Power consumer feedback analysis and processing method and system based on intelligent cognitive guidance

Abstract

The invention relates to a power consumer feedback analysis and processing method and system based on intelligent cognitive guidance. Through intelligent analysis and automatic response generation functions of a large language model, user requirements are efficiently classified and automatically processed under large-scale feedback data, and the feedback processing speed and accuracy are remarkably improved; meanwhile, through a fault prediction and early warning mechanism, potential power faults can be found in advance, preventive maintenance is carried out, the influence of sudden faults on users is reduced, and the operation and maintenance cost is reduced.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a method and system for analyzing and processing power user feedback based on intelligent cognitive guidance. Background Art

[0002] With the rapid development of the power industry, the number of power users has increased dramatically, and users' requirements for power service quality have gradually increased. Traditional power user feedback systems are mostly based on manual processing, which not only has a slow response speed, but also makes it difficult to accurately analyze and effectively classify a large amount of diverse feedback, making it difficult to improve user experience and satisfaction. In addition, traditional systems lack intelligent fault prediction functions, and cannot promptly detect, warn, and handle potential faults in the power service process, further increasing the operational pressure on the service department.

[0003] In the existing technology, some companies try to use machine learning classifiers or rule-based systems to analyze user feedback, such as using keywords or simple sentiment analysis methods to identify user dissatisfaction or complaints. However, rule-based methods have obvious limitations, especially when faced with user feedback that is rich in language and diverse in expression. These methods are difficult to adapt to complex semantic changes and subtle emotional changes. At the same time, traditional machine learning methods are highly dependent on data, and often require manual annotation of a large amount of sample data to improve the effectiveness of the model, which increases the cost and difficulty of system development.

[0004] In addition, sentiment analysis systems based on natural language processing (NLP) have begun to be applied to user feedback analysis. For example, some companies use convolutional neural networks (CNN) or recurrent neural networks (RNN) to analyze the sentiment and content tendencies of user feedback, which partially improves the level of automation in feedback processing. However, traditional NLP technology still has bottlenecks in long text understanding and multi-level semantic analysis, and it is difficult to accurately capture the deep-level needs and implicit information in user feedback. These systems find it difficult to fully utilize knowledge specific to the power industry, and therefore have significant deficiencies in understanding user feedback and personalized processing.

[0005] Therefore, the analysis and processing technology for power user feedback still needs to be improved. Summary of the invention

[0006] Based on this, it is necessary to provide a method and system for analyzing and processing power user feedback based on intelligent cognitive guidance to address the above technical problems, so as to achieve efficient and intelligent feedback processing.

[0007] In a first aspect, the present application provides a method for analyzing and processing power user feedback based on intelligent cognitive guidance. The method comprises:

[0008] Get current user feedback data;

[0009] Use large language models to analyze the sentiment and intent of current user feedback data, extract user needs and quantify the intensity of emotions; prioritize user needs based on the intensity of emotions and historical user feedback data;

[0010] Generate dynamic prompts guided by prompt word templates based on user needs, user need priorities, emotional intensity, and historical user feedback data; call the power industry knowledge base and combine dynamic prompts to generate automated responses;

[0011] Failure mode prediction and early warning are carried out based on current user feedback data and historical user feedback data.

[0012] In one embodiment, using a large language model to perform sentiment and intent analysis on current user feedback data, extracting user needs and quantifying sentiment intensity includes:

[0013] Identify user needs in current user feedback data through the contextual semantic understanding capabilities of large language models;

[0014] A large language model is used to obtain the emotional tendency in the current user feedback data, and the emotional tendency is quantified to obtain the emotional intensity.

[0015] In one embodiment, the method further comprises:

[0016] Receive user quality evaluation of automated responses, and optimize prompt word templates, power industry knowledge base, and large language models based on the quality evaluation.

[0017] In one embodiment, performing failure mode prediction and early warning based on current user feedback data and historical user feedback data includes:

[0018] Extract the characteristics of the failure mode and the failure-related keywords based on the pre-built historical data set containing historical user feedback data, failure logs and maintenance records;

[0019] Detect fault-related keywords in current user feedback data and use the contextual semantic understanding capabilities of large language models to identify fault modes;

[0020] Push warning information based on failure mode.

[0021] In one embodiment, performing failure mode prediction and early warning based on current user feedback data and historical user feedback data further includes:

[0022] Identify relevant feedback from the same area or the same device as the current user feedback data within the preset monitoring period, determine whether there is a regional failure trend, and push warning information based on the judgment results.

[0023] In one embodiment, performing failure mode prediction and early warning based on current user feedback data and historical user feedback data further includes:

[0024] Perform time series analysis on historical data sets to identify the periodic characteristics of fault occurrence and build a time series prediction model based on the periodic characteristics;

[0025] Based on current user feedback data, the time series prediction model is used to predict failure modes, and early warning information is pushed according to the prediction results.

[0026] In the second aspect, the present application also provides a power user feedback analysis and processing system based on intelligent cognitive guidance. The system includes:

[0027] Data collection module, used to obtain current user feedback data;

[0028] Feedback intelligent analysis module, which uses a large language model to analyze the sentiment and intent of current user feedback data, extract user needs and quantify the intensity of emotions; and prioritize user needs based on the intensity of emotions and historical user feedback data;

[0029] The automatic response and personalized suggestion generation module is used to generate dynamic prompts guided by prompt word templates based on user needs, user need priorities, emotional intensity and historical user feedback data; call the power industry knowledge base and combine dynamic prompts to generate automatic responses;

[0030] The fault prediction and warning module is used to predict fault modes and provide warnings based on current user feedback data and historical user feedback data.

[0031] In a third aspect, the present application further provides a computer device, which includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps in the above-mentioned method for analyzing and processing power user feedback based on intelligent cognitive guidance when executing the computer program.

[0032] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, which implements the steps in the above-mentioned method for analyzing and processing power user feedback based on intelligent cognitive guidance when executed by a processor.

[0033] In a fifth aspect, the present application also provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, the steps in the above-mentioned power user feedback analysis and processing method based on intelligent cognitive guidance are implemented.

[0034] The above-mentioned method and system for analyzing and processing power user feedback based on intelligent cognitive guidance can efficiently classify and automatically process user needs under large-scale feedback data through the intelligent analysis and automatic response generation functions of a large language model, thereby significantly improving the feedback processing speed and accuracy; at the same time, through the fault prediction and early warning mechanism, potential power faults can be discovered in advance, preventive maintenance can be carried out, the impact of sudden faults on users can be reduced, and operation and maintenance costs can be reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is a flowchart of a method for analyzing and processing power user feedback based on intelligent cognitive guidance in one embodiment;

[0036] Figure 2 A flowchart of a feedback intelligent analysis module in one embodiment;

[0037] Figure 3 A flowchart of an automated response and personalized suggestion generation module in one embodiment. DETAILED DESCRIPTION

[0038] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0039] This application aims to solve the problems of inefficiency, low precision and lack of intelligence in the processing of feedback from electricity users, especially the response speed and accuracy in a large-scale, multi-channel feedback environment. Traditional electricity user feedback systems mainly rely on manual work when processing feedback, resulting in slow response speeds and high operating costs, making it difficult to meet the increasing demand for service quality and response time from electricity users. At the same time, the existing feedback system's understanding and analysis of user emotions and needs remain at a basic level, lacking in-depth semantic analysis and emotion recognition capabilities, and unable to provide appropriate solutions based on user personalized needs. In addition, traditional systems have limitations in feedback classification and urgency judgment, making it difficult to reasonably prioritize user feedback, making it difficult for power companies to respond quickly to high-priority issues.

[0040] Another important goal of this application is to solve the problem of fault prediction and early warning in power services. Existing feedback systems usually lack the ability to comprehensively analyze historical data and user feedback, and therefore cannot effectively predict possible power failures or hidden dangers. Traditional systems only respond after a problem occurs, resulting in failures that may have unnecessary impacts on the user's power experience and may increase the maintenance costs of the power sector. Therefore, a system is needed that can use historical feedback data and real-time feedback for predictive analysis, so that power companies can intervene in advance before problems occur.

[0041] This application proposes a method for analyzing and processing power user feedback based on a large language model guided by intelligent cognition, aiming to fundamentally improve the above problems. By applying the natural language processing and semantic understanding capabilities of a large language model, it is possible to conduct a multi-dimensional analysis of the emotions, needs, and key issues in the user feedback data, and automatically generate personalized response content in combination with the user's historical data. In addition, by deeply mining historical user feedback data, it is also possible to identify potential failure modes in the feedback, analyze and predict the possibility and severity of potential failures, and then issue early warnings in a timely manner to notify relevant departments to inspect or maintain possible sources of failure to avoid more serious problems.

[0042] Specifically, the embodiment of the present application provides a method for analyzing and processing power user feedback based on intelligent cognitive guidance, such as Figure 1 As shown, the following steps are included:

[0043] S1. Obtain current user feedback data.

[0044] Collect data containing user feedback from various channels (such as telephone, social media, online messages, etc.), pre-process the raw data of user feedback by cleaning, standardizing and formatting, and obtain current user feedback data for subsequent text analysis and model application.

[0045] S2. Use large language models to analyze the sentiment and intent of current user feedback data, extract user needs and quantify the intensity of emotions; prioritize user needs based on the intensity of emotions and historical user feedback data.

[0046] The pre-processed current user feedback data is input into the Large Language Model (LLM), which uses natural language understanding and semantic analysis technology to analyze the feedback content in multiple dimensions, identify the emotional tendency and demand category in the feedback, and achieve intelligent understanding of user feedback. In order to improve service quality, user needs are prioritized according to emotional intensity and historical user feedback data, so that corresponding services can be provided in order of priority.

[0047] S3. Generate dynamic prompts guided by prompt word templates based on user needs, user need priorities, emotional intensity, and historical user feedback data; call the power industry knowledge base and generate automated responses in combination with dynamic prompts.

[0048] Based on the results of the analysis of the current user feedback data obtained in step S2, the large language model combines the user's historical records and the power industry knowledge base to generate a personalized automated response, which includes suggestions for solving user needs. The automated response can directly reply to the user or be pushed to the customer service staff, greatly improving the feedback processing efficiency and user experience.

[0049] S4. Perform failure mode prediction and early warning based on current user feedback data and historical user feedback data.

[0050] Through large-scale language models, fault pattern recognition and time series analysis are performed on historical user feedback data to identify potential fault features in the feedback, predict the occurrence of faults, generate early warnings in real time, and notify relevant departments to perform preventive maintenance.

[0051] This embodiment can realize efficient and accurate analysis and processing of large-scale feedback data from multiple channels, thereby achieving the purpose of intelligent and automated user feedback analysis and power service improvement.

[0052] In one embodiment, S1 is responsible for standardizing and structuring user feedback data from multiple channels, laying the foundation for subsequent intelligent analysis and automatic response generation. S1 mainly includes:

[0053] S11. Multi-channel data collection: Obtain raw data of user feedback from multiple user feedback channels such as phone recordings, social media messages, online forms, emails, etc. Since the data formats and content forms of each channel are different, it is necessary to establish an interface to collect various types of feedback data in a unified manner and record metadata such as feedback time, channel, user identity, etc. to ensure the integrity and accuracy of the feedback data.

[0054] S12. Data cleaning and denoising: The original data may contain noise (such as extra punctuation, spaces, advertising information, etc.), and data cleaning operations are required. Specifically, it includes removing special characters, blank characters and repeated data, eliminating irrelevant information, and retaining the core content of the feedback text. For telephone voice feedback, you can first convert the voice into text through speech recognition technology (such as ASR, Automatic Speech Recognition), and then enter the cleaning step.

[0055] S13. Text preprocessing: Perform basic preprocessing operations of natural language processing on the cleaned text data, including word segmentation, part-of-speech tagging, stop word removal, and keyword extraction, etc. Word segmentation decomposes the text into words or phrases for subsequent semantic analysis; part-of-speech tagging marks the grammatical attributes for each word or phrase to help the large language model better understand the context meaning. In addition, removing stop words (such as "de", "le", etc.) helps improve the efficiency and accuracy of text analysis.

[0056] S14. Preliminary sentiment and intention tagging: In some cases, to improve the efficiency of subsequent analysis, the system can perform preliminary sentiment and intention tagging on user feedback. Through shallow sentiment analysis algorithms or keyword matching, the system can preliminarily judge the emotional tendency (such as positive, negative, neutral) of user feedback and the general intention of the feedback (such as complaint, consultation, suggestion, etc.), providing a reference for the in-depth analysis of the subsequent large language model.

[0057] S15. Data format standardization: Format the preprocessed data to form a unified data structure, such as in JSON or database table form, to ensure that the data can be seamlessly transferred to the subsequent feedback intelligent analysis module. At the same time, metadata such as feedback channels, time, user IDs, etc. will be associated with the preprocessed text for storage, providing support for further analysis and personalized responses.

[0058] Through the above steps, valuable content can be extracted from the complex and diverse current user feedback data, laying a data foundation for subsequent intelligent analysis, automatic response generation, and fault prediction, ensuring the coherence and accuracy of the overall system process.

[0059] In one embodiment, in S2, the large language model is used to perform sentiment and intention analysis on the current user feedback data, and extract user needs and quantify the emotional intensity, including: through the context semantic understanding ability of the large language model, identifying the user needs in the current user feedback data; using the large language model to obtain the emotional tendency in the current user feedback data and quantify the intensity of the emotional tendency to obtain the emotional intensity.

[0060] S2 is the core step of the method, using the large language model to perform in-depth semantic understanding and sentiment analysis on the user feedback content. This step aims to accurately identify the core needs, emotional tendencies, and problem types of users, providing data support and analysis basis for subsequent automated responses and fault predictions. S2 mainly includes:

[0061] S21. Contextual semantic understanding: User feedback content often contains multi-level information and implicit intent. The contextual semantic understanding capabilities of large language models can be used to identify key content and semantic relationships in the current user feedback data. Specifically, the large language model will analyze the semantic structure and context of the current user feedback data, and accurately locate key issues and needs (such as power outages, billing issues, service complaints, etc.) in the current user feedback data. Preferably, when user feedback contains multiple types of information (such as power fluctuations and excessive costs), the large language model can distinguish the primary and secondary demands of the feedback to ensure the accuracy of the analysis results.

[0062] S22. Emotion recognition and sentiment analysis: The large language model judges the emotional tendency (such as positive, negative, neutral) of user feedback through multi-level sentiment analysis and quantifies the intensity of the emotion. This analysis process not only relies on the emotional weight of the keywords, but also makes a comprehensive judgment based on the overall semantics of the feedback. For example, for feedback such as "I am very dissatisfied with the power service and the power supply is always interrupted", the large language model can identify strong dissatisfaction. Preferably, the priority is marked according to the intensity of the emotion to provide support for subsequent rapid response and further analysis.

[0063] S23. Demand classification and prioritization: User needs are classified according to the analysis results of large language models, such as technical support, billing issues, complaints and suggestions, and prioritized according to factors such as sentiment intensity, keyword features, and historical data. For example, feedback containing the keywords "power outage" or "emergency" will be given priority and pushed to the relevant departments to ensure timely response. In addition, prioritization will also take into account the user's historical feedback records, such as frequently occurring issues will be marked as high priority.

[0064] S24. Historical feedback correlation analysis: The large language model combines the user's historical feedback data to identify the user's potential demands and concerns by recording and analyzing past issues. For example, if a user has repeatedly consulted about electricity bill issues, such issues will be given priority, and more targeted suggestions or solutions will be generated based on past responses. This correlation analysis can further improve the personalized processing effect of feedback and provide users with customized services.

[0065] S25. Fault feature extraction and cluster analysis: By clustering the feedback content, possible fault feature patterns can be identified. For example, the system can automatically cluster feedback containing similar descriptions such as "voltage instability" and "frequent power outages" into a group, and further analyze whether these feedbacks point to potential power equipment failures or regional power grid problems. This process combines the pattern recognition capabilities of clustering algorithms and large language models, which not only improves the accuracy of fault detection, but also provides important input for fault prediction and early warning modules.

[0066] S26, contextual prompts and knowledge base calls: During the large-scale language model analysis process, prompts will be dynamically generated based on the feedback content to guide the large-scale language model to combine with the power industry knowledge base for more accurate analysis. For example, when user feedback mentions specific power equipment or service conditions, the system will automatically call the relevant knowledge base content to enrich the contextual understanding ability of the large-scale language model. This prompt and knowledge base call mechanism can significantly improve the performance of the system in professional scenarios and ensure that the analysis results are specific and accurate to the power industry.

[0067] S27. Result output and data formatting: After in-depth analysis by a large language model, the system formats the generated analysis results into structured data, including feedback categories, emotional tendencies, priorities, key demand points, etc. The result data will be stored in the system database and used as input for subsequent modules (such as automatic response generation and fault prediction and warning) to ensure the linkage and data fluency of the entire system.

[0068] S2 achieves a comprehensive analysis of user feedback through the semantic understanding, sentiment analysis, pattern recognition and other functions of a large language model. This step ensures that the system can accurately identify user needs and emotions in a complex feedback environment, reasonably sort and classify feedback, and improve overall processing efficiency and user service quality.

[0069] S3 generates automated responses and personalized solutions based on understanding user feedback through a large language model. This step aims to improve the efficiency and quality of feedback processing and user experience through intelligent and precise automatic responses. S3 mainly includes:

[0070] S31. Receive feedback analysis results

[0071] S311, receiving analysis data: first receiving the structured data obtained according to S2, including the demand category in the feedback content (such as technical support, billing issues, fault reports, etc.), emotional tendencies (such as dissatisfaction, anxiety, doubts, etc.), feedback priorities (high, medium, low), and historical user feedback data.

[0072] S311, Strategy determination: Determine the processing strategy for the response content based on the feedback demand category and priority. For example, high-priority fault report feedback needs to be handled urgently and a clear solution should be generated; while feedback such as bill consultation requires detailed explanation and guidance.

[0073] S32. Dynamic prompt generation

[0074] S321. Design prompt word templates: Based on the different types of feedback needs and emotional tendencies, design specific prompt word templates to guide the response generation of large language models. For example, for feedback with strong complaint emotions, prompt words will use more soothing language, increase understanding vocabulary, and improve the emotional adaptability of the system.

[0075] S322. Historical record prompt integration: When a user has multiple feedback records, historical user feedback data will be integrated to dynamically adjust the prompt words. For example, for users who have consulted about billing issues multiple times, the prompt words may be added with "Regarding your previous billing inquiry...", so that the large language model can output a coherent and personalized response.

[0076] S323, dynamic prompt word generation: Combine the current user feedback data and its analysis results to generate prompt words that include emotional adaptation, personalized needs and professional knowledge, and guide the response generation of the large language model. This prompt word guides the language, tone and content of the generated response, so that the large language model can output a response that is more in line with the user's emotions and needs.

[0077] S33. Calling the power industry knowledge base

[0078] S331. Knowledge base retrieval and matching: Match the keywords identified in the prompt words with the electric power industry knowledge base, and retrieve technical support information, FAQs, bill calculation details, etc. related to the current user feedback data from the electric power industry knowledge base to ensure the accuracy and professionalism of the response.

[0079] S332. Content screening and integration: When user feedback involves power-related professional content (such as "voltage instability" or "abnormal electricity charges"), the power industry knowledge base will search and provide corresponding knowledge points (such as common causes of voltage fluctuations and coping methods). Relevant knowledge is automatically screened from the power industry knowledge base and integrated into the prompt words of the large language model, so that the generated responses have rich industry background.

[0080] S333. Real-time knowledge update: The power industry knowledge base can be updated regularly to ensure timely addition of the latest policies and technical information, so that the accuracy of the response content remains at the forefront of the industry at all times.

[0081] S34. Automated response generation

[0082] S341. Large language model response generation: After receiving the prompt words and the power industry knowledge base information, the large language model generates an automated reply based on the feedback content. The response content includes three core elements:

[0083] (1) A large language model interprets the question based on the feedback content to help users understand the current situation. For example, for frequent power outages, the system will explain the possible causes of the outage and provide relevant information.

[0084] (2) Solution suggestions: Based on the semantic generation capabilities of the power industry knowledge base and large language models, it generates solution suggestions for specific problems, such as how to query electricity bills and how to check electrical equipment.

[0085] (3) Supplementary explanation: If the user's emotions tend to be dissatisfied or anxious, a soothing supplementary explanation will be generated, such as "We understand your concerns, and the following are possible solutions...", to enhance the emotional adaptability of the response.

[0086] S342. Emotional expression adaptation: Large language models will automatically adjust the language style and tone when generating responses. For example, feedback on dissatisfaction will use soothing language, while feedback on questions will focus on explanation. This emotional adaptation makes the response more humane and contextual.

[0087] S35. Feedback content push and channel adaptation

[0088] S351, Channel Adaptation and Push: Push the generated response to the user, supporting multiple channels such as SMS, APP push, email, web page online, etc., to ensure that the user can receive the response in time. It can also automatically adjust the format and length of the response according to the channel to ensure the best reading experience in each channel.

[0089] S352. Automatic and manual review mechanism: For general questions, automatically generated replies will be sent directly; for special cases (such as questions with complex emotions and high technical requirements), automatically generated responses will be pushed to customer service personnel for review. Customer service personnel can further optimize the language and content to ensure the accuracy of the response.

[0090] S36. Response quality evaluation and feedback storage

[0091] S361. Quality evaluation and feedback improvement: Evaluate the response quality based on the user's subsequent behavior (such as whether to ask questions, feedback emotional changes), and incorporate it into the quality scoring system. By regularly summarizing the user's quality evaluation feedback, gradually optimize the prompt word template, the power industry knowledge base call method and the large language model fine-tuning strategy, and improve the quality of subsequent response content and user satisfaction.

[0092] S362, Data storage and continuous optimization: The generated responses and quality evaluation results will be stored in the user feedback database and used as the basis for subsequent response content optimization. The system regularly analyzes this data, continuously improves the prompt word generation logic, knowledge base content, and large language model fine-tuning methods to ensure that the system gradually improves the response effect in different types of feedback scenarios.

[0093] S4 uses a large language model to analyze historical data and current feedback from users, identify possible failure modes, and conduct predictive analysis and real-time warnings. This module aims to use user feedback information and feature extraction of failure modes to provide early warnings of potential power problems, helping power companies to take preventive maintenance measures in a timely manner, thereby reducing the probability of failures and maintenance costs. S4 specifically includes:

[0094] S41. Historical data analysis and pattern learning

[0095] S411. Dataset construction: First, a historical dataset containing historical user feedback data, fault logs, and maintenance records over the years is constructed. These data are used to train and fine-tune a large language model so that it can learn the characteristic patterns of different types of faults and their corresponding language descriptions.

[0096] S412, Feature extraction and tagging: Extract and tag the keywords and language patterns related to the fault from the historical data set. For example, keywords such as "frequent power outages", "voltage fluctuations", and "equipment overheating" will be used as tags, and the contextual associations of these keywords will be captured through a large language model.

[0097] S413, Time series feature learning: Based on historical data sets, time series analysis is used to identify the periodic characteristics of fault occurrence, such as equipment problems caused by seasonal power load changes. Large language models are combined with time information to learn the time series characteristics of fault occurrence to support the timeliness and accuracy of early warning.

[0098] S42, real-time feedback fault mode identification

[0099] S421, keyword detection and preliminary analysis: When new user feedback data is obtained, the large language model first performs keyword detection on the new feedback content to determine whether it contains potential fault signals. If feature words such as "voltage instability" or "power outage" are detected, the context of the feedback and the relevance of the fault will be further analyzed.

[0100] S422, Contextual pattern recognition: Large language models not only rely on keywords, but also use contextual understanding capabilities to identify fault patterns. For example, for the feedback "The equipment has been frequently powered off recently, and the voltage seems unstable", it recognizes that this is a possible fault signal and further analyzes its fault type and urgency.

[0101] S423, related feedback detection: Identify similar feedback in the same area or the same device in a short period of time to determine whether there is a regional failure trend. If multiple users report similar failure phenomena (such as frequent power outages) in the same area, it will be judged as a high priority problem and further push warning information.

[0102] S43. Fault prediction model generation

[0103] S431, Time Series Prediction Model: Based on the characteristics of historical data, a time series prediction model based on a large language model is established to predict the probability and time of certain faults. For example, if an increase in grid load is detected in the summer, and past data indicates that equipment overload will occur during this period, the possibility of equipment overload is predicted and an early warning is generated.

[0104] S432, Contextualized prediction tips: Large language models will combine time factors, emotional tendencies and other information in the feedback content to predict the possible impact range and time of the fault. For example, when users report that the voltage is unstable during the peak period and the emotions are strong, the urgency of the feedback will be weighted and the relevant departments will be advised to check in advance.

[0105] S433. Dynamic update and model optimization: The fault prediction model will be dynamically adjusted and optimized based on new feedback data and actual faults, so that the accuracy of the model in predicting different types of faults will gradually improve.

[0106] S44, Fault warning generation and push

[0107] S441. Warning information generation: When the large language model detects a potential failure mode or a high-probability failure event, a warning message is generated. The warning content includes the type of failure, the time when it may occur, the area or equipment involved, and the recommended preventive maintenance measures.

[0108] S442, Automated push and hierarchical alarm: The system will automatically push warning information to relevant departments according to the severity and urgency of the fault. For high-priority warnings, the system will push them directly to the maintenance management level; while general faults will notify front-line operation and maintenance personnel to conduct routine inspections.

[0109] S443, Multi-level alarm mechanism: Dynamically adjust the warning level according to the number and pattern changes of feedback. For example, an increase in similar feedback in the same area will trigger a higher level of alarm to ensure that high-risk faults are responded to quickly.

[0110] S45, Fault feedback and early warning improvements

[0111] S451. Fault handling result feedback: After a fault occurs, the actual results of the fault handling are recorded, including the cause of the fault, handling time, user feedback, etc. This information will be used to optimize the prediction model and improve the accuracy of the system's early warning in similar scenarios.

[0112] S452. Prediction effect evaluation: Regularly evaluate the accuracy and effect of warnings, such as the timeliness of warnings, troubleshooting rate and other indicators, and evaluate the effect of the prediction model. For warning cases with low accuracy, the system will re-analyze key features to improve the learning ability of the model.

[0113] S453. Dynamic adjustment of warning parameters: Automatically adjust warning parameters and pattern characteristics based on prediction results and feedback data to improve the accuracy of subsequent fault predictions and further reduce fault omission and false alarm rates.

[0114] This application uses large language model technology to provide new possibilities for intelligent analysis and processing of user feedback. Through pre-training on massive text data, large language models show powerful natural language understanding and generation capabilities, especially in the context processing of long texts, complex semantic analysis and multi-level emotion recognition. Compared with traditional methods, large language models can more accurately understand the implicit needs in user feedback, and combine users' historical feedback and industry knowledge base to provide users with personalized processing suggestions and problem solutions. Especially in the user feedback analysis scenario in the power industry, large language models can not only generate accurate automated responses based on users' feedback emotions and content, but also analyze potential patterns in feedback data, predict possible fault problems, and issue early warnings to help power departments respond in time and perform targeted maintenance. Therefore, the proposal of a power user feedback analysis and processing system based on a large language model can not only solve the shortcomings of traditional feedback processing systems, but also greatly improve the service quality and operational efficiency of power companies. The application of this technology will drive the power industry towards a new model of more intelligent, automated and forward-looking user services, providing users with a better user experience.

[0115] The main innovations of this application include:

[0116] 1. Through multi-dimensional semantic analysis and emotion recognition of user feedback through large-scale language models, we can deeply understand the core needs and emotional tendencies of users, generate appropriate responses, and improve the accuracy and speed of feedback processing.

[0117] 2. Personalized response generation based on multi-source information fusion: Combining current user feedback data, historical user feedback data, and the power industry knowledge base, a large language model generates contextualized and personalized automated responses, providing users with accurate and personalized solutions.

[0118] 3. Based on historical user feedback data analysis and keyword pattern recognition, it can detect potential failure modes in feedback in advance, make time series predictions and generate early warnings, helping power companies take preventive maintenance measures before problems occur.

[0119] 4. Support the integration of user feedback data from multiple channels (such as telephone, social media, online messages, etc.), and reasonably classify and prioritize the feedback content based on the classification capabilities of large language models to achieve efficient and intelligent feedback processing.

[0120] Through the above innovations, the effective effects that this application can achieve include:

[0121] 1. Improve feedback processing efficiency: Through the intelligent analysis and automatic response generation functions of large-scale language models, the system can efficiently classify and automatically process user needs under large-scale feedback data, significantly improving the speed and accuracy of feedback processing.

[0122] 2. Personalized service: The system generates personalized suggestions and solutions based on the user's historical feedback and emotional tendencies, improves user satisfaction and loyalty, and optimizes the power company's service experience.

[0123] 3. Reduce the impact of failures and maintenance costs: Through the fault prediction and early warning mechanism, power companies can detect potential power failures in advance, carry out preventive maintenance, reduce the impact of sudden failures on users, and reduce operation and maintenance costs.

[0124] 4. Enhanced data utilization and decision support: Multi-channel integrated feedback data and large-scale language model analysis results provide power companies with in-depth insights into user needs and provide reliable support for optimizing service strategies and resource allocation.

[0125] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0126] Based on the same inventive concept, the embodiment of the present application also provides a system for analyzing and processing electric power user feedback based on intelligent cognitive guidance for realizing the above-mentioned method for analyzing and processing electric power user feedback based on intelligent cognitive guidance. The implementation scheme for solving the problem provided by the system is similar to the implementation scheme recorded in the above-mentioned method, so the specific limitations in one or more embodiments of the electric power user feedback analysis and processing system based on intelligent cognitive guidance provided below can refer to the limitations of the method for analyzing and processing electric power user feedback based on intelligent cognitive guidance above, and will not be repeated here.

[0127] In one embodiment, a power user feedback analysis and processing system based on intelligent cognitive guidance is provided, including: a data acquisition module, a feedback intelligent analysis module, an automatic response and personalized suggestion generation module, and a fault prediction and early warning module, wherein:

[0128] Data collection module, used to obtain current user feedback data;

[0129] Feedback intelligent analysis module, which uses a large language model to analyze the sentiment and intent of current user feedback data, extract user needs and quantify the intensity of emotions; and prioritize user needs based on the intensity of emotions and historical user feedback data;

[0130] The automatic response and personalized suggestion generation module is used to generate dynamic prompts guided by prompt word templates based on user needs, user need priorities, emotional intensity and historical user feedback data; call the power industry knowledge base and combine dynamic prompts to generate automatic responses;

[0131] The fault prediction and warning module is used to predict fault modes and provide warnings based on current user feedback data and historical user feedback data.

[0132] like Figure 2 The following is the workflow diagram of the feedback intelligent analysis module: Figure 3 Shown is the workflow diagram of the automated response and personalized suggestion generation module.

[0133] Each module in the above-mentioned power user feedback analysis and processing system based on intelligent cognitive guidance can be implemented in whole or in part by software, hardware and their combination. Each of the above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0134] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in all the above method embodiments when executing the computer program.

[0135] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in all the above method embodiments are implemented.

[0136] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in all the above method embodiments when executed by a processor.

[0137] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards of relevant countries and regions.

[0138] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited thereto. The processors involved in each embodiment provided in this application may be general-purpose processors, central processing units, graphics processors, digital signal processors, programmable logic devices, data processing logic devices based on quantum computing, etc., but are not limited thereto.

[0139] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0140] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for analyzing and processing power user feedback based on intelligent cognitive guidance, characterized in that: The method comprises: Get current user feedback data; Using a large language model to perform sentiment and intent analysis on the current user feedback data, extract user needs and quantify sentiment intensity; and prioritize the user needs based on the sentiment intensity and historical user feedback data; Generate dynamic prompts guided by prompt word templates according to the user needs, the priorities of the user needs, the emotional intensity and the historical user feedback data; call the power industry knowledge base and generate automated responses in combination with the dynamic prompts; Fault mode prediction and early warning are performed based on the current user feedback data and the historical user feedback data.

2. The method according to claim 1, characterized in that The use of a large language model to perform sentiment and intent analysis on the current user feedback data, extract user needs and quantify sentiment intensity includes: Identifying the user needs in the current user feedback data through the contextual semantic understanding capability of the large language model; The large language model is used to obtain the emotional tendency in the current user feedback data, and the emotional tendency is quantified to obtain the emotional intensity.

3. The method according to claim 1, characterized in that The method further comprises: The user's quality evaluation of the automated response is received, and the prompt word template, the electric power industry knowledge base, and the large language model are optimized according to the quality evaluation.

4. The method according to claim 1, characterized in that: The performing of fault mode prediction and early warning according to the current user feedback data and the historical user feedback data includes: Extracting the characteristics of the failure mode and the failure-related keywords according to the pre-built historical data set including the historical user feedback data, the failure log and the maintenance record; Detecting the fault-related keywords in the current user feedback data, and identifying the fault mode by using the contextual semantic understanding capability of the large language model; Push warning information according to the failure mode.

5. The method according to claim 1, characterized in that The performing of fault mode prediction and early warning according to the current user feedback data and the historical user feedback data further includes: Identify relevant feedback from the same area or the same device as the current user feedback data within a preset monitoring period, determine whether there is a regional failure trend, and push warning information based on the judgment result.

6. The method according to claim 1, characterized in that The performing of fault mode prediction and early warning according to the current user feedback data and the historical user feedback data further includes: Performing time series analysis on the historical data set to identify periodic characteristics of fault occurrence, and building a time series prediction model based on the periodic characteristics; Based on the current user feedback data, the time series prediction model is used to predict the failure mode, and early warning information is pushed according to the prediction result.

7. A power user feedback analysis and processing system based on intelligent cognitive guidance, characterized in that: The system comprises: Data collection module, used to obtain current user feedback data; A feedback intelligent analysis module, which is used to use a large language model to perform sentiment and intent analysis on the current user feedback data, extract user needs and quantify the intensity of emotions; and prioritize the user needs according to the intensity of emotions and historical user feedback data; The automatic response and personalized suggestion generation module is used to generate dynamic prompts guided by prompt word templates according to the user needs, the priorities of the user needs, the emotional intensity and the historical user feedback data; call the power industry knowledge base, combine the dynamic prompts, and generate automatic responses; The fault prediction and early warning module is used to perform fault mode prediction and early warning according to the current user feedback data and the historical user feedback data.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images