Power supply service data processing method and system under complex constraints and electronic equipment

Abstract

The application discloses a power supply service data processing method and system under complex constraints and electronic equipment, comprising: acquiring power supply service data of different sources, analyzing the internal correlation of the power supply service data, and determining a power supply service data correlation relationship network topology graph; constructing a power supply service data fusion model, and performing fusion of the power supply service data based on the power supply service data fusion model to determine fused data; determining business constraint rules and data constraint rules based on the power supply service data correlation relationship network topology graph, performing quality evaluation on the fused data, determining problem data that does not satisfy the business constraint rules and the data constraint rules, and processing the problem data to perform data intelligent checking. The application adopts a graph structure clustering method to gather and represent data with strong internal dependence, can check data problems, simultaneously trace back and locate a problem data source, and improves source system data quality.

Description

Technical Field

[0001] This invention relates to the field of power system information application technology, and more specifically, to a power supply service data processing method, system, and electronic device under complex constraints. Background Technology

[0002] Power supply service data is distributed across multiple business systems, each with its own data definitions and descriptions. Daily power supply service operations require switching between these systems to collect information and data and process transactions. The data distribution across these systems exhibits a pattern of high dispersion and low cohesion, with business application data entities tightly linked through business relationships. In emergency situations, the efficiency of handling sudden events is low due to the discrete nature of the business data, hindering effective and transparent emergency response. Simultaneously, data from different business systems exhibits inherent dependencies, working together to serve the scheduling and application of power supply service resources. This paper calculates the connection weights between data entities based on business relationships and constructs a weighted association topology graph based on these relationships. The research focuses on constructing power supply service data association relationships based on inherent dependencies. It analyzes the inherent relationships of power supply service data using graph clustering algorithms, reduces the dimensionality of the data entity association topology graph using metadata as the basic unit, and forms an easily searchable and traversable power supply service data association network topology graph. This provides data topology support for rapid extraction and transformation of power supply service data, improving the efficiency of cross-system data flow.

[0003] The power supply command center's business involves multiple professional fields such as marketing, operation and maintenance, and dispatching. Each professional field defines and describes the organization of data and business constraint rules for its professional business systems according to its own professional needs and business development patterns. However, this also leads to certain fragmentation and differences between the various business systems. Especially in the daily operation of power supply services, business personnel and data application personnel need to frequently switch between multiple business systems to complete business processing, information collection, and data writing. In emergency situations, this complex data acquisition and cross-model data analysis mode leads to low efficiency in handling emergencies and makes it impossible to achieve effective and transparent emergency command. At the same time, due to the significant dispersion of functions and data in the cross-business systems involved in power supply services, the effectiveness and transparency of power supply service emergency command are also significantly affected. The daily operation of power supply services requires switching between multiple business systems to collect information and data to complete business processing. To address this issue, this paper studies a multi-dimensional and multi-morphic power supply service data model fusion technology for business consistency. The power supply service business is sorted out to form a business description and define a business consistency constraint rule base. A unified data model is defined and designed with the goal of ensuring business support integrity. Data with the same business application is identified and integrated into the data model of that business object. A support vector machine model is used to learn the data model characteristics of different businesses. Based on the learning results, unidentifiable and unidentifiable data models are identified and classified. Based on the classification results, business identifiers are mapped to them to achieve model fusion.

[0004] The analysis and judgment results of power supply service operations depend on the accuracy, timeliness, and completeness of its supporting data. Inaccurate or incomplete basic supporting data can lead to manual intervention or erroneous instructions, significantly impacting normal business operations and processing efficiency. The accuracy, timeliness, and completeness of power supply service data are constrained by both business operations and data requirements; therefore, the verification of power supply service data must meet both constraints.

[0005] Traditional data verification methods mainly rely on manual review and simple rule checks, which suffer from low efficiency, low accuracy, and poor real-time performance. Especially in the context of the new power system, the large-scale integration of new energy sources and the in-depth advancement of power market reforms have placed higher demands on the accuracy, consistency, and timeliness of power supply service data.

[0006] The intelligent processing method for power supply service data under complex constraints integrates business rule constraints and data characteristic constraints to achieve intelligent identification, automatic verification, and quality improvement of multi-dimensional and multi-form power supply service data.

[0007] Business constraints: This refers to the constraints and verification of data based on the business rules and procedures of power supply services. For example, in the business expansion application process, the user's application capacity, electricity address, and other information must comply with relevant business regulations; in the electricity billing process, the electricity bill calculation results must match information such as electricity pricing policies and electricity consumption.

[0008] Data constraints: These are constraints imposed on the characteristics and quality of the data itself, including its completeness, accuracy, and consistency. For example, data cannot contain missing or erroneous values, and data from different data sources must be consistent.

[0009] Business and data constraints are not independent entities, but rather an organic whole that influences and interacts with each other. Business constraints provide the business logic foundation for data constraints, while data constraints provide the technical support for the implementation of business constraints.

[0010] Therefore, a method for processing power supply service data under complex constraints is needed. Summary of the Invention

[0011] This invention proposes a method, system, and electronic device for processing power supply service data under complex constraints, in order to solve the problem of how to efficiently process power supply service data.

[0012] To address the aforementioned problems, according to one aspect of the present invention, a method for processing power supply service data under complex constraints is provided, the method comprising: Obtain power supply service data from different sources, analyze the inherent correlation of the power supply service data, and determine the network topology of the power supply service data correlation. A power supply service data fusion model is constructed, and power supply service data is fused based on the power supply service data fusion model to determine the fused data; Based on the network topology diagram of the power supply service data association, business constraint rules and data constraint rules are determined. The quality of the fused data is assessed to identify problematic data that does not meet the business constraint rules and data constraint rules. The problematic data is then processed for intelligent data verification.

[0013] Preferably, the analysis of the inherent correlation of the power supply service data to determine the network topology of the power supply service data correlation includes: Based on the associated topology and electrical connection relationships of the devices, the connection weight between devices is characterized by the consistency of the number of device connections and operating status; A data association weight matrix is ​​constructed based on the connection weights, and a device connection weight topology graph is constructed based on the data association weight matrix. Based on the device connection weight topology map, a spectral clustering algorithm is used to cluster the devices, and the device connection weight topology map is divided according to the clustering results to determine the network topology map of power supply service data association.

[0014] Preferably, the power supply service data is fused based on the power supply service data fusion model to determine the fused data, including: Determine the data fusion constraint paradigm; The unstructured data in the power supply service data is mapped to structured data; wherein, natural language processing algorithms are used to extract the topic description information of the text data; and image, video and speech recognition algorithms are used to extract the attribute and feature data of the objects described by the images, videos and speech. Based on the aforementioned data fusion constraint paradigm and structured data, business semantic tags and basic data tags are constructed for the data; wherein, each data unit is encoded using a fusion object representation; The business semantic tags and basic data tags are input into the power supply service data fusion model. The fusion degree coefficient between different data is calculated based on the fusion object identifier. Data fusion is performed based on the fusion coefficient and the preset forward and reverse recognition intervals of the fusion degree coefficient to determine the fused data.

[0015] Preferably, the method further includes: For data with multiple business semantics and uses, a high-dimensional fusion coefficient is calculated based on the composite identifier of the data.

[0016] Preferably, the processing of the problematic data for intelligent data verification includes: For problematic data that does not meet the business constraint rules and lacks business classification identifiers, a Bayesian estimation algorithm is used to calculate the expected value of the posterior distribution of the target data based on the prior distribution and likelihood function of the already classified and identified business data. Under given parameter conditions, the business classification of the data is determined and it is identified. For problem data that does not meet the business constraint rules and has missing time series data, the k-nearest neighbor and time series prediction model are used to estimate the value of the missing object by calculating the smooth curve or mean of the data of the nearest neighbor of the missing object, and the time series prediction model is used to predict the value of the missing object based on the preceding data sequence of the missing object. For group-related missing data, a joint distribution probability prediction model is used based on business association rules to calculate the value or content of the missing object by using the joint distribution conditional probability of data quality in historical data, thereby realizing the verification of group-related missing objects; For problematic data that does not meet the data constraint rules and has missing data, similarity imputation and linear fitting algorithms are used to fill in the missing values ​​and contents of the objects. For inconsistent data that does not comply with the data constraint rules, data standardization and multi-value imputation methods are used to fill in the inconsistencies. After data population, data with inconsistencies and abnormal business relationships are verified. The data verification results were compared and backfilled.

[0017] Preferably, the verification of the corrected problematic data includes: The corrected problem data is verified based on deep learning algorithms, stacking strategies, and attention mechanisms.

[0018] Preferably, the data verification based on deep learning algorithms includes: Based on the characteristics and verification requirements of power supply service data, a suitable deep learning model architecture is selected and a deep learning model is constructed. The deep learning model is trained by adjusting its parameters so that it can accurately learn the patterns and rules in the data. The corrected problematic data is input into the trained deep learning model to extract and analyze features, calculate the matching degree or similarity between data, determine whether the data is abnormal based on a preset threshold, and output an error message when the data does not conform to business rules or data rules.

[0019] Preferably, the data verification based on the stacking strategy includes: The dual constraints are transformed into features and criteria that the model can recognize, realizing the characterization of business constraints and the standardization of data constraints. Specifically, in the characterization of business constraints, explicit business rules are transformed into Boolean features, and implicit business relationships are generated into related features through domain knowledge graphs. In the standardization of data constraints, constraint terms are added to the loss function. Construct a two-level stacked model adapted to multidimensional polymorphic data and train the model; The corrected problem data is input into the trained two-level stacked model, and anomaly classification is performed based on the model's output judgment results and confidence level.

[0020] Preferably, the data verification based on the attention mechanism includes: The process involves constructing features based on business and data constraints, transforming these dual constraints into features and criteria that the model can recognize. Specifically, explicit rules are converted into Boolean features, implicit rules are used to generate associated features through a domain knowledge graph, and the priority of business rules is converted into initial values ​​for attention weights. Construct a system of quality scoring plus penalty rules; An attention mechanism model is constructed and trained; wherein the attention mechanism model adopts a four-layer structure consisting of a feature input layer, an attention feature enhancement layer, a feature fusion layer, and a classification output layer. The corrected problem data is input into the trained attention mechanism model, and the verification results are output and graded according to the degree of impact.

[0021] According to another aspect of the present invention, a power supply service data processing system under complex constraints is provided, the system comprising: The association construction unit is used to acquire power supply service data from different sources, analyze the inherent correlation of the power supply service data, and determine the network topology of the power supply service data association. The fusion unit is used to construct a power supply service data fusion model and fuse power supply service data based on the power supply service data fusion model to determine the fused data. The verification unit is used to determine business constraint rules and data constraint rules based on the network topology diagram of the power supply service data association, to perform quality assessment on the fused data, to identify problematic data that does not meet the business constraint rules and data constraint rules, and to process the problematic data for intelligent data verification.

[0022] Preferably, the relationship construction unit analyzes the inherent relationships of the power supply service data and determines the network topology of the power supply service data relationship, including: Based on the associated topology and electrical connection relationships of the devices, the connection weight between devices is characterized by the consistency of the number of device connections and operating status; A data association weight matrix is ​​constructed based on the connection weights, and a device connection weight topology graph is constructed based on the data association weight matrix. Based on the device connection weight topology map, a spectral clustering algorithm is used to cluster the devices, and the device connection weight topology map is divided according to the clustering results to determine the network topology map of power supply service data association.

[0023] Preferably, the fusion unit fuses power supply service data based on the power supply service data fusion model to determine fused data, including: Determine the data fusion constraint paradigm; The unstructured data in the power supply service data is mapped to structured data; wherein, natural language processing algorithms are used to extract the topic description information of the text data; and image, video and speech recognition algorithms are used to extract the attribute and feature data of the objects described by the images, videos and speech. Based on the aforementioned data fusion constraint paradigm and structured data, business semantic tags and basic data tags are constructed for the data; wherein, each data unit is encoded using a fusion object representation; The business semantic tags and basic data tags are input into the power supply service data fusion model. The fusion degree coefficient between different data is calculated based on the fusion object identifier. Data fusion is performed based on the fusion coefficient and the preset forward and reverse recognition intervals of the fusion degree coefficient to determine the fused data.

[0024] Preferably, the fusion unit is further configured to: For data with multiple business semantics and uses, a high-dimensional fusion coefficient is calculated based on the composite identifier of the data.

[0025] Preferably, the verification unit processes the problematic data to perform intelligent data verification, including: For problematic data that does not meet the business constraint rules and lacks business classification identifiers, a Bayesian estimation algorithm is used to calculate the expected value of the posterior distribution of the target data based on the prior distribution and likelihood function of the already classified and identified business data. Under given parameter conditions, the business classification of the data is determined and it is identified. For problem data that does not meet the business constraint rules and has missing time series data, the k-nearest neighbor and time series prediction model are used to estimate the value of the missing object by calculating the smooth curve or mean of the data of the nearest neighbor of the missing object, and the time series prediction model is used to predict the value of the missing object based on the preceding data sequence of the missing object. For group-related missing data, a joint distribution probability prediction model is used based on business association rules to calculate the value or content of the missing object by using the joint distribution conditional probability of data quality in historical data, thereby realizing the verification of group-related missing objects; For problematic data that does not meet the data constraint rules and has missing data, similarity imputation and linear fitting algorithms are used to fill in the missing values ​​and contents of the objects. For inconsistent data that does not comply with the data constraint rules, data standardization and multi-value imputation methods are used to fill in the inconsistencies. After data population, data with inconsistencies and abnormal business relationships are verified. The data verification results were compared and backfilled.

[0026] Preferably, the verification unit verifies the corrected problem data, including: The corrected problem data is verified based on deep learning algorithms, stacking strategies, and attention mechanisms.

[0027] Preferably, the verification unit performs data verification based on a deep learning algorithm, including: Based on the characteristics and verification requirements of power supply service data, a suitable deep learning model architecture is selected and a deep learning model is constructed. The deep learning model is trained by adjusting its parameters so that it can accurately learn the patterns and rules in the data. The corrected problematic data is input into the trained deep learning model to extract and analyze features, calculate the matching degree or similarity between data, determine whether the data is abnormal based on a preset threshold, and output an error message when the data does not conform to business rules or data rules.

[0028] Preferably, the verification unit performs data verification based on a stacking strategy, including: The dual constraints are transformed into features and criteria that the model can recognize, realizing the characterization of business constraints and the standardization of data constraints. Specifically, in the characterization of business constraints, explicit business rules are transformed into Boolean features, and implicit business relationships are generated into related features through domain knowledge graphs. In the standardization of data constraints, constraint terms are added to the loss function. Construct a two-level stacked model adapted to multidimensional polymorphic data and train the model; The corrected problem data is input into the trained two-level stacked model, and anomaly classification is performed based on the model's output judgment results and confidence level.

[0029] Preferably, the verification unit performs data verification based on an attention mechanism, including: The process involves constructing features based on business and data constraints, transforming these dual constraints into features and criteria that the model can recognize. Specifically, explicit rules are converted into Boolean features, implicit rules are used to generate associated features through a domain knowledge graph, and the priority of business rules is converted into initial values ​​for attention weights. Construct a system of quality scoring plus penalty rules; An attention mechanism model is constructed and trained; wherein the attention mechanism model adopts a four-layer structure consisting of a feature input layer, an attention feature enhancement layer, a feature fusion layer, and a classification output layer. The corrected problem data is input into the trained attention mechanism model, and the verification results are output and graded according to the degree of impact.

[0030] According to another aspect of the present invention, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any one of a power supply service data processing method under complex constraints.

[0031] According to another aspect of the present invention, the present invention provides an electronic device, comprising: The aforementioned computer-readable storage medium; and One or more processors for executing a program in the computer-readable storage medium.

[0032] This invention provides a method, system, and electronic device for processing power supply service data under complex constraints. The method includes: acquiring power supply service data from different sources; analyzing the inherent correlations of the power supply service data; determining a network topology diagram of the power supply service data correlations; constructing a power supply service data fusion model; fusing power supply service data based on the power supply service data fusion model to determine fused data; determining business constraint rules and data constraint rules based on the power supply service data correlation network topology diagram; performing quality assessment on the fused data; identifying problematic data that does not meet the business constraint rules and data constraint rules; and processing the problematic data for intelligent data verification. This invention uses a graph-based clustering method to aggregate and represent data with strong inherent dependencies in the power supply service data correlations, achieving a unified data model transformation. This supports cross-system data integration and cross-professional business flow. A unified business data model is designed with business support integrity as the boundary constraint condition. Intelligent processing of data and business constraint data is employed. Deep learning, stacking strategies, attention mechanisms, and other technologies are used, along with manual correction methods, to verify business and data issues. Simultaneously, the source of problematic data is traced and located, improving the data quality of the source system. Attached Figure Description

[0033] Exemplary embodiments of the present invention can be more fully understood by referring to the following figures: Figure 1 A flowchart of a power supply service data processing method 100 under complex constraints according to an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the process of constructing power supply service data association relationships based on graph clustering according to an embodiment of the present invention; Figure 3 This is a schematic diagram illustrating the data fusion process of a multidimensional polymorphic power supply service model based on business consistency according to an embodiment of the present invention. Figure 4 This is a schematic diagram of the power supply service data processing system 400 under complex constraints according to an embodiment of the present invention. Detailed Implementation

[0034] Exemplary embodiments of the invention will now be described with reference to the accompanying drawings. However, the invention may be embodied in many different forms and is not limited to the embodiments described herein. These embodiments are provided to fully and completely disclose the invention and to fully convey its scope to those skilled in the art. The terminology used in the exemplary embodiments illustrated in the drawings is not intended to limit the invention. In the drawings, the same units / elements are referred to by the same reference numerals.

[0035] Unless otherwise stated, the terms used herein (including technical terms) have their common meaning as understood by one of ordinary skill in the art. Furthermore, it is understood that terms defined in commonly used dictionaries should be understood to have a meaning consistent with the context of their relevant field, and not to be interpreted as having an idealized or overly formal meaning.

[0036] Figure 1 This is a flowchart of a power supply service data processing method 100 under complex constraints according to an embodiment of the present invention. Figure 1 As shown, the power supply service data processing method under complex constraints provided by the embodiments of the present invention uses a graph structure clustering method based on the inherent dependency relationship of power supply service data to aggregate and represent data with strong inherent dependencies, realize unified data model transformation, support cross-system data integration and cross-professional business flow, design a unified business data model with business support integrity as the boundary constraint condition, and adopt intelligent processing of data and business constraint data. It uses technologies such as deep learning, stacking strategies, attention mechanisms, and manual correction methods to verify business and data problems, and traces and locates the source of problematic data, thereby improving the data quality of the source system. The power supply service data processing method 100 under complex constraints provided by the embodiments of the present invention starts from step 101. In step 101, power supply service data from different sources is acquired, the inherent correlation of the power supply service data is analyzed, and the network topology of the power supply service data correlation relationship is determined.

[0037] Preferably, the analysis of the inherent correlation of the power supply service data to determine the network topology of the power supply service data correlation includes: Based on the associated topology and electrical connection relationships of the devices, the connection weight between devices is characterized by the consistency of the number of device connections and operating status; A data association weight matrix is ​​constructed based on the connection weights, and a device connection weight topology graph is constructed based on the data association weight matrix. Based on the device connection weight topology map, a spectral clustering algorithm is used to cluster the devices, and the device connection weight topology map is divided according to the clustering results to determine the network topology map of power supply service data association.

[0038] This invention first constructs a power supply service data association based on inherent dependencies. Power supply service data sources include EMS, OMS, distribution automation systems, electricity consumption information collection systems, municipal event management systems, Marketing 2.0 application systems, PMS3.0 distribution network emergency repair applications, and power supply service command systems. Each business system has certain differences in data definition and description, requiring switching between multiple systems to complete information and data collection and business processing in daily power supply service operations. In emergency situations, the efficiency of handling sudden events is low, as the discrete relationships of business data prevent effective and transparent emergency response. Simultaneously, the data from various business systems exhibit inherent dependencies, working together to serve the scheduling and application of power supply service resources. Based on graph clustering algorithms, the inherent relationships of power supply service data are analyzed, cross-professional data is sorted and integrated, a power supply command professional dataset is constructed, a unified data model conversion is achieved, and cross-system data connectivity and cross-professional business flow are supported.

[0039] In this invention, the process of constructing power supply service data associations based on inherent dependencies is as follows: Figure 2 As shown. Specifically, it includes: 1. Definition of device and user entity correlation. Based on the device association topology and electrical connection relationships, the connection weight between devices is characterized by the consistency of the number of device connections and operating status.

[0040] 2. Construct a data association weight matrix and build a device connection weight topology diagram based on the weight definition.

[0041] 3. Data Association Weighted Topology Clustering. A spectral clustering algorithm is used to cluster the equipment in the distribution network, and the connection topology of the equipment is divided based on the clustering results. The clustering results divide the topology into multiple sub-regions, where data has high cohesion and is more likely to be queried in densely connected areas. A graph-based clustering method is used to aggregate and represent data with strong inherent dependencies, reducing the complexity of the extensive topological connections between power supply service data.

[0042] In step 102, a power supply service data fusion model is constructed, and power supply service data is fused based on the power supply service data fusion model to determine the fused data.

[0043] Preferably, the power supply service data is fused based on the power supply service data fusion model to determine the fused data, including: Determine the data fusion constraint paradigm; The unstructured data in the power supply service data is mapped to structured data; wherein, natural language processing algorithms are used to extract the topic description information of the text data; and image, video and speech recognition algorithms are used to extract the attribute and feature data of the objects described by the images, videos and speech. Based on the aforementioned data fusion constraint paradigm and structured data, business semantic tags and basic data tags are constructed for the data; wherein, each data unit is encoded using a fusion object representation; The business semantic tags and basic data tags are input into the power supply service data fusion model. The fusion degree coefficient between different data is calculated based on the fusion object identifier. Data fusion is performed based on the fusion coefficient and the preset forward and reverse recognition intervals of the fusion degree coefficient to determine the fused data.

[0044] Preferably, the method further includes: For data with multiple business semantics and uses, a high-dimensional fusion coefficient is calculated based on the composite identifier of the data.

[0045] In this invention, multidimensional and polymorphic power supply service data are fused based on business consistency.

[0046] The power supply command center's operations involve multiple professional fields such as marketing, operation and maintenance, and dispatching. Each professional field defines and describes the organization of data and business constraints for its professional business systems according to its own professional needs and business development patterns. However, this has led to certain fragmentation and differences between the various business systems. Especially in the daily operation of power supply services, business personnel and data application personnel need to frequently switch between multiple business systems to complete business processing, information collection, and data writing. In emergency situations, this complex data acquisition and cross-model data analysis mode leads to low efficiency in handling emergencies. At the same time, due to the significant dispersion of functions and data in the cross-business systems involved in power supply services, the effectiveness and transparency of power supply service emergency command are also significantly affected.

[0047] Combination Figure 3 As shown, in this invention, the process of fusing multi-dimensional and multi-morphic power supply service data based on business consistency includes: 1. User-defined data fusion constraint paradigm for business consistency.

[0048] 2. Mapping between unstructured and structured data: Natural language processing algorithms are used to extract topic description information from text data through text topic analysis, word segmentation and matching, etc.; image, video and speech recognition algorithms based on artificial intelligence technology are used to extract the attribute and feature data of the descriptive objects in video, audio and image data.

[0049] 3. After mapping unstructured and structured data, business semantic labels and basic data labels are constructed based on a data fusion constraint paradigm that ensures business consistency. Each data unit is encoded using a fusion object identifier. A unified data model is designed with business support integrity as the boundary constraint. A support vector regression model is used to calculate the fusion degree coefficient between different data based on the fusion object identifier. For data models with multiple business semantics and uses, a high-dimensional fusion degree coefficient is calculated based on its composite identifier. Forward and reverse recognition intervals of the fusion degree coefficient are set to fuse data model pairs.

[0050] 4. Based on the fusion result, identify and classify the data models or updated model results that cannot be identified or identified, and map them to business identifiers according to the classification results to achieve model fusion.

[0051] In step 103, business constraint rules and data constraint rules are determined based on the power supply service data association network topology diagram. The quality of the fused data is evaluated to identify problematic data that does not meet the business constraint rules and data constraint rules. The problematic data is then processed to perform intelligent data verification.

[0052] Preferably, the processing of the problematic data for intelligent data verification includes: For problematic data that does not meet the business constraint rules and lacks business classification identifiers, a Bayesian estimation algorithm is used to calculate the expected value of the posterior distribution of the target data based on the prior distribution and likelihood function of the already classified and identified business data. Under given parameter conditions, the business classification of the data is determined and it is identified. For problem data that does not meet the business constraint rules and has missing time series data, the k-nearest neighbor and time series prediction model are used to estimate the value of the missing object by calculating the smooth curve or mean of the data of the nearest neighbor of the missing object, and the time series prediction model is used to predict the value of the missing object based on the preceding data sequence of the missing object. For group-related missing data, a joint distribution probability prediction model is used based on business association rules to calculate the value or content of the missing object by using the joint distribution conditional probability of data quality in historical data, thereby realizing the verification of group-related missing objects; For problematic data that does not meet the data constraint rules and has missing data, similarity imputation and linear fitting algorithms are used to fill in the missing values ​​and contents of the objects. For inconsistent data that does not comply with the data constraint rules, data standardization and multi-value imputation methods are used to fill in the inconsistencies. After data population, data with inconsistencies and abnormal business relationships are verified. The data verification results were compared and backfilled.

[0053] Preferably, the verification of the corrected problematic data includes: The corrected problem data is verified based on deep learning algorithms, stacking strategies, and attention mechanisms.

[0054] Preferably, the data verification based on deep learning algorithms includes: Based on the characteristics and verification requirements of power supply service data, a suitable deep learning model architecture is selected and a deep learning model is constructed. The deep learning model is trained by adjusting its parameters so that it can accurately learn the patterns and rules in the data. The corrected problematic data is input into the trained deep learning model to extract and analyze features, calculate the matching degree or similarity between data, determine whether the data is abnormal based on a preset threshold, and output an error message when the data does not conform to business rules or data rules.

[0055] Preferably, the data verification based on the stacking strategy includes: The dual constraints are transformed into features and criteria that the model can recognize, realizing the characterization of business constraints and the standardization of data constraints. Specifically, in the characterization of business constraints, explicit business rules are transformed into Boolean features, and implicit business relationships are generated into related features through domain knowledge graphs. In the standardization of data constraints, constraint terms are added to the loss function. Construct a two-level stacked model adapted to multidimensional polymorphic data and train the model; The corrected problem data is input into the trained two-level stacked model, and anomaly classification is performed based on the model's output judgment results and confidence level.

[0056] Preferably, the data verification based on the attention mechanism includes: The process involves constructing features based on business and data constraints, transforming these dual constraints into features and criteria that the model can recognize. Specifically, explicit rules are converted into Boolean features, implicit rules are used to generate associated features through a domain knowledge graph, and the priority of business rules is converted into initial values ​​for attention weights. Construct a system of quality scoring plus penalty rules; An attention mechanism model is constructed and trained; wherein the attention mechanism model adopts a four-layer structure consisting of a feature input layer, an attention feature enhancement layer, a feature fusion layer, and a classification output layer. The corrected problem data is input into the trained attention mechanism model, and the verification results are output and graded according to the degree of impact.

[0057] In this invention, intelligent verification of multidimensional and multi-state power supply service data is performed under the dual constraints of industry data.

[0058] Specifically, in combination Figure 3 As shown, the verification process includes: 1. Based on the network topology diagram of the power supply service data relationship, define the business constraint rules for the power supply service data to form a data business constraint rule library. Back up the business constraint rules that have cross-differences. Similarly, define the data constraint rules for the power supply service data from the perspective of data basic compliance to form a data constraint rule library.

[0059] 2. Evaluate the quality of multi-source power supply service data, identify problematic data that contradicts business constraint rules and data constraint rules, and conduct quality assessment, data supplementation, data correction, verification, and backfilling of the results for problematic data.

[0060] For data lacking business classification identifiers that do not meet business constraints, a Bayesian estimation algorithm is used. Based on the prior distribution and likelihood function of the already classified business data, the expected posterior distribution of the target data is calculated. Given parameters, the business classification of the data is determined and it is then identified. For time-series data missing that does not meet business constraints, a k-nearest neighbor and time-series prediction model is used. The value of the missing object is estimated by calculating the smooth curve or mean of its nearest neighbors. The time-series prediction model then predicts the value of the missing object based on its preceding data sequence. For group-related missing data, since traditional methods based on preceding or nearest values ​​cannot be used to estimate the value of the target missing object, a joint distribution probability prediction model is used based on business association rules. This model calculates the value or content of the missing object by using the joint distribution conditional probability of data quality in historical data, thus verifying group-related missing objects. For data missing that does not meet data constraints, similarity imputation and linear fitting algorithms are used to impute the value and content of the missing object. For data inconsistency that does not meet data constraints, data standardization and multi-value imputation methods are used to impute inconsistent data. After data population, data with business inconsistencies and business-related anomalies are corrected. At the same time, technologies such as deep learning, stacking strategies, and attention mechanisms, as well as manual correction methods, are used to verify data with business and data inconsistencies.

[0061] 3. Achieve high-precision intelligent verification through technologies such as deep learning, stacking strategies, and attention mechanisms. Specifically, this includes: (1) Intelligent verification method for multidimensional and polymorphic power supply service data under the dual constraints of business number based on deep learning Constructing a deep learning model: Based on the characteristics of the power supply service data and the verification requirements, select a suitable deep learning model architecture, such as a convolutional neural network (CNN) or a gated recurrent unit (GRU). For example, a multi-interleaved deep learning model can be constructed, which consists of multiple single-interleaved networks stacked on top of each other. Each single-interleaved network first captures the temporal correlation of the data through a bidirectional gated recurrent unit (Bi-GRU), and then extracts the correlation features between the data within a single sampling time point through a CNN.

[0062] Model Training and Optimization: The deep learning model is trained using a large amount of historical data. By adjusting the model's parameters, it is made possible to accurately learn the patterns and regularities in the data. During training, appropriate loss functions and optimization algorithms, such as marginal loss function and Adam optimization algorithm, are employed to improve the model's accuracy and generalization ability.

[0063] Data verification and anomaly detection: Real-time collected power supply service data is input into a trained deep learning model. The model extracts and analyzes features from the data, calculates the matching degree or similarity between data points, and determines whether the data is abnormal based on preset thresholds. If the data does not conform to business rules or data constraints, the model will output an anomaly alert so that staff can conduct further verification and processing.

[0064] Model Evaluation and Updates: Regularly evaluate the performance of deep learning models, using new sample data to test metrics such as accuracy and recall. Based on the evaluation results, update and optimize the models promptly to adapt to changes in power supply service data and adjustments in business needs.

[0065] (2) Intelligent verification method for multidimensional and polymorphic power supply service data under dual constraints of stacking strategy 1) Embedding of features constrained by the number of industries The dual constraints are transformed into features and criteria that the model can recognize, realizing "constraints are features": Business constraint characterization: Explicit business rules are transformed into Boolean features (such as "whether the installed capacity exceeds the upper limit of the substation area"), and implicit business relationships are generated into related features through domain knowledge graphs (such as "the degree of matching between equipment model and rated power"). Data constraint criterionization: Add constraint terms to the loss function, such as assigning higher penalty weights to samples with large data consistency deviations, to ensure that the model prioritizes learning the patterns of samples that meet data quality standards.

[0066] 2) Stacked verification model construction and training Construct a two-level stacked model adapted to multidimensional and polymorphic data, and integrate dual constraints to achieve accurate verification: Base model layer design: Select 3-5 differential base models to cover different data feature types: Decision Tree (DT): Quickly captures the relationships between explicit business rules such as electricity billing; Support Vector Machine (SVM): Processes non-linear features in device operation data; Bi-Gated Cyclic Unit (Bi-GRU): Mining temporal dependencies in telemetry data; Random Forest (RF): Reduces noise interference in user electricity consumption behavior data.

[0067] K-fold cross-validation is used to train the base model, and its prediction results on the training set are concatenated into a "constraint-feature-prediction" matrix, which is used as the input to the meta-model.

[0068] Meta-model layer optimization: Logistic regression was chosen as the meta-model due to its strong interpretability, low overfitting risk, and ability to clearly quantify the contribution weights of each base model. During training, business constraint thresholds (such as electricity billing error ≤ 0.1%) were introduced as hard constraints to ensure that the output results meet business requirements.

[0069] Model training strategy: The Adam optimization algorithm is used to minimize the mixed loss function (cross-entropy loss + constraint penalty term), and the parameters of the base model and meta-model are iteratively adjusted until the verification accuracy and recall on the validation set tend to stabilize.

[0070] 3) Intelligent verification and model iteration Real-time verification and anomaly classification: Preprocessed real-time data is input into the trained stacked model, which outputs a "normal / abnormal" judgment and confidence level. Anomaly classification is achieved using a dual-constraint system. Level 1 anomaly (business violation): such as electricity bill calculation violating electricity pricing policy, directly triggering manual verification; Level 2 anomaly (data anomaly): If telemetry data is missing and there is no reasonable business reason, the automatic data completion process will be triggered.

[0071] Model evaluation and iteration: The model performance is evaluated using a three-dimensional index of "accuracy-recall-F1 score", with a focus on the verification accuracy of core business scenarios (such as electricity billing and business expansion application).

[0072] (3) Intelligent verification method for multidimensional and polymorphic power supply service data under the dual constraints of attention mechanism and business number 1) Construction of industry number constraint features: connecting constraints and models The dual constraints of business numbers are transformed into features and criteria that the model can recognize, realizing "constraints are features, criteria are objectives": A "rule → feature" transformation mechanism is adopted: explicit rules (such as "applied capacity ≤ remaining capacity of the transformer area") are transformed into Boolean features (1 for compliance, 0 for non-compliance); implicit rules (such as "positive correlation between load and temperature") generate associated features (such as "load-temperature correlation coefficient") through the domain knowledge graph. At the same time, the priority of business rules is transformed into initial values ​​of attention weights, such as setting the initial weight of the feature related to "electricity billing logic" to 0.8 and the auxiliary feature to 0.2.

[0073] Construct a "quality score + penalty rule" system: Calculate a quality score for each data point (completeness 20 points, consistency 30 points, accuracy 30 points, timeliness 20 points), and mark samples with scores below 60 points as "low-quality samples"; during model training, multiply the loss value of low-quality samples by a penalty coefficient of 1.5 to force the model to learn the feature patterns of high-quality data.

[0074] 2) Attention mechanism model training: accurate extraction of core features It adopts a four-layer structure: "feature input layer → attention feature enhancement layer → feature fusion layer → classification output layer". Feature input layer: Receives preprocessed multimodal features (numerical, text, graph embedding vectors); Attention Feature Enhancement Layer: Three attention modules are deployed in parallel: a time-series attention module (capturing key time node features for time-series data such as load), a cross-modal attention module (calculating the association weights of features from different modalities, such as the fusion of numerical and textual features), and a business attention module (adjusting feature weights according to business rules to strengthen core business features). Feature fusion layer: The enhanced multi-dimensional features are fused into a unified feature vector through a fully connected layer; Category output layer: Outputs "normal / abnormal" judgment and abnormality type (such as business violation, data inconsistency, numerical abnormality).

[0075] The "step-by-step training + adaptive tuning" strategy is adopted: the first step is to train the basic model with high-quality historical data to ensure that the model grasps the basic feature rules; the second step is to introduce business constraint samples (such as business violation data and data anomaly samples) for fine-tuning, and strengthen constraint cognition through the business attention module; during the training process, the Adam optimization algorithm is used, and the learning rate is adaptively adjusted according to the accuracy of the validation set (the learning rate is reduced when the accuracy increases, and training is paused and parameters are adjusted when the accuracy decreases) to avoid overfitting.

[0076] 3) Intelligent verification and iterative optimization: forming a closed-loop improvement Through a closed-loop mechanism of "real-time verification → anomaly handling → model update", the accuracy of verification and business adaptability are continuously improved. Regularly acquire power supply service data, preprocess it, input it into the trained model, output verification results, and classify them according to "impact level": Level 1 anomalies (business violations that affect safety, such as the applied capacity far exceeding the area capacity) directly trigger the manual verification process and issue an alert; Level 2 anomalies (data quality issues, such as missing electricity consumption) trigger automatic completion or source tracing correction; Level 3 anomalies (minor deviations, such as power factor slightly lower than the standard) are recorded and regularly summarized and analyzed.

[0077] Establish a dual-dimensional evaluation system encompassing both technical and business metrics: technical metrics include accuracy (≥95%), recall (≥92%), and false positive rate (≤5%); business metrics include anomaly handling efficiency (Level 1 anomaly response time ≤1 hour) and business compliance rate (verified data business compliance rate ≥98%). Regularly use new real-time data and newly added business rules to incrementally train the model, updating attention weights and constraint parameters to adapt to changes in data distribution and business upgrades.

[0078] 4. Finally, the data verification results are compared and backfilled to ensure the usability of the basic data. Feedback is provided based on the above data verification methods to analyze the accuracy, validity, and timeliness of the source data.

[0079] The methods of this invention include constructing power supply service data associations based on inherent dependencies, fusing multidimensional and polymorphic power supply service data models for business consistency, and intelligent verification of multidimensional and polymorphic power supply service data under dual constraints of business and data. The method of constructing power supply service data associations based on inherent dependencies forms a network topology diagram of power supply service data associations that is easy to retrieve and traverse, providing data topology support for rapid extraction and transformation of power supply service data and improving the efficiency of cross-system data flow. The method of fusing multidimensional and polymorphic power supply service data models for business consistency ensures data consistency and coherence between different business links and systems by fusing multidimensional and polymorphic data models. The fused multidimensional and polymorphic data model provides a more comprehensive and richer data perspective, helping to discover potential relationships and patterns in the data. The intelligent verification of multidimensional and polymorphic power supply service data under dual constraints of business and data analyzes the accuracy, validity, and real-time performance of data from various source systems, data platforms, and measurement centers, verifies data problems, traces and locates the source of problematic data, and improves the data quality of the real-time measurement center and various source systems.

[0080] The following specific examples illustrate the embodiments of the present invention. In an embodiment of the present invention, taking the equipment management data involved in the power supply service command business as an example, there are certain differences in the degree of correlation between equipment in different regions within the same business system.

[0081] In power supply service command and maintenance work, a power supply service data association relationship construction method based on intrinsic dependency is adopted to establish a close topological representation relationship between the operating conditions of equipment in the fault area and the equipment data in the fixed area of ​​electricity consumption information, so as to facilitate the rapid extraction and analysis of subsequent data.

[0082] The multidimensional and polymorphic power supply service data model fusion method based on business consistency identifies and classifies unidentifiable and unidentifiable data models or updated model results, and maps them to business identifiers based on the classification results to achieve model fusion.

[0083] The intelligent verification method for multidimensional and multi-state power supply service data under the dual constraints of business and data constraints evaluates the quality of power supply service equipment management data and identifies problematic data that contradicts business and data constraint rules. Data cleaning techniques are used to remove noise, appropriate methods are employed to fill in missing values ​​and correct erroneous data, and normalization and standardization methods are used to eliminate dimensional differences. Features closely related to faults, such as voltage and current fluctuations at the time of fault occurrence, are extracted.

[0084] A multi-dimensional, multi-morphic power supply service data intelligent verification method based on deep learning and dual constraints is used to calculate the matching degree or similarity between data points, and to determine whether the data is abnormal based on a preset threshold. If the data does not conform to business rules or data constraints, the model will output an anomaly message so that staff can conduct further verification and processing.

[0085] A multi-dimensional, multi-state intelligent verification method for power supply service data, based on the dual constraints of stacked strategy and operational data, combines the distribution network topology and collected real-time data, utilizing a fault location method based on stacked topology. By acquiring the current of each node in the distribution network, it judges current anomalies according to preset current thresholds, and then quickly determines the location of the fault node through information interaction between intelligent terminals. For example, if the current of a node exceeds the threshold and no overcurrent information is detected in adjacent nodes, it can be determined that there is a fault downstream of that node. This method significantly improves the efficiency and accuracy of distribution network emergency repairs. Fault location time is reduced by more than 30% compared to traditional methods, effectively reducing power outage time and improving power supply reliability and user satisfaction.

[0086] The intelligent verification method for multi-dimensional and multi-state power supply service data based on the dual constraints of attention mechanism and business data prioritizes data with high fault correlation (such as switch position signals, distribution transformer power failure data, and load drop curves) through attention weight allocation, filters redundant information, and strengthens the support of core features for fault judgment. It embeds business rules such as distribution network operation procedures and fault handling standards, including ground fault line selection principles and load transfer logic, to ensure that the verification results meet the requirements of actual emergency repair scenarios. Through data consistency verification (such as matching topology relationships with actual wiring and comparing measurement data with historical baselines), abnormal data is eliminated to avoid judgment deviations caused by erroneous data. It achieves precise fault location and range isolation: combining verified distribution transformer voltage, power failure signals, and other data, it accurately locates the fault area (such as pole sections or meter boxes), guiding emergency repair personnel to directly investigate and avoiding blind line patrols.

[0087] Figure 4 This is a schematic diagram of the power supply service data processing system 400 under complex constraints according to an embodiment of the present invention. Figure 4 As shown, the power supply service data processing system 400 under complex constraints provided in this embodiment of the invention includes: a relation construction unit 401, a fusion unit 402, and a verification unit 403.

[0088] Preferably, the association construction unit 401 is used to acquire power supply service data from different sources, analyze the inherent correlation of the power supply service data, and determine the power supply service data association network topology.

[0089] Preferably, the association construction unit 401 analyzes the inherent correlation of the power supply service data and determines the network topology of the power supply service data association, including: Based on the associated topology and electrical connection relationships of the devices, the connection weight between devices is characterized by the consistency of the number of device connections and operating status; A data association weight matrix is ​​constructed based on the connection weights, and a device connection weight topology graph is constructed based on the data association weight matrix. Based on the device connection weight topology map, a spectral clustering algorithm is used to cluster the devices, and the device connection weight topology map is divided according to the clustering results to determine the network topology map of power supply service data association.

[0090] Preferably, the fusion unit 402 is used to construct a power supply service data fusion model and perform power supply service data fusion based on the power supply service data fusion model to determine the fused data.

[0091] Preferably, the fusion unit 402 fuses power supply service data based on the power supply service data fusion model to determine fused data, including: Determine the data fusion constraint paradigm; The unstructured data in the power supply service data is mapped to structured data; wherein, natural language processing algorithms are used to extract the topic description information of the text data; and image, video and speech recognition algorithms are used to extract the attribute and feature data of the objects described by the images, videos and speech. Based on the aforementioned data fusion constraint paradigm and structured data, business semantic tags and basic data tags are constructed for the data; wherein, each data unit is encoded using a fusion object representation; The business semantic tags and basic data tags are input into the power supply service data fusion model. The fusion degree coefficient between different data is calculated based on the fusion object identifier. Data fusion is performed based on the fusion coefficient and the preset forward and reverse recognition intervals of the fusion degree coefficient to determine the fused data.

[0092] Preferably, the fusion unit 402 is further configured to: For data with multiple business semantics and uses, a high-dimensional fusion coefficient is calculated based on the composite identifier of the data.

[0093] Preferably, the verification unit 403 is used to determine business constraint rules and data constraint rules based on the power supply service data association network topology diagram, perform quality assessment on the fused data, identify problematic data that does not meet the business constraint rules and data constraint rules, and process the problematic data to perform intelligent data verification.

[0094] Preferably, the verification unit 403 processes the problematic data to perform intelligent data verification, including: For problematic data that does not meet the business constraint rules and lacks business classification identifiers, a Bayesian estimation algorithm is used to calculate the expected value of the posterior distribution of the target data based on the prior distribution and likelihood function of the already classified and identified business data. Under given parameter conditions, the business classification of the data is determined and it is identified. For problem data that does not meet the business constraint rules and has missing time series data, the k-nearest neighbor and time series prediction model are used to estimate the value of the missing object by calculating the smooth curve or mean of the data of the nearest neighbor of the missing object, and the time series prediction model is used to predict the value of the missing object based on the preceding data sequence of the missing object. For group-related missing data, a joint distribution probability prediction model is used based on business association rules to calculate the value or content of the missing object by using the joint distribution conditional probability of data quality in historical data, thereby realizing the verification of group-related missing objects; For problematic data that does not meet the data constraint rules and has missing data, similarity imputation and linear fitting algorithms are used to fill in the missing values ​​and contents of the objects. For inconsistent data that does not comply with the data constraint rules, data standardization and multi-value imputation methods are used to fill in the inconsistencies. After data population, data with inconsistencies and abnormal business relationships are verified. The data verification results were compared and backfilled.

[0095] Preferably, the verification unit 403 verifies the corrected problem data, including: The corrected problem data is verified based on deep learning algorithms, stacking strategies, and attention mechanisms.

[0096] Preferably, the verification unit 403 performs data verification based on a deep learning algorithm, including: Based on the characteristics and verification requirements of power supply service data, a suitable deep learning model architecture is selected and a deep learning model is constructed. The deep learning model is trained by adjusting its parameters so that it can accurately learn the patterns and rules in the data. The corrected problematic data is input into the trained deep learning model to extract and analyze features, calculate the matching degree or similarity between data, determine whether the data is abnormal based on a preset threshold, and output an error message when the data does not conform to business rules or data rules.

[0097] Preferably, the verification unit 403 performs data verification based on a stacking strategy, including: The dual constraints are transformed into features and criteria that the model can recognize, realizing the characterization of business constraints and the standardization of data constraints. Specifically, in the characterization of business constraints, explicit business rules are transformed into Boolean features, and implicit business relationships are generated into related features through domain knowledge graphs. In the standardization of data constraints, constraint terms are added to the loss function. Construct a two-level stacked model adapted to multidimensional polymorphic data and train the model; The corrected problem data is input into the trained two-level stacked model, and anomaly classification is performed based on the model's output judgment results and confidence level.

[0098] Preferably, the verification unit 403 performs data verification based on an attention mechanism, including: The process involves constructing features based on business and data constraints, transforming these dual constraints into features and criteria that the model can recognize. Specifically, explicit rules are converted into Boolean features, implicit rules are used to generate associated features through a domain knowledge graph, and the priority of business rules is converted into initial values ​​for attention weights. Construct a system of quality scoring plus penalty rules; An attention mechanism model is constructed and trained; wherein the attention mechanism model adopts a four-layer structure consisting of a feature input layer, an attention feature enhancement layer, a feature fusion layer, and a classification output layer. The corrected problem data is input into the trained attention mechanism model, and the verification results are output and graded according to the degree of impact.

[0099] The power supply service data processing system 400 under complex constraints in an embodiment of the present invention corresponds to the power supply service data processing method 100 under complex constraints in another embodiment of the present invention, and will not be described again here.

[0100] According to another aspect of the present invention, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any one of a power supply service data processing method under complex constraints.

[0101] According to another aspect of the present invention, the present invention provides an electronic device, comprising: The aforementioned computer-readable storage medium; and One or more processors for executing a program in the computer-readable storage medium.

[0102] The present invention has been described with reference to a few embodiments. However, it will be apparent to those skilled in the art that other embodiments besides those disclosed above fall equivalently within the scope of the present invention.

[0103] Generally, all terms used in this invention are interpreted according to their ordinary meaning in the art, unless otherwise expressly defined herein. All references to “a / the / the [device, component, etc.]” ​​are openly interpreted as at least one instance of said device, component, etc., unless otherwise expressly stated. The steps of any method disclosed herein need not be performed in the exact order disclosed unless explicitly stated otherwise.

[0104] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0105] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0106] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0107] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0108] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the present invention.

Claims

1. A method for processing power supply service data under complex constraints, characterized in that, The method includes: Obtain power supply service data from different sources, analyze the inherent correlation of the power supply service data, and determine the network topology of the power supply service data correlation. A power supply service data fusion model is constructed, and power supply service data is fused based on the power supply service data fusion model to determine the fused data; Based on the network topology diagram of the power supply service data association, business constraint rules and data constraint rules are determined. The quality of the fused data is assessed to identify problematic data that does not meet the business constraint rules and data constraint rules. The problematic data is then processed for intelligent data verification.

2. The method according to claim 1, characterized in that, Analyze the inherent correlations of the power supply service data to determine the network topology of the power supply service data correlation, including: Based on the associated topology and electrical connection relationships of the devices, the connection weight between devices is characterized by the consistency of the number of device connections and operating status; A data association weight matrix is ​​constructed based on the connection weights, and a device connection weight topology graph is constructed based on the data association weight matrix. Based on the device connection weight topology map, a spectral clustering algorithm is used to cluster the devices, and the device connection weight topology map is divided according to the clustering results to determine the network topology map of power supply service data association.

3. The method according to claim 1, characterized in that, The power supply service data is fused based on the aforementioned power supply service data fusion model to determine the fused data, including: Determine the data fusion constraint paradigm; The unstructured data in the power supply service data is mapped to structured data; wherein, natural language processing algorithms are used to extract the topic description information of the text data; and image, video and speech recognition algorithms are used to extract the attribute and feature data of the objects described by the images, videos and speech. Based on the aforementioned data fusion constraint paradigm and structured data, business semantic tags and basic data tags are constructed for the data; wherein, each data unit is encoded using a fusion object representation; The business semantic tags and basic data tags are input into the power supply service data fusion model. The fusion degree coefficient between different data is calculated based on the fusion object identifier. Data fusion is performed based on the fusion coefficient and the preset forward and reverse recognition intervals of the fusion degree coefficient to determine the fused data.

4. The method according to claim 3, characterized in that, The method further includes: For data with multiple business semantics and uses, a high-dimensional fusion coefficient is calculated based on the composite identifier of the data.

5. The method according to claim 1, characterized in that, The problematic data is processed to perform intelligent data verification, including: For problematic data that does not meet the business constraint rules and lacks business classification identifiers, a Bayesian estimation algorithm is used to calculate the expected value of the posterior distribution of the target data based on the prior distribution and likelihood function of the already classified and identified business data. Under given parameter conditions, the business classification of the data is determined and it is identified. For problem data that does not meet the business constraint rules and has missing time series data, k-nearest neighbor and time series prediction model are used to estimate the value of the missing object by calculating the smooth curve or mean of the data of the nearest neighbor of the missing object, and the time series prediction model is used to predict the value of the missing object based on the preceding data sequence of the missing object. For group-related missing data, a joint distribution probability prediction model is used based on business association rules to calculate the value or content of the missing object by using the joint distribution conditional probability of data quality in historical data, thereby realizing the verification of group-related missing objects; For problematic data that does not meet the data constraint rules and has missing data, similarity imputation and linear fitting algorithms are used to fill in the missing values ​​and contents of the objects. For inconsistent data that does not comply with the data constraint rules, data standardization and multi-value imputation methods are used to fill in the inconsistencies. After data population, data with inconsistencies and abnormal business relationships are verified. The data verification results were compared and backfilled.

6. The method according to claim 5, characterized in that, The corrected problematic data was verified, including: The corrected problem data is verified based on deep learning algorithms, stacking strategies, and attention mechanisms.

7. The method according to claim 6, characterized in that, Data verification based on deep learning algorithms includes: Based on the characteristics and verification requirements of power supply service data, a suitable deep learning model architecture is selected and a deep learning model is constructed. The deep learning model is trained by adjusting its parameters so that it can accurately learn the patterns and rules in the data. The corrected problematic data is input into the trained deep learning model to extract and analyze features, calculate the matching degree or similarity between data, determine whether the data is abnormal based on a preset threshold, and output an error message when the data does not conform to business rules or data rules.

8. The method according to claim 6, characterized in that, Data verification based on a stacking strategy includes: The dual constraints are transformed into features and criteria that the model can recognize, realizing the characterization of business constraints and the standardization of data constraints. Specifically, in the characterization of business constraints, explicit business rules are transformed into Boolean features, and implicit business relationships are generated into related features through domain knowledge graphs. In the standardization of data constraints, constraint terms are added to the loss function. Construct a two-level stacked model adapted to multidimensional polymorphic data and train the model; The corrected problem data is input into the trained two-level stacked model, and anomaly classification is performed based on the model's output judgment results and confidence level.

9. The method according to claim 6, characterized in that, Data verification based on attention mechanisms includes: The process involves constructing features based on business and data constraints, transforming these dual constraints into features and criteria that the model can recognize. Specifically, explicit rules are converted into Boolean features, implicit rules are used to generate associated features through a domain knowledge graph, and the priority of business rules is converted into initial values ​​for attention weights. Construct a system of quality scoring plus penalty rules; An attention mechanism model is constructed and trained; wherein the attention mechanism model adopts a four-layer structure consisting of a feature input layer, an attention feature enhancement layer, a feature fusion layer, and a classification output layer. The corrected problem data is input into the trained attention mechanism model, and the verification results are output and graded according to the degree of impact.

10. A power supply service data processing system under complex constraints, characterized in that, The system includes: The association construction unit is used to acquire power supply service data from different sources, analyze the inherent correlation of the power supply service data, and determine the network topology of the power supply service data association. The fusion unit is used to construct a power supply service data fusion model and fuse power supply service data based on the power supply service data fusion model to determine the fused data. The verification unit is used to determine business constraint rules and data constraint rules based on the network topology diagram of the power supply service data association, to perform quality assessment on the fused data, to identify problematic data that does not meet the business constraint rules and data constraint rules, and to process the problematic data for intelligent data verification.

11. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps of the method as described in any one of claims 1-9.

12. An electronic device, characterized in that, include: The computer-readable storage medium as described in claim 11; as well as One or more processors for executing a program in the computer-readable storage medium.

Images