Intelligent work order research and judgment system and method based on geographic information and grid service data integration

Abstract

The invention discloses an intelligent work order research and judgment system and method based on geographic information and grid service data integration, and relates to the field of power customer service and data analysis. The system comprises a data access layer, a data fusion layer and an intelligent research and judgment layer, the data access layer is used for obtaining multi-source heterogeneous core data and past work order data, and the data fusion layer forms a structural work order data resource pool through standardization integration of the multi-source heterogeneous core data. The intelligent research and judgment layer realizes accurate recognition, priority ranking, root tracing and hierarchical closed-loop management and control of core demands through BERT semantic recognition, internal and external collaborative research and judgment, an association rule algorithm and the like; the method correspondingly comprises the steps of data access, data fusion and intelligent research and judgment. According to the method, data islands are broken, work order research and judgment accuracy and response efficiency are improved, repeated work orders are reduced, conversion of power services from passive response to active pre-judgment is realized, and the method is suitable for fine processing of power grid enterprise work orders.

Description

Technical Field

[0001] This invention relates to the field of digital customer service technology in the power system, specifically to an intelligent work order analysis system and method based on the integration of geographic information and grid service data. Background Technology

[0002] With the deepening of digital transformation in the power industry, customer service scenarios for power grid companies are becoming increasingly diversified. Customer request channels have expanded from the traditional 95598 hotline to multiple platforms, including the State Grid APP, WeChat official accounts, Alipay service accounts, the 12345 government service hotline, and grid-based on-site feedback, forming a multi-source, heterogeneous work order data ecosystem. This data encompasses geospatial data from the "Power Grid Map" platform of the power grid resource business middleware, electricity consumption data from the Marketing 2.0 system, electricity bill settlement data, user profile information, on-site request records from the grid-based platform, call transcripts from the 95598 business support system, and customer interaction messages from various online channels. The data volume is growing exponentially and is characterized by diverse formats, dispersed sources, complex relationships, and high real-time requirements.

[0003] However, the current work order processing model of power grid companies still faces many technical bottlenecks that urgently need to be addressed, making it difficult to adapt to the refined service needs under digital transformation. First, multi-source data is stored in different systems, lacking standardized integration and correlation mechanisms, resulting in poor data interoperability and difficulty in comprehensively accessing relevant information during analysis. Second, demand identification relies on manual interpretation, leading to inaccurate core intent uncovering, subjective and arbitrary priority judgments, and delayed responses to urgent demands from sensitive users. Third, the analysis process relies solely on single-dimensional data, lacking collaborative analysis of internal and external data, making misjudgments and omissions likely. Fourth, the root cause of demands is often vaguely identified, often remaining at the level of surface-level problem handling, resulting in frequent duplicate work orders. Fifth, order dispatch lacks precision, response times are long, closed-loop management lacks effective verification mechanisms, and the processing effectiveness is difficult to guarantee.

[0004] These problems not only lead to low efficiency in work order processing and serious waste of resources, but also affect the service quality and brand image of power grid companies, making it difficult to meet customers' demands for precise, efficient, and personalized power services in the new era. Therefore, there is an urgent need to build a technical solution that can integrate multi-source geographic information and grid service data to achieve intelligent work order analysis, accurate work order dispatch, root cause tracing, and closed-loop management, thereby addressing the pain points of the existing model and promoting the transformation of power customer service from passive response to proactive prediction, and from extensive management to refined operation. Summary of the Invention

[0005] The purpose of this invention is to provide a work order intelligent analysis system and method based on the integration of geographic information and grid service data, so as to achieve standardization, intelligence and precision in work order processing, improve service response efficiency, analysis accuracy and customer satisfaction, and reduce operating costs and service risks.

[0006] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:

[0007] A work order intelligent analysis system based on the integration of geographic information and grid service data includes a data access layer, a data fusion layer and an intelligent analysis layer that are connected in sequence.

[0008] The data access layer is used to acquire multi-source heterogeneous core data related to power customer service and power grid grid to support intelligent work order analysis, and also to acquire past work order data accumulated by the system. The data fusion layer is used to standardize, integrate and correlate the effective data output by the data access layer. The resulting structured data includes request text data, user tag data, internal business data, external channel data, geospatial data and historical work order data, forming a structured work order data resource pool. The internal business data includes equipment status, power outage records and arrears information, the external channel data includes grid requests, government feedback and customer interactions, and the geospatial data includes grid worker real-time location data.

[0009] The intelligent analysis layer comprises, in sequence, a core demand identification and priority response module, an internal and external data collaborative analysis module, a demand root cause tracing and source early warning module, and a hierarchical early warning and control module. The core demand identification and priority response module first analyzes the demand text data and user tag data in the structured work order data resource pool to uncover the customer's core intent and construct a priority model, outputting a priority ranking result for highly sensitive / urgent demands. Based on the priority ranking result, the internal and external data collaborative analysis module calls upon internal business data and external channel data in the structured work order data resource pool to conduct cross-dimensional collaborative analysis, outputting multi-dimensional analysis results. Based on the multi-dimensional analysis results, the demand root cause tracing and source early warning module correlates equipment status and historical work order data in the structured work order data resource pool to uncover the core root cause of the demand, constructs a source early warning model, and outputs root cause risk early warning information. Based on the root cause risk early warning information, the hierarchical early warning and control module combines geospatial data and grid worker real-time location data in the structured work order data resource pool to generate hierarchical early warnings, achieve accurate order dispatch, and verify processing effectiveness.

[0010] Furthermore, the multi-source heterogeneous core data includes the original geospatial data of the "Power Grid One Map" of the power grid resource business platform, electricity consumption collection data, electricity bill data and user profile data of the Marketing 2.0 system, on-site demand data of the grid platform, call translation text data of the 95598 business support system, and customer interaction data of online channels; the electricity consumption collection data includes equipment status and power outage records, the electricity bill data includes overdue payment information, the user profile data includes sensitive tags related to people's livelihood, and the customer interaction data includes government feedback information; the online channels include the State Grid website, WeChat, and Alipay; the data access layer adopts a primary and backup dual-link backup transmission mechanism, and after data collection, invalid data is removed through format verification, integrity verification, and consistency verification algorithms, and the valid original data before standardization is output to ensure the continuity, integrity, and effectiveness of data transmission.

[0011] Furthermore, the standardized integration and correlation processing of the data fusion layer specifically includes the following steps:

[0012] a1) Classification and Transformation: Based on the valid raw data output from the data access layer, the on-site request data and call transcript data are transformed into request text data through word segmentation, deduplication, and semantic normalization; sensitive tags related to people's livelihood and electricity consumption feature tags are extracted from user profile data and formatted uniformly to transform into user tag data; electricity collection data and electricity bill data are transformed into internal business data through field standardization and association; customer interaction data and the above-mentioned standardized on-site request data are integrated into external channel data; raw geospatial data and grid worker real-time location data are transformed into geospatial data through WGS84 coordinate system standardization and address binding; at the same time, past work order data are transformed into historical work order data through field completion and index association.

[0013] a2) Association mapping: Based on the "station-line-transformer-customer" topology, an association index for the above six types of structured data is established through the association keys of the transformer area number, user number and equipment number;

[0014] a3) Secure storage: Sensitive data is encrypted using the SM2 algorithm and data integrity is verified using the SM3 algorithm. The data is stored in a distributed storage node according to a unified schema of "work order basic information - user information - associated device information - geospatial information - request content information", forming a structured work order data resource pool.

[0015] a4) Data dictionary specifications: The unified schema includes fields for basic work order information, user information, associated equipment information, geospatial information, and request content information. The value range of each field is defined according to the State Grid digital data standard.

[0016] a5) Missing field completion: When a critical business field is missing, the data is removed. When a non-critical field is missing, it is completed by interpolation based on the default value or historical data. The completed data is marked with a "completion identifier" and stored in the structured order data resource pool.

[0017] Furthermore, the core request identification and priority response module is implemented as follows: It calls the request text data, user tag data, and historical work order data from the structured work order data resource pool, and executes the following steps:

[0018] b1) The BERT semantic understanding model is used to perform in-depth analysis on the call translation text and work order description in the appeal text data. The BERT semantic understanding model adopts the bert-base-chinese pre-trained model, the input text length is limited to 512 characters, the learning rate in the fine-tuning stage is 5e-5, batch_size=32, the number of iterations is 100, and the dropout rate is 0.1. The core semantic points extracted include frequent power outages, electricity bill disputes and voltage anomalies, and the difference between the surface appeal and the real needs is identified.

[0019] b2) Read the livelihood-sensitive tags from the user profile in the user tag data. The livelihood-sensitive tags include schools, nursing homes and oxygen users. Assign a score to the sensitive tags. Among them, livelihood-sensitive users get 8-10 points and ordinary users get 0-5 points.

[0020] b3) Based on the type of request (fault reporting, consultation, complaint, etc.), scope of impact (single household, multiple households, area) in the request text data and historical feedback data in the historical work order data, construct an urgency score, in which urgent requests are scored 8-10 points, general requests are scored 4-7 points, and ordinary requests are scored 0-3 points.

[0021] b4) Calculate the score according to the priority scoring formula: Priority score = Sensitive tag score × 0.4 + Urgency score × 0.3 + Historical feedback score × 0.3. The historical feedback score is assigned as follows: 8-10 points for no duplicate requests, 4-7 points for one duplicate request, and 0-3 points for two or more duplicate requests. Priority scores ≥ 8 points are high priority, 5-7 points are medium priority, and < 5 points are low priority. High priority requests are pushed to local management personnel first. The warning information includes the core intent, user profile, and historical service records.

[0022] Furthermore, the internal and external data collaborative analysis module is implemented by calling internal business data, external channel data, and geospatial data from the structured single data resource pool, and executing the following steps:

[0023] c1) Internal data collaboration: Extract internal business data, including equipment status data, power outage record data, and overdue payment information data, from the structured single data resource pool, and construct an equipment-user-demand association graph by combining the "station-line-transformer-customer" topology relationship;

[0024] c2) External data collaboration: Access the on-site verification data of the grid platform, the 12345 government feedback data and the online State Grid user interaction data from the external channel data of the structured chemical single data resource pool to supplement customer-side scenario information;

[0025] c3) Collaborative analysis rules: When internal business data detects a regional power outage and there is no planned power outage information, if there are ≥5 similar requests per hour in external channel data, a red-level warning for multiple power outages will be triggered; when internal business data shows that users are ≥30 days in arrears, combined with the analysis data of active payment areas in external channel data, a customized collection plan including nearby payment outlets and online payment guides will be pushed.

[0026] Furthermore, the implementation method of the request root cause tracing and source early warning module is as follows: calling the internal business data, historical work order data and request text data in the structured work order data resource pool, and executing the following steps:

[0027] d1) An association rule algorithm is used to analyze the correlation between equipment fault records in internal business data, historical power outage work orders in historical work order data, and user request content in request text data. The minimum support of the association rule algorithm is 0.2 (support = number of work orders including a certain root cause / total number of work orders), and the minimum confidence is 0.7 (confidence = number of work orders including a certain root cause and having the same type of request / number of work orders including that root cause). The algorithm identifies high-frequency problem root causes, including aging lines, insufficient convenience of payment channels, and equipment overload.

[0028] d2) The identified root cause types are classified using a decision tree model. The maximum depth of the decision tree model is 10, the minimum number of sample splits is 2, the minimum number of sample leaf nodes is 1, the criterion uses the Gini coefficient, and the root cause priority is output.

[0029] d3) Set root cause risk thresholds, which include equipment failures ≥3 times per month, similar requests ≥10 times per week, line failure intervals ≤7 days, and payment channel complaints ≥15 times per month. Once the thresholds are reached, a source governance warning will be automatically triggered and pushed to the operation and maintenance department or business optimization department.

[0030] d4) Track the changes in work order data after source governance, and verify the effectiveness of root cause resolution by comparing the frequency of similar requests before and after governance.

[0031] Furthermore, the implementation method of the hierarchical early warning and control module is as follows: It calls the geospatial data, appeal text data, and various analysis result data from the structured chemical unit data resource pool, and executes the following steps:

[0032] e1) Generation of graded early warnings: Based on the aforementioned analysis results, graded early warnings are generated. Red-level early warnings correspond to scenarios including large-scale power outages and emergency requests from users with sensitive livelihoods. Orange-level early warnings correspond to scenarios including requests from multiple households and overdue unprocessed work orders. Yellow-level early warnings correspond to scenarios including ordinary requests from a single household and potential risks.

[0033] e2) Precise task assignment: Based on the real-time location information of grid workers in geospatial data (no need for additional map system integration, echoing the standardized results of data fusion layer), the real-time latitude and longitude of grid workers are obtained, the straight-line distance and reachability time between grid workers and work order addresses are calculated, and work orders are dynamically assigned in combination with the workload of grid workers (current unfinished work orders × 1 + estimated work order time × 0.5), with a distance weight of 0.5 and a workload weight of 0.5, giving priority to assigning to the grid workers who are closest and have the least workload;

[0034] e3) Enhanced control: Red-level warning work orders are notified to local management personnel through both in-station messages and SMS, and are marked as key supervision work orders in the power supply command system; Orange-level warning work orders are reminded once every 2 hours, and Yellow-level warning work orders are processed according to the regular procedure.

[0035] e4) Closed-loop verification: After the work order is processed, a semantic analysis algorithm is called to calculate the semantic similarity between the processing result and the initial request in the request text data. The semantic similarity is calculated using cosine similarity, with a threshold of ≥0.8. If the similarity is ≥0.8, the problem is considered resolved, and if it is <0.8, a secondary warning is triggered. Work orders that are not processed within the time limit are automatically escalated to the superior management department according to the time limit requirements of red level ≤2 hours, orange level ≤4 hours, and yellow level ≤8 hours.

[0036] e5) Early warning push: Early warning information is automatically pushed through the WeChat robot platform and SMS interface, and the progress of work order processing is fed back to the system in real time.

[0037] Furthermore, the system also includes an application output layer; the application output layer is communicatively connected to the intelligent analysis layer and is used to provide functions such as visual display of work order risks, hierarchical early warning notification, closed-loop processing of work orders, and automatic generation of analysis reports based on the output results of the intelligent analysis layer; the work order risk visualization display function realizes the distribution of work order risks, the presentation of heat maps of high-incidence complaint areas, and real-time updates of grid member work status based on the "power grid map", and supports multi-dimensional filtering and query by region, time, complaint type, and early warning level; the automatic analysis report generation function can generate monthly reports on key issues of power supply services and periodic hot complaint analysis reports, and supports export in PDF and Excel formats.

[0038] A work order intelligent assessment method based on the integration of geographic information and grid service data, applied to the aforementioned work order intelligent assessment system, includes the following steps:

[0039] S1. Data Access: Acquire multi-source heterogeneous core data related to power customer service and power grid grid to support intelligent analysis of work orders, and call up past work order data accumulated by the system;

[0040] S2. Data Fusion: The multi-source heterogeneous core data and past work order data are standardized, integrated, and correlated to transform them into structured data containing request text data, user tag data, internal business data, external channel data, geospatial data, and historical work order data, forming a structured work order data resource pool; the internal business data includes equipment status, power outage records, and overdue payment information, the external channel data includes grid requests, government feedback, and customer interactions, and the geospatial data includes grid worker real-time location data;

[0041] S3. Intelligent Analysis: First, analyze the request text data and user tag data in the structured work order data resource pool to obtain priority ranking results; then, based on the priority ranking results, call internal business data and external channel data to carry out cross-dimensional collaborative analysis to obtain multi-dimensional analysis results; next, based on the multi-dimensional analysis results, link equipment status and historical work order data to mine the core root causes of the requests and obtain root cause risk warning information; finally, based on the root cause risk warning information, combine geospatial data and grid member real-time location data to generate hierarchical warnings, achieve accurate work order dispatch, and verify the processing effect.

[0042] Furthermore, the method also includes the following steps:

[0043] S4. Application Output: Based on the processing results of the intelligent assessment steps, visualize the work order risk, issue tiered early warnings, perform closed-loop processing of work orders, and automatically generate analysis reports.

[0044] The beneficial effects of this invention are:

[0045] This application transforms heterogeneous data from multiple sources, such as geospatial data, equipment data, user data, and demand data, scattered across various systems including the power grid map, marketing 2.0, 95598, and grid platform, into a unified, tightly linked structured data resource pool through standardized processing of classification, transformation, association mapping, and secure storage at the data fusion layer. By establishing an association index based on the "station-line-transformer-customer" topology, it achieves deep binding of four-dimensional data of equipment, users, demands, and geography, completely breaking down data silos and solving the problem of data interoperability.

[0046] Invalid data is eliminated through a triple algorithm of format verification, integrity verification, and consistency verification. Combined with rules for removing missing key fields and completing and annotating non-key fields, the data verification pass rate is improved. Sensitive data is encrypted using the SM2 algorithm, and data integrity is verified using the SM3 algorithm, which complies with the national information security level protection level three standard, ensuring data security compliance. At the same time, the geospatial data has small positioning errors after being standardized by the WGS84 coordinate system, providing high-quality data support for subsequent accurate dispatching and regional analysis.

[0047] Automated and standardized processing replaces the traditional manual data processing mode, reducing the data integration cycle from several hours to minutes (real-time synchronization of 95598 call data, hourly synchronization of electricity consumption data, and daily incremental synchronization of gridded data). The data processing efficiency is greatly improved compared to the traditional mode, meeting the needs of real-time analysis of massive work orders.

[0048] A BERT semantic understanding model pre-trained with BERT-based Chinese (input length 512 characters, fine-tuned learning rate 5e-5) is used to perform deep analysis on transcribed text and work order descriptions, significantly improving the accuracy of core semantic recognition. It can accurately extract core semantic points such as frequent power outages, electricity bill disputes, and voltage anomalies, effectively distinguishing between surface-level requests and real needs. For example, it can identify multiple households experiencing unstable voltage behind the inability to use appliances, avoiding subjective errors from human interpretation. A quantified priority determination model is constructed, assigning 8 to users with sensitive livelihood needs (schools, nursing homes, oxygen users, etc.). -10 points are given high weight, and emergency requests (large-scale power outages, affecting multiple households) are graded into levels of 8-10 points, completely replacing the traditional subjective experience-based judgment; high-priority requests are pushed to local management personnel first, and the response time for emergency requests from users with sensitive livelihoods is shortened compared to the traditional model, effectively avoiding delays in emergency requests; by prioritizing, urgent and difficult requests are handled first, and less important requests are diverted to less important ones. High-priority work orders are focused on core resources, such as experienced grid workers and emergency repair teams, while ordinary requests are processed according to the regular process, improving resource utilization and reducing the contradiction in the traditional model where ordinary requests occupy emergency resources and emergency requests have insufficient resources.

[0049] Internal and external data collaboration leverages a structured work order data resource pool. Internally, it utilizes internal business data such as equipment status, power outage records, and overdue payment information to construct a device-user-request relationship graph. Externally, it integrates grid-based on-site verification data, 12345 government feedback data, and online user interaction data to supplement scenario information. The analysis dimensions have expanded from single work orders to internal and external collaboration, significantly improving the accuracy of analysis compared to traditional single-data analysis. Scenario-based early warning responses are timely. Precise rules are set for core scenarios such as multi-household power outages and overdue payment reminders. When there is a power outage in the internally monitored area with no planned information and ≥5 similar external requests per hour, a multi-household power outage red-level warning is triggered in real time, shortening the warning response time and buying time for large-scale power outage emergency response. When users are ≥30 days overdue, customized reminder plans are pushed based on data from areas with active payment, improving payment conversion rates and reducing the risk of power outages due to overdue payments.

[0050] By analyzing the correlation between equipment fault records, historical power outage work orders, and user complaints using an association rule algorithm (minimum support 0.2, minimum confidence 0.7), and combining this with a decision tree model (maximum depth 10, Gini coefficient classification) to categorize and rank root causes, the accuracy of identifying high-frequency problem root causes (such as aging lines, equipment overload, and inconvenient payment channels) has been improved, addressing the pain point of ambiguous root cause localization in traditional models. Proactive early warning replaces passive response, setting risk thresholds such as ≥3 equipment faults per month and ≥10 similar complaints per week. Upon reaching these thresholds, automatic source governance warnings are pushed to the operations and maintenance or business optimization departments, forming a closed loop of discovery-early warning-governance-verification. By proactively rectifying aging lines and optimizing payment channels, the number of duplicate work orders has been reduced, lowering operations and maintenance manpower costs. Root cause tracing results quantitatively present the core issues for different regions and equipment types, providing precise data for line upgrades, equipment updates, and service process optimization. For example, for regions with ≥15 monthly complaints about payment channels, online payment entry points have been optimized or offline service points have been added, reducing similar complaints at their source.

[0051] Leveraging real-time location information of grid workers in geospatial data and combining it with a dynamic dispatch model, work orders are prioritized for allocation to the nearest grid worker with the least workload, shortening on-site response time and the average processing time for multiple power outages and requests from sensitive users. A three-tiered early warning and control mechanism (red, orange, and yellow) has been established. Red-level warning work orders are notified via both internal station messages and SMS, and are given priority supervision by the power supply command system. Orange-level work orders receive reminders every 2 hours, and yellow-level work orders are processed as usual, ensuring priority protection for key requests and orderly handling of ordinary requests. Work orders that have not been processed within the specified time are automatically escalated according to the following timeframes: red ≤ 2 hours, orange ≤ 4 hours, and yellow ≤ 8 hours, avoiding delays in processing. A cosine similarity algorithm (threshold ≥ 0.8) is used to verify the matching degree between the processing result and the initial request, replacing the traditional subjective user feedback verification method and improving the substantive problem resolution rate. For work orders with a similarity < 0.8, a secondary warning is triggered, avoiding superficial closure and ensuring a genuine response to user requests.

[0052] Based on the "Power Grid One Map," the system enables the tracking of work order risk distribution, heat maps of high-complaint areas, and real-time updates on grid worker work status. It supports multi-dimensional filtering and querying by region, time, complaint type, and warning level, allowing managers to quickly identify service weaknesses and improve decision-making and response efficiency. The system automatically generates monthly reports on key power supply service issues and periodic hot-button complaint analysis reports, covering core indicators such as complaint type distribution, high-frequency root cause statistics, response efficiency ranking, and regional service quality comparisons. It supports exporting to PDF and Excel formats, providing data support for adjustments to operation and maintenance strategies, allocation of service resources, and optimization of product functions, thus driving the transformation of power customer service from experience-driven to data-driven approaches. Attached Figure Description

[0053] Figure 1 This is a schematic diagram of the intelligent work order analysis system of the present invention. Detailed Implementation

[0054] To facilitate understanding of the present invention, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Those skilled in the art should understand that the embodiments described are merely illustrative of the invention and should not be considered as specific limitations thereof.

[0055] In the following description of the embodiments, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods are omitted so as not to obscure the description of this application with unnecessary detail.

[0056] It should be understood that, when used in this specification and the appended claims, the term "comprising" indicates the presence of a described feature, integral, step, operation, element, and / or component, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or collections thereof. It should also be understood that, as used in this specification and the appended claims, the term "and / or" refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0057] As used in this application specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if detected [the described condition or event]" may be interpreted, depending on the context, as "once determined," "in response to determination," "once detected [the described condition or event]," or "in response to detection [the described condition or event]."

[0058] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance. References to "one embodiment" or "some embodiments" in this application mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0059] Example 1: Intelligent Work Order Analysis System Based on Geographic Information and Grid Service Data Integration

[0060] like Figure 1 As shown, the system comprises a data access layer, a data fusion layer, an intelligent analysis layer, and an application output layer, which are connected in sequence. Each layer works collaboratively according to the logic of data access, fusion preprocessing, intelligent analysis, and application output.

[0061] Data Access Layer: This layer acquires multi-source heterogeneous core data related to power customer service and power grid grid to support intelligent work order analysis, while also acquiring past work order data accumulated in the system. It employs a primary / backup dual-link backup transmission mechanism, using format verification, integrity verification, and consistency verification algorithms to remove invalid data and output valid raw data before standardization. The multi-source heterogeneous core data includes the original geospatial data from the "Power Grid Map" of the power grid resource business platform, electricity consumption data (including equipment status and power outage records) from the Marketing 2.0 system, electricity billing data (including overdue payment information), user profile data (including sensitive public service tags), on-site request data from the grid-based platform, call transcript data from the 95598 business support system, and customer interaction data (including government feedback information) from online channels such as State Grid, WeChat, and Alipay.

[0062] Data Fusion Layer: This layer standardizes, integrates, and correlates the valid raw data output from the data access layer and past work order data to form a structured work order data resource pool. The resulting structured data includes request text data, user tag data, internal business data (including equipment status, power outage records, and overdue payment information), external channel data (including grid requests, government feedback, and customer interactions), geospatial data (including real-time location data of grid workers), and historical work order data. Specific processing steps include:

[0063] 1) Classification and Transformation: On-site complaint data and call transcript data are transformed into complaint text data through word segmentation, deduplication, and semantic normalization; sensitive tags related to people's livelihood and electricity consumption characteristics are extracted from user profile data and formatted uniformly to transform into user tag data; electricity consumption data and electricity bill data are transformed into internal business data through field standardization and association; customer interaction data and standardized on-site complaint data are integrated into external channel data; raw geospatial data and grid worker real-time location data are transformed into geospatial data through WGS84 coordinate system standardization and address binding; past work order data are transformed into historical work order data through field completion and index association.

[0064] 2) Association Mapping: Based on the "station-line-transformer-customer" topology, an association index for six types of structured data is established through the association keys of transformer area number, user number, and equipment number;

[0065] 3) Secure storage: Sensitive data is encrypted using the SM2 algorithm and data integrity is verified using the SM3 algorithm. The data is stored in a distributed storage node according to a unified schema of work order basic information, user information, associated device information, geospatial information, and request content information.

[0066] 4) Data dictionary standard: The value range of each field in the unified schema shall be defined according to the State Grid digital data standard;

[0067] 5) Missing field completion: When a critical business field is missing, the data is removed; when a non-critical field is missing, it is completed by interpolation based on the default value or historical data and marked with a "completion mark".

[0068] The data dictionary specifications include: basic work order information fields including work order ID, submission time, source channel and processing status; user information fields including user ID, name, contact information, electricity address and public welfare sensitive tags; associated equipment information fields including equipment number, equipment type, substation area and operating status; geospatial information fields including latitude and longitude, administrative division code and geographical range of the substation area; and request content information fields including request type, description text, core semantic points and urgency level.

[0069] The rules for filling in missing fields are as follows: When key business fields (such as user ID, device number, and request type) are missing, they are judged as invalid data and removed; when non-key fields (such as supplementary contact information for users and supplementary description of requests) are missing, they are filled in by default value of "none" or by interpolation based on the historical average of users of the same type in the same area.

[0070] Intelligent Analysis Layer: This is the core of the system, comprising a core demand identification and priority response module, an internal and external data collaborative analysis module, a demand root cause tracing and source early warning module, and a hierarchical early warning and control module, which process data according to a progressive logic.

[0071] 1) Core Request Identification and Priority Response Module: This module calls upon the request text data, user tag data, and historical work order data from the structured work order data resource pool. It uses the BERT semantic understanding model pre-trained with bert-base-chinese to extract core semantic points. Combining the scores of sensitive tags related to people's livelihood, urgency scores, and historical feedback scores, it calculates the priority according to the formula "Priority Score = Sensitive Tag Score × 0.4 + Urgency Score × 0.3 + Historical Feedback Score × 0.3" and outputs the ranking results.

[0072] 2) Internal and external data collaborative analysis module: Based on the priority ranking results, it calls internal business data, external channel data and geospatial data to build a device-user-demand relationship map, supplement customer-side scenario information, and trigger early warnings or push customized solutions according to preset rules;

[0073] 3) Request Root Cause Tracing and Source Early Warning Module: Based on multi-dimensional analysis results, it associates internal business data, historical work order data and request text data, uses association rule algorithm and decision tree model to identify and classify high-frequency root causes, sets risk thresholds to trigger source governance early warning, and tracks and verifies the governance effect;

[0074] 4) Tiered early warning and control module: Based on root cause risk early warning information, combined with geospatial data and grid member real-time location data, a tiered early warning (red, orange and yellow levels) is generated. Tasks are dynamically assigned according to distance and workload, and differentiated control is implemented. The processing effect is verified through semantic similarity.

[0075] Application output layer: Provides functions such as visual display of work order risks, hierarchical early warning notification, closed-loop processing of work orders, and automatic generation of analysis reports; the visualization display is based on "one map of the power grid" to realize multi-dimensional filtering and query, and the analysis report supports export in PDF and Excel formats.

[0076] Example 2: Intelligent Work Order Analysis Method Based on Geographic Information and Grid Service Data Integration

[0077] This method, when applied to the aforementioned system, includes the following steps:

[0078] S1. Data Access: Acquire multi-source heterogeneous core data related to power customer service and power grid grid to support intelligent work order analysis, and call the past work order data accumulated by the system; after invalid data is removed by format verification, integrity verification and consistency verification algorithms, the valid original data and past work order data are transmitted to the data fusion layer;

[0079] S2. Data Fusion: Standardize, integrate, and associate the received data, and form a structured single data resource pool through operations such as classification transformation, association mapping, and secure storage;

[0080] S3. Intelligent Analysis: The four core modules of the intelligent analysis layer process the data in a progressive manner: First, analyze the request text data and user tag data to output priority ranking results. Then, call internal business data and external channel data to carry out collaborative analysis and output multi-dimensional analysis results. Next, associate device status and historical work order data to mine the root cause and output early warning information. Finally, combine geospatial data to generate hierarchical early warning, achieve accurate order dispatch, and verify the processing effect.

[0081] S4. Application Output: Receive intelligent analysis results and provide visualization, early warning notifications, work order processing, and analysis report generation services.

[0082] Example 3: Specific implementation of the above system and method

[0083] I. Implementation Background

[0084] This embodiment uses the power grid customer service work order processing in a district of Beijing as an application scenario. Given the wide coverage of the power grid in this area, the diverse customer types (including schools, nursing homes, residential communities, commercial complexes, etc.), and the multiple channels for requesting assistance (95598 hotline, State Grid APP, 12345 government hotline, grid-based on-site feedback, etc.), the intelligent analysis system and method proposed in this application are deployed to solve the problems of scattered data, low analysis accuracy, slow response, and numerous duplicate work orders in traditional work order processing. The implementation scope covers 12 power supply grids in the district, involving approximately 800,000 electricity customers, 30 grid workers, and an average daily processing volume of 200-300 work orders.

[0085] II. System Deployment Environment

[0086] (a) Hardware deployment

[0087] The system is based on the State Grid Cloud (Alibaba) platform and deployed in the data center of the company's management information region. The hardware configuration includes: 10 8-core 16G cloud servers (for data access, fusion, analysis and application output), 5 distributed storage nodes (total storage capacity of 10TB, used for structured single data resource pool storage), gigabit fiber optic network (to ensure real-time synchronization of data from multiple systems), and GPS positioning module (integrated into the grid worker terminal to collect the grid worker's latitude and longitude in real time).

[0088] (II) Software and Algorithm Configuration

[0089] Operating system: CentOS 7.9;

[0090] Databases: MySQL 8.0 (for storing structured data), Neo4j 4.4 (for storing the relationship between storage devices, users, and requests), and ElasticSearch 7.17 (for supporting fast retrieval of work order data).

[0091] Algorithm framework: PyTorch 1.12 (supports BERT semantic understanding model, decision tree model training and inference);

[0092] Visualization tool: ECharts 5.4 (for visualization of heatmaps, distribution points, etc.);

[0093] Security components: SM2 / SM3 encryption module (compliant with the State Cryptography Administration standard), access control system (based on the RBAC model).

[0094] III. Specific Implementation Process of Each Layer of the System and its Methods

[0095] (I) Implementation of the data access layer

[0096] 1. Data Collection Scope

[0097] Multi-source heterogeneous core data:

[0098] The original geospatial data of the "Power Grid Map" in the power grid resource business platform includes: the 10kV line routes, transformer locations (latitude and longitude accuracy to 10 meters), and transformer area boundaries of 12 grids in a certain area.

[0099] Marketing 2.0 system data: electricity consumption data (load rate, voltage value, and current value of more than 1,000 transformers, collected once every 15 minutes), electricity bill data (user payment records, amount owed, and duration of arrears), and user profile data (information on 800,000 customers, including sensitive tags related to people's livelihood such as 20 schools, 15 nursing homes, and 5 users receiving oxygen therapy).

[0100] Data from the grid management platform: Equipment malfunction records, photos of potential line hazards, and textual descriptions of customer requests reported on-site by grid workers (50-80 items collected daily).

[0101] 95598 Business Support System Data: Call Transcription Text (120-150 pieces per day);

[0102] Online channel data: Customer interaction data from the State Grid APP, WeChat official account, and Alipay service account (including payment inquiries, fault reporting, complaints and suggestions, averaging 80-100 messages per day), and 12345 government feedback data (averaging 30-50 messages per day, mostly involving requests for power outages and unstable voltage in the area).

[0103] The system has accumulated historical work order data: 120,000 historical work order records from January 2023 to June 2024, including fields such as work order number, request content, processing result, and duplicate request marker.

[0104] 2. Data transmission and verification

[0105] Transmission mechanism: A dual-link configuration is adopted with the business connection platform as the main link and the data middle platform as the backup link. 95598 call data is synchronized in real time, electricity collection data and power outage records are synchronized once per hour, and grid-based demand data and online channel data are incrementally synchronized daily (synchronization time is 2:00 am).

[0106] Data validation: Valid data is filtered using a triple algorithm.

[0107] Format validation: Remove data with incorrect field types, such as text data in the "voltage value" field;

[0108] Integrity verification: Remove work orders that are missing key fields such as "user ID, device number, and request type" (accounting for approximately 0.3% of the total data).

[0109] Consistency check: Compare cross-system data (such as whether the power outage records of the Marketing 2.0 system are consistent with the power outage request time of the 95598 work order) and remove logically contradictory data (accounting for approximately 0.2% of the total data volume).

[0110] Output results: After verification, 200-280 valid raw data entries are output daily, and 120,000 historical work order data entries are synchronously retrieved and transmitted to the data fusion layer.

[0111] (II) Implementation of the Data Fusion Layer

[0112] 1. Classification and Conversion

[0113] The text data of the appeals was processed by using the jieba word segmentation tool to segment the grid-based on-site appeal data and the 95598 call transcript. After deduplication, semantic normalization was achieved by replacing synonyms (e.g., "unstable voltage" and "abnormal voltage" were unified as "abnormal voltage"). For example, "The voltage at home fluctuates and the air conditioner cannot start" was transformed into "Abnormal voltage, appliances cannot be used".

[0114] User tag data: Extract livelihood-sensitive tags (such as “nursing home in a certain district of Beijing” labeled as “livelihood-sensitive user - nursing home”) and electricity consumption characteristic tags (such as “average monthly electricity consumption ≥ 5000 kWh” labeled as “high energy consumption user”) from user profile data, with a unified format of “tag type - specific content”;

[0115] Internal business data: The "transformer load rate 85%, voltage value 200V" in the electricity consumption data and the "amount in arrears 300 yuan, duration of arrears 35 days" in the electricity bill data are linked through field standardization to form integrated data of "equipment status - power outage record - arrears information";

[0116] External channel data: Integrate the "payment consultation" data from the State Grid APP and the "regional power outage complaint" data from the 12345 government service feedback with the standardized grid-based on-site demand data (such as "potential hazards in the grid lines of Jianguo Road in a certain district"), and label them with "channel source - demand type".

[0117] Geospatial data: The original geospatial data of the "Power Grid Map" (such as the transformer's latitude and longitude "116.48°E, 39.92°N") and the real-time location data of grid workers (collected through the grid workers' mobile phone terminals via GPS, with a positioning error of ≤5 meters) are standardized using the WGS84 coordinate system and bound to specific addresses (such as "Building 1, No. 88 Jianguo Road, a certain district") to form unified geospatial data;

[0118] Historical work order data: For 120,000 past work order records, missing fields (non-critical fields) such as "associated equipment number" and "geospatial coordinates" were supplemented. Historical data from the same transformer area were used for interpolation to supplement the missing fields (e.g., if a work order is missing "transformer number", the transformer number of other work orders in the same transformer area is matched according to its address). The supplemented data is marked with "supplementation mark" and finally transformed into structured historical work order data.

[0119] 2. Association Mapping

[0120] Based on the "station-line-transformer-customer" topology, six types of structured data are linked using association keys such as "transformer area number = BJ-CY-001", "customer number = CY-80001", and "equipment number = BZ-1200". For example:

[0121] User ID CY-80001 (resident of No. 88 Jianguo Road, a certain district) → Substation ID BJ-CY-001 → Equipment ID BZ-1200 (transformer) → Geographic coordinates (116.48°E, 39.92°N) → Request text data "voltage abnormality" → Historical work order data "reported for repair due to voltage abnormality in March 2024".

[0122] 3. Secure Storage and Data Standards

[0123] Secure storage: Sensitive data such as user ID numbers and contact information are encrypted using the SM2 algorithm, and data integrity is verified using the SM3 algorithm (verification code length 32 bits). The data is stored in distributed storage nodes according to a unified schema of "work order basic information - user information - associated device information - geospatial information - request content information".

[0124] Data dictionary specifications: Define the range of field values ​​according to the State Grid digital data standard. For example, the value of "request type" is {fault reporting, payment consultation, complaint and suggestion, business processing}, and the value of "early warning level" is {red level, orange level, yellow level}.

[0125] Missing field completion: If a work order is missing "User contact information" (a non-critical field), fill it in as "None" and mark it with "Completion mark"; If a work order is missing "Equipment number" (a critical business field), remove it directly.

[0126] The final result is a structured data resource pool containing approximately 125,000 structured data entries, with a data interoperability rate of 96% and a data verification pass rate of 99.5%.

[0127] (III) Implementation of the Intelligent Analysis Layer

[0128] 1. Core Request Identification and Priority Response Module

[0129] To access the request text data, user tag data, and historical work order data from the structured work order data resource pool, follow these steps:

[0130] Step b1: Using the BERT semantic understanding model pre-trained with bert-base-chinese (input text length limited to 512 characters, learning rate 5e-5 during fine-tuning, batch size=32, iterations 100, dropout rate 0.1), the translated text of the 95598 call, "Our community experienced a power outage starting at 3 pm today. The elderly in my family need oxygen, and all appliances are unusable," was analyzed. The core semantic points were extracted: "power outage in the community, sensitive user (oxygen needing), appliances unusable." The actual need was identified as "emergency power restoration" (the surface request was "power outage repair report"). The semantic recognition accuracy was 93%.

[0131] Step b2: Read the "People's Livelihood Sensitive Tag - Oxygen Inhalation User" from the user's tag data and assign a score of 9 points to the sensitive tag;

[0132] Step b3: Based on the "request type = fault report, scope of impact = community (multiple households)" in the request text data, and combined with the "no duplicate request records" in the historical work order data, construct an urgency score of 10 points;

[0133] Step b4: Calculate the priority score according to the formula: 9×0.4+10×0.3+10×0.3=9.6 points, which is determined to be high priority and is pushed to the local management personnel of the Jianguo Road grid in a certain district. The warning information also includes "core intent: emergency power restoration, user profile: oxygen user (people's livelihood sensitive), historical service record: no duplicate repair requests".

[0134] 2. Internal and external data collaborative analysis module

[0135] Based on the above priority ranking results, retrieve internal business data, external channel data, and geospatial data, and execute the following steps:

[0136] Step c1: Extract internal business data from the structured chemical single data resource pool (transformer load rate of equipment number BZ-1200 is 105%, power outage record "15:00-16:30 unplanned power outage", no outstanding payment information), and combine the "station-line-transformer-customer" topology relationship to construct an equipment-user-request association graph in the Neo4j graph database, which is "transformer BZ-1200 → transformer area BJ-CY-001 → user CY-80001 (and 200 customers in the same transformer area) → request 'power outage'".

[0137] Step c2: Access the on-site verification data of the grid platform "Transformer BZ-1200 fault trip", the 12345 government feedback data "More than 10 households in the No. 88 Jianguo Road community complained about power outages" from external channels, and the online State Grid APP user interaction data "5 households in the same community reported power outages" to supplement the customer-side scenario information.

[0138] Step c3: Execute collaborative analysis rules: Internal business data monitoring detects "regional power outages with no planned power outages," and external channels receive 18 similar requests per hour (≥5 per hour), triggering a multi-household power outage red-level warning; simultaneously, for another user CY-90002, internal data shows "35 days of arrears, arrears amount of 500 yuan," combined with external channel payment activity area analysis data (the user's area has an online payment rate of 60% and 3 offline outlets), a customized collection plan is pushed, including "nearby payment outlets (Jianguo Road Business Hall, Guanghua Road self-service payment machine) and online payment guide (screenshot of the State Grid APP payment entrance)."

[0139] 3. Module for Tracing the Root Causes of Complaints and Providing Early Warning

[0140] Based on the above multi-dimensional analysis results, internal business data, historical work order data, and request text data are accessed, and the following steps are performed:

[0141] Step d1: Use the association rule algorithm (minimum support 0.2, minimum confidence 0.7) to analyze the association between "equipment fault records - historical power outage work orders - user request content": extract 3 fault records of transformer BZ-1200 in the past 3 months, 45 power outage work orders of the same type, and 38 user requests for "abnormal voltage" and "power outage". Calculate the support = 38 / 150 (total number of work orders in this area in the past 3 months) = 0.25 (≥0.2), and the confidence = 38 / 45≈0.84 (≥0.7). Identify the root cause of the high-frequency problem as "transformer BZ-1200 aging overload".

[0142] Step d2: Use a decision tree model (maximum depth 10, minimum number of sample splits 2, minimum number of sample leaf nodes 1, and Gini coefficient as the criterion) to classify the root cause and output the root cause priority "Level 1 (high risk)";

[0143] Step d3: Trigger risk threshold judgment: If transformer BZ-1200 has ≥3 failures per month (meets the threshold), an automatic source governance warning will be triggered and pushed to the operation and maintenance department. The warning information includes "Root cause: transformer aging and overload, Affected area: transformer substation BJ-CY-001, Affected users: 200 households, Recommended measures: replace the transformer";

[0144] Step d4: The maintenance department replaced the transformer 3 days later and tracked the frequency of similar complaints one month after the treatment: only 2 complaints of abnormal voltage occurred, a decrease of 87% compared with before the treatment (an average of 15 complaints per month), verifying that the root cause was effectively resolved.

[0145] 4. Tiered Early Warning and Control Module

[0146] Based on the root cause risk early warning information, combined with geospatial data and real-time location data of grid workers, the following steps shall be followed:

[0147] Step e1: Generate tiered warnings: The above-mentioned multi-household power outage event is judged as a red-level warning (corresponding to the scenario of "large-scale power outage + emergency requests from users with sensitive livelihoods"); a user's "payment consultation" request is judged as a yellow-level warning (corresponding to the scenario of "single household ordinary request"); a community's "multi-household voltage instability but no power outage" request is judged as an orange-level warning (corresponding to the scenario of "multi-household requests").

[0148] Step e2: Precise Dispatch: Obtain the real-time latitude and longitude of 30 grid workers, calculate the straight-line distance and reach time between the grid worker and No. 88 Jianguo Road Community (work order address): Grid worker Zhang XX (latitude and longitude 116.47°E, 39.91°N) is the closest (1.2 km, reach time 10 minutes), currently has 2 uncompleted work orders, the estimated time for each work order is 60 minutes, workload = 2×1 + 60×0.5 = 32; the workload of other grid workers is ≥40, therefore, the red-level power outage work order is assigned to Zhang XX, the on-site response time is shortened by 25% compared to traditional random dispatch;

[0149] Step e3: Upgraded control: Red-level power outage work orders are notified to local management personnel through a combination of "station message + SMS" and marked as "key supervision work orders" in the power supply command system; Orange-level warning work orders remind grid members every 2 hours; Yellow-level warning work orders are processed according to the regular procedure.

[0150] Step e4: Closed-loop verification: Grid worker Zhang XX arrives at the scene 15 minutes later, repairs the transformer fault and restores power. After the work order is processed, the semantic analysis algorithm is called to calculate the cosine similarity between the processed result ("The faulty transformer has been replaced and power supply has been restored") and the initial request ("The community is experiencing a power outage, and the elderly need power for oxygen therapy") and 0.92 (≥0.8). The problem is determined to be resolved. Another orange-level work order has been unprocessed for 3 hours (orange-level threshold ≤4 hours). The system automatically reminds the grid worker. If the grid worker does not respond in time, the work order will be escalated to the superior management department after 4 hours.

[0151] Step e5: Early warning push: Push the work order information to grid member Zhang XX through the WeChat robot platform, push the red-level early warning notification to the local management personnel via SMS, and provide real-time feedback on the work order processing progress ("dispatching in progress → processing in progress → completed") to the system.

[0152] (iv) Implementation of the application output layer

[0153] 1. Visualized display of work order risks

[0154] Based on the "one map of the power grid":

[0155] Work order risk distribution markers: Red, orange, and yellow dots are marked on the electronic map, corresponding to red, orange, and yellow warning work orders, respectively. Clicking on a dot will display the work order details (request content, user information, and processing progress).

[0156] Heat map of high-complaint areas: The Jianguo Road and Guanghua Road areas in a certain district are shown in red heat (high-complaint area), corresponding to the root cause of "aging transformers and potential line hazards", which helps management personnel to accurately allocate resources;

[0157] The grid worker's work status is updated in real time: it displays the grid worker's current location, the number of completed work orders, and the number of incomplete work orders, and supports multi-dimensional filtering and querying by "warning level, processing status, and region".

[0158] 2. Tiered early warning notification and closed-loop processing of work orders

[0159] Red alert: Notification will be sent via both SMS and in-site message, with a 100% response rate within 10 minutes;

[0160] Orange alert: Automatic reminders every 2 hours, with a 98% response rate within 4 hours;

[0161] Yellow-level alert: Standard procedures are being followed, with a 95% response rate within 8 hours;

[0162] Work order closed-loop processing: The completion rate of work orders is automatically calculated before the end of each workday. The completion rate of red-level work orders is 100%, orange-level work orders are 97%, and yellow-level work orders are 93%.

[0163] 3. Analysis reports are generated automatically.

[0164] The monthly report on key issues in power supply services in a certain district is automatically generated before the 5th of each month, and includes:

[0165] Distribution of request types: Fault reports accounted for 45%, payment inquiries accounted for 30%, complaints and suggestions accounted for 15%, and business processing accounted for 10%;

[0166] High-frequency root cause statistics: transformer aging (32%), inconvenient payment channels (25%), line overload (20%).

[0167] Response efficiency ranking: Jianguo Road grid has the shortest response time (average 25 minutes).

[0168] Supports exporting in PDF and Excel formats, with an export response time of 2 seconds.

[0169] IV. Implementation Results Verification

[0170] 1. Breaking down data silos: It has achieved unified integration of data from multiple systems such as power grid, marketing, government affairs and grid management, allowing managers to obtain full-dimensional work order information through a single platform;

[0171] 2. Resource allocation optimization: High-priority work orders are focused on core resources, avoiding the consumption of urgent resources by ordinary requests, thus improving the workload balance of grid workers by 40%;

[0172] 3. From passive response to proactive prevention: By addressing problems at the source and providing early warnings, issues such as transformer aging and line overload can be resolved in advance, reducing unexpected work orders;

[0173] 4. Improved customer satisfaction: Customer satisfaction surveys after implementation showed that the percentage of very satisfied customers increased from 68% to 92%, and no major complaints occurred.

[0174] This embodiment fully replicates the entire process of the system and method of this application, verifying the feasibility, stability and effectiveness of the technical solution, and can be promoted and applied in the field of customer service of the national power grid.

[0175] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0176] In various embodiments, the hardware implementation of the technology can directly utilize existing smart devices, including but not limited to industrial control computers, PCs, smartphones, handheld devices, and floor-standing devices. Its input device preferably uses an on-screen keyboard, its data storage and computing modules utilize existing memory, calculators, and controllers, its internal communication modules utilize existing communication ports and protocols, and its remote communication utilizes existing GPRS networks, the World Wide Web, etc.

[0177] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0178] In the embodiments provided by this invention, it should be understood that the disclosed apparatus / terminal devices and methods can be implemented in other ways. For example, the apparatus / terminal device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the shown or discussed mutual couplings or direct couplings or communication connections may be through some interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms. Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, i.e., they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0179] In the various embodiments of the present invention, the functional units can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated units can be implemented in hardware or as software functional units. If the integrated module / unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention can implement all or part of the processes in the methods of the above embodiments, or it can be accomplished by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

Claims

1. A work order intelligent analysis system based on the integration of geographic information and grid service data, characterized in that, It includes a data access layer, a data fusion layer, and an intelligent analysis layer that are connected sequentially. The data access layer is used to acquire multi-source heterogeneous core data related to power customer service and power grid grid to support intelligent work order analysis, and also to acquire past work order data accumulated by the system. The data fusion layer is used to standardize, integrate and correlate the effective data output by the data access layer. The resulting structured data includes request text data, user tag data, internal business data, external channel data, geospatial data and historical work order data, forming a structured work order data resource pool. The internal business data includes equipment status, power outage records and arrears information, the external channel data includes grid requests, government feedback and customer interactions, and the geospatial data includes grid worker real-time location data. The intelligent analysis layer includes a core demand identification and priority response module, an internal and external data collaborative analysis module, a demand root cause tracing and source early warning module, and a hierarchical early warning and control module, which are connected in sequence. The core demand identification and priority response module first analyzes the demand text data and user tag data in the structured order data resource pool, mines the customer's core intent and builds a priority model, and outputs the priority ranking results of highly sensitive / highly urgent demands. The internal and external data collaborative analysis module, based on the above priority ranking results, calls internal business data and external channel data in the structured work order data resource pool to conduct cross-dimensional collaborative analysis and output multi-dimensional analysis results. The demand root cause tracing and source early warning module, based on the above multi-dimensional analysis results, associates equipment status and historical work order data in the structured work order data resource pool to mine the core root cause of the demand, constructs a source early warning model, and outputs root cause risk early warning information. The hierarchical early warning and control module, based on the above root cause risk early warning information, combines geospatial data and grid member real-time location data in the structured work order data resource pool to generate hierarchical early warnings, achieve accurate order dispatch, and verify the processing effect.

2. The intelligent work order analysis system based on the integration of geographic information and grid service data as described in claim 1, characterized in that, The multi-source heterogeneous core data includes the original geospatial data from the "Power Grid One Map" platform of the power grid resource business middleware, electricity consumption data, electricity bill data and user profile data from the Marketing 2.0 system, on-site demand data from the grid management platform, call transcript data from the 95598 business support system, and customer interaction data from online channels. The electricity consumption data includes equipment status and power outage records, the electricity bill data includes overdue payment information, the user profile data includes sensitive tags related to people's livelihoods, and the customer interaction data includes government feedback information. The online channels include the State Grid website, WeChat, and Alipay. The data access layer adopts a primary and backup dual-link transmission mechanism. After data collection, invalid data is removed through format verification, integrity verification, and consistency verification algorithms, and the valid original data before standardization is output.

3. The intelligent work order analysis system based on the integration of geographic information and grid service data as described in claim 2, characterized in that, The standardized integration and correlation processing of the data fusion layer specifically includes the following steps: a1) Classification and Transformation: Based on the valid raw data output from the data access layer, the on-site request data and call transcript data are transformed into request text data through word segmentation, deduplication, and semantic normalization; sensitive tags related to people's livelihood and electricity consumption feature tags are extracted from user profile data and formatted uniformly to transform into user tag data; electricity collection data and electricity bill data are transformed into internal business data through field standardization and association; customer interaction data and the above-mentioned standardized on-site request data are integrated into external channel data; raw geospatial data and grid worker real-time location data are transformed into geospatial data through WGS84 coordinate system standardization and address binding; at the same time, past work order data are transformed into historical work order data through field completion and index association. a2) Association mapping: Based on the "station-line-transformer-customer" topology, an association index for the above six types of structured data is established through the association keys of the transformer area number, user number and equipment number; a3) Secure storage: Sensitive data is encrypted using the SM2 algorithm and data integrity is verified using the SM3 algorithm. Data is stored in a distributed storage node according to a unified schema of work order basic information, user information, associated device information, geospatial information and request content information, forming a structured work order data resource pool. a4) Data dictionary specifications: The unified schema includes fields for basic work order information, user information, associated equipment information, geospatial information, and request content information. The value range of each field is defined according to the State Grid digital data standard. a5) Missing field completion: When a key business field is missing, the data is removed. When a non-key field is missing, it is completed by interpolation based on the default value or historical data. The completed data is marked with a completion identifier and stored in the structured chemical order data resource pool.

4. The intelligent work order analysis system based on the integration of geographic information and grid service data as described in claim 1, characterized in that, The core request identification and priority response module is implemented by calling the request text data, user tag data, and historical work order data from the structured work order data resource pool, and executing the following steps: b1) The BERT semantic understanding model is used to perform in-depth analysis on the call translation text and work order description in the appeal text data. The BERT semantic understanding model adopts the bert-base-chinese pre-trained model, the input text length is limited to 512 characters, the learning rate in the fine-tuning stage is 5e-5, batch_size=32, the number of iterations is 100, and the dropout rate is 0.

1. The core semantic points extracted include frequent power outages, electricity bill disputes and voltage anomalies, and the difference between the surface appeal and the real needs is identified. b2) Read the livelihood-sensitive tags from the user profile in the user tag data. The livelihood-sensitive tags include schools, nursing homes and oxygen users. Assign a score to the sensitive tags. Among them, livelihood-sensitive users get 8-10 points and ordinary users get 0-5 points. b3) Based on the type of request and scope of impact in the text data of the request and the historical feedback data in the historical work order data, construct an urgency score, in which urgent requests are scored 8-10 points, general requests are scored 4-7 points, and ordinary requests are scored 0-3 points. b4) Calculate the score according to the priority scoring formula: Priority score = Sensitive tag score × 0.4 + Urgency score × 0.3 + Historical feedback score × 0.

3. The historical feedback score is assigned as follows: 8-10 points for no duplicate requests, 4-7 points for one duplicate request, and 0-3 points for two or more duplicate requests. Priority scores ≥ 8 points are high priority, 5-7 points are medium priority, and < 5 points are low priority. High priority requests are pushed to local management personnel first. The warning information includes the core intent, user profile, and historical service records.

5. The intelligent work order analysis system based on the integration of geographic information and grid service data as described in claim 1, characterized in that, The internal and external data collaborative analysis module is implemented by calling internal business data, external channel data, and geospatial data from the structured single data resource pool, and executing the following steps: c1) Internal data collaboration: Extract internal business data, including equipment status data, power outage record data, and overdue payment information data, from the structured single data resource pool, and construct an equipment-user-demand association graph by combining the "station-line-transformer-customer" topology relationship; c2) External data collaboration: Access the on-site verification data of the grid platform, the 12345 government feedback data and the online State Grid user interaction data from the external channel data of the structured chemical single data resource pool to supplement customer-side scenario information; c3) Collaborative analysis rules: When internal business data detects a regional power outage and there is no planned power outage information, if there are ≥5 similar requests per hour in external channel data, a red-level warning for multiple power outages will be triggered; when internal business data shows that users are ≥30 days in arrears, combined with the analysis data of active payment areas in external channel data, a customized collection plan including nearby payment outlets and online payment guides will be pushed.

6. The intelligent work order analysis system based on the integration of geographic information and grid service data as described in claim 1, characterized in that, The implementation method of the request root cause tracing and source early warning module is as follows: call the internal business data, historical work order data and request text data in the structured work order data resource pool, and execute the following steps: d1) The association rule algorithm is used to analyze the correlation between equipment failure records in internal business data, historical power outage work orders in historical work order data, and user request content in request text data. The minimum support of the association rule algorithm is 0.2 and the minimum confidence is 0.

7. The algorithm identifies the root causes of high-frequency problems, including aging lines, insufficient convenience of payment channels, and equipment overload. d2) The identified root cause types are classified using a decision tree model. The maximum depth of the decision tree model is 10, the minimum number of sample splits is 2, the minimum number of sample leaf nodes is 1, the criterion uses the Gini coefficient, and the root cause priority is output. d3) Set root cause risk thresholds, which include equipment failures ≥3 times per month, similar requests ≥10 times per week, line failure intervals ≤7 days, and payment channel complaints ≥15 times per month. Once the thresholds are reached, a source governance warning will be automatically triggered and pushed to the operation and maintenance department or business optimization department. d4) Track the changes in work order data after source governance, and verify the effectiveness of root cause resolution by comparing the frequency of similar requests before and after governance.

7. The intelligent work order analysis system based on the integration of geographic information and grid service data as described in claim 1, characterized in that, The implementation method of the hierarchical early warning and control module is as follows: calling geospatial data, appeal text data, and various judgment result data from the structured chemical single data resource pool, and executing the following steps: e1) Generation of graded early warnings: Based on the aforementioned analysis results, graded early warnings are generated. Red-level early warnings correspond to scenarios including large-scale power outages and emergency requests from users with sensitive livelihoods. Orange-level early warnings correspond to scenarios including requests from multiple households and overdue unprocessed work orders. Yellow-level early warnings correspond to scenarios including ordinary requests from a single household and potential risks. e2) Precise work order assignment: Based on the real-time location information of grid workers in geospatial data, the grid workers' real-time latitude and longitude are obtained, the straight-line distance and reach time between the grid worker and the work order address are calculated, and the work orders are dynamically assigned in combination with the grid workers' workload. The distance weight is 0.5 and the workload weight is 0.5, and the work orders are assigned to the grid workers who are closest and have the least workload first. e3) Enhanced control: Red-level warning work orders are notified to local management personnel through both in-station messages and SMS, and are marked as key supervision work orders in the power supply command system; Orange-level warning work orders are reminded once every 2 hours, and Yellow-level warning work orders are processed according to the regular procedure. e4) Closed-loop verification: After the work order is processed, a semantic analysis algorithm is called to calculate the semantic similarity between the processing result and the initial request in the request text data. The semantic similarity is calculated using cosine similarity, with a threshold of ≥0.

8. If the similarity is ≥0.8, the problem is considered resolved, and if it is <0.8, a secondary warning is triggered. Work orders that are not processed within the time limit are automatically escalated to the superior management department according to the time limit requirements of red level ≤2 hours, orange level ≤4 hours, and yellow level ≤8 hours. e5) Early warning push: Early warning information is automatically pushed through the WeChat robot platform and SMS interface, and the progress of work order processing is fed back to the system in real time.

8. The intelligent work order analysis system based on the integration of geographic information and grid service data according to claim 1, characterized in that, The system also includes an application output layer; the application output layer is communicatively connected to the intelligent analysis layer and is used to provide functions such as visual display of work order risks, hierarchical early warning notification, closed-loop processing of work orders, and automatic generation of analysis reports based on the output results of the intelligent analysis layer; the work order risk visualization display function realizes the distribution of work order risks, presents heat maps of areas with high complaint rates, and updates the work status of grid members in real time based on the "power grid map"; it supports multi-dimensional filtering and querying by region, time, complaint type, and early warning level; the automatic analysis report generation function can generate monthly reports on key issues in power supply services and periodic hot complaint analysis reports.

9. A work order intelligent analysis method based on the integration of geographic information and grid service data, characterized in that, The work order intelligent analysis system based on the integration of geographic information and grid service data, as described in any one of claims 1-8, includes the following steps: S1. Data Access: Acquire multi-source heterogeneous core data related to power customer service and power grid grid to support intelligent analysis of work orders, and call up past work order data accumulated by the system; S2. Data Fusion: The multi-source heterogeneous core data and past work order data are standardized, integrated, and correlated to transform them into structured data containing request text data, user tag data, internal business data, external channel data, geospatial data, and historical work order data, forming a structured work order data resource pool; the internal business data includes equipment status, power outage records, and overdue payment information, the external channel data includes grid requests, government feedback, and customer interactions, and the geospatial data includes grid worker real-time location data; S3. Intelligent Analysis: First, analyze the request text data and user tag data in the structured work order data resource pool to obtain priority ranking results; then, based on the priority ranking results, call internal business data and external channel data to carry out cross-dimensional collaborative analysis to obtain multi-dimensional analysis results; next, based on the multi-dimensional analysis results, link equipment status and historical work order data to mine the core root causes of the requests and obtain root cause risk warning information; finally, based on the root cause risk warning information, combine geospatial data and grid member real-time location data to generate hierarchical warnings, achieve accurate work order dispatch, and verify the processing effect.

10. The intelligent work order analysis method based on the integration of geographic information and grid service data according to claim 9, characterized in that, The method further includes the following steps: S4. Application Output: Based on the processing results of the intelligent assessment steps, visualize the work order risk, issue tiered early warnings, perform closed-loop processing of work orders, and automatically generate analysis reports.

Images