CN116596574A - Power grid user portrait construction method and system - Google Patents

Power grid user portrait construction method and system Download PDF

Info

Publication number
CN116596574A
CN116596574A CN202310677681.4A CN202310677681A CN116596574A CN 116596574 A CN116596574 A CN 116596574A CN 202310677681 A CN202310677681 A CN 202310677681A CN 116596574 A CN116596574 A CN 116596574A
Authority
CN
China
Prior art keywords
user
phase space
power grid
data
space reconstruction
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
CN202310677681.4A
Other languages
Chinese (zh)
Inventor
骆晨
程道卫
冯玉
吴凯
吴少雷
汪柏松
郭小东
周建军
陈振宁
胡钰杰
左宇翔
李博
邵珺伟
卞真旭
张晨晨
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.)
State Grid Anhui Electric Power Co Ltd Anqing Power Supply Co
Electric Power Research Institute of State Grid Anhui Electric Power Co Ltd
Huanshang Power Supply Co of State Grid Anhui Electric Power Co Ltd
Hefei Power Supply Co of State Grid Anhui Electric Power Co Ltd
Original Assignee
State Grid Anhui Electric Power Co Ltd Anqing Power Supply Co
Electric Power Research Institute of State Grid Anhui Electric Power Co Ltd
Huanshang Power Supply Co of State Grid Anhui Electric Power Co Ltd
Hefei Power Supply Co of State Grid Anhui Electric Power 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 State Grid Anhui Electric Power Co Ltd Anqing Power Supply Co, Electric Power Research Institute of State Grid Anhui Electric Power Co Ltd, Huanshang Power Supply Co of State Grid Anhui Electric Power Co Ltd, Hefei Power Supply Co of State Grid Anhui Electric Power Co Ltd filed Critical State Grid Anhui Electric Power Co Ltd Anqing Power Supply Co
Priority to CN202310677681.4A priority Critical patent/CN116596574A/en
Publication of CN116596574A publication Critical patent/CN116596574A/en
Pending legal-status Critical Current

Links

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
    • 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
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/27—Regression, e.g. linear or logistic regression
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/048—Activation functions
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • 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
    • G06Q10/06395—Quality analysis or management
    • 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—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Electricity, gas or water supply
    • 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

Abstract

The invention discloses a method and a system for constructing a user portrait of a power grid, comprising the steps of obtaining historical moment data of different data sources of the power grid; combining and mapping the historical moment data of each data source into a shared representation to integrate the correlation among modes, and extracting the time sequence characteristics of the initial power grid user; reversely constructing the time sequence characteristics of the initial power grid user into a phase space structure of an original system by adopting phase space reconstruction to obtain a phase space reconstruction matrix; extracting an abstract user feature set from the phase space reconstruction matrix based on mutual information correlation of joint entropy; and carrying out regression prediction on the user portrait labels based on the abstract user feature set, clustering the user portrait labels to generate user clusters, and constructing user cluster portraits. Accurate electricity user image is realized, and the power grid is assisted to improve the service quality of the operation of the power grid.

Description

Power grid user portrait construction method and system
Technical Field
The invention relates to the technical field of big data mining of power distribution networks, in particular to a method and a system for constructing a user portrait of a power grid.
Background
Along with the rapid development of information technologies such as various intelligent cloud computing, internet service, internet of things and the like, the informatization construction level is increasingly improved, and big data is a revolution of modern information technology in the field. The big data technology mainly excavates and acquires valuable potential information from different types of mass data, and under the trend of innovation development of the big data technology, more and more fields gradually introduce big data application, so that the high-efficiency sustainable development of enterprises is promoted to a certain extent.
In the power field, to accommodate the ever-increasing power demand and long-term development of the power grid, the development of smart power grids has become an irreversible trend. Along with the comprehensive construction of digitalization, informatization and intellectualization of a power grid, various data generated in the operation of the power grid are exponentially increased, and how to provide corresponding judgment basis for the maintenance, operation, planning and construction of the power grid by utilizing a big data technology becomes a primary problem.
With gradual transformation of the power grid to the innovative service type enterprise, the electricity consumption experience of the user and corresponding user behavior analysis are paid attention to gradually, and meanwhile, the promotion of automatic construction of the power distribution network and the application of the large data technology of the power distribution network are possible, so that user portraits are possible. Based on a large amount of user electricity consumption data, the user behavior is analyzed, and the characteristics of different user groups can be identified, so that the purposes of user cognition, risk management and service are achieved. Compared with the traditional user behavior analysis, the user behavior analysis based on data mining can improve the accuracy of the user behavior analysis and realize quantitative description of the user electricity consumption behavior; the user is accurately portrayed through the big data technology, the analysis is more focused on the prediction of the electricity consumption risk of the user and the mining of the interests of the big user, the establishment of a targeted personalized service scheme is facilitated, and the user experience is improved.
In recent years, with the rapid development of the Internet, new energy and a distributed power grid are gradually advanced, and the number of access users of related load side equipment is greatly increased; the power market reforms continuously and deeply, the activity of power users is increasingly prominent, and the requirements of the power users on the power grid service level are also continuously improved. On the other hand, the intelligent construction of the power grid is gradually advanced, the construction of data acquisition systems such as an electric energy management system, a fault maintenance system, a load side operator management system and the like is accelerated and continuously perfected, the management system is gradually built, and the intelligent construction of the power grid is urgently needed to be converted into production value, so that the fine application efficiency of the power big data is improved. While in traditional power big data research, the previous power data has a standardized format, and the network equipment data which needs to be introduced at present needs to further explore a more efficient data user abstract user feature set extraction method.
Electrical energy is one of the main driving forces for human activity, and demand response is critical to maintaining reliable and efficient operation of smart grid systems, which can manage power delivery from the power system to the consumer and smooth system loads in residential power distribution. The construction of the grid image of the power grid user can predict the recent demands of the user and provide key information for the response decision of the residential demands. On the other hand, the household network image built in real time by the power grid can help to meet the auxiliary judging fault decision requirement and reduce a series of risks such as power failure. Although the user portrait technology is widely applied to the fields of credit evaluation, social media, electronic commerce and the like in the medical and health fields, the application scene of the current power grid user portrait technology is not clear enough, the technology and algorithm are simple, the monitoring data are too trivial, and the user attribute is difficult to describe accurately and efficiently.
In the related technology, the Chinese patent application document with the application publication number of CN114119057A discloses a construction system of a user portrait model, wherein the user attribute characteristics are extracted by acquiring historical data of a user to be analyzed through a word bag model, and index tag characteristics are selected from a user key characteristic sub-library through a characteristic selection module, so that a characteristic set and a user portrait of the visible fixed attribute level characteristics are constructed; and training the index tag of the training data of each time interval according to the weight values of the training data of a plurality of time intervals is needed.
The Chinese patent application document with the application publication number of CN112801207A discloses a power user portrait construction method based on big data, which focuses on the characteristic of mining time sequence evolution. Specifically, the machine learning information mining method is utilized to carry out information mining on the power user data, and then the feature vector of the power user is processed through a feature vector processing algorithm based on time sequence evolution. Since the scheme is more consistent with respect to attribute representation of time for single power source data, a simple time-sequential evolving eigenvector processing algorithm is used.
Disclosure of Invention
The technical problem to be solved by the invention is how to accurately image the power consumer, and assist the power grid to improve the service quality of the power grid operation.
The invention solves the technical problems by the following technical means:
on one hand, the invention provides a power grid user portrait construction method, which comprises the following steps:
acquiring historical moment data of different data sources of a power grid;
combining and mapping the historical moment data of each data source into a shared representation to integrate the correlation among modes, and extracting the time sequence characteristics of the initial power grid user;
reversely constructing the time sequence characteristics of the initial power grid user into a phase space structure of an original system by adopting phase space reconstruction to obtain a phase space reconstruction matrix;
extracting an abstract user feature set from the phase space reconstruction matrix based on mutual information correlation of joint entropy;
and carrying out regression prediction on the user portrait labels based on the abstract user feature set, clustering the user portrait labels to generate user clusters, and constructing user cluster portraits.
Further, the historical time data of the different data sources of the power grid include historical time data of a communication network side and historical time data of a power grid side, and after the historical time data of the different data sources of the power grid are obtained, the method further includes:
preprocessing historical moment data of different data sources of the power grid, wherein the preprocessing operation comprises filtering, classified summarization, error compensation, normalization and data dimension reduction.
Further, the mapping the historical time data combination of each data source into the shared representation to integrate the correlation between modalities, and extracting the initial power grid user time sequence feature comprises the following steps:
processing the historical moment data of each data source by adopting a pre-trained natural language processing model to generate initial user text data source embedding characteristics of default dimension;
and fusing the embedded characteristics of the text data sources of the initial users from a plurality of data sources by using a multi-layer neural network to obtain the time sequence characteristics of the users of the initial power grid.
Further, the reversely constructing the initial power grid user time sequence characteristic into the phase space structure of the original system by adopting phase space reconstruction to obtain a phase space reconstruction matrix, which comprises the following steps:
C-C method is adopted to determine the optimal embedding dimension and the optimal delay time of the phase space reconstruction;
and reversely constructing the initial power grid user time sequence characteristic into a phase space structure of the original system based on the optimal embedding dimension and the optimal delay time to obtain a phase space reconstruction matrix.
Further, the determining the optimal embedding dimension and the optimal delay time of the phase space reconstruction by adopting the C-C method comprises the following steps:
defining test statistics S and delta S based on a probability representation that a distance between any two points in a phase space is less than a given distance range;
rounding the time value of the first zero of the average value of the test statistic S or the first minimum of the average value of the test statistic Δs as the optimal delay time;
determining a global minimum value as an optimal embedding window based on the average value of the test statistic S and the average value of the test statistic delta S;
and calculating the optimal embedding dimension of the phase space reconstruction based on the optimal delay time and the optimal embedding window.
Further, the initial power grid user time sequence feature is reversely constructed into a phase space structure of an original system based on the optimal embedding dimension and the optimal delay time to obtain a phase space reconstruction matrix, and the formula is expressed as follows:
wherein:representing a phase space reconstruction matrix, x= { x 1 ,x 2 …x N And the initial grid user time sequence characteristic is represented, M represents the optimal embedding dimension, t represents the optimal delay time, and M=N- (M-1) t.
Further, the extracting the abstract user feature set from the phase space reconstruction matrix based on mutual information correlation of joint entropy includes:
for two discrete random variables, mutual information based on joint entropy is defined as information obtained when two discrete random variables are observed simultaneously, and the formula is expressed as:
wherein: MI (E, E) t ) Representing the extracted mutual information as an abstract set of user features,representing the joint probability of two discrete random variables, E i Representing the input of a discrete random variable,/->Represents a target value, p (E i ) Representation E i Is a single point probability of (a),representation->Single point probability of (c).
Further, the extracting the abstract user feature set from the phase space reconstruction matrix based on mutual information correlation of joint entropy includes:
for two discrete random variables, mutual information based on joint entropy is defined as information obtained when two discrete random variables are observed simultaneously, and the formula is expressed as:
wherein: MI (E, E) t ,E m ,E n ) Representing the extracted mutual information as an abstract set of user features,representing joint probabilities of four discrete random variables, E i Is to input discrete random variables, ">Is a target value, & lt & gt>Is the average value>Is the mode, p (E i ) Representation E i Single point probability of->Representation->Single point probability of->Representation->Is a single point probability of (a),representation->Single point probability of (c).
Further, the performing regression prediction on the user portrait label based on the abstract user feature set, generating a user cluster for the user portrait label cluster, and constructing a user cluster portrait, including:
carrying out regression prediction on the abstract user feature set by adopting an LSTM time sequence regression unit to obtain a user prediction attribute tag;
coding the simple label fixed by the user by adopting a coding mode to generate a label with fixed attribute of the user;
splicing the user prediction attribute tag and the user fixed attribute tag to obtain a user portrait tag;
and clustering the user portrait labels by adopting a K-means clustering algorithm to construct user cluster portraits.
In a second aspect, the present invention further provides a system for constructing a user portrait of a power grid, where the system includes:
the acquisition module is used for acquiring historical moment data of different data sources of the power grid;
the initial feature extraction module is used for combining and mapping the historical moment data of each data source into a shared representation to integrate the correlation among the modes and extract the time sequence features of the initial power grid user;
the characteristic phase space reconstruction module is used for reversely constructing the time sequence characteristics of the initial power grid user into a phase space structure of the original system by adopting phase space reconstruction to obtain a phase space reconstruction matrix;
the mutual information abstract feature extraction module is used for extracting an abstract user feature set from the phase space reconstruction matrix based on mutual information correlation of joint entropy;
and the user cluster portrait construction module is used for carrying out regression prediction on the user portrait labels based on the abstract user feature set, clustering the user portrait labels to generate user clusters and constructing user cluster portraits.
The invention has the advantages that:
(1) According to the method, the historical moment data of each data source are combined and mapped into the shared representation to integrate the correlation among the modes, the time sequence characteristics of the initial power grid user are extracted, and the correlation among the attributes among the data sources can provide additional clues in the integration process; after the initial power grid user time sequence characteristics, based on random and nonlinear behaviors of users, user attribute behaviors are difficult to accurately describe through a traditional method, dynamic characteristics can be better described by carrying out phase space reconstruction on the power grid user time sequence characteristics and reversely constructing one-dimensional time sequence data into a phase space structure of an original system by adopting the phase space reconstruction pair; based on mutual information correlation of the joint entropy, abstract user feature sets are extracted from the phase space reconstruction matrix, abstract semantics of user features are enhanced, and processing speed of the power grid under multi-source data is improved; and carrying out regression prediction on the user portrait labels based on the abstract user feature set, and establishing a power user portrait label system.
(2) Meanwhile, a multi-source data abstraction feature extraction technology of the operation and maintenance data of the power grid equipment and the equipment data of the user load side is combined, and an electric power user portrait tag system is established, so that the power grid can be further assisted to improve the operation service quality of the power grid.
(3) When the mutual information features are extracted, candidate inputs are ordered according to the information values, abstract features are selected by maximizing correlation and minimizing redundancy, meanwhile, the number and dimension of the features are reduced, the processing of big data in practical application is quickened, instantaneity is achieved while the most representative abstract features are obtained, and a household network image built in real time by a power grid can help to meet the requirements of auxiliary judging fault decision and reducing a series of risks such as power failure.
Additional aspects and advantages of the invention will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be learned by practice of the invention.
Drawings
FIG. 1 is a flow diagram of a method for constructing a user portrait of a power grid;
FIG. 2 is a schematic block diagram of a power grid user portrait construction method proposed by the present invention;
FIG. 3 is a schematic block diagram of feature extraction in the process of constructing the user portrait of the power grid;
fig. 4 is a schematic structural diagram of the system for constructing the user portrait of the power grid.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions in the embodiments of the present invention will be clearly and completely described in the following in conjunction with the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
As shown in fig. 1 to 3, a first embodiment of the present invention proposes a method for constructing a user portrait of a power grid, the method comprising the following steps:
s10, acquiring historical moment data of different data sources of a power grid;
s20, combining and mapping historical time data of each data source into a shared representation to integrate correlation among modes, and extracting initial power grid user time sequence characteristics;
s30, reversely constructing the time sequence characteristics of the initial power grid user into a phase space structure of an original system by adopting phase space reconstruction to obtain a phase space reconstruction matrix;
s40, extracting an abstract user feature set from the phase space reconstruction matrix based on mutual information correlation of joint entropy;
s50, carrying out regression prediction on the user portrait labels based on the abstract user feature set, clustering the user portrait labels to generate user clusters, and constructing user cluster portraits.
In the embodiment, the historical moment data combination of each data source is mapped into the shared representation to integrate the correlation among the modes, the initial power grid user time sequence characteristics are extracted, and the correlation among the attributes among the data sources can provide additional clues in the integration process; after the initial power grid user time sequence characteristics, based on random and nonlinear behaviors of users, user attribute behaviors are difficult to accurately describe through a traditional method, dynamic characteristics can be better described by carrying out phase space reconstruction on the power grid user time sequence characteristics and reversely constructing one-dimensional time sequence data into a phase space structure of an original system by adopting the phase space reconstruction pair; based on mutual information correlation of the joint entropy, abstract user feature sets are extracted from the phase space reconstruction matrix, abstract semantics of user features are enhanced, and processing speed of the power grid under multi-source data is improved; and carrying out regression prediction on the user portrait labels based on the abstract user feature set, and establishing a power user portrait label system.
In an embodiment, the historical time data of the different data sources of the power grid includes historical time data of a communication network side and historical time data of a power grid side, and after the obtaining the historical time data of the different data sources of the power grid, the method further includes:
preprocessing historical moment data of different data sources of the power grid, wherein the preprocessing operation comprises filtering, classified summarization, error compensation, normalization and data dimension reduction.
In the embodiment, by carrying out data collection and text data preprocessing on historical moment data from two different data sources of a communication network and a power network, the state update of each user in the data is combined into one document of each user in order to construct the initial characteristic embedding of the text data source. Based on the data of the power grid full-service unified data center, the key data of the power distribution network are extracted from the original data such as monitoring data, standing account data, defect and fault data, load data, overhaul test data and the like, and the key internal characteristic data of the power distribution network are extracted. External data such as user geographic attribute data, user load end flow monitoring data and the like are collected in a targeted manner.
Because of inconsistent data sources, irregular data formats, data redundancy, missing and the like, the data preprocessing operations such as filtering, classified summarization, error compensation, normalization, data dimension reduction and the like are adopted to process the data of different data sources, so that subsequent feature extraction processing is facilitated.
The application scene of the current power grid user portrait technology is not clear enough, the user attribute problem is difficult to describe accurately and efficiently, the key technical thought of the realization of the power grid user portrait technology is provided, and meanwhile, a power grid operation and maintenance data and a multi-source data abstract feature extraction technology of user load side equipment data are combined to establish a power grid user portrait label system, so that the power grid operation service quality can be further assisted.
In one embodiment, as shown in fig. 2, the step S20: the method comprises the steps of combining and mapping historical time data of each data source into a shared representation to integrate correlations among modes, extracting initial power grid user time sequence characteristics, and specifically comprises the following steps:
s21, processing historical moment data of each data source by adopting a pre-trained natural language processing model, and generating initial user text data source embedding characteristics of default dimension;
it should be noted that, in this embodiment, the attribute group extracted from the state update of the user represents each user, and then the feature extraction is performed on the historical time data of each data source by using the FastText pre-training model based on natural language processing, so as to generate the default dimension pre-training FastText vector as the embedded feature of the initial user text data source.
S22, fusing the embedded characteristics of the text data sources of the initial users from a plurality of data sources by using a multi-layer neural network to obtain the time sequence characteristics of the users of the initial power grid.
It should be noted that, in this embodiment, the embedded features of the text data sources of the initial users from the multiple data sources are primarily fused by using the multi-layer neural network to obtain the shallow user features of the historical moment T, and further the deep user features of the historical moment T are obtained by fusing as the time sequence features of the user of the initial power grid.
In one embodiment, the step S30: the initial power grid user time sequence characteristics are reversely constructed into a phase space structure of an original system by adopting phase space reconstruction, and a phase space reconstruction matrix is obtained, and the method specifically comprises the following steps:
s31, determining the optimal embedding dimension and the optimal delay time of phase space reconstruction by adopting a C-C method;
the phase space reconstruction is considered to be a component generated by a nonlinear power system. Because of the random and nonlinear behaviors of the user, the user attribute behaviors are difficult to accurately describe through a traditional method, and the equivalent Gao Weixiang space of the behavior analysis of the user in the power system network can be reconstructed through correlation change rules of different data sources in the power system. The key of phase space reconstruction is to determine the optimal embedding dimension and the optimal delay time.
S32, reversely constructing the initial power grid user time sequence characteristic into a phase space structure of the original system based on the optimal embedding dimension and the optimal delay time to obtain a phase space reconstruction matrix.
According to the embodiment, based on random and nonlinear behaviors of the user, the user attribute behaviors are difficult to accurately describe through a traditional method, and the C-C method is used for reconstructing the phase space of the time sequence characteristics of the power grid, so that the dynamic characteristics can be better described, and the most representative abstract semantic characteristics are extracted; meanwhile, the C-C method is adopted to reconstruct the phase space of the time sequence characteristics of the power grid, so that the data of different data sources can be mapped to a high-level abstract space, and the purpose of multi-source data characteristic extraction is achieved.
By using the advantages of optimal embedding dimensions and optimal delay times: because data from different data sources need to be fused, and the feature dimensions and time representations of the different data sources are not identical, such data cannot be fused and processed. Therefore, the optimal embedding dimension and the optimal delay time of the phase space reconstruction are determined through the C-C method, so that the optimal semantic information of each data can be reserved while ensuring that the data from different data sources can be fused and processed normally, and the situation that the data of one data source has redundancy or serious deletion can not exist.
In one embodiment, the step S31: the method for determining the optimal embedding dimension and the optimal delay time of the phase space reconstruction by adopting the C-C method specifically comprises the following steps:
s311, defining test statistics S and delta S based on probability representation that the distance between any two points in the phase space is smaller than a given distance range;
specifically, if a set of time series features is x= { x 1 ,x 2 …x N If the embedding dimension is M and the time delay is t, m=n- (M-1) t, then the set of points in the reconstructed phase space can be expressed by the following formula:
at this time, the correlation integral indicates that the distance between any two points in the phase space is smaller than r k Probability of (2):
wherein:according to the statistical conclusion, when N>At 3000, M and r can be obtained k Is a value range of m epsilon {2,3,4,5}, r k Expressed as given distance =kx0.5σThe real number of the off-range, sigma, is the standard deviation of the time sequence, k epsilon {1,2,3,4}.
Test statistics S and Δs are defined and a block averaging strategy is employed, formulated as follows:
wherein:and->Representing that the correlation integral represents that the distance between any two points in the phase space is less than r k Where a block averaging strategy is performed on the phase space.
S312, rounding the moment value of the first zero of the average value of the test statistic S or the first minimum value of the average value of the test statistic delta S as the optimal delay time;
specifically, the formula of the average of the test statistic S and the average of the test statistic Δs is as follows:
s313, determining a global minimum value as an optimal embedded window based on the average value of the test statistic S and the average value of the test statistic delta S;
specifically, S is cor The global minimum of (t) is the optimal embedding window t ω The formula is:
s314, calculating the optimal embedding dimension of the phase space reconstruction based on the optimal delay time and the optimal embedding window. Specifically, the optimal delay t opt And an optimal embedding window t ω Substituting into the following formula, rounding to obtain the optimal embedded dimension m opt ：
t ω ＝(m opt -1)t opt 。
In one embodiment, the step S32: based on the optimal embedding dimension and the optimal delay time, reversely constructing the time sequence characteristics of the initial power grid user into a phase space structure of an original system to obtain a phase space reconstruction matrix, wherein the formula is expressed as follows:
wherein:representing a phase space reconstruction matrix, x= { x 1 ,x 2 …x N And the initial grid user time sequence characteristic is represented, M represents the optimal embedding dimension, t represents the optimal delay time, and M=N- (M-1) t.
It should be noted that, in this embodiment, after the deep semantic feature fusion is performed by using the neural network to obtain the semantic features of the power grid user at the historical moment, based on the random and nonlinear behaviors of the user, the user attribute behavior is difficult to accurately describe by the traditional method, the C-C method is used to reconstruct the phase space of the power grid time sequence features, and the phase space reconstruction is used to reversely construct the one-dimensional time sequence data into the phase space structure of the original system, so that the dynamic characteristics can be better described.
In one embodiment, the step S40: extracting an abstract user feature set from the phase space reconstruction matrix based on mutual information correlation of joint entropy, comprising:
for two discrete random variables, mutual information based on joint entropy is defined as information obtained when two discrete random variables are observed simultaneously, and the formula is expressed as:
wherein: MI (E, E) t ) Representing the extracted mutual information as an abstract set of user features,representing the joint probability of two discrete random variables, E i Representing the input of a discrete random variable,/->Represents a target value, p (E i ) Representation E i Is a single point probability of (a),representation->Single point probability of (c).
It should be noted that, in machine learning, function extraction/selection is a process of selecting a subset of abstract features from a given dataset to avoid explosion of dimensions, and for two discrete random variables, mutual information based on joint entropy is defined as information obtained when two discrete random variables are observed simultaneously, and the mathematical description is as follows:
in feature selection, information common to two variables is indispensable, and the formula is expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,is common information between two variables, called mutual information, in which case candidate inputs are ordered between the input and target values by mutual information techniques.
From the entropy-based mutual information technique, the following three inferences can be drawn:
(1) If it isIndicating discrete random variables and independence;
(2) If it isThe larger the value of (2), the more highly relevant the discrete random variable is explained;
(3) If it isThe smaller the value of (c), the less discrete variable and slightly correlated.
Further, the mutual information feature selection method proposed by the above embodiment may cause a problem in mutual information prediction. Thus, this embodiment uses an average value in addition to the target value, as the average value and the target value are equally important, while the average value is very low if some of the selected features are very small. Adding the average to the other two parameters is not enough and can also lead to mutual information prediction problems. The present increase is therefore four variables as follows:
wherein: MI (E, E) t ,E m ,E n ) Representing the extracted mutual information as an abstract set of user features,representing joint probabilities of four discrete random variables, E i Is to input discrete random variables, ">Is a target value, & lt & gt>Is the average value>Is the mode, p (E i ) Representation E i Single point probability of->Representation->Single point probability of->Representation->Is a single point probability of (a),representation->Single point probability of (c).
It should be noted that the mutual information feature extraction technique adopted in this embodiment is used for solving E i ,E t ,E n ,E m Mutual information among the four variables is improved in accuracy and convergence rate by a mutual information feature selection technology, and input is ranked according to the importance of the information based on the improved function extraction technology; filtering the ranked inputs using an indifferent and redundant filter to eliminate indifferent redundant information; the subset of selected functions contains the best and more relevant information that contributes significantly to accuracy and convergence. The improved mutual information feature extraction technique provides two benefits: (a) Selecting appropriate and relevant features may minimize the prediction error, and (b) selecting a subset of features may increase the convergence speed; the above-mentioned user feature space reconstruction and abstract feature extraction flow are shown in fig. 3.
In one embodiment, the step S50: the regression prediction user portrait label is carried out based on the abstract user feature set, a user cluster is generated for clustering the user portrait label, and a user cluster portrait is constructed, and the method specifically comprises the following steps:
s51, carrying out regression prediction on the abstract user feature set by adopting an LSTM time sequence regression unit to obtain a user prediction attribute tag;
s52, coding the simple label fixed by the user in a coding mode to generate a label with fixed attribute of the user;
s53, splicing the user prediction attribute tag and the user fixed attribute tag to obtain a user portrait tag;
s54, clustering the user portrait labels by adopting a K-means clustering algorithm to construct user cluster portraits.
It should be noted that, in this embodiment, the user portrait tag is divided into a fixed simple tag and a predictive tag, and the fixed simple tag is used for the user, for example: the electricity utilization cell, the street number, the number of electric equipment and the like where the user is located can be generated by simple statistical coding or one-hot coding. And for predictive labels, such as: whether the user is at home, predicts the electricity consumption of the user in the month, the traffic use condition and the like; regression or classification predictions need to be made using the abstract features of the user. Therefore, in this embodiment, regression prediction is performed through the extracted abstract user features, so that a more accurate user tag can be obtained.
Specifically, in this embodiment, the LSTM timing regression unit is introduced to accurately predict the user tag, so that { x } 1 ,x 2 ,…,x t "represents a typical input sequence of LSTM cells, where x t ∈R k Representing a K-dimensional real value vector at time step t. To establish a time connection, the LSTM cell defines and maintains an internal memory cell state throughout the life cycle, which is the most important element of the LSTM cell. Memory cell state s t-1 And intermediate output h t-1 And subsequent input x t And interacting, namely outputting the previous time step and the current time step according to the previous time step. In addition to the internal states, the LSTM cell also defines an input node g t Input gate i t Forgetting door f t And an output gate O t The formulas for all nodes in the LSTM cell are as follows:
f t ＝σ(W fx x t +W fh h t-1 +b f )
i t ＝σ(W ix x t +W ih h t-1 +b i )
g t ＝Φ(W gx x t +W gh h t-1 +b g )
o t ＝σ(W ox x t +W oh h t-1 +b o )
s t ＝g t ⊙i t +s t-1 ⊙f t
h t ＝Φ(s t )⊙o t
wherein W is fx ,W fh ,W ix ,W ih ,W gx ,W gh ,W ox And W is oh Is a weight matrix input by the network activation function, wherein, the weight matrix is a weight matrix input by the network activation function, the weight matrix is a multiplicative element by element, the sigma is a sigmoid activation function, and the phi is a tanh activation function; the user abstract label is then predicted over the fully connected network.
Furthermore, the user portraits are clustered through a K-means clustering algorithm, and the algorithm is simple and easy to operate and is one of the most widely applied clustering methods. The method comprises the steps of firstly, randomly selecting a group of initial cluster centers through iteration, wherein the cluster centers are independent and mutually close, and taking the average value of all data samples in a cluster subset as the center of a class.
The embodiment is based on multi-source heterogeneous user data, and besides the fixed attribute characteristics, the scheme further excavates the abstract label characteristics of the relationship between different data source attributes and between users through a C-C method, so that more accurate user portrait relationship can be further constructed, the relationship is not limited to fixed index labels, and the user portrait association degree is deeper. In addition, the embodiment mines the information of the flow sources from the power source and the external user side to further construct the user figure about the power network, so that the processing is more complex in the data characteristic extraction and fusion stage, different data sources are mined through a C-C method, the optimal delay time of the phase space reconstruction is determined, the time sequence information of each data can be correctly represented while ensuring the normal fusion and processing of the data from the different data sources, and the purpose of heterogeneous data fusion is achieved.
Furthermore, as shown in fig. 4, a second embodiment of the present invention proposes a system for constructing a grid customer portrait, the system comprising:
the acquisition module 10 is used for acquiring historical moment data of different data sources of the power grid;
the initial feature extraction module 20 is configured to map the historical time data combinations of the data sources into a shared representation to integrate correlations between modalities, and extract initial grid user timing features;
the characteristic phase space reconstruction module 30 is configured to reversely construct the initial grid user time sequence characteristic into a phase space structure of the original system by adopting phase space reconstruction, so as to obtain a phase space reconstruction matrix;
a mutual information abstract feature extraction module 40, configured to extract an abstract user feature set from the phase space reconstruction matrix based on mutual information correlation of joint entropy;
and the user cluster portrait construction module 50 is used for carrying out regression prediction on the user portrait labels based on the abstract user feature set, clustering the user portrait labels to generate user clusters and constructing user cluster portraits.
In the embodiment, the historical moment data combination of each data source is mapped into the shared representation to integrate the correlation among the modes, the initial power grid user time sequence characteristics are extracted, and the correlation among the attributes among the data sources can provide additional clues in the integration process; after the initial power grid user time sequence characteristics, based on random and nonlinear behaviors of users, user attribute behaviors are difficult to accurately describe through a traditional method, dynamic characteristics can be better described by carrying out phase space reconstruction on the power grid user time sequence characteristics and reversely constructing one-dimensional time sequence data into a phase space structure of an original system by adopting the phase space reconstruction pair; based on mutual information correlation of the joint entropy, abstract user feature sets are extracted from the phase space reconstruction matrix, abstract semantics of user features are enhanced, and processing speed of the power grid under multi-source data is improved; and carrying out regression prediction on the user portrait labels based on the abstract user feature set, and establishing a power user portrait label system.
In an embodiment, the historical time data of the different data sources of the power grid includes historical time data of a communication network side and historical time data of a power grid side, and the system further includes:
the preprocessing module is used for preprocessing the historical moment data of different data sources of the power grid, and the preprocessing operation comprises filtering, classified summarization, error compensation, normalization and data dimension reduction.
In one embodiment, the initial feature extraction module 20 includes:
the initial feature extraction unit is used for processing the historical moment data of each data source by adopting a pre-trained natural language processing model to generate initial user text data source embedded features with default dimensions;
and the characteristic fusion unit is used for fusing the embedded characteristics of the initial user text data sources from the plurality of data sources by using the multi-layer neural network to obtain the user time sequence characteristics of the initial power grid.
In an embodiment, the characteristic phase space reconstruction module 30 includes:
a reconstruction parameter determining unit, configured to determine an optimal embedding dimension and an optimal delay time for phase space reconstruction by using a C-C method;
and the reconstruction unit is used for reversely constructing the initial power grid user time sequence characteristic into a phase space structure of the original system based on the optimal embedding dimension and the optimal delay time to obtain a phase space reconstruction matrix.
In an embodiment, the reconstruction parameter determination unit is specifically configured to perform the following steps:
defining test statistics S and delta S based on a probability representation that a distance between any two points in a phase space is less than a given distance range;
rounding the time value of the first zero of the average value of the test statistic S or the first minimum of the average value of the test statistic Δs as the optimal delay time;
determining a global minimum value as an optimal embedding window based on the average value of the test statistic S and the average value of the test statistic delta S;
and calculating the optimal embedding dimension of the phase space reconstruction based on the optimal delay time and the optimal embedding window.
In an embodiment, the reconstruction unit obtains a phase space reconstruction matrix, expressed as:
wherein:representing a phase space reconstruction matrix, x= { x 1 ,x 2 …x N And the initial grid user time sequence characteristic is represented, M represents the optimal embedding dimension, t represents the optimal delay time, and M=N- (M-1) t.
In one embodiment, the mutual information abstract feature extraction module 40 is specifically configured to:
for two discrete random variables, mutual information based on joint entropy is defined as information obtained when two discrete random variables are observed simultaneously, and the formula is expressed as:
wherein: MI (E, E) t ) Representing the extracted mutual information as an abstract set of user features,representing the joint probability of two discrete random variables, E i Representing the input of a discrete random variable,/->Represents a target value, p (E i ) Representation E i Single point probability of (v)，Representation->Single point probability of (c).
In one embodiment, the mutual information abstract feature extraction module 40 is specifically configured to:
for two discrete random variables, mutual information based on joint entropy is defined as information obtained when two discrete random variables are observed simultaneously, and the formula is expressed as:
wherein: MI (E, E) t ,E m ,E n ) Representing the extracted mutual information as an abstract set of user features,representing joint probabilities of four discrete random variables, E i Is to input discrete random variables, ">Is a target value, & lt & gt>Is the average value>Is the mode, p (E i ) Representation E i Single point probability of->Representation->Single point probability of->Representation->Is a single point probability of (a),representation->Single point probability of (c).
In one embodiment, the user cluster portrait construction module 50 specifically includes:
the prediction unit is used for carrying out regression prediction on the abstract user feature set by adopting the LSTM time sequence regression unit to obtain a user prediction attribute tag;
the coding unit is used for coding the simple labels fixed by the user in a coding mode to generate the labels with the fixed attributes of the user;
the splicing unit is used for splicing the user prediction attribute tag and the user fixed attribute tag to obtain a user portrait tag;
and the clustering unit is used for clustering the user portrait labels by adopting a K-means clustering algorithm to construct user cluster portraits.
It should be noted that, other embodiments of the system for constructing a user portrait of a power grid or the implementation method thereof according to the present invention may refer to the above embodiments of the method, and are not repeated here.
In the description of the present specification, a description referring to terms "one embodiment," "some embodiments," "examples," "specific examples," or "some examples," etc., means that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiments or examples. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
Furthermore, the terms "first," "second," and the like, are used for descriptive purposes only and are not to be construed as indicating or implying a relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defining "a first" or "a second" may explicitly or implicitly include at least one such feature. In the description of the present invention, the meaning of "plurality" means at least two, for example, two, three, etc., unless specifically defined otherwise.
While embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and not to be construed as limiting the invention, and that variations, modifications, alternatives and variations may be made to the above embodiments by one of ordinary skill in the art within the scope of the invention.

Claims (10)

1. The utility model provides a power grid user portrait construction method which is characterized in that the method comprises the following steps:
acquiring historical moment data of different data sources of a power grid;
combining and mapping the historical moment data of each data source into a shared representation to integrate the correlation among modes, and extracting the time sequence characteristics of the initial power grid user;
reversely constructing the time sequence characteristics of the initial power grid user into a phase space structure of an original system by adopting phase space reconstruction to obtain a phase space reconstruction matrix;
extracting an abstract user feature set from the phase space reconstruction matrix based on mutual information correlation of joint entropy;
and carrying out regression prediction on the user portrait labels based on the abstract user feature set, clustering the user portrait labels to generate user clusters, and constructing user cluster portraits.
2. The method for constructing a user figure of a power grid according to claim 1, wherein the historical moment data of the different data sources of the power grid comprises historical moment data of a communication network side and historical moment data of a power grid side, and after the historical moment data of the different data sources of the power grid is obtained, the method further comprises:
preprocessing historical moment data of different data sources of the power grid, wherein the preprocessing operation comprises filtering, classified summarization, error compensation, normalization and data dimension reduction.
3. The grid consumer representation construction method of claim 1, wherein the mapping of historical time of day data combinations of data sources into a shared representation to integrate correlations between modalities, extracting initial grid consumer timing features, comprises:
processing the historical moment data of each data source by adopting a pre-trained natural language processing model to generate initial user text data source embedding characteristics of default dimension;
and fusing the embedded characteristics of the text data sources of the initial users from a plurality of data sources by using a multi-layer neural network to obtain the time sequence characteristics of the users of the initial power grid.
4. The method for constructing a grid user portrait of claim 1, wherein the reversely constructing the initial grid user time sequence characteristic into a phase space structure of an original system by adopting phase space reconstruction, to obtain a phase space reconstruction matrix, comprises the following steps:
C-C method is adopted to determine the optimal embedding dimension and the optimal delay time of the phase space reconstruction;
and reversely constructing the initial power grid user time sequence characteristic into a phase space structure of the original system based on the optimal embedding dimension and the optimal delay time to obtain a phase space reconstruction matrix.
5. The grid consumer representation construction method of claim 4, wherein said determining the optimal embedding dimension and the optimal delay time for phase space reconstruction using the C-C method comprises:
defining test statistics S and delta S based on a probability representation that a distance between any two points in a phase space is less than a given distance range;
rounding the time value of the first zero of the average value of the test statistic S or the first minimum of the average value of the test statistic Δs as the optimal delay time;
determining a global minimum value as an optimal embedding window based on the average value of the test statistic S and the average value of the test statistic delta S;
and calculating the optimal embedding dimension of the phase space reconstruction based on the optimal delay time and the optimal embedding window.
6. The method for constructing a user portrait of a power grid according to claim 4, wherein the initial power grid user time sequence feature is reversely constructed into a phase space structure of an original system based on an optimal embedding dimension and an optimal delay time to obtain a phase space reconstruction matrix, and the formula is as follows:
wherein:representing a phase space reconstruction matrix, x= { x 1 ,x 2 …x N And the initial grid user time sequence characteristic is represented, M represents the optimal embedding dimension, t represents the optimal delay time, and M=N- (M-1) t.
7. The grid consumer representation construction method of claim 1, wherein the extracting the abstract set of user features from the phase space reconstruction matrix based on mutual information correlation of joint entropy comprises:
for two discrete random variables, mutual information based on joint entropy is defined as information obtained when two discrete random variables are observed simultaneously, and the formula is expressed as:
wherein: MI (E, E) t ) Representing extracted mutual information as abstract user characteristicsThe collection of the symptoms is carried out,representing the joint probability of two discrete random variables, E i Representing the input of a discrete random variable,/->Represents a target value, p (E i ) Representation E i Single point probability of->Representation->Single point probability of (c).
8. The grid consumer representation construction method of claim 1, wherein the extracting the abstract set of user features from the phase space reconstruction matrix based on mutual information correlation of joint entropy comprises:
for two discrete random variables, mutual information based on joint entropy is defined as information obtained when two discrete random variables are observed simultaneously, and the formula is expressed as:
wherein: MI (E, E) t ,E m ,E n ) Representing the extracted mutual information as an abstract set of user features,representing joint probabilities of four discrete random variables, E i Is to input discrete random variables, ">Is a target value, & lt & gt>Is the average value>Is the mode, p (E i ) Representation E i Single point probability of->Representation->Single point probability of->Representation->Single point probability of->Representation->Single point probability of (c).
9. The method for constructing a user figure of a power grid according to claim 1, wherein the performing regression prediction on the user figure label based on the abstract user feature set, clustering the user figure labels to generate a user cluster, and constructing a user cluster figure comprises:
carrying out regression prediction on the abstract user feature set by adopting an LSTM time sequence regression unit to obtain a user prediction attribute tag;
coding the simple label fixed by the user by adopting a coding mode to generate a label with fixed attribute of the user;
splicing the user prediction attribute tag and the user fixed attribute tag to obtain a user portrait tag;
and clustering the user portrait labels by adopting a K-means clustering algorithm to construct user cluster portraits.
10. A grid consumer representation construction system, the system comprising:
the acquisition module is used for acquiring historical moment data of different data sources of the power grid;
the initial feature extraction module is used for combining and mapping the historical moment data of each data source into a shared representation to integrate the correlation among the modes and extract the time sequence features of the initial power grid user;
the characteristic phase space reconstruction module is used for reversely constructing the time sequence characteristics of the initial power grid user into a phase space structure of the original system by adopting phase space reconstruction to obtain a phase space reconstruction matrix;
the mutual information abstract feature extraction module is used for extracting an abstract user feature set from the phase space reconstruction matrix based on mutual information correlation of joint entropy;
and the user cluster portrait construction module is used for carrying out regression prediction on the user portrait labels based on the abstract user feature set, clustering the user portrait labels to generate user clusters and constructing user cluster portraits.
CN202310677681.4A 2023-06-07 2023-06-07 Power grid user portrait construction method and system Pending CN116596574A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310677681.4A CN116596574A (en) 2023-06-07 2023-06-07 Power grid user portrait construction method and system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310677681.4A CN116596574A (en) 2023-06-07 2023-06-07 Power grid user portrait construction method and system

Publications (1)

Publication Number Publication Date
CN116596574A true CN116596574A (en) 2023-08-15

Family

ID=87606288

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310677681.4A Pending CN116596574A (en) 2023-06-07 2023-06-07 Power grid user portrait construction method and system

Country Status (1)

Country Link
CN (1) CN116596574A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117435904A (en) * 2023-12-20 2024-01-23 电子科技大学 Single feature ordering and composite feature extraction method

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117435904A (en) * 2023-12-20 2024-01-23 电子科技大学 Single feature ordering and composite feature extraction method
CN117435904B (en) * 2023-12-20 2024-03-15 电子科技大学 Single feature ordering and composite feature extraction method

Similar Documents

Publication Publication Date Title
CN111444236B (en) Mobile terminal user portrait construction method and system based on big data
Wu et al. Knowledge engineering with big data
CN108764984A (en) A kind of power consumer portrait construction method and system based on big data
CN112330078B (en) Power consumption prediction method and device, computer equipment and storage medium
Raghavendra et al. Artificial humming bird with data science enabled stability prediction model for smart grids
Shaheen et al. Mining sustainability indicators to classify hydrocarbon development
Bedi et al. CitEnergy: A BERT based model to analyse Citizens’ Energy-Tweets
CN116596574A (en) Power grid user portrait construction method and system
Park et al. Study on the future sign detection in areas of academic interest related to the digitalization of the energy industry
Kong et al. RETRACTED ARTICLE: Multimodal interface interaction design model based on dynamic augmented reality
CN113094448B (en) Analysis method and analysis device for residence empty state and electronic equipment
Khan et al. Power grid stability analysis using pipeline machine
Zhu et al. Multiscale deep network based multistep prediction of high‐dimensional time series from power transmission systems
Li et al. An improved genetic-XGBoost classifier for customer consumption behavior prediction
CN116450827A (en) Event template induction method and system based on large-scale language model
CN113342988B (en) Method and system for constructing service knowledge graph to realize service combination optimization based on LDA cross-domain
Li et al. A study on customer churn of commercial banks based on learning from label proportions
Li et al. A machine learning-based decision support framework for energy storage selection
Ghebriout et al. Harmonic-NAS: Hardware-Aware Multimodal Neural Architecture Search on Resource-constrained Devices
CN113987808A (en) Electricity user complaint early warning method of feature weighted Bayesian network
CN113779248A (en) Data classification model training method, data processing method and storage medium
CN108446724B (en) Fusion feature classification method
Li et al. Multi-source heterogeneous log fusion technology of power information system based on big data and imprecise reasoning theory
Xiao et al. Similarity matching method of power distribution system operating data based on neural information retrieval
CN111242520A (en) Feature synthesis model generation method and device and electronic equipment

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