A large model-based full-channel customer service work order early warning method and system

Abstract

The application provides a large model-based full-channel customer service order early warning method and system, the method comprises the following steps: acquiring full-channel multi-source customer service order data; using a large language model, classifying customer service orders into sensitive orders and non-sensitive orders according to the channels of the customer service order data and sensitive words in the orders, and classifying customer service orders into different levels according to the attribute contents of the customer service orders; analyzing the orders through an order analysis intelligent agent module, outputting the problem judgment and risk level corresponding to the orders; the order analysis intelligent agent module comprises an industry expansion intelligent agent for analyzing the industry expansion process of the customer service orders, and an electricity price and electricity charge intelligent agent for analyzing the customer service orders with the first classification result of electricity price and electricity charge; based on the risk warning level of the order and the order classification type, order warning information is sequentially generated, and the efficiency and reliability of the customer service order warning are improved.

Description

Technical Field

[0001] This invention relates to the field of customer service work order processing, and in particular to an omnichannel customer service work order early warning method and system based on a large model. Background Technology

[0002] The timely and effective processing of customer service work orders in power supply companies not only allows business managers to conveniently obtain more reliable business support data and analysis results, reducing human error and improving the stability of power supply service operation; but also significantly shortens the response cycle for the electricity needs of enterprises, institutions and residents, improves the efficiency of handling problems such as fault repair and business consultation, and reduces production stoppages and inconvenience caused by power problems.

[0003] However, the current status and existing problems in the field of work order early warning analysis mainly include: First, the work orders that the supply and dispatch center handles on a daily basis are scattered across various systems, lacking unified means and methods for dispatching, early warning and handling. Currently, customer complaint work orders across all channels include 95598 fault reporting work orders, 95598 non-emergency repair and expedited work orders, 12398 work orders, public opinion work orders, and operation and maintenance verification work orders in the internal network system; and 12345 work orders and vehicle networking work orders in the external network system.

[0004] Secondly, work order data is scattered, work order characteristics are difficult to identify, historical work order data is incomplete, and it is impossible to comprehensively consider factors such as work order type, responsible party and on-site conditions to formulate work order service strategies, determine the cause of work orders, and work order processing efficiency needs to be improved.

[0005] Third, it is difficult to determine the risk of work orders. Currently, the risk of new work orders is identified by human experience. Work orders with duplicate or related requests are also screened manually. In-depth analysis of work orders also relies on manual work. The quality and efficiency of the work need to be further improved.

[0006] To address this issue, the present invention provides a method and system for early warning of omnichannel customer service work orders based on a large model, in order to solve the above problems. Summary of the Invention

[0007] To address the problems existing in the prior art, this invention proposes a method and system for omnichannel customer service work order early warning based on a large model. This effectively solves the problem of low efficiency and reliability of customer service work order early warning caused by the prior art, and effectively improves the efficiency and reliability of customer service work order early warning.

[0008] The first aspect of this invention provides a method for early warning of omnichannel customer service work orders based on a large model, comprising: Acquire customer service work order data from multiple channels and sources; Using a large language model, customer service work orders are classified into sensitive work orders and non-sensitive work orders based on the channel of the customer service work order data and the sensitive words in the work orders. Customer service work orders are also classified into different levels based on the attribute content of the customer service work orders. The first-level classification results are business expansion and installation and electricity price and electricity fee. A work order analysis and judgment module is created to analyze and judge work orders, outputting the corresponding problem judgment and risk level. This module includes two parts: a business expansion and installation intelligence agent and an electricity price and fee intelligence agent. The business expansion and installation intelligence agent, based on the power supply service knowledge base and linking historical work order data and solutions, analyzes the business expansion process of customer service work orders and outputs a risk warning level. The electricity price and fee intelligence agent, based on the power supply service knowledge base, analyzes customer service work orders whose primary classification result is electricity price and fee, and outputs a risk warning level when the primary classification result is electricity price and fee. Work order warning information is generated sequentially based on the risk warning level and work order classification type of the work order.

[0009] A second aspect of this invention provides an omnichannel customer service work order early warning system based on a large model, comprising: The acquisition module retrieves customer service work order data from multiple sources across all channels. The classification module uses a large language model to classify customer service work orders into sensitive and non-sensitive work orders based on the channel of the customer service work order data and the sensitive words in the work orders. It also classifies customer service work orders into different levels based on the attribute content of the customer service work orders. The first-level classification results are business expansion and installation and electricity price and electricity fee. The system includes a creation and analysis module, and a work order analysis intelligent agent module. This module analyzes and judges work orders, outputting the corresponding problem judgment and risk level. The work order analysis intelligent agent module includes two parts: a business expansion application intelligent agent and an electricity price and fee intelligent agent. The business expansion application intelligent agent, based on the power supply service knowledge base and linking historical work order data and solutions, analyzes the business expansion process of customer service work orders and outputs a risk warning level. The electricity price and fee intelligent agent, based on the power supply service knowledge base, analyzes customer service work orders whose primary classification result is electricity price and fee, and outputs a risk warning level when the primary classification result is electricity price and fee. The generation module generates work order warning information sequentially based on the risk warning level and work order classification type of the work order.

[0010] The technical solution adopted in this invention has the following technical effects: 1. This invention utilizes a large language model to classify customer service work orders into sensitive and non-sensitive work orders based on the channel of the work order data and the sensitive words in the work orders. It further classifies customer service work orders into different levels based on their attribute content. A work order analysis intelligent agent module analyzes the work orders and outputs the corresponding problem judgment and risk level. This module includes a business expansion application intelligent agent and an electricity price and fee intelligent agent. The business expansion application intelligent agent, based on a power supply service knowledge base and in conjunction with historical work order data and solutions, analyzes the business expansion process of customer service work orders and outputs a risk warning level. The electricity price and fee intelligent agent, based on a power supply service knowledge base, analyzes customer service work orders whose primary classification result is electricity price and fee, and outputs a risk warning level. Based on the risk warning level and work order classification type, work order warning information is generated sequentially, effectively solving the problem of low efficiency and reliability of customer service work order warnings caused by existing technologies, and effectively improving the efficiency and reliability of customer service work order warnings.

[0011] 2. In the technical solution of this invention, the intelligent agent module for work order analysis reduces the time spent by business personnel in verifying the authenticity of work order content and analyzing root causes in the processing of customer request work orders. It proactively identifies sensitive work orders and work orders of different first-level categories, reduces the workload of grassroots staff in reviewing and analyzing work orders, shortens the traditional handling process, reduces the burden on grassroots staff, saves labor costs for enterprises, and improves company efficiency.

[0012] 3. The technical solution of this invention utilizes intelligent analysis and judgment of work orders across all channels. This allows business management personnel to conveniently obtain more reliable business support data and analysis results, reducing human error, mitigating various power supply service issues, and improving the stability of power supply service operations. In terms of management models, it breaks down the hierarchical barriers between "municipal companies, county companies, and power supply stations," improving the efficiency of grassroots business response. Regarding talent development, the system's built-in "disposal plan recommendation" function acts as a "digital mentor" for new employees, shortening the business proficiency period.

[0013] 4. The technical solution of this invention has a significant role in improving the efficiency of social public services through intelligent judgment and analysis of work orders across all channels. By integrating power service requests from multiple channels, it enables early identification and graded warning of work order risks, greatly shortens the response cycle for the power needs of enterprises, institutions and residents, improves the efficiency of handling issues such as fault repair and business consultation, effectively reduces production stoppages and inconvenience caused by power problems, and truly enhances residents' sense of security and happiness.

[0014] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit the invention. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart illustrating the method of Embodiment 1 in the present invention; Figure 2 This is a schematic diagram of the architecture of the method in Embodiment 1 of the present invention; Figure 3 This is a schematic diagram of the work order analysis intelligent agent module in the method of Embodiment 1 of the present invention; Figure 4 This is a schematic diagram of the intelligent judgment logic for work orders in Embodiment 1 of the present invention, where the primary classification result is electricity price and electricity charge, and the secondary classification result is electricity price execution and electricity charge settlement. Figure 5 This is a schematic diagram of the intelligent judgment logic for work orders in Embodiment 1 of the present invention, where the primary classification result is electricity price and the secondary classification result is power outage / restoration due to arrears. Figure 6 This is a schematic diagram of the intelligent judgment logic for work orders in Embodiment 1 of the present invention when the primary classification result is either abnormal electricity or electricity bill and personnel service behavior. Figure 7 This is a schematic diagram of the system structure in Embodiment 2 of the present invention. Detailed Implementation

[0017] To clearly illustrate the technical features of this solution, the invention will be described in detail below through specific embodiments and in conjunction with the accompanying drawings. The following disclosure provides many different embodiments or examples for implementing different structures of the invention. To simplify the disclosure of the invention, components and arrangements of specific examples are described below. Furthermore, reference numerals and / or letters may be repeated in different examples. This repetition is for simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or arrangements discussed. It should be noted that the components illustrated in the drawings are not necessarily drawn to scale. Descriptions of well-known components, processing techniques, and processes are omitted in this invention to avoid unnecessarily limiting the invention.

[0018] Example 1 like Figures 1-2 As shown, this invention provides a method for omnichannel customer service work order early warning based on a large model, including: S1, acquire customer service work order data from multiple sources across all channels; S2 utilizes a large language model to classify customer service work orders into sensitive and non-sensitive work orders based on the channel of the customer service work order data and the sensitive words in the work orders. It also classifies customer service work orders into different levels based on the attribute content of the customer service work orders. The first-level classification results are business expansion and installation and electricity price and electricity fee. S3, Create a work order analysis and judgment intelligent agent module. This module analyzes and judges work orders, outputting the corresponding problem judgment and risk level. The work order analysis and judgment intelligent agent module includes: a business expansion and installation intelligent agent and an electricity price and fee intelligent agent. The business expansion and installation intelligent agent, based on the power supply service knowledge base and linking historical work order data and solutions, analyzes the business expansion process of customer service work orders and outputs a risk warning level. The electricity price and fee intelligent agent, based on the power supply service knowledge base, analyzes customer service work orders whose primary classification result is electricity price and fee, and outputs a risk warning level when the primary classification result is electricity price and fee. S4 generates work order warning information sequentially based on the risk warning level and work order classification type of the work order.

[0019] In step S1, which involves work order data aggregation, the system can connect to platforms such as Marketing 2.0, 95598, and 12345 to integrate work order data from multiple channels. This enables centralized management and service of marketing work order data and achieves unified aggregation of multi-source work order data. Work order data cleaning and standardization involve identifying and correcting non-standard data using a large model to construct a standardized work order dataset and improve data quality.

[0020] By applying a technical approach that integrates "RPA + MCP service + small model + large model + RAG + prompt word engineering + intelligent agent + embedded system", problems such as time-consuming and labor-intensive work order collection, cumbersome classification and organization, complex analysis logic, difficulty in root cause analysis, and untimely handling can be solved.

[0021] Knowledge base construction: In the knowledge base construction phase, the State Grid Knowledge Base "Zhizhi" is used to build a vector knowledge base for work orders across all power supply and service channels. Relevant normative and institutional documents are collected based on the needs of power supply quality business. The construction of the power supply and service knowledge base involves "transferring" relevant content from the State Grid Knowledge Base "Zhizhi" to the vector knowledge base for work orders across all power supply and service channels, and supporting relevant intelligent agents to call it through API.

[0022] In the work order collection stage, process automation technology (RPA) is used to automatically collect six types of work orders: marketing 95598 work orders, fault repair work orders, 12398 work orders, public opinion work orders, expedited work orders, and 12345 work orders. It can also collect 95598 fault repair work orders, 95598 non-emergency repair and expedited work orders, 12398 work orders, public opinion work orders, and operation and maintenance verification work orders from the intranet system. The work orders are stored in the omnichannel work order business data center, and the omnichannel work order intelligent early warning analysis application uses its own system interface to realize centralized monitoring of work orders, laying the data foundation for subsequent work order analysis and handling.

[0023] In step S2, a large language model is used to classify customer service work orders into sensitive work orders and non-sensitive work orders based on the channel of the customer service work order data and the sensitive words in the work orders. Customer service work orders are classified into different levels according to the attribute content of the customer service work orders. The first-level classification results are business expansion application and electricity price and electricity fee.

[0024] Work order classification can be achieved through a work order classification intelligent agent. The construction of the work order classification intelligent agent involves establishing a business research template, including a business requirement summary, work order data collection, and intelligent analysis logic. Based on this template, relevant data such as customer profiles and work orders are retrieved from the omnichannel work order business data center via the MCP service. Using a large model and intelligent work order classification prompts, work orders are classified, customer entity analysis is performed, and sensitive work orders are identified. The classification results are stored in the "Analysis Support Database" of the omnichannel business data center via API service, supporting subsequent detailed analysis by the work order analysis intelligent agent.

[0025] Sensitive work order identification can be achieved through a sensitive work order sub-agent. The input parameters of the sensitive work order sub-agent are the work order number and channel type, and the output parameters are whether it is a sensitive work order and the sensitive type. The sensitive types are specifically 95598 opinions, 95598 reminders, 95598 repeated requests, public opinion, 12398, etc.

[0026] The specific algorithm for identifying sensitive work order agents is shown in Table 1 below: Table 1: Algorithm for Judging Sensitive Work Order Agents

[0027] Sensitive words can include "spring irrigation", "electrification of wells", "agricultural irrigation", "college entrance examination", "key livelihood demands", "coal-to-electricity conversion", "personal injury or death", "people trapped in elevators", "patients receiving oxygen", "upgrading 12398, it is recommended to contact customers on the first day", "multiple channels of feedback have not been resolved, it is recommended to contact customers on the first day", and "intention to petition".

[0028] The most prominent issues include "designated three categories for business expansion and installation", "arbitrary charges", "exceeding time limits", "staff working under the influence of alcohol", "service attitude of business hall staff", "service attitude of emergency repair personnel", and "frequent power outages".

[0029] Multi-level classification of work orders can be achieved through a multi-level classification sub-agent, which classifies customer service work orders into different levels based on their attribute content. The first-level classification results are business expansion application and electricity price / fee categories.

[0030] Classification can be achieved through keywords. A mapping relationship can be established between keywords at different levels of classification. For example, in the category of "business expansion and installation," common electricity usage scenarios and their corresponding business types and keywords could be used as primary classifications, such as "business processing" and "business expansion." Then, secondary classifications could be based on business types, which could be: (1) New installations for low-voltage residential use: residential electricity, lighting electricity, and household electricity; (2) New installations and capacity increases for low-voltage non-residential applications: Low-voltage electricity use situations such as residential charging piles, agricultural production, agricultural irrigation and drainage, and general industrial and commercial applications that are not considered residential self-use electricity. (3) High-voltage new installation and capacity expansion: high-voltage power supply and installation of dedicated transformers; (4) Low-voltage batch installation: Batch installation of low-voltage residential and non-residential new installations; (5) New installation of supporting facilities in the community: electricity application for newly built residential communities.

[0031] On the other hand, keywords corresponding to common electricity usage scenarios related to electricity prices and fees: for example, keywords such as "electricity price", "electricity fee", "electricity cost", etc. are classified as primary categories, and the corresponding secondary categories include electricity price execution, electricity fee settlement, and power suspension and restoration due to arrears. Then, secondary categories are made according to business type.

[0032] Electricity price refers to the price at which power companies sell electricity to users; it is the monetary expression of the value of electricity as a commodity. Electricity charge, on the other hand, is the fee a user pays based on their electricity consumption and the corresponding electricity price. Electricity prices vary depending on factors such as electricity usage category, voltage level, and time of day. Examples include residential electricity prices, non-residential lighting electricity prices, and industrial and commercial electricity prices. Residential electricity prices may also be subject to tiered pricing mechanisms.

[0033] (I) Electricity Price Implementation: This includes work order classification, data-assisted analysis, and matching analysis of electricity price implementation rules and regulations. It also involves verifying customer file types, checking implemented electricity prices, verifying electricity usage change records (checking electricity price strategies), quantitative and proportional calculations (within electricity price information), and electricity consumption curves (monthly electricity consumption comparison: monitoring recent comparisons and year-on-year comparisons; daily electricity consumption comparison: monthly daily electricity consumption change - year-on-year comparison). Meter Reading (incorrect / missing readings): Incorrect readings (check the last meter replacement date; within 2 years is considered suspicious), missing readings (check for zero electricity sales in the last 12 consecutive months). Electricity Bill Collection: This involves verifying customer type (fee-controlled / postpaid), and checking for outstanding payments (checking electricity bill balances).

[0034] (II) Electricity bill settlement: Whether the electricity bill has been received (customer file / electricity consumption information / recent payment record). Electricity bill refund: suspected fraudulent requests (based on past work order records, using phone numbers as the reference, check whether special customer markings indicate fraudulent requests, and whether there are multiple (3 or more) similar refund requests from the same number). Invoices: Business hall invoice service / invoice information inquiry.

[0035] (III) Power outage and restoration due to arrears: Power outage due to arrears (compliance of power outage, whether the power outage approval process has exceeded the time limit or been suspended, whether the IoT power outage strategy is standardized (verify arrears); whether SMS messages have been sent (all those implementing fee control must send SMS messages), fee control below the threshold, 3 days and half an hour in advance notice, non-fee control accounts, 7 days in advance notice (temporarily not verified); verification of power outage signs (Marketing 2.0 customer information, 360 homepage); verification of data on power outages that should have been stopped but were not (landlord reports power outages due to arrears), check power outage and restoration records in the Marketing 2.0 system). Power restoration due to arrears: whether the restoration conditions have been met (whether there are new arrears, balance greater than zero), verification of unsuccessful power restoration (collect data on abnormal automatic power restoration).

[0036] (iv) Abnormal electricity consumption and charges (when the primary classification result is electricity price and charge, the secondary classification is preferred), curve data verification, monthly electricity consumption comparison: monitor recent comparisons and year-on-year comparisons. Daily electricity consumption comparison: monthly daily electricity consumption changes (find patterns, year-on-year comparisons). 24-hour curves, check electricity consumption behavior (high-consumption equipment left on / equipment turned off). Electricity price execution / peak-valley tiered calculation, check miscellaneous business work orders, check peak-valley change records. Check whether the special identifier in the customer file is electric heating (mainly work order feedback indicates it, but the actual file does not). Query the annual cumulative electricity consumption, calculate whether it meets the tiered upgrade standard (there are clear calculation rules, clarify according to the facts), and consider transfer of ownership, coal-to-electricity conversion, and multiple people in one household.

[0037] (v) Personnel service behavior (when the primary classification result is electricity price and electricity fee, the secondary classification is preferred, mainly for labeling and in-depth analysis in the later stage of service), collection frequency, collection channels, and collection methods.

[0038] Preferably, in addition to business expansion and installation categories and electricity price and fee categories, the primary classification may also include power supply quality, service channels, etc., which are not limited in this invention.

[0039] In step S3, such as Figure 3 As shown, the work order analysis and judgment intelligent agent module includes: a business expansion and installation intelligent agent and an electricity price and fee intelligent agent. The business expansion and installation intelligent agent is used to analyze the business expansion process and personnel behavior norms of customer service work orders based on the power supply service knowledge base, and to output the risk warning level. The electricity price and fee intelligent agent is used to analyze customer service work orders whose primary classification result is electricity price and fee based on the power supply service knowledge base, and to output the risk warning level when the primary classification result of the work order is electricity price and fee.

[0040] The intelligent agent for work order analysis can analyze the following processes through sub-agents: 1. Intelligent power supply quality assessment (optimization) Power supply quality assessment focuses on scenarios involving multiple power outages and voltage quality issues. It leverages multi-source information, including work orders from all channels, 95598 repair requests, marketing user information, power outage and restoration records, and transformer area power failure data, to analyze outages from dimensions such as nature, scope, restoration timeliness, and correlation matching. The final conclusions encompass outage nature determination, outage scope, voltage curves, voltage quality, and risk levels, providing comprehensive support for the efficient handling of power supply quality problems.

[0041] 2. Intelligent analysis and selection of service channels Service channel analysis focuses on core scenarios such as abnormal SMS sending and receiving, and service standards of grid and business hall personnel. Relying on multi-source data such as SMS records, subscription orders, and customer information, it analyzes from dimensions such as SMS sending, subscription status, information consistency, and service compliance, and outputs the authenticity of work orders, root causes of problems, risk warning levels, and integrated conclusions, providing comprehensive support for the accurate and efficient handling of service channel complaints and problems.

[0042] 3. Intelligent analysis of business expansion and installation applications The business expansion application analysis focuses on the business expansion process (business expansion time limit, process handling) and personnel service behavior scenarios. It is supported by business expansion work order data, electricity meter data and knowledge base time limit requirements. It analyzes from three dimensions: process compliance, process standardization and personnel service. Finally, it outputs business type, problem judgment and risk level (level 1 / level 2 warning) to support the accurate and efficient handling of business expansion application related issues.

[0043] 4. Intelligent analysis of electricity prices and charges The electricity price and bill analysis covers core scenarios such as electricity price execution, electricity bill settlement, power outages and restorations due to arrears, abnormal electricity consumption and billing, and personnel service behavior. It is supported by multi-source data such as business processing records, electricity bills, power outage and restoration records, electricity meter data, and contract information. It analyzes from dimensions such as electricity price compliance, billing accuracy, power outage and restoration process standardization, reasonableness of electricity consumption and billing fluctuations, and personnel service standardization. Finally, it outputs problem type determination, anomaly root cause analysis, and work order risk level, providing comprehensive support for the accurate and efficient handling of electricity price and bill related complaints and issues.

[0044] Specifically, the business expansion process assessment includes assessment of work orders exceeding time limits, assessment of external circulation, assessment of non-standard process handling, assessment of suspected illegal charges, and assessment of suspected compliance with preset specified types.

[0045] The following will provide a detailed explanation through the analysis of different types of business expansion processes: 1. Analysis of Overdue Work Orders Data Acquisition: (1) In the Energy Internet Marketing Service System (hereinafter referred to as Marketing 2.0 System), query the work order information for at least one month before and after the date in the customer's description of the request: Select the unit corresponding to the electricity location described by the customer for "Power Supply Unit"; Select one of "Business Type" according to the customer's needs: "High Voltage New Installation Capacity Increase", "Low Voltage Residential New Installation Capacity Increase", "Low Voltage Non-Residential New Installation Capacity Increase", "Low Voltage Batch New Installation", "Community Supporting New Installation"; Select "All" for "Work Order Status".

[0046] Common electricity consumption scenarios and their corresponding business types and keywords: ① New installations for low-voltage residential use: residential electricity, lighting electricity, and household electricity; ② New installations and capacity increases for low-voltage non-residential applications: Low-voltage electricity use situations such as residential charging piles, agricultural production, agricultural irrigation and drainage, and general industrial and commercial applications that are not considered residential self-use electricity. ③ High-voltage new installation and capacity expansion: High-voltage power supply and installation of dedicated transformers; ④Low-voltage bulk installation: Carry out bulk installations of low-voltage residential and non-residential units; ⑤ New community amenities: Electricity connection application for newly built residential communities; Access path: Business Expansion Access / Statistics Query / Query Topic / Work Order Query (2) In the work order details retrieved, determine whether there are any work orders related to the customer based on the user name, electricity address, contact information, electricity demand and other information in the work order description.

[0047] Judgment logic: If no application for business expansion related to the user is found, it indicates that the user's electricity application has not been accepted, which is suspected to be "off-balance-sheet" and is judged as "Level 1 warning"; if the application for business expansion related to the user can be found, the work order needs to be opened to further check the start time, processing time and completion time of the specific steps of the work order.

[0048] The processing time for each service is as follows: All time limits are calculated on a working day basis, excluding statutory holidays and weekends, and business hours are calculated in hours, minutes, and seconds, not whole days. If the calculated time exceeds the business processing requirements, it will be classified as "Level 1 Warning"; for residential new installation and capacity expansion business, if the time from the start of the business application to the completion of the meter installation and power connection exceeds 5 working days, it will be classified as "Level 1 Warning"; for low-voltage non-residential new installation and capacity expansion business of residential self-use charging piles, if the time from the start of the business application to the completion of the meter installation and power connection exceeds 15 working days, it will be classified as "Level 1 Warning".

[0049] 2. The assessment of extracorporeal circulation specifically includes: Data Acquisition: (1) In the Energy Internet Marketing Service System (hereinafter referred to as Marketing 2.0 System), query the work order information for at least one month before and after the date in the customer's description of the request: Select the unit corresponding to the electricity location described by the customer for "Power Supply Unit"; Select one of "Business Type" according to the customer's needs: "High Voltage New Installation Capacity Increase", "Low Voltage Residential New Installation Capacity Increase", "Low Voltage Non-Residential New Installation Capacity Increase", "Low Voltage Batch New Installation", "Community Supporting New Installation"; Select "All" for "Work Order Status".

[0050] (2) In the work order details retrieved, determine whether there are any work orders related to the customer based on the user name, electricity address, contact information, electricity demand and other information in the work order description.

[0051] Judgment Logic: If the following situation occurs, it is judged as "extracorporeal circulation" and classified as "Level 1 Warning": (1) The customer clearly stated the date of acceptance of the business, and could be accurate to the day, but the user's electricity business application, including the termination work order, could not be found.

[0052] (2) The customer stated that no electricity meter was installed, but the user's business process could be found to be archived, but no data was collected by the electricity meter. The specific steps are as follows: ① The query process has been archived; ② Check the user's electricity meter; In the "Work Order Inquiry" query results, click on the user number under "User Number". The page will jump to the customer information interface. On this interface, click on "Metering Information". The page will jump to the metering information interface. Scroll down to the bottom of the metering information interface to find the electricity meter number and copy the number.

[0053] ③ Query user electricity meter data in the new generation electricity information collection system Access path: New Generation Electricity Information Collection System - Batch Data Query; The redirected page is as follows. On the redirected page, enter the queried electricity meter number, adjust the start time to 1 day before the customer's business expansion process is archived, and adjust the end time to 3 days after the customer's business expansion process is archived to query the customer's electricity meter data. If the user's electricity meter data cannot be found, or the data retrieved is 0 for 3 consecutive days, it is determined to be an external circulation.

[0054] 3. Analysis of non-standard process handling Data acquisition: Customer work order description; Judgment Logic: If a customer reports that staff have required them to make multiple trips to the business hall to supplement documents, sign, stamp, or run errands, it is judged as a Level 1 warning.

[0055] 4. Suspected illegal charges Data acquisition: Customer work order description; Judgment Logic: The following situations are judged as “Level 1 Warning”: (1) Power supply company staff require customers to pay for materials such as electricity meters, transformers, cables, and meter boxes, as well as construction fees, lost work fees, and overtime fees; (2) Among residential users, power supply company staff require customers to purchase the cable before the electricity meter themselves; (3) Require customers to pay other business fees besides high reliability power supply fees.

[0056] 5. Suspected "Three-Designation" (Suspected to meet the preset designation type judgment) Data acquisition: Customer work order description; Judgment Logic: The following situations are judged as “Level 1 Warning”: (1) Customers report that the power supply company staff recommend, introduce, or designate construction, design, testing, and material supply units to them; (2) Power supply company staff restrict customers from choosing construction, design, testing, and material supply units by means of not installing meters or not accepting them.

[0057] Preferably, the business expansion process assessment can also include charging pile assessment and personnel behavior assessment. The business related to charging piles and the low-voltage non-residential new installation / capacity expansion and high-voltage new installation / capacity expansion in the assessment of overdue work orders have the same service risks and the same query methods. Personnel behavior assessment includes suspected refusal to accept service (with slightly lower priority for reminders), suspected evasiveness, and suspected poor service attitude. Assessment logic: Customer complaints of evasiveness and suspected poor service attitude from power company staff are classified as "Level 1 Warning." The risk warning level of Level 1 Warning is higher than that of Level 2 Warning.

[0058] Specifically, based on the power supply service knowledge base, when the primary classification result of a work order is electricity price and charge, the customer service work orders with this primary classification result are analyzed, and a risk warning level is output. When the primary classification result is electricity price and charge, the corresponding secondary classification results include electricity price execution, electricity charge settlement, and power outage / restoration due to arrears.

[0059] The following will provide a detailed explanation of the analysis based on customer service work orders with different secondary classification results (the primary classification result is electricity price and electricity bill): (I) Electricity Price Implementation: This includes work order classification, data-assisted analysis, and matching analysis of electricity price implementation rules and regulations. It also involves verifying customer file types, checking implemented electricity prices, verifying electricity usage change records (checking electricity price strategies), quantitative and proportional calculations (within electricity price information), and electricity consumption curves (monthly electricity consumption comparison: monitoring recent comparisons and year-on-year comparisons; daily electricity consumption comparison: monthly daily electricity consumption change - year-on-year comparison). Meter Reading (incorrect / missing readings): Incorrect readings (check the last meter replacement date; within 2 years is considered suspicious), missing readings (check for zero electricity sales in the last 12 consecutive months). Electricity Bill Collection: This involves verifying customer type (fee-controlled / postpaid), and checking for outstanding payments (checking electricity bill balances).

[0060] (II) Electricity bill settlement: Whether the electricity bill has been received (customer file / electricity consumption information / recent payment record). Electricity bill refund: suspected fraudulent requests (based on past work order records, using phone numbers as the reference, check whether special customer markings indicate fraudulent requests, and whether there are multiple (3 or more) similar refund requests from the same number). Invoices: Business hall invoice service / invoice information inquiry.

[0061] (III) Power outage and restoration due to arrears: Power outage due to arrears (compliance of power outage, whether the power outage approval process has exceeded the time limit or been suspended, whether the IoT power outage strategy is standardized (verify arrears); whether SMS messages have been sent (all those implementing fee control must send SMS messages), fee control below the threshold, 3 days and half an hour in advance notice, non-fee control accounts, 7 days in advance notice (temporarily not verified); verification of power outage signs (Marketing 2.0 customer information, 360 homepage); verification of data on power outages that should have been stopped but were not (landlord reports power outages due to arrears), check power outage and restoration records in the Marketing 2.0 system). Power restoration due to arrears: whether the restoration conditions have been met (whether there are new arrears, balance greater than zero), verification of unsuccessful power restoration (collect data on abnormal automatic power restoration).

[0062] (iv) (Preferred) Abnormal electricity consumption and charges: Verify curve data, monthly electricity consumption comparison: monitor recent comparisons and comparisons with the same period last year. Daily electricity consumption comparison: monthly daily electricity consumption changes (find patterns, compare with the same period last year). 24-hour curves: check electricity consumption behavior (high-consumption equipment left on / equipment turned off). Electricity price execution / peak-valley tiered calculation: verify miscellaneous business work orders, check peak-valley change records. Verify whether the special markings in the customer file indicate electric heating (mainly work order feedback indicates it, but the actual file does not). Query the annual cumulative electricity consumption, calculate whether it meets the tiered upgrade standard (there are clear calculation rules, clarify according to facts), and consider transfer of ownership, coal-to-electricity conversion, and multiple people in one household.

[0063] (v) Personnel service behavior (mainly labeling and in-depth analysis after service), collection frequency, collection channels, and collection methods.

[0064] Warning rules: Level 1 warning is issued for errors in electricity pricing, incorrect / missed meter readings, errors in peak / valley electricity pricing, poor service attitude, power outages due to unpaid bills in non-directly supplied communities, and power restoration for unpaid bills exceeding 24 hours. Level 2 warning is issued for abnormal meter readings with fluctuating curves, repeated complaints about electricity bills, and power not restored within 6 hours (less than 24 hours) after payment.

[0065] Triggering conditions for the electricity price and charge assessment intelligent agent: When the "primary category" of the work order is "electricity price and charge", the process of assessing the electricity price and charge intelligent agent will be automatically initiated.

[0066] The specific input parameters for the intelligent agent for judging electricity prices and charges are shown in Table 2 below: Table 2: Input Parameters for the Intelligent Agent for Electricity Price and Fee Judgment

[0067] The specific business parameters of the intelligent agent for electricity price and charge analysis are shown in Table 3 below: Table 3: Business Parameters of the Intelligent Agent for Electricity Price and Fee Analysis

[0068] The specific output parameters of the intelligent agent for judging electricity prices and charges are shown in Table 4 below: Table 4: Output Parameters of the Intelligent Agent for Electricity Price and Fee Judgment

[0069] The specific algorithm for the intelligent agent that analyzes electricity prices and charges is shown in Table 5 below. Figures 4-6 As shown: Table 5: Algorithm for Electricity Price and Fee Judgment Agent

[0070] The MCP service for electricity price and charge analysis includes: business processing record query service (Table 6), metering point query service (Table 7), refund information query service (Table 8), contract type query service (Table 9), electricity bill information query service (Table 10), power outage and restoration record query service (Table 11), payment record query service (Table 12), and electricity meter status word query service (Table 13).

[0071] Table 6: Business Processing Record Inquiry Service Table

[0072] Table 7: Metering Point Inquiry Service Table

[0073] Table 8: Refund Information Inquiry Service Table

[0074] Table 9: Contract Type Inquiry Service Table

[0075] Table 10: Electricity Bill Information Inquiry Service Table

[0076] Table 11: Power Outage and Restoration Record Inquiry Service Table

[0077] Table 12: Payment Record Inquiry Service Table

[0078] Table 13: Status Word Inquiry Service Table for Electricity Meters

[0079] Table 14: Data Sources for Electricity Price and Fee Assessment

[0080] For each type of generated conclusion, an analysis panel is displayed, which mainly includes three parts: basic work order information, analysis information, and auxiliary supporting data.

[0081] In summary, the work order analysis intelligent agent module, based on historical work order data and utilizing large language model technology, performs professional matching and responsibility identification of work orders to achieve "precise sorting." Work order anomaly diagnosis leverages the semantic understanding capabilities of the large language model, focusing on scenarios such as power supply quality, business expansion applications, electricity prices and charges, service channels, changes in electricity usage, electricity metering, and emergency repair services. It integrates data from business systems to achieve proactive review of business data and intelligent root cause diagnosis of work orders. For handling solution recommendation, based on the power supply service knowledge base, it intelligently links historical work order data and solutions, comprehensively considering factors such as work order type, responsible party, and on-site conditions. Utilizing the reasoning capabilities of the large language model, it assists in formulating work order service strategies, improving work order processing efficiency.

[0082] In step S4, based on the risk warning level and classification type of the work order, work order warning information is generated sequentially, i.e., work order risk assessment. For new work orders, large-scale model technology is used to screen work orders with duplicate requests according to business and service analysis rules, and the work order risk is classified into levels (Level 1 warning or Level 2 warning) to achieve proactive warning for high-risk work orders, generating work order risk warning information (whether it is a sensitive work order, and whether the risk warning is Level 1 or Level 2 warning). The processing order can be sensitive work orders with Level 1 warning, sensitive work orders with Level 2 warning, non-sensitive work orders with Level 1 warning, and non-sensitive work orders with Level 2 warning.

[0083] By calling relevant data from the MCP service through intelligent agents, the results are integrated using a large model. For information in the integrated results that triggers the early warning rules, an early warning status is marked, and the judgment results and early warning information are stored in the omnichannel business data center.

[0084] Preferably, in the work order processing stage, based on the intelligent analysis results of the work order, a browser plugin can be used to "embed" the results of the intelligent analysis of the work order into the user's work order processing process without changing the original work order business process, thereby assisting grassroots users in analyzing power outage events.

[0085] Specifically, for sensitive work orders such as 95598 complaints, 12398 work orders, public opinion work orders, prominent issues work orders in power supply services, escalated and overflowed and complaint risk work orders, multi-channel demands and repeated demands, etc., they can be automatically identified and actively warned. The important sensitive work order prompt words are fixed in the database, such as issues like "three designations" in business expansion and installation, arbitrary charges, over-time limit, staff going on duty after drinking alcohol, service attitudes of staff in business halls / repair teams / copying and collection teams / construction teams, multiple power outages, voltage quality, indifference to infringing on the interests of the masses, etc., as well as work orders with [Complaint Demotion], [Escalation 12398 Risk] in the label or with the words "customer has the intention to complain" in the content. After the work order arrives, the system conducts research and judgment. If it meets the important sensitive customer demand judgment rules, the system identifies it and marks the word "sensitive" in the Marketing 2.0 system. The dispatching staff can visually mark important sensitive work orders through the icon with the word "sensitive" in the lower right corner. For important sensitive work orders that need to be upgraded for control, the system can automatically generate a supervision QR code, and the supervisor can generate supervision information by scanning the code through "i State Grid".

[0086] Preferably, it is also possible to conduct personalized time limit warning for work orders. The personalized time limit warning supports business personnel to freely set the warning time limit, and through the dynamic coloring of approaching due work orders, visually display the risk of overdue, which is convenient for carrying out personalized control.

[0087] Preferably, based on the semantic understanding and reasoning ability of the large model, combined with various institutional specification documents, work order data, business data and statistical indicators of power grid business, it is possible to provide intelligent data analysis and question-answering functions for power supply service business. Achieve quick replies to power supply service-related questions and quick analysis of work order data and form statistical charts.

[0088] Preferably, it is also possible to support in-depth analysis of work orders. Based on the large model to analyze historical work order data, use the heat map mode to display the work order data of the entire company in terms of region, specialty and location. Focus on warning for power supply quality, business expansion and installation, electricity price and electricity fee, service channels, electricity use change, electric energy metering, repair service, etc. Warn key areas and personnel of concern, support multi-dimensional analysis according to region, transformer area, grid manager, specialty category, etc., support personalized customized analysis of work order data, and assist management decision-making. Customer demand portrait, integrate data such as user basic information, electricity use information, and historical work order information to construct an overall user service portrait, deeply explore user demand characteristics and laws, generate preventive service suggestions, eliminate hidden dangers, take the initiative to care, and reduce the incidence of work orders. Work order report generation, for the report forms of daily customer work orders, the system automatically conducts data summary and calculation, and automatically generates an analysis report to provide decision support for management.

[0089] Specifically, the work order analysis can include the following aspects: 1. Chart statistical analysis Statistical analysis can be performed from multiple dimensions, including the trend of work orders across all channels, the number of work orders in each channel, various outstanding issues of work orders, monitoring of in-transit fault reports from various units, and the distribution of sensitive work orders. The distribution of work orders in each district and county can be displayed in the form of a map.

[0090] 2. Analysis of service delivery status across various channels The analysis of service requests across various channels focuses on core indicators such as the number of work orders, customer satisfaction rate, timeliness of processing, and the number of requests per 10,000 households. It involves statistical analysis from multiple dimensions, including problem trends, work order classification, district / county problem distribution, and channel work order monitoring, presenting the data intuitively using GIS maps and data visualization. Currently, thematic analyses of service requests from eight channels have been completed, including 95598 work orders, the 12345 hotline, marketing service public opinion, 12398 work orders, the vehicle networking platform, operational verification, and reporting of important matters.

[0091] 3. Analysis of Outstanding Issues in Power Supply Services The power supply service problem analysis focuses on core indicators such as the number of work orders for various problems, the number of requests per 10,000 households, and customer satisfaction during follow-up visits. It conducts statistical analysis from multiple dimensions, including trends in work orders for various problems, the distribution of problems across different units, the sources of problems, and the manifestations of problems (categorization), and presents the data intuitively in the form of GIS maps and data visualizations. Currently, the design, development, and trial operation of a large-screen display for analyzing 13 hot issues, including power grid construction, power supply quality, and business expansion applications, have been completed.

[0092] 4. One-click report generation The power supply service report is generated intelligently by dividing the analysis report into business analysis modules, and then combining them according to the analysis modules required for various types of reports. The corresponding report is generated after analysis by the large model.

[0093] This invention utilizes a large language model to classify customer service work orders into sensitive and non-sensitive orders based on the channel of the data and sensitive words within the work orders. It further categorizes customer service work orders into different levels based on their attribute content. A work order analysis module analyzes and judges the work orders, outputting the corresponding problem judgment and risk level. This module includes a business expansion application intelligent agent and an electricity price and fee intelligent agent. The business expansion application intelligent agent, based on a power supply service knowledge base and in conjunction with historical work order data and solutions, analyzes the business expansion process of customer service work orders and outputs a risk warning level. The electricity price and fee intelligent agent, based on the power supply service knowledge base, analyzes customer service work orders whose primary classification result is electricity price and fee, and outputs a risk warning level. Based on the risk warning level and work order classification type, work order warning information is generated sequentially, effectively solving the problem of low efficiency and reliability of customer service work order warnings caused by existing technologies, and significantly improving the efficiency and reliability of customer service work order warnings.

[0094] In the technical solution of this invention, the intelligent agent module for work order analysis reduces the time spent by business personnel in verifying the authenticity of work order content and analyzing root causes in the processing of customer request work orders. It proactively identifies sensitive work orders and work orders of different first-level categories, reduces the workload of front-line staff in reviewing and analyzing work orders, shortens the traditional handling process, reduces the burden on front-line staff, saves labor costs for enterprises, and improves company efficiency.

[0095] This invention's technical solution utilizes intelligent analysis and judgment of work orders across all channels. Business managers can conveniently obtain more reliable business support data and analysis results, reducing human error, mitigating various power supply service issues, and improving the stability of power supply service operations. In terms of management, it breaks down the hierarchical barriers between "municipal company, county company, and power supply station," improving the efficiency of grassroots business response. Regarding talent development, the system's built-in "disposal plan recommendation" function acts as a "digital mentor" for new employees, shortening the business proficiency period.

[0096] The technical solution of this invention has a significant role in improving the efficiency of social public services through intelligent judgment and analysis of work orders across all channels. By integrating power service requests from multiple channels, it enables early identification and graded warning of work order risks, greatly shortens the response cycle for the power needs of enterprises, institutions and residents, improves the efficiency of handling issues such as fault repair and business consultation, effectively reduces production stoppages and inconvenience caused by power problems, and truly enhances residents' sense of security and happiness.

[0097] Example 2 like Figure 7As shown, the technical solution of the present invention also provides an omnichannel customer service work order early warning system based on a large model, including: The acquisition module retrieves customer service work order data from multiple sources across all channels. The classification module uses a large language model to classify customer service work orders into sensitive and non-sensitive work orders based on the channel of the customer service work order data and the sensitive words in the work orders. It also classifies customer service work orders into different levels based on the attribute content of the customer service work orders. The first-level classification results are business expansion and installation and electricity price and electricity fee. The system includes a creation and analysis module, and a work order analysis intelligent agent module. This module analyzes and judges work orders, outputting the corresponding problem judgment and risk level. The work order analysis intelligent agent module includes two parts: a business expansion application intelligent agent and an electricity price and fee intelligent agent. The business expansion application intelligent agent, based on the power supply service knowledge base and linking historical work order data and solutions, analyzes the business expansion process of customer service work orders and outputs a risk warning level. The electricity price and fee intelligent agent, based on the power supply service knowledge base, analyzes customer service work orders whose primary classification result is electricity price and fee, and outputs a risk warning level when the primary classification result is electricity price and fee. The generation module generates work order warning information sequentially based on the risk warning level and work order classification type of the work order.

[0098] It should be noted that the implementation methods of the acquisition module, classification module, creation and analysis module, and generation module in this embodiment correspond to the method steps in Embodiment 1, and this embodiment does not impose any limitations on them.

[0099] This invention utilizes a large language model to classify customer service work orders into sensitive and non-sensitive orders based on the channel of the data and sensitive words within the work orders. It further categorizes customer service work orders into different levels based on their attribute content. A work order analysis module analyzes and judges the work orders, outputting the corresponding problem judgment and risk level. This module includes a business expansion application intelligent agent and an electricity price and fee intelligent agent. The business expansion application intelligent agent, based on a power supply service knowledge base and in conjunction with historical work order data and solutions, analyzes the business expansion process of customer service work orders and outputs a risk warning level. The electricity price and fee intelligent agent, based on the power supply service knowledge base, analyzes customer service work orders whose primary classification result is electricity price and fee, and outputs a risk warning level. Based on the risk warning level and work order classification type, work order warning information is generated sequentially, effectively solving the problem of low efficiency and reliability of customer service work order warnings caused by existing technologies, and significantly improving the efficiency and reliability of customer service work order warnings.

[0100] In the technical solution of this invention, the intelligent agent module for work order analysis reduces the time spent by business personnel in verifying the authenticity of work order content and analyzing root causes in the processing of customer request work orders. It proactively identifies sensitive work orders and work orders of different first-level categories, reduces the workload of front-line staff in reviewing and analyzing work orders, shortens the traditional handling process, reduces the burden on front-line staff, saves labor costs for enterprises, and improves company efficiency.

[0101] This invention's technical solution utilizes intelligent analysis and judgment of work orders across all channels. Business managers can conveniently obtain more reliable business support data and analysis results, reducing human error, mitigating various power supply service issues, and improving the stability of power supply service operations. In terms of management, it breaks down the hierarchical barriers between "municipal company, county company, and power supply station," improving the efficiency of grassroots business response. Regarding talent development, the system's built-in "disposal plan recommendation" function acts as a "digital mentor" for new employees, shortening the business proficiency period.

[0102] The technical solution of this invention has a significant role in improving the efficiency of social public services through intelligent judgment and analysis of work orders across all channels. By integrating power service requests from multiple channels, it enables early identification and graded warning of work order risks, greatly shortens the response cycle for the power needs of enterprises, institutions and residents, improves the efficiency of handling issues such as fault repair and business consultation, effectively reduces production stoppages and inconvenience caused by power problems, and truly enhances residents' sense of security and happiness.

[0103] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A method for early warning of omnichannel customer service work orders based on a large model, characterized in that, include: Acquire customer service work order data from multiple channels and sources; Using a large language model, customer service work orders are classified into sensitive work orders and non-sensitive work orders based on the channel of the customer service work order data and the sensitive words in the work orders. Customer service work orders are also classified into different levels based on the attribute content of the customer service work orders. The first-level classification results are business expansion and installation and electricity price and electricity fee. A work order analysis and judgment module is created to analyze and judge work orders, outputting the corresponding problem judgment and risk level. This module includes two parts: a business expansion and installation intelligence agent and an electricity price and fee intelligence agent. The business expansion and installation intelligence agent, based on the power supply service knowledge base and linking historical work order data and solutions, analyzes the business expansion process of customer service work orders and outputs a risk warning level. The electricity price and fee intelligence agent, based on the power supply service knowledge base, analyzes customer service work orders whose primary classification result is electricity price and fee, and outputs a risk warning level when the primary classification result is electricity price and fee. Work order warning information is generated sequentially based on the risk warning level and work order classification type of the work order.

2. The method for early warning of omnichannel customer service work orders based on a large model according to claim 1, characterized in that, The business expansion process assessment includes assessment of work orders exceeding time limits, assessment of external circulation, assessment of non-standard process handling, assessment of suspected illegal charges, and assessment of suspected compliance with preset specified types.

3. The method for early warning of omnichannel customer service work orders based on a large model according to claim 2, characterized in that, The specific analysis of overdue work orders is as follows: To retrieve work order information for at least one month before and after the date in the customer's complaint description: In the retrieved work order details, determine whether there are any work orders related to the customer based on the user name, electricity address, contact information, and electricity demand information in the work order description; If a user's business expansion application is found in a work order related to a customer, the start time, completion time, and processing time of the business expansion process in that work order are queried. Based on the different processing time limits for business expansion, the corresponding risk warning level is output.

4. The method for early warning of omnichannel customer service work orders based on a large model according to claim 2, characterized in that, The assessment of extracorporeal circulation is as follows: To retrieve work order information for at least one month before and after the date in the customer's complaint description: In the retrieved work order details, determine whether there are any work orders related to the customer based on the user name, electricity address, contact information, and electricity demand information in the work order description; If the customer's business expansion application is not found in the work order related to the customer, it indicates that the customer's electricity application has not been accepted, it is suspected to be an off-line circulation, and a risk warning is issued. If a user's business expansion application is found, but the customer description in the work order indicates that an electricity meter has not been installed, and the user's business process has been archived and no data has been collected from the electricity meter, then it is determined to be a suspected off-line circulation, and a risk warning is issued.

5. The method for early warning of omnichannel customer service work orders based on a large model according to claim 2, characterized in that, The following are specific criteria for identifying irregularities in the handling process:

1. If a customer's work order describes a problem where staff require multiple trips to the business hall, this is considered irregularity and a risk warning is issued.

2. If a customer's work order describes a problem where power company staff require the customer to pay for materials, construction, lost work time, and overtime; or requires the customer to purchase the cable before the electricity meter; or requires the customer to pay for services other than electricity fees, this is considered suspected illegal charging and a risk warning is issued.

3. If a customer's work order describes a problem where power company staff recommend, introduce, or designate construction, design, testing, or material suppliers; or where power company staff restrict the customer's choice of construction, design, testing, or material suppliers by not installing the meter or conducting acceptance checks, this is considered suspected violation of a pre-defined type and a risk warning is issued.

6. The method for early warning of omnichannel customer service work orders based on a large model according to claim 1, characterized in that, Customer service work orders are categorized into primary and secondary categories. The secondary categories are subcategories of the primary categories. When the primary category is electricity price and charge, the corresponding secondary categories include electricity price execution, electricity charge settlement, and power outage / restoration due to arrears.

7. The method for early warning of omnichannel customer service work orders based on a large model according to claim 6, characterized in that, When the secondary classification result is electricity price implementation category, based on the acceptance time, it is determined whether the user number has handled any changes in category, changes in residential peak-valley electricity rates, or adjustments to the tiered base rate for households with multiple residents within the three months prior to the acceptance time. If so, a record of handling such matters for the user is generated. If not, a record of not handling such matters for the user is generated. The customer service work order is checked to see if the user has a metering point with a fixed or proportional electricity calculation method. If so, a record of a fixed or proportional metering point for the user is generated. If not, a record of not having a fixed or proportional metering point for the user is generated.

8. The method for early warning of omnichannel customer service work orders based on a large model according to claim 6, characterized in that, When the secondary classification result is electricity bill settlement, check whether the user has applied for a refund three or more times in the customer service work order. If so, generate a record of multiple refund applications by the customer; otherwise, generate a record of no multiple refund applications by the customer.

9. A method for early warning of omnichannel customer service work orders based on a large model as described in claim 6, characterized in that, When the secondary classification result is the "Out-of-Payment Power Outage" category within the "Out-of-Payment Power Outage and Resumption" category, check whether the electricity bill calculation agreement in the customer's contract information in the customer service work order is a postpaid agreement, and whether the customer's overdue monthly payment deadline is less than 30 days from the date the work order was accepted. If so, generate a suspected postpaid customer's power outage due to insufficient time limit and activate the secondary warning. When the secondary classification result is "Power outage notification not received" under the category of "Power outage and restoration due to overdue payment", check whether the customer's prepayment control flag is "yes" in the customer service work order and whether there is any power outage information in the power outage and restoration record. If so, generate a suspected case of inadequate notification before prepayment control power outage. When the secondary classification result is "Power outage and restoration due to overdue payment" under the category of "Power outage and restoration due to overdue payment", check whether the customer's account balance in the customer service work order is greater than 0 and whether the status of the energy meter's status relay is "off". Then compare the customer's most recent payment time with the work order acceptance time. If the customer's most recent payment time and the work order acceptance time exceed 6 hours, generate a suspected case of "Power outage not restored after 6 hours of payment" and activate a secondary warning. If the customer's most recent payment time and the work order acceptance time exceed 24 hours, generate a suspected case of "Power outage not restored after 24 hours of payment" and activate a primary warning.

10. A large-scale model-based omnichannel customer service work order early warning system, characterized in that, include: The acquisition module retrieves customer service work order data from multiple sources across all channels. The classification module uses a large language model to classify customer service work orders into sensitive and non-sensitive work orders based on the channel of the customer service work order data and the sensitive words in the work orders. It also classifies customer service work orders into different levels based on the attribute content of the customer service work orders. The first-level classification results are business expansion and installation and electricity price and electricity fee. The system includes a creation and analysis module, and a work order analysis intelligent agent module. This module analyzes and judges work orders, outputting the corresponding problem judgment and risk level. The work order analysis intelligent agent module includes two parts: a business expansion application intelligent agent and an electricity price and fee intelligent agent. The business expansion application intelligent agent, based on the power supply service knowledge base and linking historical work order data and solutions, analyzes the business expansion process of customer service work orders and outputs a risk warning level. The electricity price and fee intelligent agent, based on the power supply service knowledge base, analyzes customer service work orders whose primary classification result is electricity price and fee, and outputs a risk warning level when the primary classification result is electricity price and fee. The generation module generates work order warning information sequentially based on the risk warning level and work order classification type of the work order.

Images