CN-122022873-A - Potential demand mining and value added service recommendation method based on multi-source power data

CN122022873ACN 122022873 ACN122022873 ACN 122022873ACN-122022873-A

Abstract

The invention discloses a potential demand mining and value-added service recommendation method based on multi-source power data, which relates to the field of demand service recommendation and comprises the steps of forming a unified data set by integrating heterogeneous data and performing cleaning, standardization and desensitization treatment, performing bidirectional cluster analysis on customers and products based on a demand-product-benefit trinity model, establishing an adaptation relation between customer portraits and product portraits, designing 3 sets of differential packages for homogenous customer groups by combining a scene benefit model to quantify service potential, comprehensively evaluating factors such as demand forced tangency, consumption capability and the like to generate value scores, establishing a dynamic sequencing mechanism, and automatically adjusting recommendation strategies when electricity utilization characteristics or benefit fluctuation exceeds a threshold value. The method has the advantages that the multi-source power data are integrated, the differentiated package is generated by means of the trinity model, the bidirectional grouping and the scene revenue model, the recommendation strategy is adjusted by combining a dynamic ordering mechanism, and the client requirements and the enterprise revenue are matched.

Inventors

WU LIXIAN
WANG XINGBO
PANG WEILIN
LIN HAO
HUANGFU HANCONG
GUAN ZHAOXIONG
SONG CAIHUA
DU JIABING
CAO XINYI
Li Muxu

Assignees

广东电网有限责任公司佛山供电局

Dates

Publication Date: 20260512
Application Date: 20260410

Claims (9)

1. The potential demand mining and value-added service recommending method based on the multi-source power data is characterized by comprising the following steps of: Collecting multi-type heterogeneous data sources in an electric power system, and generating a standardized multi-source electric power data set through outlier rejection, missing value interpolation filling, standardized dimension and format processing and privacy data desensitization operation; Constructing a trinity value mining support model, wherein the trinity value mining support model consists of a demand library, a product library and a benefit library, and the three are mutually associated and cooperatively supported to carry out full-flow coverage of potential demand mining and value-added service recommendation; Adopting a data layer fusion strategy, matching electricity consumption, load and resource data based on customer numbers, overlapping interaction and external environment data, performing duplication elimination, complementation and conflict elimination, and generating a fusion data set for integrating electricity consumption characteristics of a customer side and resource conditions of a power grid side; Taking the fusion data set as input, carrying out bidirectional clustering based on a K-means clustering algorithm, carrying out client clustering, dividing an homogenized client group and generating a proprietary image, carrying out product clustering, dividing a product group and generating a product image, and obtaining the adaptation relation between different client groups and service products; based on the fusion data set and the two-way portrait, carrying out electric field scene classification, and quantifying the profit potential of various services in different scenes by constructing a scene profit model aiming at customer experience improvement and enterprise profit growth; screening the matched basic service and the value-added service for each homogeneous customer group based on the two-way portrait and the scene profit model, and combining to generate 3 sets of differential packages; Based on the profit library and the scenerized profit model, measuring and calculating the direct profit and the indirect profit of each set of packages, and comprehensively evaluating the service value by combining the customer demand urgent degree and the consumption capability to generate a quantized score; And establishing a dynamic ordering mechanism, monitoring the electricity utilization characteristics of the clients and the change of the income data in real time, setting a threshold value, and re-measuring and calculating package income and value scores when the electricity utilization of the clients or the fluctuation of the income data exceeds the threshold value, and adjusting the recommendation sequence.
2. The method for mining potential requirements and recommending value-added services based on multi-source power data according to claim 1, wherein the step of collecting multi-type heterogeneous data sources in the power system, and the step of generating the standardized multi-source power data set by outlier rejection, missing value interpolation filling, standardized dimension and format processing, and privacy data desensitization operation specifically comprises the steps of: collecting multi-type heterogeneous data sources in an electric power system, wherein the multi-type heterogeneous data sources comprise customer electricity utilization basic data, power grid load data, power grid resource data, customer interaction data and external environment data; preprocessing the collected multi-type heterogeneous data sources, and identifying and removing abnormal values and noise data generated in the data collection and transmission processes through an abnormal value detection method; filling the missing data by adopting an interpolation method, carrying out standardized processing on data with different formats and different dimensions, and unifying the data formats and the measurement standards; and hiding the customer privacy information through data desensitization processing, and obtaining the standardized multi-source power data set.
3. The method for potential demand mining and value added service recommendation based on multi-source power data according to claim 1, wherein the step of constructing a trinity value mining support model, the trinity value mining support model is composed of a demand library, a product library and a profit library, the three are mutually associated and cooperatively supported, and the step of performing full-flow coverage of potential demand mining and value added service recommendation specifically comprises the following steps: Constructing a demand library based on a standardized multi-source power data set, mining dominant demands and potential demands of clients through data association analysis, classifying and storing according to electricity consumption types and demand types, and recording occurrence frequencies of various demands, association factors and demand urgency; Integrating power basic service and value added service resources to construct a product library, and labeling attributes of each type of service product, including service content, applicable customer groups, service cost, service period and implementation conditions; And constructing a profit library by combining the cost data, the electricity consumption data and the market quotation data of various service products in the product library, and recording the unit profits, expected profits and profit periods of the various service products.
4. The method for mining potential demands and recommending value-added services based on multi-source power data according to claim 1, wherein the step of adopting a data layer fusion strategy to match power consumption, load and resource data based on customer numbers, and the step of superposing interaction and external environment data to perform duplication elimination, complementation and conflict elimination, and the step of generating a fusion data set for integrating power consumption characteristics of a customer side and resource conditions of a power grid side specifically comprises the following steps: carrying out fusion processing on the preprocessed standardized multi-source power data set by adopting a data layer fusion strategy, and carrying out association matching on customer power consumption basic data, power grid load data and power grid resource data by establishing a unified data association rule; The data one-to-one correspondence is carried out through the customer power utilization user numbers, and the customer interaction data, the external environment data and the associated data are overlapped and fused, so that the data redundancy and the data conflict are eliminated; and (3) through data deduplication and data completion operations, key effective data are reserved, and a unified-format and unified-dimension fusion data set is generated.
5. The method for mining potential demands and recommending value-added services based on multi-source power data according to claim 1, wherein the method for mining potential demands and recommending value-added services based on multi-source power data is characterized in that the fused data set is used as input, bidirectional clustering is performed based on a K-means clustering algorithm, clients are clustered to divide homogeneous client groups and generate exclusive images, products are clustered to divide product groups and generate product images, and the method for acquiring the adaptation relation between different client groups and service products comprises the following steps: Taking the fusion data set as input, and respectively carrying out client clustering and product clustering by applying a K-means clustering algorithm to construct a client-product bi-directional portrait; in the customer grouping process, selecting customer electricity load, electricity consumption period, electricity consumption capacity, electricity fee payment level and interaction behavior as clustering indexes, dividing customers into a plurality of homogenized customer groups, and generating exclusive customer portraits; In the product grouping process, selecting an applicable scene, service cost, expected benefits, implementation difficulty and customer suitability of service products as clustering indexes, and dividing the service products into a plurality of product clusters through a K-means algorithm to generate product images; And acquiring the adaptation relation between different customer groups and different product groups through the association matching of the customer portraits and the product portraits.
6. The method for mining potential requirements and recommending value-added services based on multi-source power data according to claim 1, wherein the step of classifying the electric field scenes based on the fusion data set and the two-way portrait, and the step of quantifying the revenue potential of various services in different scenes by constructing a scene revenue model targeting customer experience improvement and enterprise revenue growth comprises the following steps: Based on the fusion data set and the two-way portrait, different electricity utilization scenes are divided by combining load data, power grid resource data and customer electricity utilization data; aiming at each electricity utilization scene, constructing a scene-based income model by combining load fluctuation characteristics, power grid resource utilization rate and customer electricity utilization requirements in the scene; the model takes the promotion of customer electricity consumption experience and the gain growth of power enterprises as dual targets, quantifies the gain potential of different service products in each scene, associates the gain association degree of customer electricity consumption characteristics and service products, and obtains the suitability and gain level of various service products in different scenes.
7. The method for mining potential requirements and recommending value-added services based on multi-source power data according to claim 1, wherein the screening of adapted basic services and value-added services for each homogenous customer group based on the bi-directional portrait and scenerized revenue model, and the generating of 3 sets of differentiated packages by combination specifically comprises: screening the matched basic service and value-added service aiming at each homogenized customer group based on the customer-product bi-directional portrait and the scene revenue model, and combining to generate a personalized basic and value-added service package; The basic service is a necessary service, the value-added service is determined to be an optional service according to the electricity consumption type and the basic requirement of the customer, and the value-added service screening is carried out by combining the core requirement of the customer group and the measuring and calculating result of the scenic profit model; Each customer group corresponds to 3 sets of differentiated packages, and the package content marks service items, service periods, charging standards, expected benefits and customer benefits.
8. The method for mining and recommending value-added services based on potential demand of multi-source power data according to claim 1, wherein the model for revenue based on a revenue library and a scenerized revenue, measuring and calculating the direct benefit and the indirect benefit of each package, and comprehensively evaluating the service value by combining the client demand urgency degree and the consumption capability, wherein the generating of the quantitative score specifically comprises the following steps: Based on the profit library and the scenerized profit model, carrying out profit calculation on each generated set of basic plus added value service packages, wherein the calculation content comprises direct profit and indirect profit; The direct benefit is the net benefit after the charging income of the service package is deducted from the service cost, and the indirect benefit is the indirect economic benefit brought by improving the client retention rate, improving the power grid resource utilization rate and reducing the power grid load pressure through the service package; and comprehensively evaluating the service value of each package by combining the urgent degree of the client demands and the client consumption capability in the client portrait to generate a service value score.
9. The method for mining potential demands and recommending value-added services based on multi-source power data according to claim 1, wherein the establishing a dynamic ordering mechanism, monitoring the changes of the power consumption characteristics and the profit data of the clients in real time, setting a threshold value, and re-measuring package profit and value scores when the fluctuation of the power consumption or the profit data of the clients exceeds the threshold value, and adjusting the recommending sequence specifically comprises: Based on the profit measurement result and the service value score, combining the change of the real-time electricity utilization characteristics of the clients, establishing a dynamic ordering mechanism, and adjusting the recommendation priority of the service packages in real time; And acquiring power utilization data change data and income data update of the client in real time, establishing a priority update trigger mechanism, setting a trigger threshold, and performing priority adjustment when the power utilization characteristic change amplitude or the income data update amplitude of the client exceeds the threshold.

Description

Potential demand mining and value added service recommendation method based on multi-source power data Technical Field The invention relates to the field of demand service recommendation, in particular to a potential demand mining and value-added service recommendation method based on multi-source power data. Background With the deep advancement of energy structure transformation and electric power market innovation, an electric power system is faced with new challenges of supply-demand interaction refinement and service diversification. Traditional power data analysis focuses on load prediction and equipment monitoring, and user behavior patterns and potential requirements implicit in massive power consumption data cannot be fully mined. The popularization of novel power elements such as intelligent ammeter, distributed energy, electric automobile and the like enables multi-source power data to show explosive growth, but the data value is not released effectively yet. Most of the similar methods on the market lack system integration and standard preprocessing of multi-source heterogeneous power data, so that the problems of data redundancy, privacy disclosure or insufficient data quality are easy to occur, and the accurate requirement mining is difficult to support. Most methods do not establish a supporting model with strong cooperativity, the demands, the products and the benefits are disjointed, the whole flow coverage can not be realized, the grouping mode is single, the bidirectional adaptation of the clients and the products is not realized, and the service recommendation pertinence is insufficient. Meanwhile, a scene benefit quantification and dynamic adjustment mechanism is lacking, recommended packages are seriously homogenized, customer experience and enterprise benefit are difficult to be considered, the changes of customer electricity utilization characteristics and benefit data cannot be responded in real time, a recommendation strategy is stiff, practicability and operability are weak, and bidirectional value improvement of customers and power enterprises is difficult to be achieved. Disclosure of Invention In order to perfect the existing method, a potential demand mining and value-added service recommending method based on multi-source power data is provided, and the method is used for generating a differential package by integrating the multi-source power data and relying on a trinity model, a bidirectional grouping and a scene revenue model, and adjusting a recommending strategy in real time by combining a dynamic ordering mechanism, so that the customer demands and enterprise profits are accurately matched. In order to achieve the above purpose, the invention adopts the following technical scheme: The potential demand mining and value added service recommending method based on the multi-source power data comprises the following steps: Collecting multi-type heterogeneous data sources in an electric power system, and generating a standardized multi-source electric power data set through outlier rejection, missing value interpolation filling, standardized dimension and format processing and privacy data desensitization operation; Constructing a trinity value mining support model, wherein the trinity value mining support model consists of a demand library, a product library and a benefit library, and the three are mutually associated and cooperatively supported to carry out full-flow coverage of potential demand mining and value-added service recommendation; Adopting a data layer fusion strategy, matching electricity consumption, load and resource data based on customer numbers, overlapping interaction and external environment data, performing duplication elimination, complementation and conflict elimination, and generating a fusion data set for integrating electricity consumption characteristics of a customer side and resource conditions of a power grid side; Taking the fusion data set as input, carrying out bidirectional clustering based on a K-means clustering algorithm, carrying out client clustering, dividing an homogenized client group and generating a proprietary image, carrying out product clustering, dividing a product group and generating a product image, and obtaining the adaptation relation between different client groups and service products; based on the fusion data set and the two-way portrait, carrying out electric field scene classification, and quantifying the profit potential of various services in different scenes by constructing a scene profit model aiming at customer experience improvement and enterprise profit growth; screening the matched basic service and the value-added service for each homogeneous customer group based on the two-way portrait and the scene profit model, and combining to generate 3 sets of differential packages; Based on the profit library and the scenerized profit model, measuring and calculating the direct profit and the indirect profit of each set