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

Abstract

The application discloses a potential demand mining and value-added service recommendation method based on multi-source power data and relates to the demand service recommendation field, which comprises the following steps: forming a unified data set by integrating heterogeneous data and performing cleaning, standardization and desensitization processing; based on a demand-product-revenue trinity model, adopting a K-means algorithm to perform bidirectional clustering analysis on customers and products, establishing an adaptation relationship between customer portraits and product portraits; combining a scenario-based revenue model to quantify service potential, designing three sets of differentiated packages for homogeneous customer groups, and comprehensively evaluating demand urgency, consumption capacity and other factors to generate a value score; and establishing a dynamic sorting mechanism to automatically adjust the recommendation strategy when the electricity consumption characteristics or revenue fluctuation exceeds a threshold value. The application has the advantages that multi-source power data is integrated, differentiated packages are generated relying on the trinity model, bidirectional grouping and the scenario-based revenue model, the recommendation strategy is adjusted in combination with the dynamic sorting mechanism, and customer demand and enterprise revenue are matched.

Description

Technical Field

[0001] This invention relates to the field of demand service recommendation, and in particular to a method for mining potential demand and recommending value-added services based on multi-source power data. Background Technology

[0002] With the deepening of energy structure transformation and power market reform, the power system is facing new challenges such as more refined supply and demand interaction and diversified services. Traditional power data analysis focuses mainly on load forecasting and equipment monitoring, failing to fully explore the user behavior patterns and potential needs hidden in massive amounts of electricity consumption data. The popularization of new power elements such as smart meters, distributed energy, and electric vehicles has led to an explosive growth in multi-source power data, but the value of this data has not yet been effectively realized.

[0003] Most existing methods lack systematic integration and standardized preprocessing of multi-source heterogeneous power data, leading to issues such as data redundancy, privacy leaks, or insufficient data quality, making it difficult to support accurate demand mining. Most methods lack a robust supporting model, resulting in a disconnect between demand, products, and revenue, failing to achieve full-process coverage. Their segmentation methods are simplistic, failing to achieve bidirectional matching between customers and products, resulting in insufficiently targeted service recommendations. Furthermore, they lack scenario-based revenue quantification and dynamic adjustment mechanisms, leading to highly homogenized recommended packages that fail to balance customer experience and corporate profits. They also cannot respond in real-time to changes in customer electricity consumption characteristics and revenue data, resulting in rigid recommendation strategies with weak practicality and operability, hindering the achievement of mutual value enhancement for both customers and power companies. Summary of the Invention

[0004] To improve existing methods, this paper proposes a method for potential demand mining and value-added service recommendation based on multi-source power data. This method integrates multi-source power data, generates differentiated packages based on a three-in-one model, two-way clustering, and scenario-based revenue model, and adjusts the recommendation strategy in real time with a dynamic ranking mechanism to accurately match customer needs with enterprise profits.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] Methods for identifying potential demand and recommending value-added services based on multi-source power data include:

[0007] Collect data from various heterogeneous data sources in the power system, and generate a standardized multi-source power dataset through outlier removal, missing value interpolation and imputation, standardization of units and formats, and desensitization of privacy data.

[0008] A three-in-one value mining support model is constructed, which consists of a demand library, a product library, and a revenue library. The three are interconnected and support each other to cover the entire process of potential demand mining and value-added service recommendation.

[0009] A data layer fusion strategy is adopted to match electricity consumption, load and resource data based on customer account number, and to overlay interactive and external environment data for deduplication, completion and conflict elimination, to generate a fusion dataset that integrates customer-side electricity consumption characteristics and grid-side resource status.

[0010] Using the fused dataset as input, bidirectional clustering is performed based on the K-means clustering algorithm. Customer clustering divides homogeneous customer groups and generates exclusive profiles, while product clustering divides product clusters and generates product profiles, thereby obtaining the matching relationship between different customer groups and service products.

[0011] Based on the fusion dataset and two-way profile, electricity consumption scenarios are classified. By building a scenario-based revenue model aimed at improving customer experience and increasing corporate revenue, the revenue potential of various services in different scenarios is quantified.

[0012] Based on two-way profiling and scenario-based revenue models, we select suitable basic and value-added services for each homogeneous customer group and combine them to generate three differentiated packages.

[0013] Based on the revenue pool and scenario-based revenue model, the direct and indirect revenue of each package is calculated, and the service value is comprehensively evaluated by combining the urgency of customer needs and spending power, generating a quantitative score.

[0014] Establish a dynamic ranking mechanism to monitor changes in customer electricity consumption characteristics and revenue data in real time, and set thresholds. When customer electricity consumption or revenue data fluctuates beyond the threshold, recalculate the package revenue and value score, and adjust the recommendation order.

[0015] Preferably, the process of collecting data from multiple heterogeneous data sources in the power system, and generating a standardized multi-source power dataset through outlier removal, missing value interpolation and imputation, standardization of units and formats, and desensitization of privacy data, specifically includes:

[0016] Collect data from multiple heterogeneous data sources in the power system, including customer electricity consumption data, power grid load data, power grid resource data, customer interaction data, and external environment data.

[0017] The collected heterogeneous data sources are preprocessed, and outlier values ​​and noise data generated during data collection and transmission are identified and removed using outlier detection methods.

[0018] Interpolation is used to fill in missing data, and data of different formats and units are standardized to unify data formats and measurement standards.

[0019] By de-identifying data and hiding customer privacy information, a standardized multi-source power dataset is obtained.

[0020] Preferably, the construction of the three-in-one value mining support model, which consists of a demand database, a product database, and a revenue database, is interconnected and collaboratively supportive, providing full-process coverage for potential demand mining and value-added service recommendation. Specifically, this includes:

[0021] A demand database is built based on standardized multi-source power datasets. Through data correlation analysis, explicit and potential customer demands are mined, and the data is classified and stored according to electricity consumption type and demand type. The frequency of occurrence, correlation factors, and urgency of each type of demand are recorded.

[0022] Integrate basic power services and value-added services resources to build a product library, and label each type of service product with attributes, including service content, applicable customer groups, service costs, service cycle, and implementation conditions;

[0023] A revenue database is constructed by combining cost data, electricity consumption data, and market data of various service products in the product library, and the unit revenue, expected revenue, and revenue cycle of various service products are recorded.

[0024] Preferably, the data layer fusion strategy, which matches electricity consumption, load, and resource data based on customer account numbers, and overlays interactive and external environment data for deduplication, completion, and conflict elimination, to generate a fused dataset integrating customer-side electricity consumption characteristics and grid-side resource conditions, specifically includes:

[0025] A data layer fusion strategy is adopted to fuse the preprocessed standardized multi-source power dataset. By establishing unified data association rules, customer electricity consumption data is associated and matched with power grid load data and power grid resource data.

[0026] By matching customer electricity account numbers one-to-one, customer interaction data, external environment data and related data are overlaid and integrated to eliminate data redundancy and data conflicts.

[0027] By performing data deduplication and data completion operations, key and valid data are retained to generate a fusion dataset with a unified format and unified dimensions.

[0028] Preferably, the step of using the fused dataset as input, performing bidirectional clustering based on the K-means clustering algorithm, dividing homogeneous customer groups into customer groups and generating exclusive profiles, and dividing product groups into product clusters and generating product profiles, and obtaining the adaptation relationship between different customer groups and service products specifically includes:

[0029] Using the fused dataset as input, the K-means clustering algorithm is applied to segment customers and products respectively, and a two-way customer-product profile is constructed.

[0030] During the customer segmentation process, customer electricity load, electricity usage time, electricity capacity, electricity payment level, and interaction behavior are selected as clustering indicators to divide customers into several homogeneous customer groups and generate exclusive customer profiles.

[0031] During the product clustering process, the applicable scenarios, service costs, expected benefits, implementation difficulty, and customer suitability of service products are selected as clustering indicators. The K-means algorithm is used to divide the service products into several product clusters and generate product profiles.

[0032] By matching customer profiles with product profiles, we can obtain the compatibility between different customer groups and different product clusters.

[0033] Preferably, the electricity consumption scenario classification based on the fused dataset and bidirectional profile, and the quantification of the revenue potential of various services in different scenarios by building a scenario-based revenue model aimed at improving customer experience and increasing enterprise revenue, specifically includes:

[0034] Based on the fusion dataset and two-way profile, combined with load data, power grid resource data, and customer electricity consumption data, different electricity consumption scenarios are divided.

[0035] For each electricity consumption scenario, a scenario-based revenue model is constructed by combining the load fluctuation characteristics, grid resource utilization rate, and customer electricity demand within the scenario.

[0036] The model aims to improve the customer's electricity experience and increase the revenue of power companies. It quantifies the revenue potential of different service products in various scenarios, correlates the revenue of customer electricity consumption characteristics with that of service products, and obtains the adaptability and revenue level of various service products in different scenarios.

[0037] Preferably, the step of selecting suitable basic and value-added services for each homogeneous customer group based on two-way profiling and scenario-based revenue models, and combining them to generate three differentiated packages, specifically includes:

[0038] Based on customer-product dual profiles and scenario-based revenue models, suitable basic and value-added services are selected for each homogeneous customer group, and personalized "basic + value-added" service packages are generated.

[0039] The basic services are mandatory and are determined based on the customer's electricity usage type and basic needs. The value-added services are optional and are selected based on the core needs of the customer group and the calculation results of the scenario-based revenue model.

[0040] Each customer group has three differentiated packages, with each package detailing the services offered, service period, pricing, expected returns, and customer benefits.

[0041] Preferably, the calculation of direct and indirect revenue for each package based on the revenue pool and scenario-based revenue model, combined with a comprehensive assessment of service value considering customer needs and spending power, to generate a quantitative score specifically includes:

[0042] Based on the revenue library and scenario-based revenue model, the revenue of each generated "basic + value-added" service package is calculated, including direct revenue and indirect revenue.

[0043] The direct revenue is the net profit after deducting service costs from the service package fee revenue, while the indirect revenue is the indirect economic benefits brought about by improving customer retention rate, increasing power grid resource utilization rate, and reducing power grid load pressure through the service package.

[0044] By combining the urgency of customer needs and their spending power in the customer profile, the service value of each package is comprehensively evaluated, and a service value score is generated.

[0045] Preferably, the establishment of a dynamic ranking mechanism, which monitors changes in customer electricity consumption characteristics and revenue data in real time, and sets thresholds, involves recalculating package revenue and value scores and adjusting the recommendation order when customer electricity consumption or revenue data fluctuations exceed the thresholds. Specifically, this includes:

[0046] Based on revenue calculation results and service value scores, and combined with changes in customers' real-time electricity consumption characteristics, a dynamic ranking mechanism is established to adjust the recommendation priority of service packages in real time.

[0047] Real-time collection of customer electricity consumption data and revenue data updates; establishment of a priority update trigger mechanism; setting trigger thresholds; priority adjustment when the magnitude of changes in customer electricity consumption characteristics or revenue data updates exceeds the thresholds.

[0048] Compared with the prior art, the advantages of the present invention are:

[0049] This system comprehensively integrates diverse and heterogeneous power data, ensuring data quality and privacy through standardized preprocessing and anonymization, while also laying a solid foundation for demand analysis. Leveraging a three-in-one model encompassing a demand database, product database, and revenue database, it achieves end-to-end collaborative support, addressing the pain points of fragmented data and a disconnect between demand and products. Through data layer fusion and K-means bidirectional clustering, it accurately matches customers with service products, quantifies revenue potential using a scenario-based revenue model, and generates three differentiated packages that balance personalized customer needs with corporate profits. Simultaneously, through quantitative scoring and dynamic ranking mechanisms, it responds in real-time to changes in customer electricity consumption and revenue data, adjusting recommendation strategies promptly to effectively improve the accuracy, flexibility, and effectiveness of service recommendations. This achieves the dual goals of enhanced customer experience and increased power company revenue, demonstrating strong practicality and operability. Attached Figure Description

[0050] Figure 1 This is a schematic diagram of the method proposed in this invention;

[0051] Figure 2 This is a schematic diagram of the power data acquisition and preprocessing proposed in this invention;

[0052] Figure 3 This is a schematic diagram of the three-in-one value mining support model proposed in this invention;

[0053] Figure 4 This is a schematic diagram of the multi-source data fusion processing proposed in this invention;

[0054] Figure 5 This is a schematic diagram of the customer-product bidirectional profiling and grouping proposed in this invention;

[0055] Figure 6 This is a schematic diagram illustrating the construction of a scenario-based revenue model proposed in this invention;

[0056] Figure 7 This is a schematic diagram of the service package generation proposed in this invention;

[0057] Figure 8 This is a schematic diagram illustrating the revenue calculation and service value assessment proposed in this invention;

[0058] Figure 9 This is a schematic diagram illustrating the service recommendation priority adjustment proposed in this invention. Detailed Implementation

[0059] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0060] See Figure 1 As shown, the method for potential demand mining and value-added service recommendation based on multi-source power data includes:

[0061] Step 1: Collect data from multiple heterogeneous data sources in the power system, and generate a standardized multi-source power dataset through outlier removal, missing value interpolation and imputation, standardization of units and formats, and desensitization of privacy data;

[0062] Step 2: Construct a three-in-one value mining support model, which consists of a demand library, a product library, and a revenue library. The three are interconnected and support each other to cover the entire process of potential demand mining and value-added service recommendation.

[0063] Step 3: Adopt a data layer fusion strategy to match electricity consumption, load, and resource data based on customer account number, and overlay interactive and external environment data for deduplication, completion, and conflict elimination to generate a fusion dataset that integrates customer-side electricity consumption characteristics and grid-side resource status.

[0064] Step 4: Using the fused dataset as input, perform bidirectional clustering based on the K-means clustering algorithm. Customer clustering divides homogeneous customer groups and generates exclusive profiles. Product clustering divides product clusters and generates product profiles, obtaining the matching relationship between different customer groups and service products.

[0065] Step 5: Classify electricity usage scenarios based on the fused dataset and bidirectional profiles. Quantify the revenue potential of various services in different scenarios by building a scenario-based revenue model aimed at improving customer experience and increasing corporate revenue.

[0066] Step Six: Based on two-way profiling and scenario-based revenue models, select suitable basic and value-added services for each homogeneous customer group and combine them to generate three differentiated packages;

[0067] Step 7: Based on the revenue pool and scenario-based revenue model, calculate the direct and indirect revenue of each package, and comprehensively evaluate the service value by combining the urgency of customer needs and spending power, and generate a quantitative score.

[0068] Step 8: Establish a dynamic ranking mechanism to monitor changes in customer electricity consumption characteristics and revenue data in real time, and set thresholds. When customer electricity consumption or revenue data fluctuates beyond the threshold, recalculate the package revenue and value score, and adjust the recommendation order.

[0069] See Figure 2 As shown, a standardized multi-source power dataset is generated by collecting data from multiple heterogeneous data sources in the power system. This is achieved through outlier removal, missing value interpolation and imputation, standardization of units and formats, and desensitization of privacy data. Specifically, this includes:

[0070] Collect data from multiple heterogeneous data sources in the power system, including customer electricity consumption data, power grid load data, power grid resource data, customer interaction data, and external environment data.

[0071] The collected heterogeneous data sources are preprocessed, and outlier values ​​and noise data generated during data collection and transmission are identified and removed using outlier detection methods.

[0072] Interpolation is used to fill in missing data, and data of different formats and units are standardized to unify data formats and measurement standards.

[0073] By de-identifying data and hiding customer privacy information, a standardized multi-source power dataset is obtained.

[0074] Specifically, outlier and noise data processing is performed. Outliers are identified using the 3σ principle combined with box plots. Data exceeding the normal range and noise generated during transmission are marked and removed after manual verification. Next, missing data is imputed. For continuous data, linear interpolation is used to accurately calculate missing values ​​based on the changing trends of adjacent valid data. For discrete data, mode imputation is used, selecting the most frequent value to fill in missing values, ensuring data integrity. Then, data standardization is performed, using normalization methods to convert data of different formats and dimensions to the 0-1 range. Electricity capacity and load data are normalized to the maximum value, while electricity bill data is normalized to the industry average level, eliminating the impact of dimension differences. The data format is unified to JSON. Finally, data anonymization is performed, using hash encryption to encrypt customer names, ID numbers, contact numbers, and other private information. Electricity account numbers are partially hidden by character replacement, while preserving data relevance and usability. The result is a clean, standardized, and secure multi-source power dataset.

[0075] See Figure 3 As shown, a three-in-one value mining support model is constructed. This model consists of a demand database, a product database, and a revenue database. These three databases are interconnected and support each other, providing full-process coverage for potential demand mining and value-added service recommendation. Specifically, this includes:

[0076] A demand database is built based on standardized multi-source power datasets. Through data correlation analysis, explicit and potential customer demands are mined, and the data is classified and stored according to electricity consumption type and demand type. The frequency of occurrence, correlation factors, and urgency of each type of demand are recorded.

[0077] Integrate basic power services and value-added services resources to build a product library, and label each type of service product with attributes, including service content, applicable customer groups, service costs, service cycle, and implementation conditions;

[0078] A revenue database is constructed by combining cost data, electricity consumption data, and market data of various service products in the product library, and the unit revenue, expected revenue, and revenue cycle of various service products are recorded.

[0079] Specifically, the demand database is constructed using standardized multi-source power datasets as the core data source. Through data correlation analysis, combined with customer electricity consumption behavior, interaction records, and external environmental data, it comprehensively mines both explicit and potential customer needs. Explicit needs are extracted through direct customer interaction data; potential needs are mined through correlation analysis. After mining, the database is categorized and stored according to two dimensions: first, the electricity consumption type dimension, clearly distinguishing between three major categories: residential electricity consumption, industrial electricity consumption, and commercial electricity consumption, with each category further subdivided into specific scenarios; second, the demand type dimension, dividing the database into five major categories: basic electricity guarantee needs, energy-saving needs, energy storage needs, electricity consumption optimization needs, and customized service needs. Core information is labeled for each type of demand, forming a structured demand database.

[0080] The product database comprehensively integrates the existing basic and value-added service resources of power companies to form a complete service product system. Each type of service product is labeled with detailed attributes. Basic services are clearly divided into three categories: stable power supply, electricity bill inquiry, and fault repair. Value-added services include six categories: load optimization, energy-saving renovation, energy storage leasing, distributed power source access, electricity consultation, and customized electricity pricing packages. The product database adopts a dynamic update mechanism, regularly adding new service products and updating the attributes of existing products.

[0081] The revenue database is constructed by combining service cost data from the product database, standardized electricity consumption data, and current electricity market data to establish a comprehensive revenue calculation system. It collects detailed cost information for various service products and clarifies unit service costs. Based on customer electricity consumption data, it calculates the revenue generated from different customer groups using various service products. Finally, by deducting costs from revenue, it obtains the unit revenue, expected revenue, and revenue cycle for each service product, establishing a correlation between customer electricity consumption characteristics and revenue. For example, high-capacity customers receive higher revenue from energy storage leasing services than ordinary customers. It records the revenue differences for different customer groups across various service products, forming a structured revenue database.

[0082] See Figure 4 As shown, a data layer fusion strategy is adopted, which matches electricity consumption, load, and resource data based on customer account numbers, and overlays interactive and external environment data for deduplication, completion, and conflict elimination to generate a fused dataset that integrates customer-side electricity consumption characteristics and grid-side resource conditions. Specifically, it includes:

[0083] A data layer fusion strategy is adopted to fuse the preprocessed standardized multi-source power dataset. By establishing unified data association rules, customer electricity consumption data is associated and matched with power grid load data and power grid resource data.

[0084] By matching customer electricity account numbers one-to-one, customer interaction data, external environment data and related data are overlaid and integrated to eliminate data redundancy and data conflicts.

[0085] By performing data deduplication and data completion operations, key and valid data are retained to generate a fusion dataset with a unified format and unified dimensions.

[0086] Specifically, a layered fusion approach is adopted, first completing the fusion of core data, and then overlaying auxiliary data. The first step involves linking and integrating basic customer electricity consumption data with grid load data and grid resource data. Seamless connection of the three types of data is achieved through account number matching, integrating the electricity consumption characteristics of the same customer, the load status of the corresponding grid node, and resource utilization into a set of associated data, marking the data correspondence and collection timestamp to ensure the consistency of data time sequence. The second step involves overlaying customer interaction data into the above associated data, linking customer service consultation, fault reporting, and value-added service interaction records with customer electricity consumption data and grid data by account number, and supplementing information related to customer demand preferences. The third step involves overlaying and integrating external environmental data, linking seasonal, temperature, and industry policy data with grid load data and customer electricity consumption data of the corresponding region by region, clarifying the impact of the external environment on electricity consumption behavior and grid load.

[0087] See Figure 5 As shown, using a fused dataset as input, bidirectional clustering is performed based on the K-means clustering algorithm. Customer clustering divides homogeneous customer groups and generates exclusive profiles, while product clustering divides product clusters and generates product profiles. The specific steps to obtain the compatibility relationship between different customer groups and service products include:

[0088] Using the fused dataset as input, the K-means clustering algorithm is applied to segment customers and products respectively, and a two-way customer-product profile is constructed.

[0089] During the customer segmentation process, customer electricity load, electricity usage time, electricity capacity, electricity payment level, and interaction behavior are selected as clustering indicators to divide customers into several homogeneous customer groups and generate exclusive customer profiles.

[0090] During the product clustering process, the applicable scenarios, service costs, expected benefits, implementation difficulty, and customer suitability of service products are selected as clustering indicators. The K-means algorithm is used to divide the service products into several product clusters and generate product profiles.

[0091] By matching customer profiles with product profiles, we can obtain the compatibility between different customer groups and different product clusters.

[0092] Specifically, customer segmentation employs the K-means clustering algorithm to perform cluster analysis on the screened customer-related data. First, customer electricity load, electricity usage time, electricity capacity, electricity payment level, demand tendency, and interaction behavior are selected as core clustering indicators, with clear weight allocation for each indicator. Then, an iterative clustering process is initiated, initially setting a certain number of clusters. Through multiple iterations, the similarity between each customer's data and the cluster centers is calculated, continuously adjusting the cluster center positions until the clustering results stabilize, the similarity of customer characteristics within the same group is highest, and the differences in characteristics between different groups are most significant. This determines the optimal number of clusters. Finally, customers are divided into several homogeneous customer groups, each corresponding to specific electricity usage characteristics and demand preferences. A unique customer profile is generated for each customer group, detailing the group's core needs, electricity usage habits, and revenue potential.

[0093] The product segmentation implementation employs the K-means clustering algorithm to perform cluster analysis on service product-related data. Applicable scenarios, service costs, expected returns, implementation difficulty, and customer suitability are selected as clustering indicators, with customer suitability determined by combining information on the applicable customer groups of products in the product library. Similarly, through multiple rounds of iterative optimization, the optimal number of product clusters is determined, dividing all service products into several product clusters, each cluster corresponding to a suitable customer group. A product profile is generated for each product cluster, detailing the product's core advantages, suitable customer characteristics, and return levels.

[0094] The formula for the clustering objective function of MESS is:

[0095]

[0096] in, This represents the total clustering error. Let be the number of clusters, be the j-th cluster, and be the n-dimensional feature vectors of customers and products. Let j be the cluster center vector. Let be the distance between the i-th sample and the j-th cluster center;

[0097] Establish a matching mechanism between customer profiles and product profiles. By comparing the core attributes of the two, clarify the compatibility between different customer groups and different product clusters. For example, match the profile of high-energy-consuming industrial customers with the profile of energy-saving renovation and energy storage leasing product clusters and mark the compatibility between the two; match the profile of residential peak and off-peak electricity users with the profile of customized electricity pricing and electricity consulting product clusters and clarify the compatibility priority.

[0098] See Figure 6As shown, electricity consumption scenarios are classified based on fused datasets and bidirectional profiles. By building a scenario-based revenue model aimed at improving customer experience and increasing enterprise revenue, the revenue potential of various services in different scenarios is quantified, specifically including:

[0099] Based on the fusion dataset and two-way profile, combined with load data, power grid resource data, and customer electricity consumption data, different electricity consumption scenarios are divided.

[0100] For each electricity consumption scenario, a scenario-based revenue model is constructed by combining the load fluctuation characteristics, grid resource utilization rate, and customer electricity demand within the scenario.

[0101] The model aims to improve the customer's electricity experience and increase the revenue of power companies. It quantifies the revenue potential of different service products in various scenarios, correlates the revenue of customer electricity consumption characteristics with that of service products, and obtains the adaptability and revenue level of various service products in different scenarios.

[0102] Specifically, for each segmented electricity consumption scenario, core data is extracted from the fused dataset. This core data includes three categories: first, load data, focusing on extracting real-time load, historical load fluctuation curves, peak and valley loads, and their corresponding durations for analyzing the changing patterns of electricity load within the scenario; second, grid resource data, extracting line capacity, transformer utilization, and remaining energy storage capacity for the corresponding power supply area for assessing the carrying capacity and utilization efficiency of grid resources; and third, customer electricity consumption data, extracting electricity capacity, electricity consumption time distribution, and demand preferences for the corresponding customer group within the scenario, combined with customer profiles to clarify the core needs and electricity consumption habits of customers in that scenario.

[0103] For each electricity consumption scenario, a dedicated scenario-based revenue model is constructed, with the dual objectives of improving customer electricity consumption experience and increasing power company revenue. The core components and calculation logic of the model are clearly defined. During the model construction process, the revenue potential of different service products in the scenario is quantified by combining the load fluctuation characteristics, grid resource utilization rate, and customer electricity demand: For peak load scenarios, the indirect benefits of reducing grid load pressure through services such as energy storage leasing and load optimization are emphasized; for industrial production scenarios, the energy-saving and consumption-reducing benefits of services such as energy-saving renovation and distributed power source access, as well as the benefits of enterprise electricity cost savings, are emphasized; for residential daily electricity consumption scenarios, the customer retention benefits and service fee revenue of services such as customized electricity pricing and electricity consultation are emphasized, establishing the correlation between customer electricity consumption characteristics and service product revenue, and clarifying the revenue differences of different customer groups using various service products in the scenario.

[0104] See Figure 7As shown, based on two-way profiling and a scenario-based revenue model, suitable basic and value-added services are selected for each homogeneous customer group, and three differentiated packages are generated, specifically including:

[0105] Based on customer-product dual profiles and scenario-based revenue models, suitable basic and value-added services are selected for each homogeneous customer group, and personalized "basic + value-added" service packages are generated.

[0106] The basic services are mandatory and are determined based on the customer's electricity usage type and basic needs. The value-added services are optional and are selected based on the core needs of the customer group and the calculation results of the scenario-based revenue model.

[0107] Each customer group has three differentiated packages, with each package detailing the services offered, service period, pricing, expected returns, and customer benefits.

[0108] See Figure 8 As shown, based on the revenue pool and scenario-based revenue model, the direct and indirect revenue of each package is calculated. The service value is comprehensively evaluated by considering the urgency of customer needs and their spending power, generating a quantitative score that includes:

[0109] Based on the revenue library and scenario-based revenue model, the revenue of each generated "basic + value-added" service package is calculated, including direct revenue and indirect revenue.

[0110] The direct revenue is the net profit after deducting service costs from the service package fee revenue, while the indirect revenue is the indirect economic benefits brought about by improving customer retention rate, increasing power grid resource utilization rate, and reducing power grid load pressure through the service package.

[0111] By combining the urgency of customer needs and their spending power in the customer profile, the service value of each package is comprehensively evaluated, and a service value score is generated.

[0112] Specifically, the core of direct revenue calculation is to calculate the net profit of the service package. First, calculate the package fee revenue. Based on the marked package fee standard, combined with the package service period and the number of suitable customer groups, calculate the total fee revenue of a single package over the entire service period. Then, collect the package service costs. Combining the service product cost data in the revenue database, collect the labor costs, equipment costs, and operating costs of each service included in the package, according to the basic services and value-added services. Finally, subtract the total service cost from the total fee revenue to obtain the direct revenue of each package.

[0113] Indirect revenue calculation focuses on the long-term, implicit economic benefits brought by the packages. Combining the needs of power grid operation and customer management, two core types of indirect revenue are identified and quantified: First, customer retention revenue. Based on customer profiles, the long-term benefits of improved customer satisfaction and retention rates from the packages are calculated. By comparing the retention rates of customers who do not use the packages with those who do, the increase in subsequent service revenue resulting from improved retention rates is quantified. Second, power grid operation revenue. Using scenario-based revenue models, the indirect benefits of packages reducing power grid load pressure and improving power grid resource utilization are calculated. Energy storage leasing packages improve the utilization rate of energy storage equipment and reduce power grid investment costs. These implicit benefits are transformed into quantifiable economic value, and the indirect revenue of each package is summarized.

[0114] See Figure 9 As shown, a dynamic ranking mechanism is established to monitor changes in customer electricity consumption characteristics and revenue data in real time. Thresholds are set, and when customer electricity consumption or revenue data fluctuations exceed these thresholds, the package revenue and value score are recalculated, and the recommendation order is adjusted. Specifically, this includes:

[0115] Based on revenue calculation results and service value scores, and combined with changes in customers' real-time electricity consumption characteristics, a dynamic ranking mechanism is established to adjust the recommendation priority of service packages in real time.

[0116] Real-time collection of customer electricity consumption data and revenue data updates; establishment of a priority update trigger mechanism; setting trigger thresholds; priority adjustment when the magnitude of changes in customer electricity consumption characteristics or revenue data updates exceeds the thresholds.

[0117] Specifically, based on the initial baseline data, a multi-dimensional dynamic ranking mechanism is established. The core ranking indicators include service value score, customer real-time electricity adaptability, and revenue data update status. The ranking mechanism adopts a hierarchical ranking logic, first ranking by service value score, and then fine-tuning the order by combining real-time adaptability and revenue update status.

[0118] Establish a priority update trigger mechanism, combining historical data and industry standards to set reasonable trigger thresholds. When the change in a customer's electricity consumption characteristics or the update of revenue data exceeds the set threshold, the priority adjustment process is immediately initiated. During the adjustment process, the service value score of the corresponding package is recalculated and re-ranked according to a dynamic sorting mechanism, prioritizing packages with high service value scores that are suitable for the customer's current electricity needs.

[0119] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0120] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0121] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for potential demand mining and value-added service recommendation based on multi-source power data, characterized in that, include: Collect data from various heterogeneous data sources in the power system, and generate a standardized multi-source power dataset through outlier removal, missing value interpolation and imputation, standardization of units and formats, and desensitization of privacy data. A three-in-one value mining support model is constructed, consisting of a demand database, a product database, and a revenue database. These three databases are interconnected and supportive, providing full-process coverage for potential demand mining and value-added service recommendation. Specifically, this includes: constructing a demand database based on standardized multi-source power datasets; mining explicit and potential customer needs through data correlation analysis; classifying and storing these needs according to electricity consumption type and demand type; and recording the frequency of occurrence, related factors, and urgency of each type of demand. A product database is constructed by integrating basic power services and value-added services resources; each service product is labeled with attributes including service content, applicable customer groups, service cost, service cycle, and implementation conditions. A revenue database is constructed by combining cost data, electricity consumption data, and market data of various service products in the product database; and recording the unit revenue, expected revenue, and revenue cycle of each service product. A data layer fusion strategy is adopted to match electricity consumption, load and resource data based on customer account number, and to overlay interactive and external environment data for deduplication, completion and conflict elimination, to generate a fusion dataset that integrates customer-side electricity consumption characteristics and grid-side resource status. Using the fused dataset as input, bidirectional clustering is performed based on the K-means clustering algorithm. Customer clustering divides homogeneous customer groups and generates exclusive profiles, while product clustering divides product clusters and generates product profiles, thereby obtaining the matching relationship between different customer groups and service products. Based on the fusion dataset and two-way profile, electricity consumption scenarios are classified. By building a scenario-based revenue model aimed at improving customer experience and increasing corporate revenue, the revenue potential of various services in different scenarios is quantified. Based on two-way profiling and scenario-based revenue models, we select suitable basic and value-added services for each homogeneous customer group and combine them to generate three differentiated packages. Based on the revenue pool and scenario-based revenue model, the direct and indirect revenue of each package is calculated, and the service value is comprehensively evaluated by combining the urgency of customer needs and spending power, generating a quantitative score. Establish a dynamic ranking mechanism to monitor changes in customer electricity consumption characteristics and revenue data in real time, and set thresholds. When customer electricity consumption or revenue data fluctuates beyond the threshold, recalculate the package revenue and value score, and adjust the recommendation order.

2. The method for potential demand mining and value-added service recommendation based on multi-source power data according to claim 1, characterized in that, The process of collecting data from multiple heterogeneous data sources in the power system, including outlier removal, missing value interpolation and imputation, standardization of units and formats, and desensitization of privacy data, generates a standardized multi-source power dataset. Specifically, this includes: Collect data from multiple heterogeneous data sources in the power system, including customer electricity consumption data, power grid load data, power grid resource data, customer interaction data, and external environment data. The collected heterogeneous data sources are preprocessed, and outlier values ​​and noise data generated during data collection and transmission are identified and removed using outlier detection methods. Interpolation is used to fill in missing data, and data of different formats and units are standardized to unify data formats and measurement standards. By de-identifying data and hiding customer privacy information, a standardized multi-source power dataset is obtained.

3. The method for potential demand mining and value-added service recommendation based on multi-source power data according to claim 1, characterized in that, The aforementioned data layer fusion strategy, which matches electricity consumption, load, and resource data based on customer account numbers, and overlays interactive and external environment data for deduplication, completion, and conflict elimination, generates a fused dataset integrating customer-side electricity consumption characteristics and grid-side resource conditions. Specifically, this includes: A data layer fusion strategy is adopted to fuse the preprocessed standardized multi-source power dataset. By establishing unified data association rules, customer electricity consumption data is associated and matched with power grid load data and power grid resource data. By matching customer electricity account numbers one-to-one, customer interaction data, external environment data and related data are overlaid and integrated to eliminate data redundancy and data conflicts. By performing data deduplication and data completion operations, key and valid data are retained to generate a fusion dataset with a unified format and unified dimensions.

4. The method for potential demand mining and value-added service recommendation based on multi-source power data according to claim 1, characterized in that, The process of using a fused dataset as input, performing bidirectional clustering based on the K-means clustering algorithm, segmenting customers into homogeneous customer groups and generating unique profiles, and segmenting products into product clusters and generating product profiles, and obtaining the compatibility relationship between different customer groups and service products specifically includes: Using the fused dataset as input, the K-means clustering algorithm is applied to segment customers and products respectively, and a two-way customer-product profile is constructed. During the customer segmentation process, customer electricity load, electricity usage time, electricity capacity, electricity payment level, and interaction behavior are selected as clustering indicators to divide customers into several homogeneous customer groups and generate exclusive customer profiles. During the product clustering process, the applicable scenarios, service costs, expected benefits, implementation difficulty, and customer suitability of service products are selected as clustering indicators. The K-means algorithm is used to divide the service products into several product clusters and generate product profiles. By matching customer profiles with product profiles, we can obtain the compatibility between different customer groups and different product clusters.

5. The method for potential demand mining and value-added service recommendation based on multi-source power data according to claim 1, characterized in that, The electricity consumption scenario classification based on fused datasets and bidirectional profiles, and the construction of a scenario-based revenue model aimed at improving customer experience and increasing enterprise revenue, quantify the revenue potential of various services in different scenarios, specifically including: Based on the fusion dataset and two-way profile, combined with load data, power grid resource data, and customer electricity consumption data, different electricity consumption scenarios are divided. For each electricity consumption scenario, a scenario-based revenue model is constructed by combining the load fluctuation characteristics, grid resource utilization rate, and customer electricity demand within the scenario. The model aims to improve the customer's electricity experience and increase the revenue of power companies. It quantifies the revenue potential of different service products in various scenarios, correlates the revenue of customer electricity consumption characteristics with that of service products, and obtains the adaptability and revenue level of various service products in different scenarios.

6. The method for potential demand mining and value-added service recommendation based on multi-source power data according to claim 1, characterized in that, The method, based on two-way profiling and a scenario-based revenue model, selects suitable basic and value-added services for various homogeneous customer groups, and combines them to generate three differentiated packages, specifically including: Based on customer-product dual profiles and scenario-based revenue models, suitable basic and value-added services are selected for each homogeneous customer group, and personalized "basic + value-added" service packages are generated. The basic services are mandatory and are determined based on the customer's electricity usage type and basic needs. The value-added services are optional and are selected based on the core needs of the customer group and the calculation results of the scenario-based revenue model. Each customer group has three differentiated packages, with each package detailing the services offered, service period, pricing, expected returns, and customer benefits.

7. The method for potential demand mining and value-added service recommendation based on multi-source power data according to claim 1, characterized in that, The method, based on a revenue pool and a scenario-based revenue model, calculates the direct and indirect revenue of each package. It then comprehensively assesses the service value by considering the urgency of customer needs and their spending power, generating a quantitative score. Specifically, this includes: Based on the revenue pool and scenario-based revenue model, the revenue of each generated "basic + value-added" service package is calculated, including direct and indirect revenue. The direct revenue is the net profit after deducting service costs from the service package fee revenue, while the indirect revenue is the indirect economic benefits brought about by improving customer retention rate, increasing power grid resource utilization rate, and reducing power grid load pressure through the service package. By combining the urgency of customer needs and their spending power in the customer profile, the service value of each package is comprehensively evaluated, and a service value score is generated.

8. The method for potential demand mining and value-added service recommendation based on multi-source power data according to claim 1, characterized in that, The establishment of a dynamic ranking mechanism, which monitors customer electricity consumption characteristics and revenue data changes in real time, and sets thresholds, involves recalculating package revenue and value scores and adjusting the recommendation order when customer electricity consumption or revenue data fluctuations exceed the thresholds. Specifically, this includes: Based on revenue calculation results and service value scores, and combined with changes in customers' real-time electricity consumption characteristics, a dynamic ranking mechanism is established to adjust the recommendation priority of service packages in real time. Real-time collection of customer electricity consumption data and revenue data updates; establishment of a priority update trigger mechanism; setting trigger thresholds; priority adjustment when the magnitude of changes in customer electricity consumption characteristics or revenue data updates exceeds the thresholds.

Images