Risk management and control method for power marketing business

Abstract

The invention discloses a risk management and control method for power marketing business. Comprising the following steps: a provincial customer service center classifies work orders; the provincial customer service center obtains an emotion resonance index and intention ambiguity of the customer in real time through a voice analysis and text emotion analysis engine; based on the emotion resonance index, the intention ambiguity, the historical work order data and the customer behavior data, constructing a risk prediction model, and outputting the risk level of the work order; the risk level of the work order is primarily reviewed by provincial customer service center personnel, professional classification is carried out, and the risk appeal work order is automatically distributed; tracking and accessing the risk appeal work order; according to the invention, multi-modal emotion and intention analysis is fused, voice and text data are combined, a multi-dimensional risk prediction model is constructed, and intelligent classification and priority distribution of work orders are realized; through linkage of provincial-level first review and professional review, RPA automation and double telephone early warning, the risk identification accuracy and processing efficiency are improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of power marketing, and more particularly to a risk management and control method for power marketing business. BACKGROUND

[0002] In recent years, the internal and external environment of power marketing has changed greatly, especially the rapid development of new businesses such as distributed photovoltaic and charging piles, which has brought new pressure to the standard management of power companies. In order to strengthen the key areas of power marketing business, establish and improve the efficient coordination mechanism of risk prevention and disposal and quality supervision and inspection;

[0003] First, the service risk of the current work order is controlled, the risk classification and monitoring and early warning of hotline customer appeal are carried out, second, the risk event management and control application is built, the risk events are collected from bottom to top, and the risk early perception and early warning are achieved, third, the RPA technology is applied to realize the automatic order receiving and order returning of work order, reduce the work load of the operator staff, and improve the efficiency of work order receiving and dispatching.

[0004] At present, risk control is mainly carried out by each professional, which is relatively discrete and cannot form a combined force. The problems found in the appeal work order are not disposed in place, and the control is not closed loop. In particular, some risks are not included in the control, there are problems such as incomplete risk entry, unlayered and unclassified management, unsystematic and unclosed control process, non-auditing and non-standard disposal result reporting. SUMMARY

[0005] In view of the problems in the prior art, the purpose of the present application is to provide a risk management and control method for power marketing business, which converts risk identification from passive response after the event to active early warning in advance and in the process by building a multi-specialty coordinated joint defense and control data model, realizes automatic discovery of risks, and significantly reduces customer complaint rate; through the preset collaborative disposal workflow, the risk disposal process across departments is solidified into a standardized online task, breaking down departmental barriers, realizing intelligent distribution of disposal tasks and transparent control of the process, and greatly improving risk disposal efficiency; a complete closed loop from risk identification to disposal, and then to evaluation feedback is established. Through quantitative evaluation of the disposal result, continuous learning and self-optimization of the risk rule model are realized, ensuring the dynamic adaptability and continuous improvement of the risk control system, and finally realizing the traceability of service risk source, the controllability of process, and the evaluation of result.

[0006] To achieve the above purpose, the present application provides the following technical scheme: a risk management and control method for power marketing business, comprising the following steps:

[0007] The risk appeal work order is transmitted to the provincial customer service center, and the provincial customer service center classifies the work order;

[0008] The provincial customer service center obtains the emotional resonance index and the intention ambiguity of the customer in real time through voice analysis and text sentiment analysis engine;

[0009] Based on the emotional resonance index, the intention ambiguity, the historical work order data and the customer behavior data, a risk prediction model is constructed to output the risk level of the work order;

[0010] The risk level of the work order is preliminarily reviewed by the staff of the provincial customer service center, and is classified professionally, and the risk appeal work order is automatically distributed;

[0011] The risk appeal work order is tracked and visited, and supervision and early warning are carried out, and the customer satisfaction is obtained.

[0012] Further, the emotional resonance index is obtained by weighting and summing the emotional sub-index of the voice feature and the emotional sub-index of the text sentiment according to the weight;

[0013] The emotional sub-index of the voice feature is obtained by extracting the features of the fundamental frequency change rate, the speech rate and the volume intensity, calculating the deviation of the fundamental frequency change rate, the speech rate and the volume intensity from the historical average value, and then weighting and summing the deviation;

[0014] The emotional sub-index of the text sentiment is obtained by extracting the emotional vocabulary density and the sentiment polarity score through the text sentiment analysis engine, calculating the deviation of the emotional vocabulary density and the sentiment polarity score from the historical average value, and then weighting and summing the deviation.

[0015] Further, the intention ambiguity is obtained by weighting and averaging the intention matching degree, the intention clarity and the intention missing rate;

[0016] The cosine similarity or the Euclidean distance between the customer text and the service type is calculated, and the cosine similarity is taken as the intention matching degree;

[0017] The intention clarity is obtained by analyzing the average sentence length, the number of dependent clauses and the complexity of the grammatical structure of the customer text, calculating the deviation of the average sentence length, the number of dependent clauses and the complexity of the grammatical structure from the historical average value, and then weighting and summing the deviation;

[0018] The intention missing rate is obtained by detecting the number of missing key fields in the customer text and calculating the ratio of the number of missing key fields to the total number of preset key fields.

[0019] Further, the risk prediction model is a fusion model based on a multi-modal deep learning network and a Bayesian network;

[0020] The historical work order data and the customer behavior data are taken as the training data set to train the risk prediction model, and the weight distribution of the emotional resonance index and the intention ambiguity is optimized;

[0021] Set the risk level division rule, adjust the output threshold of the risk prediction model;

[0022] The risk level division rule includes a low risk level, a medium risk level, a high risk level, and a red line risk level.

[0023] Further, the steps of the provincial customer service center personnel performing preliminary examination are as follows:

[0024] The work orders of the low risk level, the medium risk level, the high risk level, and the red line risk level are subjected to provincial preliminary examination;

[0025] Then, the work orders are distributed to the next link by the provincial customer service center personnel after preliminary examination, or are distributed to the city / county power supply service command center;

[0026] The work orders subjected to preliminary examination by the provincial customer service center and determined to have a risk level are transferred to the provincial professional audit link, distributed to the corresponding professional auditors according to the work order type, and added to the supply service to-be-done and the i-state grid to-be-done, and a short message is sent for reminding;

[0027] If the audit is passed, the work order is archived, and if it is not passed, it is returned to the city / single back order confirmation link.

[0028] Further, the city / single back order confirmation link is set at the front end of the provincial customer service center, used for receiving the risk appeal work order, and classifying and pushing the work order according to the preliminary examination result of the work order.

[0029] Further, the provincial customer service center personnel preliminarily examine the work order in combination with the emotional resonance index and the intention ambiguity, and dynamically adjust the work order classification logic:

[0030] The work order with the emotional resonance index exceeding a first preset value is preferentially distributed to the provincial professional audit link;

[0031] The work order with the intention ambiguity exceeding a second preset value is preferentially distributed to the city / county power supply service command center, and after the city / county power supply service command center supplements the information, it is submitted to the provincial level for audit.

[0032] Further, the professional classification divides the work order into different categories according to different sources, different subcategories, and keywords, and pushes different PP messages and short messages to the corresponding personnel according to different categories of professionals.

[0033] Further, the automatic distribution is realized by RPA robots to automatically open, log in, ocr recognize, and transfer the work order, and the RPA robots replace manual operation to realize the abnormal operation warning function;

[0034] The RPA robot automatically archives the work order and synchronizes it to the i-state grid system, while monitoring the processing timeliness, and if the time is exceeded, a warning is triggered;

[0035] The RPA robot monitors the running state in real time, logs in, fails, OCR recognition error, work order flow timeout, automatically triggers abnormal early warning and notifies manual intervention; Realize the collaborative processing of RPA robot and manual audit.

[0036] Further, the tracking access adopts a two-way telephone early warning mechanism:

[0037] First telephone early warning: after the work order is assigned, generate the data to be processed in the first telephone interface, monitor whether the city / country company contacts the customer within the specified time; If not on time, trigger an early warning message, and record the follow-up results and customer satisfaction; Support filtering first telephone data according to work order type, time, risk level conditions, generate visual analysis report;

[0038] Second telephone early warning: after processing the fault work order, generate the data to be processed on the second telephone page, monitor whether the second follow-up is completed within the specified time; If not on time, trigger an early warning message, and record the follow-up results and customer satisfaction, form a closed-loop management.

[0039] Technical effects and advantages of the present application:

[0040] The present application introduces a multi-modal emotion-intention fusion analysis technology, combines voice data and text data, realizes the quantitative evaluation of customer emotion resonance index and intention ambiguity, and significantly improves the identification accuracy of risk work order; Based on the multi-dimensional data such as fundamental frequency change rate, speech rate, volume intensity, emotion vocabulary density and syntax complexity, a risk prediction model is constructed to realize intelligent division of work order risk level;

[0041] Through the linkage mechanism of provincial preliminary examination and professional examination, combined with RPA automatic processing and two-way telephone early warning, the accurate matching of resources and the closed-loop management of service process are realized, and the overall business management efficiency is improved; The provincial customer service center personnel dynamically adjust the work order classification logic according to the emotion resonance index and the intention ambiguity, and preferentially allocate high-risk work orders to the corresponding professional team to avoid resource waste and processing delay; Through the automatic operation of OCR identification, work order flow, reduce the repetitive work of manual, at the same time, real-time monitoring of operation exception, ensure the timeliness of work order processing; Through the double early warning mechanism of first telephone and second telephone, it is ensured that customer demands are responded within the specified time;

[0042] The power supply service risk research and application control method focuses on power marketing service, builds a digital empowerment business quality control platform, constructs a professional linkage and joint effort power marketing whole business quality control guarantee system, develops power supply service risk research and application development research, strengthens the management and control of key fields and key links of power marketing whole business, establishes and improves the efficient coordination mechanism of risk prevention and disposal and quality supervision and inspection, and realizes the goal of service first, prevention first, whole process supervision and comprehensive management.

[0043] Other features and advantages of the present application will become apparent from the following detailed description of illustrative embodiments thereof, taken in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF DRAWINGS

[0044] Fig. 1 is a schematic diagram of the system structure provided by the present application;

[0045] Fig. 2 is a flowchart provided by the present application. DETAILED DESCRIPTION

[0046] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and should not be used to limit the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0047] As shown in Figs. 1-2 , the risk management and control method for power marketing business provided by the embodiment of the present application comprises the following steps:

[0048] The risk appeal work order is transmitted to the provincial customer service center, and the provincial customer service center classifies the work order;

[0049] The provincial customer service center obtains the emotional resonance index and the intention ambiguity of the customer in real time through a voice analysis and text sentiment analysis engine;

[0050] Based on the emotional resonance index, the intention ambiguity, historical work order data and customer behavior data, a risk prediction model is constructed, and the risk level of the work order is output;

[0051] The risk level of the work order is preliminarily reviewed by the personnel of the provincial customer service center, and is classified professionally, and the risk appeal work order is automatically distributed;

[0052] The risk appeal work order is tracked and visited, and supervision and early warning are performed and the follow-up result is well done, and the customer satisfaction is obtained.

[0053] In the embodiment, preferably, the emotional resonance index is obtained by weighting and summing and averaging the emotional sub-index of the voice feature and the emotional sub-index of the text sentiment according to the weight;

[0054] The emotional sub-index of the voice feature is obtained by extracting the features of the fundamental frequency variation rate, the speech rate and the volume intensity, calculating the deviation degree of the fundamental frequency variation rate, the speech rate and the volume intensity from the historical average value, and then weighting and summing the deviation degrees;

[0055] The emotional sub-index of text sentiment is calculated by extracting the emotional vocabulary density and sentiment polarity score through the text sentiment analysis engine, calculating the deviation from the historical average, and then weighting and summing the deviation;

[0056] It should be noted that the emotional resonance index is constructed by weighting and averaging the emotional sub-index of the voice feature and the emotional sub-index of the text sentiment according to the weight, which accurately quantifies the customer emotional state; The deviation of the fundamental frequency change rate, speech rate, and volume intensity from the historical average is calculated, combined with the deviation of the emotional vocabulary density and sentiment polarity score from the historical average in text sentiment analysis, through multi-modal data fusion and dynamic weight distribution, which improves the accuracy and intelligence level of risk ticket identification; It can capture the intensity and persistence of customer emotional fluctuations in real time, and effectively distinguish the clarity and intention ambiguity of emotional expression, thereby optimizing the input parameters of the risk prediction model, improving the scientificity of ticket classification and priority allocation; Through the quantitative evaluation of the emotional resonance index, dynamically adjust the ticket processing strategy, reduce manual intervention, improve processing efficiency, and provide data support for provincial preliminary review and professional review, ensure high-risk ticket priority processing, reduce customer dissatisfaction risk, improve service response speed and customer satisfaction, and ultimately realize the precision, intelligence, and closed-loop of power marketing business risk control, significantly enhance the service quality and customer experience of enterprises;

[0057] The emotional sub-index of the voice feature is calculated by the following formula:

[0058] ;

[0059] Wherein, represents the fundamental frequency change rate, represents the historical average fundamental frequency, represents the speech rate, represents the historical average speech rate, represents the volume intensity, represents the preset volume threshold, , , respectively represent the weight coefficient of the voice feature;

[0060] The emotional sub-index of text sentiment is calculated by the following formula:

[0061] ;

[0062] Wherein, represents the emotional vocabulary density, represents the historical average emotional vocabulary density, represents the sentiment polarity score, represents the preset sentiment polarity threshold, , respectively represent the weight coefficients of the text features;

[0063] The emotional resonance index is calculated by the following formula:

[0064] ;

[0065] wherein, , respectively represent the weight coefficients of each modality, and the sum is 1, such as , .

[0066] In this embodiment, preferably, the intention ambiguity is calculated by weighted average of the intention matching degree, the intention clarity and the intention missing rate;

[0067] The cosine similarity or Euclidean distance between the customer text and the service type is calculated, and the cosine similarity is taken as the intention matching degree;

[0068] The average sentence length, the number of dependent clauses and the complexity of the syntax structure of the customer text are analyzed, the deviation of the average sentence length, the number of dependent clauses and the complexity of the syntax structure from the historical average value is calculated, and the deviation is weighted and summed to obtain the intention clarity;

[0069] The number of missing key fields in the customer text is detected, and the number of missing key fields is compared with the preset total number of key fields to obtain the intention missing rate;

[0070] It should be noted that by calculating the intention matching degree, the intention clarity and the intention missing rate by weighted average, the intention ambiguity is constructed, the precise quantitative evaluation of the customer's appeal intention is obtained, the cosine similarity between the customer text and the service type is calculated to quickly judge the matching degree of the appeal and the predefined service type; at the same time, the average sentence length, the number of dependent clauses and the complexity of the syntax structure of the text are analyzed, the deviation of the average sentence length, the number of dependent clauses and the complexity of the syntax structure from the historical average value is calculated, and the clarity of the intention expression is evaluated; in addition, the key field missing rate is detected to quantify the completeness of the customer information; the multi-modal data is combined with the business rules to improve the accuracy and automation level of the work order intention recognition, and to solve the classification errors and processing delays caused by the ambiguity of the appeal or the missing information in the traditional work order processing; through the dynamic calculation of IAS, the work order with unclear intention can be accurately identified, and is preferentially assigned to the city / country company to supplement information, reducing manual intervention and improving processing efficiency;

[0071] The calculation formula of the intention matching degree is as follows:

[0072] ;

[0073] wherein, represent the intention matching degree, represents the semantic similarity between the customer text and the service type, ranging from 0 to 1, and the higher the value, the higher the matching degree; represents the maximum semantic similarity between service types, used for normalization processing;

[0074] The calculation formula of the intention clarity is as follows:

[0075] ;

[0076] wherein, represents the intention clarity, represents the average number of words or characters in each sentence in the customer text; represents the average sentence length of the historical work order, used for normalization processing; represents the number of sentences in the text; represents the total number of sentences in the customer text; represents the score of the syntactic tree depth or the syntactic structure evaluation model; represents a preset syntactic complexity threshold, used to determine whether the intention expression is clear; , , respectively represent the weight coefficients of the syntactic features, , , .

[0077] The calculation formula of the intention missing rate is as follows:

[0078] ;

[0079] wherein, represents the intention missing rate, represents the number of key information fields not provided by the customer; represents the total number of predefined key information fields;

[0080] The calculation formula of the intention ambiguity is as follows:

[0081] ;

[0082] wherein, represents the intention ambiguity, , , respectively represent the weight coefficients of each modality, and the sum is 1, , , .

[0083] In this embodiment, preferably, the risk prediction model is a fusion model based on a multi-modal deep learning network and a Bayesian network.

[0084] The risk prediction model is trained by taking historical work order data and customer behavior data as a training data set, and the weight distribution of the emotional resonance index and the intention ambiguity is optimized;

[0085] The risk level division rule is set to adjust the output threshold of the risk prediction model;

[0086] The risk level division rule includes low risk level, medium risk level, high risk level and red line risk level;

[0087] It should be noted that by constructing a risk prediction model based on the fusion of multi-modal deep learning network and Bayesian network, the intelligent and accurate division of the risk level of power marketing business is realized; based on historical work order data and customer behavior data, through the dynamic weight distribution of the emotional resonance index and the intention ambiguity, the accuracy and adaptability of risk prediction are significantly improved; at the same time, by setting the low, medium, high and red line risk level division rule, combined with the model output threshold adjustment, the risk characteristics in different business scenarios can be flexibly responded to, realizing the graded early warning and priority processing of risk work orders; among them, the red line risk work order automatically triggers the supervision process, and the high risk work order is preferentially assigned to the professional audit link, so as to optimize the resource allocation and improve the processing efficiency; in addition, the model quantitatively analyzes the correlation between emotion and intention, reduces the subjectivity of manual judgment, and enhances the objectivity and traceability of risk identification.

[0088] In this embodiment, preferably, the step of preliminary examination by the provincial customer service center personnel is as follows:

[0089] The work orders of low risk level, medium risk level, high risk level and red line risk level are subjected to provincial preliminary examination;

[0090] Then the work orders are dispatched to the next link by the provincial customer service center personnel after preliminary examination, or are dispatched to the city / county power supply service command center;

[0091] The work orders that have been preliminarily examined by the provincial customer service center and determined to have risk level are transferred to the provincial professional audit link, distributed to the corresponding professional auditors in the supply service to-be-done and the i-state grid to-be-done according to the work order type, and a short message is sent for reminding;

[0092] If the audit is passed, the work order is archived, and if it is not passed, it is returned to the city back single confirmation link;

[0093] It should be noted that through the preliminary examination of the work order by the provincial customer service center personnel and the dynamic dispatching mechanism, the processing efficiency and accuracy of the risk appeal work order are improved. Based on the risk level division rule, the provincial customer service center personnel can quickly identify the risk attribute of the work order, and through the dispatching logic of the RPA robot, the work order is accurately transferred to the city / county power supply service command center or the professional audit link, avoiding resource waste and processing delay; automatically assign high-risk work orders to corresponding professional auditors, and ensure the timeliness of the audit through the short message reminder function. If the audit is passed, the work order is automatically archived and synchronized to the i-state grid system to ensure that the processing results are traceable; if the audit is not passed, it is returned to the city back single confirmation link, triggering the supplementary information process, avoiding processing deviation caused by missing information or ambiguous intent; through the collaborative audit of artificial and system, combined with dynamic adjustment of risk level, both the subjectivity of artificial intervention is reduced and the response speed of high-risk work orders is improved, ultimately realizing the efficiency, accuracy and closed-loop of the risk control of power marketing business, effectively reducing customer complaint rate and improving service satisfaction and enterprise operation efficiency;

[0094] Risk level division:

[0095] Low risk level: emotional resonance index ≤ 30 and intent ambiguity ≤ 0.3;

[0096] Medium risk level: 30 < emotional resonance index ≤ 60 or 0.3 < intent ambiguity ≤ 0.6;

[0097] High risk level: emotional resonance index > 60 or intent ambiguity > 0.6;

[0098] Red line risk level: emotional resonance index ≥ 80 and intent ambiguity ≥ 0.8.

[0099] In this embodiment, preferably, the city back single confirmation link is arranged at the front end of the provincial customer service center, used for receiving the risk appeal work order, and classifying and pushing the work order according to the result of the preliminary examination of the work order;

[0100] It should be noted that by setting the city back single confirmation link at the front end of the provincial customer service center, efficient reception and classification of the risk appeal work order is realized, and the collaborative efficiency and information integrity of the work order processing flow are optimized; and according to the dynamic adjustment of the work order classification logic according to the provincial preliminary examination result, the low-risk work order is directly assigned to the city / county power supply service command center, while the high-risk or information supplement work order triggers the targeted processing flow, ensuring accurate resource matching; through the classification pushing mechanism, the system can automatically assign the work order to the corresponding professional team according to the type and risk level, reducing artificial intervention and improving processing efficiency; at the same time, a closed-loop linkage is formed with the provincial preliminary examination, avoiding repeated processing or delay caused by missing information or ambiguous intent, ensuring that the work order can be checked and classified at the front end, providing complete and clear appeal information for subsequent professional audit.

[0101] In this embodiment, preferably, the provincial customer service center personnel preliminary review of the work order is combined with the emotional resonance index and the intention ambiguity to dynamically adjust the work order classification logic:

[0102] The work order with the emotional resonance index exceeding the first preset value is preferentially assigned to the provincial professional review link;

[0103] The work order with the intention ambiguity exceeding the second preset value is preferentially assigned to the city / county power supply service command center, and after the city / county power supply service command center supplements the information, it is submitted to the provincial level for review;

[0104] It should be noted that through the dynamic analysis of the emotional resonance index and the intention ambiguity by the provincial customer service center personnel during the preliminary review of the work order, the intelligent optimization of the work order classification logic is realized, and the accuracy and efficiency of the risk demand processing are improved. Specifically, the work order with the emotional resonance index exceeding the first preset value, the first preset value being the emotional resonance index ≥ 60, indicating that the customer's emotional fluctuation is intense or there is potential dissatisfaction, is preferentially assigned to the provincial professional review link to ensure that high-risk demands are quickly responded and deeply processed. The work order with the intention ambiguity exceeding the second preset value, the second preset value being the intention ambiguity ≥ 0.6, indicating that the customer's demand expression is not clear or key information is missing, is preferentially assigned to the city / county power supply service command center, and after the grassroots unit supplements the information, it is resubmitted to the provincial level for review to avoid processing delay or misjudgment due to incomplete information. By quantifying the relevance of emotions and intentions, combining manual preliminary review with system analysis, and dynamically adjusting the work order classification strategy, the subjectivity of manual judgment is reduced, and the priority processing capability of high-risk work orders is improved.

[0105] In this embodiment, preferably, the professional classification divides the work order into different categories according to different sources, different subcategories, and keywords, and pushes different PP messages and short messages to corresponding personnel according to different professional categories;

[0106] It should be noted that through the professional classification mechanism based on the multi-dimensional characteristics of the source, subclass and keyword of the work order, the accurate classification and intelligent pushing of risk appeal work orders are realized, and the automation level of risk control and resource allocation efficiency are improved; Specifically, by dynamically analyzing the attributes of the work order, different categories of work orders are matched to the corresponding professional team, such as complaints, reports, service applications, customer service, operation and maintenance, and law, and combined with multi-channel notification methods such as PP messages and short messages, it is ensured that the work order is quickly transferred to the person in charge, reducing manual intervention and processing delay; Through the intelligent optimization of classification logic, the problem of wrong or missed dispatch caused by incomplete information or subjective judgment in traditional manual classification is avoided, and at the same time, structured and standardized work order information is provided for subsequent professional review, improving processing efficiency and accuracy; In addition, the classification pushing mechanism is linked with the risk prediction model, the emotion resonance index and the intention ambiguity analysis to realize the whole process management, shorten the work order processing cycle, reduce the repetitive workload, and ultimately improve the risk response speed and service quality of power marketing business, enhance customer satisfaction and the intelligent level of enterprise operation.

[0107] In the embodiment, preferably, the automatic dispatch is realized by RPA robots to automatically open, log in, ocr identify, and transfer the work order. The RPA robots replace manual operation to realize the abnormal early warning function;

[0108] The RPA robot automatically archives the work order and synchronizes it to the i-state grid system, while monitoring the processing timeliness. If the time is exceeded, the early warning is triggered;

[0109] The RPA robot monitors the running state in real time, and automatically triggers abnormal early warning and notifies manual intervention in case of login failure, OCR identification error, and work order transfer timeout. The collaborative processing of RPA robots and manual review is realized;

[0110] It should be noted that the whole process automation of work order processing is realized by RPA robots, which improves the operation efficiency and system reliability of power marketing business; Specifically, the RPA robot can automatically complete system login, OCR identification, work order transfer and other operations, replacing manual execution of repetitive tasks, greatly reducing manual operation cost and error rate, while realizing real-time early warning function of operation abnormality; By automatically archiving the work order and synchronizing it to the i-state grid platform, the data consistency and traceability are ensured; If the work order processing timeliness is exceeded, the early warning mechanism is automatically triggered to remind the relevant personnel to handle it in time to avoid delay risk; In addition, the RPA robot monitors the running state in real time, such as login failure, OCR identification error, work order transfer timeout, etc., automatically identifies the abnormality and notifies manual intervention to form a closed-loop management; Through the efficient cooperation of RPA robots and manual review, the intelligence and standardization of work order processing are guaranteed, and the flexibility of manual intervention is retained, which significantly optimizes the response speed and processing quality of risk appeal work orders.

[0111] In this embodiment, preferably, the tracking access adopts a two-call telephone early warning mechanism:

[0112] The first call telephone early warning: after the work order is issued, the first call interface generates data to be processed, monitors whether the city / country company contacts the customer within the specified time; if it is not timely, an early warning message is triggered, and the follow-up result and customer satisfaction are recorded; the first call data can be filtered according to the work order type, time, risk level, and the like, and a visual analysis report is generated;

[0113] The second call telephone early warning: after the fault work order is processed, the second call page generates data to be processed, and monitors whether the second follow-up is completed within the specified time; if it is not timely, an early warning message is triggered, and the follow-up result and customer satisfaction are recorded, forming a closed-loop management;

[0114] It should be noted that by introducing the two-call telephone early warning mechanism, the standardization and closed-loop management capability of customer follow-up in power marketing business are improved; the first call telephone early warning monitors in real time whether the city / country company contacts the customer within the specified time after the work order is issued, and if it is not timely, an early warning message is automatically triggered, and the follow-up result and customer satisfaction are recorded in synchronization, ensuring the timeliness and transparency of service response; at the same time, the data can be filtered according to the work order type, time, risk level, and the like, and a visual analysis report is generated, providing data support for management to optimize resource allocation and service quality; the second call telephone early warning is for the second follow-up link after the fault work order is processed, which monitors the timeliness of follow-up, and if it is not timely, an early warning is triggered again, ensuring the completeness of problem solving and the closed-loop nature of customer feedback; through the automatic monitoring and early warning function, the lag and subjectivity of manual intervention are reduced;

[0115] When the work order is issued, the first interface generates data to be processed, and the first call telephone has five states: to be processed, normally archived, overdue and archived, overdue and not processed, and not processed and archived; when the fault work order reaches the processing link, the second call page generates data to be processed after the non-urgent work order processing link is submitted, and the second call telephone has three states: to be processed, normally archived, and overdue and archived; the two-call telephone details are counted according to the work order issuance time, the two-call telephone data details are displayed in blocks, and the first call telephone data is displayed by default, and the second call telephone data can be switched to view.

[0116] The specific operation process of the present application is as follows:

[0117] The i State Grid customer center system assigns the work order to the provincial network, the provincial marketing service center automatically receives the i State Grid work order, automatically assigns to the branch unit according to the automatic assignment mechanism, the provincial marketing service center classifies the work order, and obtains the emotional resonance index and the intention ambiguity of the customer in real time through voice analysis and text sentiment analysis engine; based on the emotional resonance index, the intention ambiguity, historical work order data and customer behavior data, a risk prediction model is constructed, and the risk level of the work order is output;

[0118] The city order receiving branch processes the work order through work order preliminary examination, archives the work order that can be processed, assigns the work order that needs to be processed to the city work order processing department, the work order processing department completes the work order processing according to the work order content and the work order processing requirement, fills in the work order processing process, analyzes and researches the customer's appeal and marks the key customer; after the work order processing is completed, the work order is submitted to the city-level audit department;

[0119] The city order receiving branch receives the processed work order, and returns the processing unit to continue to improve and supplement the work order content if the processing result is not agreed; if the processing result is agreed, the address leader audits, the city leader audit is not passed, and is transferred to the city order receiving branch, and the city leader audit is passed, and is submitted to the provincial audit department;

[0120] The provincial preliminary examination unit receives the work order audited by the city, and according to the risk level of the work order, the work order with low risk that does not need to be audited by the leader is transferred to the follow-up unit after the audit is passed; the work order that does not pass the provincial preliminary examination is transferred to the city preliminary examination unit; the work order that needs to be audited by the provincial professional unit is transferred to the provincial professional audit department after the provincial preliminary examination is passed; the work order is transferred to the work order follow-up unit after the provincial professional audit is passed, and is transferred to the provincial preliminary examination unit if it is not passed;

[0121] The work order inlet and outlet are monitored, the work order is monitored and followed up through two-way telephone, the work order service quality is provided, and the work order risk is reduced;

[0122] The processed work order is archived and synchronized to the i State Grid system.

[0123] Finally, it should be noted that: the above only describes the preferred embodiments of the present application and is not intended to limit the present application, although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent replacement of some technical features, any modification, equivalent replacement, improvement, etc. within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A risk management method for electricity marketing operations, characterized in that, The method includes the following steps: The risk request work order is transmitted to the provincial customer service center, and the provincial customer service center classifies the work order. The provincial customer service center uses voice analysis and text sentiment analysis engines to obtain customers' emotional resonance index and intent ambiguity in real time. Based on the emotional resonance index, intent ambiguity, historical work order data, and customer behavior data, a risk prediction model is constructed to output the risk level of the work order. The risk level of the work order is initially reviewed by the provincial customer service center staff and professionally classified, and the risk-related work orders are automatically dispatched. Track and visit risk-related work orders, monitor and issue early warnings, and conduct follow-up visits to obtain customer satisfaction.

2. The risk management method for electricity marketing operations according to claim 1, characterized in that: The emotional resonance index is obtained by weighting and averaging the emotional sub-indices of speech features and the emotional sub-indices of text sentiment. The emotion sub-index of speech features is obtained by extracting features such as fundamental frequency change rate, speech rate, and volume intensity, calculating the deviation of the fundamental frequency change rate, speech rate, and volume intensity from the historical average, and then summing the deviations by weight. The sentiment sub-index of text sentiment is obtained by extracting the density of sentiment words and sentiment polarity scores from the text sentiment analysis engine, calculating their deviation from the historical average, and then summing the deviations by weight.

3. The risk management method for electricity marketing operations according to claim 2, characterized in that: The intent ambiguity is calculated by weighting the intent matching degree, intent clarity, and intent missing rate. By calculating the cosine similarity or Euclidean distance between customer text and service type, and using the cosine similarity as the intent matching degree; By analyzing the average sentence length, number of clauses, and grammatical complexity of customer texts, the deviation of the average sentence length, number of clauses, and grammatical complexity from the historical average is calculated. The deviations are then weighted and summed to obtain the intent clarity. The intent missing rate is obtained by detecting the number of missing key fields in the customer's text and calculating the ratio of the number of missing key fields to the preset total number of key fields.

4. The risk management method for electricity marketing operations according to claim 3, characterized in that: The risk prediction model is a fusion model based on multimodal deep learning networks and Bayesian networks; By using historical work order data and customer behavior data as training datasets, the risk prediction model is trained, and the weight allocation of sentiment resonance index and intent ambiguity is optimized. Set risk level classification rules and adjust the output threshold of the risk prediction model; The risk level classification rules include low risk level, medium risk level, high risk level, and red line risk level.

5. The risk management method for electricity marketing operations according to claim 4, characterized in that: The steps for the initial review by the provincial customer service center personnel are as follows: Work orders classified as low-risk, medium-risk, high-risk, and red-line risk levels will undergo preliminary review at the provincial level. Then, the provincial customer service center staff will conduct a preliminary review of the work order and dispatch it to the next stage, or assign it to the municipal / county power supply service command center. Work orders that have undergone preliminary review by the provincial customer service center and are deemed to have a risk level are transferred to the provincial professional review stage. They are then assigned to the corresponding professional reviewers' service pending tasks and iGuoWang pending tasks according to the work order type, and SMS reminders are sent. If the review is approved, the work order will be archived; otherwise, it will be returned to the city / prefecture-level city receipt confirmation stage.

6. The risk management method for electricity marketing operations according to claim 5, characterized in that: The city-level receipt confirmation process is set up at the front end of the provincial customer service center. It is used to receive risk complaint work orders and to classify and push work orders according to the results of the initial review.

7. The risk management method for electricity marketing business according to claim 6, characterized in that: The provincial customer service center staff, in their initial review of work orders, combined emotional resonance index and intent ambiguity to dynamically adjust the work order classification logic. Work orders with an emotional resonance index exceeding the first preset value will be prioritized for provincial professional review. Work orders with intent ambiguity exceeding the second preset value are prioritized for assignment to the municipal / county power supply service command center, where they will supplement information before being submitted to the provincial level for review.

8. The risk management method for electricity marketing business according to claim 7, characterized in that: The professional classification divides work orders into different categories based on different sources, subcategories, and keywords, and pushes different PP messages and SMS messages to the corresponding personnel according to the professional category.

9. A risk management method for electricity marketing operations according to claim 8, characterized in that: The automatic dispatch is achieved by using RPA robots to automatically open, log in, perform OCR recognition, and process work orders. The RPA robots replace manual operations and provide early warning of operational anomalies. The RPA robot automatically archives work orders and synchronizes them to the iGuowang system, while monitoring the processing timeliness. If the timeout is exceeded, an early warning is triggered. The RPA robot monitors the operational status in real time, automatically triggering anomaly warnings and notifying human intervention in cases of login failure, OCR recognition errors, or work order timeouts; enabling collaborative processing between the RPA robot and human review.

10. A risk management method for electricity marketing operations according to claim 9, characterized in that: The tracking access uses a two-telephone alert mechanism: First call alert: After a work order is dispatched, pending data is generated on the first call interface to monitor whether the city / county company contacts the customer within the specified time; if the contact is not made on time, an alert SMS is triggered, and the follow-up results and customer satisfaction are recorded. Supports filtering of the first call data by work order type, time, and risk level, and generates a visual analysis report; Second call alert: After the fault work order is processed, pending data is generated on the second call page to monitor whether the second follow-up call is completed within the specified time. If contact is not made on time, an alert SMS will be triggered, and the follow-up results and customer satisfaction will be recorded to form a closed-loop management system.

Images