λ, the greater the weight of the user's recent purchase behavior when generating the user representation.
Step 3: the similarity between the morphology and the load level of the two load curves is evaluated by respectively adopting the pearson correlation coefficient and the Euclidean distance, and the comprehensive similarity distance of the load curves is constructed, wherein the expression is as follows:
Figure BDA0003997258440000035
Figure BDA0003997258440000036
Figure BDA0003997258440000037
Figure BDA0003997258440000038
wherein Z is u,v Representing the combined similarity distance of the user u and v load curves.
Figure BDA0003997258440000039
And->
Figure BDA00039972584400000310
The pearson distance and euclidean distance between the user u and user v load curves, respectively. Epsilon P And epsilon D The scaling coefficients of the pearson distance and the euclidean distance in the integrated similarity distance are represented, respectively. Q (Q) u,τ And Q v,τ Representing the power consumption load of user u and user v at time τ, +.>
Figure BDA00039972584400000311
And->
Figure BDA00039972584400000312
The average daily power consumption of user u and user v is represented. Ω denotes the number of load periods taken into account.
Step 4: and classifying the training set users by adopting a spectral clustering algorithm.
Taking the load curve double-scale comprehensive similar distance derived in the step 3 as the distance between user sample points, and calculating the edge weight between the user sample points by adopting a Gaussian kernel function, wherein the expression is as follows:
Figure BDA0003997258440000041
wherein w is u,v Representing the edge weights between the two points of users u and v. σ is the kernel parameter of the gaussian kernel function. By edge weights w between any two points u,v And forming an adjacency matrix W, and generating an undirected weighted graph according to the adjacency matrix W.
In order to realize optimal cutting of the graph, an optimal clustering result is obtained, and the graph cutting of the undirected weighted graph is carried out by adopting a normalized cutting method. Representing the set of all points in the undirected weighted graph by G 1 ,G 2 ,···,G M Representing the set of points contained in the M subgraphs obtained after graph cutting, the objective function of the N-Cut graph cutting method can be expressed as follows:
Figure BDA0003997258440000042
Figure BDA0003997258440000043
wherein G is m Representing the set of points contained in the mth subgraph,
Figure BDA0003997258440000044
represents G m Is a complement of (a). vol (G) m ) For set G m Edge weight sum of the points contained in (a). By minimizing the objective function N-Cut (G 1 ,G 2 ,···,G M ) And (3) enabling the edge weight sum of different subgraphs after graph cutting to be minimum and enabling the edge weight sum in the subgraphs to be maximum, and obtaining a final clustering result.
And clustering training users into M clusters by using a spectral clustering algorithm based on an N-Cut graph cutting method.
Step 5: and evaluating the association degree of each label and the user behavior by adopting a contour coefficient index according to the distribution condition of the user portrait labels in the user clustering result based on the load data. The specific definition of the contour coefficients is as follows:
Figure BDA0003997258440000045
Figure BDA0003997258440000046
Figure BDA0003997258440000047
in the method, in the process of the invention,
Figure BDA0003997258440000048
a contour coefficient representing tag j, and +.>
Figure BDA0003997258440000049
m u Indicating the cluster number in which user u is located. I U tr Representing training user set U tr The number of users, |U tr (m u ) I represents the cluster U where user U is located tr (m u ) The number of users in the system. />
Figure BDA0003997258440000051
Representing the average distance between the score of user u for tag j and the scores of other users in all the clusters to which it belongs,/-, for tag j>
Figure BDA0003997258440000052
Representing the average distance that user u scores label j from the user's scores for label j in all other clusters.
For package label j, the smaller the difference between the portrait score of user u and the portrait score of user in the same cluster, the larger the difference between the portrait scores of users in other clusters, the larger the cluster profile coefficient of label j, which indicates that the higher the correlation between label j and user load characteristics. Thus, the greater the profile factor of a tag, the greater its weight should be in measuring user similarity. The weight of tag j can be expressed as:
Figure BDA0003997258440000053
users with similar usability habits have similar preferences and needs for power packages, so neighbors with similar behavior to the target user can be found from the user-package tag portrayal. And evaluating the similarity degree between users by adopting a weighted Euclidean distance, wherein the expression is as follows:
Figure BDA0003997258440000054
in the method, in the process of the invention,
Figure BDA0003997258440000055
representing weighted euclidean similarity between users u and v. Omega j Is the weight coefficient of tag j, and
Figure BDA0003997258440000056
the label weight coefficient quantifies the correlation between different labels and the habit of the user, influences the evaluation result of the similarity among the users to different degrees, and further influences the package recommendation result.
Step 6: knowing the load data of the training user, the corresponding package cost can be calculated, and the economic score of the package can be calculated according to the ranking of the package cost in all packages, which can be expressed as:
Figure BDA0003997258440000057
wherein r is u,i Representing the score of user u for package i and r u,i ∈(0,1]. I is a set of all alternative retail packages, |i| represents the number of packages in set I. h is a u,i Ranking the cost of package i to user u in all packages, h u,i ∈[1,|I|]。
Due to the lack of load data for the test user, the potential scoring of packages by the test user needs to be predicted by means of their economic scores of their neighbors in the training user. According to the step 4, the similarity between the test user and the training set user can be obtained, and k training users with highest similarity are used as nearest neighbors of the test user to form a setFit, user predictive score r for package u,i Can be expressed as:
Figure BDA0003997258440000061
after the economic scores of the users on all packages are obtained, the N packages with the highest scores are recommended to the users.
The power package recommendation results are shown in table 1.
TABLE 1 user Package recommendation Table
Figure BDA0003997258440000062
Note that: in order to make the recommended result more visual, the package is named as "package type-corresponding month power consumption grading". For example, F-1000 represents a flat rate-flat rate package suitable for users with a monthly power usage of 1000 kWh; FT-750 represents a flat rate-time-of-use electricity price package suitable for users with a monthly electricity usage of 750 kWh.
What is not described in detail in this specification is all that is known to those skilled in the art.
Although the present invention has been described with reference to the above embodiments, it should be understood that the invention is not limited to the embodiments described above, but is capable of modification and variation without departing from the spirit and scope of the present invention.

Claims (2)

1. A power package recommendation method based on virtual power plant user-package label portrait is characterized by comprising the following steps:
step 1: selecting common tag attributes in the electric package to establish a characteristic tag set, and preprocessing package tag data by adopting a normalization scoring mode;
step 2: acquiring implicit scores of the users on package labels through historical package purchase records, and constructing a user-package label portrait model considering time weights;
step 3: according to load data of a sample user, comprehensively considering the power consumption level and the power consumption habit, and calculating the similarity of a training user load curve;
step 4: clustering training users according to a spectral clustering algorithm;
step 5: determining a weight coefficient of each label according to a cluster contour coefficient of a sample user clustering result, and constructing a user similarity evaluation model based on the differentiated label weights;
step 6: and predicting potential scores of the target user on the packages according to the economic scores of the neighbor users on the packages, and recommending the most economical electric packages for the user.
2. The virtual power plant user-package label portrait-based electric package recommendation method according to claim 1, characterized in that:
step 1: constructing an electric package feature tag set from a pricing mechanism, electric energy cost, clean energy proportion, star grade grading of an electric company and package exit cost dimension, and preprocessing package tag data by adopting normalized grading according to different data types of package feature tags;
the star rating and clean energy ratio of electricity-selling companies belong to the numerical-benefit type label, and the normalization is expressed as:
Figure FDA0003997258430000011
in the middle of
Figure FDA0003997258430000012
Normalized score, x, for benefit tag j representing package i i,j An actual value representing tag j of package i;
the electric energy cost and the package exit cost belong to a numerical value-cost type label, and the normalization is expressed as follows:
Figure FDA0003997258430000013
in the method, in the process of the invention,
Figure FDA0003997258430000014
normalized score for cost tag j representing package i;
the pricing mode label and the time-of-use electricity price label are nominal labels, the method sets the label normalization scores of the flat rate and the unified electricity price attribute to 0, and the label normalization scores of the variable rate and the time-of-use electricity price attribute to 1;
step 2: obtaining an electric power package set purchased by a user according to the historical electric power package purchase record of the user; considering that recently purchased packages can more accurately reflect the needs and preferences of users than packages purchased early, when generating user tag portraits by utilizing historical package tag information, the influence of time on the real needs of users is corrected by introducing a time decay function; the user-package tab representation taking into account the time weights is expressed as:
Figure FDA0003997258430000021
γ(t)＝e -λ(T-t)
wherein L is u,j Representing a representation score for user u with respect to tag j; t represents the historical package purchase times of user u; t represents the time sequence of the user purchasing the package, and the closer T is to T, the closer the time the user purchases the package is to the current time;
Figure FDA0003997258430000022
package i representing user u t purchase u(t) Scoring of tag j; lambda is a time attenuation coefficient, lambda is more than or equal to 0, and the larger the lambda value is, the larger the weight of the user in recent purchasing behavior is when the user portrait is generated;
step 3: the similarity between the morphology and the load level of the two load curves is evaluated by respectively adopting the pearson correlation coefficient and the Euclidean distance, and the comprehensive similarity distance of the load curves is constructed, wherein the expression is as follows:
Figure FDA0003997258430000023
Figure FDA0003997258430000024
Figure FDA0003997258430000025
Figure FDA0003997258430000026
wherein Z is u,v Representing the comprehensive similarity distance of the load curves of the users u and v;
Figure FDA0003997258430000027
and->
Figure FDA0003997258430000028
Representing the pearson distance and euclidean distance between the user u and user v load curves, respectively; epsilon P And epsilon D The scaling factors of the pearson distance and the euclidean distance in the comprehensive similar distance are respectively represented; q (Q) u,τ And Q v,τ Representing the power consumption of user u and user v at time τ, respectively, < >>
Figure FDA0003997258430000029
And->
Figure FDA00039972584300000210
Daily average electricity consumption of the user u and the user v are respectively represented; omega represents the number of load periods counted;
step 4: classifying the training set users by adopting a spectral clustering algorithm;
taking the load curve double-scale comprehensive similar distance derived in the step 3 as the distance between user sample points, and calculating the edge weight between the user sample points by adopting a Gaussian kernel function, wherein the expression is as follows:
Figure FDA00039972584300000211
wherein w is u,v Representing the edge weight between the two points of the user u and v; sigma is a kernel parameter of a gaussian kernel function; by edge weights w between any two points u,v Forming an adjacent matrix W, and generating an undirected weighted graph according to the adjacent matrix W;
in order to realize optimal cutting of the graph, obtaining an optimal clustering result, and cutting the graph with the undirected weighted graph by adopting a normalized cutting method; representing the set of all points in the undirected weighted graph by G 1 ,G 2 ,···,G M Representing a set of points contained in M subgraphs obtained after graph cutting, and representing an objective function of an N-Cut graph cutting method as follows:
Figure FDA0003997258430000031
Figure FDA0003997258430000032
wherein G is m Representing the set of points contained in the mth subgraph,
Figure FDA0003997258430000033
represents G m Is a complement of (a); vol (G) m ) For set G m Edge weight sum of the points contained in the (B); by minimizing the objective function N-Cut (G 1 ,G 2 ,···,G M ) The edge weights in the subgraphs are maximized while the edge weights among the different subgraphs are minimized after the graph is cut, and a final clustering result is obtained;
clustering training users into M clusters by a spectral clustering algorithm based on an N-Cut graph cutting method;
step 5: evaluating the association degree of each label and the user behavior by adopting a contour coefficient index according to the distribution condition of the user portrait labels in the user clustering result based on the load data; the specific definition of the contour coefficients is as follows:
Figure FDA0003997258430000034
Figure FDA0003997258430000035
Figure FDA0003997258430000036
in the method, in the process of the invention,
Figure FDA0003997258430000037
a contour coefficient representing tag j, and +.>
Figure FDA0003997258430000038
m u The cluster number of the user u is represented; u (U) tr Representing training user set U tr The number of users in (U) tr (m u ) Indicating the cluster U where user U is located tr (m u ) The number of users in (a); />
Figure FDA0003997258430000039
Representing the average distance between the score of user u for tag j and the scores of other users in all the clusters to which it belongs,/-, for tag j>
Figure FDA00039972584300000310
Representing the average distance between the score of user u for tag j and the scores of users in all other clusters for tag j;
for package label j, the smaller the difference between the portrait score of user u and the portrait score of user in the same cluster is, the larger the difference between the portrait score of user in other clusters is, the larger the cluster profile coefficient of label j is, which shows that the higher the correlation between label j and user load feature is; therefore, the larger the profile factor of the tag, the greater the weight it should be in measuring user similarity; the weight of tag j is expressed as:
Figure FDA0003997258430000041
users with similar usability habits have similar preferences and requirements for power packages, so that neighbors with similar usability behaviors to the target user are found according to the user-package label portraits; and evaluating the similarity degree between users by adopting a weighted Euclidean distance, wherein the expression is as follows:
Figure FDA0003997258430000042
in the method, in the process of the invention,
Figure FDA0003997258430000043
representing weighted European similarity between users u and v; omega j Is the weight coefficient of tag j, and
Figure FDA0003997258430000044
the label weight coefficient quantifies the correlation between different labels and the habit of the user, influences the evaluation result of the similarity among the users to different degrees and further influences the package recommendation result;
step 6: the load data of the known training user can calculate corresponding package cost, and the economic score of the package is calculated according to the ranking of the package cost in all packages, which is expressed as:
Figure FDA0003997258430000045
wherein r is u,i Representing user u's score for package iAnd r is u,i ∈(0,1]The method comprises the steps of carrying out a first treatment on the surface of the I is all the alternative power retail package sets, and I represents the number of packages in the set I; h is a u,i Ranking the cost of package i to user u in all packages, h u,i ∈[1,I]；
Due to the lack of load data for the test user, the potential scoring of packages by the test user needs to be predicted by means of the economic scores of their neighbors in the training user; according to the step 4, the similarity between the test user and the training set user can be obtained, k training users with highest similarity are used as nearest neighbors of the test user to form a set, and then the prediction score r of the user to the package is calculated u,i Expressed as:
Figure FDA0003997258430000046
after the economic scores of the users on all packages are obtained, the N packages with the highest scores are recommended to the users.
CN202211595823.4A 2022-12-13 2022-12-13 Electric power package recommendation method based on virtual power plant user-package label portrait Pending CN116187808A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202211595823.4A CN116187808A (en) 2022-12-13 2022-12-13 Electric power package recommendation method based on virtual power plant user-package label portrait

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202211595823.4A CN116187808A (en) 2022-12-13 2022-12-13 Electric power package recommendation method based on virtual power plant user-package label portrait

Publications (1)

Publication Number Publication Date
CN116187808A true CN116187808A (en) 2023-05-30

Family

ID=86433459

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202211595823.4A Pending CN116187808A (en) 2022-12-13 2022-12-13 Electric power package recommendation method based on virtual power plant user-package label portrait

Country Status (1)

Country Link
CN (1) CN116187808A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116596640A (en) * 2023-07-19 2023-08-15 国网山东省电力公司营销服务中心(计量中心) Recommendation method, system, equipment and storage medium for electric power retail electric charge package

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116596640A (en) * 2023-07-19 2023-08-15 国网山东省电力公司营销服务中心(计量中心) Recommendation method, system, equipment and storage medium for electric power retail electric charge package

Similar Documents

Publication Publication Date Title
Han et al. Segmentation of telecom customers based on customer value by decision tree model
CN110490729B (en) Financial user classification method based on user portrait model
US7577579B2 (en) Method of predicting sales based on triple-axis mapping of customer value
Ngai et al. Application of data mining techniques in customer relationship management: A literature review and classification
Albadvi et al. A hybrid recommendation technique based on product category attributes
CN109063945B (en) Value evaluation system-based 360-degree customer portrait construction method for electricity selling company
CN110222267A (en) A kind of gaming platform information-pushing method, system, storage medium and equipment
CN108280541A (en) Customer service strategies formulating method, device based on random forest and decision tree
CN108388955A (en) Customer service strategies formulating method, device based on random forest and logistic regression
CN108388974A (en) Top-tier customer Optimum Identification Method and device based on random forest and decision tree
CN108389069A (en) Top-tier customer recognition methods based on random forest and logistic regression and device
CN110826886A (en) Electric power customer portrait construction method based on clustering algorithm and principal component analysis
CN108364191A (en) Top-tier customer Optimum Identification Method and device based on random forest and logistic regression
CN108921602A (en) A kind of user&#39;s buying behavior prediction technique based on integrated neural network
CN114782076B (en) Online mall consumption platform lottery integral exchange intelligent management method, system and computer storage medium
CN107886241A (en) Resource analysis method, apparatus, medium and electronic equipment
CN116187808A (en) Electric power package recommendation method based on virtual power plant user-package label portrait
Hasheminejad et al. Clustering of bank customers based on lifetime value using data mining methods
Khajvand et al. Analyzing customer segmentation based on customer value components (case study: a private bank)
Lewaaelhamd Customer segmentation using machine learning model: an application of RFM analysis
Chuang et al. A study on the applications of data mining techniques to enhance customer lifetime value—based on the department store industry
Chou et al. The RFM Model Analysis for VIP Customer: A case study of golf clothing brand
CN116739652A (en) Clothing e-commerce sales prediction modeling method
WO2002027621A1 (en) Genetic algorithm method for aggregating electricity consumption and optimizing electric buying groups
CN115829683A (en) Power integration commodity recommendation method and system based on inverse reward learning optimization

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