Method and system for dynamically updating large model knowledge for operation and maintenance technical services

Abstract

This application provides a method and system for dynamic knowledge updating of a large-scale model for operation and maintenance (O&M) technical services. It relates to the field of O&M technology, and models the knowledge generation triggering factors, iteration impact range, and knowledge association strength in multi-source O&M knowledge update sources to obtain a knowledge full-link evolution set including knowledge generation trigger records, iteration impact assessment results, and cross-knowledge association networks. Then, dynamic coupling rules are generated based on the real-time scenario characteristics of O&M tasks and this knowledge full-link evolution set. These rules are used to filter and integrate the knowledge full-link evolution set to obtain a scenario-coupled knowledge set. This set is embedded into the knowledge architecture of the large-scale O&M technical service model, and storage node parameters and retrieval logic are adjusted to obtain the basic updated model. Finally, the basic updated model is dynamically adapted in multiple dimensions, and the knowledge call priority sequence and storage node parameters are adjusted in conjunction with changes in scenario characteristics to generate an adapted updated model, thereby improving the adaptability of the large-scale O&M technical service model.

Description

Technical Field

[0001] This application relates to the field of operation and maintenance technology, and more specifically, to a method and system for dynamic updating of large-scale model knowledge for operation and maintenance technical services. Background Technology

[0002] With the rapid development of information technology and the widespread application of various complex systems, operation and maintenance (O&M) knowledge is experiencing explosive growth. Multiple sources of O&M knowledge updates, such as updated technical documents from equipment manufacturers, experience summaries accumulated by O&M personnel in practical operations, and revisions to industry standards, are constantly generating new O&M knowledge. However, existing methods for updating O&M technical service knowledge have many limitations.

[0003] On the one hand, traditional methods often view each source of knowledge updates in isolation, lacking systematic modeling of knowledge generation triggers, the scope of iterative impact, and the strength of connections between different knowledge sources. This makes it difficult to fully grasp the evolution of knowledge during the knowledge update process, and to accurately assess the impact of new knowledge on the original knowledge system, thus leading to incomplete or inaccurate knowledge updates.

[0004] On the other hand, when applying updated knowledge to operations and maintenance tasks, existing methods fail to fully consider the real-time scenario characteristics of these tasks. Different operations and maintenance tasks differ in terms of task objective type, equipment operating status parameters, and business impact scope. Traditional methods cannot dynamically adjust knowledge retrieval strategies based on these real-time scenario characteristics, resulting in the knowledge provided to operations and maintenance personnel lacking relevance and practicality, thus reducing operations and maintenance efficiency and quality. Summary of the Invention

[0005] In view of this, the purpose of this application is to provide a method and system for dynamic updating of large model knowledge for operation and maintenance technical services.

[0006] According to a first aspect of this application, a method for dynamically updating large-scale model knowledge for operation and maintenance technical services is provided, the method comprising:

[0007] The knowledge generation triggering factors, iteration impact range, and correlation strength between different knowledge in the multi-source operation and maintenance knowledge update source are modeled and processed to obtain the knowledge full-link evolution set, which includes knowledge generation triggering records, iteration impact assessment results, and cross-knowledge association network;

[0008] Based on the real-time scenario features and knowledge full-link evolution set of operation and maintenance tasks, dynamic coupling rules between scenarios and knowledge are generated. The dynamic coupling rules include scenario knowledge matching conditions, knowledge call priority sequence and related knowledge linkage logic. The real-time scenario features of operation and maintenance tasks include task target type, equipment operating status parameters and business impact range.

[0009] The dynamic coupling rules are used to couple and filter the knowledge content in the knowledge full-link evolution set, extract the knowledge content that meets the matching conditions and integrate them to obtain the scene-coupled knowledge set.

[0010] The scenario-coupled knowledge set is embedded into the hierarchical knowledge architecture of the large-scale operation and maintenance technical service model. Simultaneously, the storage node parameters and retrieval logic of the hierarchical knowledge architecture are adjusted based on the knowledge full-link evolution set to obtain the large-scale operation and maintenance technical service model after basic update.

[0011] The basic updated operation and maintenance technical service big model is dynamically adapted in multiple dimensions. The knowledge call priority sequence and storage node parameters are adjusted in combination with the real-time scenario feature changes of operation and maintenance tasks to generate the adapted updated operation and maintenance technical service big model.

[0012] According to a second aspect of this application, a system for dynamically updating large model knowledge for operation and maintenance technical services is provided. The system includes a machine-readable storage medium and a processor. The machine-readable storage medium stores machine-executable instructions. When the processor executes the machine-executable instructions, the system implements the aforementioned method for dynamically updating large model knowledge for operation and maintenance technical services.

[0013] According to a third aspect of this application, a computer-readable storage medium is provided, wherein computer-executable instructions are stored therein, and when the computer-executable instructions are executed, the aforementioned method for dynamically updating large model knowledge for operation and maintenance technical services is implemented.

[0014] Based on any of the above aspects, the technical effect of this application is as follows:

[0015] By comprehensively and deeply modeling the multi-source O&M knowledge update sources, a knowledge end-to-end evolution set is constructed. This allows for the capture of knowledge generation triggering factors, clear definition of the scope of iterative impact, and accurate characterization of the correlation strength between different knowledge sets. Dynamic coupling rules are generated based on the real-time scenario characteristics of O&M tasks and the knowledge end-to-end evolution set, ensuring that knowledge invocation closely aligns with actual O&M scenario needs. Based on the real-time characteristics of different O&M tasks, such as task objective types, device operating status parameters, and business impact scope, scenario knowledge matching conditions, knowledge invocation priority sequences, and related knowledge linkage logic are dynamically adjusted, effectively improving O&M efficiency and quality. The scenario-coupled knowledge set is embedded into the O&M technical service big model and divided into hierarchical knowledge architectures. Storage node parameters and retrieval logic are adjusted synchronously to achieve dynamic optimization of the knowledge architecture. Furthermore, through multi-dimensional dynamic adaptation, combined with changes in real-time scenario characteristics of O&M tasks, the knowledge invocation priority sequence and storage node parameters are further adjusted to generate an adapted and updated O&M technical service big model. This enables the big model to quickly respond to changes in O&M scenarios, continuously providing accurate and efficient knowledge services for O&M tasks, significantly enhancing the adaptability of the O&M technical service big model. Attached Figure Description

[0016] Figure 1 A flowchart illustrating the method for dynamic updating of large model knowledge for operation and maintenance technical services provided in an embodiment of this application is shown.

[0017] Figure 2 This illustration shows a component structure diagram of a system for dynamically updating large model knowledge for operation and maintenance technical services, provided in an embodiment of this application, for implementing the above-described method for dynamically updating large model knowledge for operation and maintenance technical services. Detailed Implementation

[0018] Figure 1 This paper illustrates a flowchart of a method and system for dynamically updating large-scale model knowledge for operation and maintenance technical services, as provided in an embodiment of this application. The detailed steps include:

[0019] Step S110: Model the knowledge generation triggering factors, iteration impact range, and correlation strength between different knowledge in the multi-source operation and maintenance knowledge update source to obtain the knowledge full-link evolution set, which includes knowledge generation triggering records, iteration impact assessment results, and cross-knowledge association networks.

[0020] This embodiment uses a smart operation and maintenance service scenario for chain stores as an example for illustration. In this scenario, the large-scale operation and maintenance technical service model needs to handle the operation and maintenance tasks of various devices in the store, such as air conditioning equipment, information publishing systems, and IoT control systems. Therefore, it is necessary to integrate knowledge from multiple aspects to support accurate and efficient operation and maintenance services.

[0021] Step S111: Access the multi-source operation and maintenance knowledge update source, which includes the operation and maintenance service historical task case library, the equipment manufacturer technical upgrade document library, the industry operation and maintenance standard revision library, the operation and maintenance event real-time processing library, and the operation and maintenance knowledge user feedback library.

[0022] In the smart operation and maintenance (O&M) scenario of chain stores, accessing multi-source O&M knowledge update sources is the foundation for dynamic knowledge updates. The O&M service historical task case library stores detailed records of various O&M tasks performed on different chain stores, such as air conditioning malfunction repairs, information display system black screen handling, and IoT control system response delays. The equipment manufacturer technical upgrade document library regularly receives technical upgrade documents from air conditioning manufacturers, information display equipment manufacturers, and IoT control module manufacturers, such as updated maintenance manuals for new air conditioning models and software version upgrade instructions for information display systems. The industry O&M standard revision library includes content such as revisions to energy efficiency standards for commercial air conditioning operation and updates to information system security O&M standards issued by national or industry organizations. The real-time O&M event processing library receives real-time reports of sudden O&M events from various stores, such as air conditioning cooling failures, abnormal information display screens, and IoT control command execution failures, along with their handling processes and solutions. The O&M knowledge user feedback library collects feedback from on-site engineers and store managers on applied O&M knowledge; for example, engineers may find the steps for handling certain air conditioning malfunctions cumbersome, or store managers may feel that the information display system maintenance knowledge is not detailed enough.

[0023] Step S112: Extract the knowledge generation process record of each historical operation and maintenance task from the historical task case library of the operation and maintenance service, including the initial knowledge when the task starts, the knowledge supplement during the execution of the task, and the knowledge summary after the task ends, arrange them in the order of task execution time, and generate a knowledge generation timeline for a single task.

[0024] For the historical task case library of operation and maintenance services, we take an air conditioner malfunction repair task from a chain store as an example. We extract the knowledge generation process record of this historical operation and maintenance task from the case library. The initial knowledge at the start of the task may include common fault types of this brand of air conditioner and corresponding preliminary troubleshooting methods, such as insufficient refrigerant or compressor failure if the air conditioner is not cooling. During task execution, the engineer discovers on-site that the condenser is blocked; this new knowledge from the repair process is added as supplementary knowledge. After the task is completed, the engineer's knowledge summary of the cause of the fault, handling steps, and preventive measures is also extracted. Arranging the initial knowledge, supplementary knowledge, and knowledge summary in chronological order of task execution forms the knowledge generation timeline for this single air conditioner malfunction repair task. This timeline clearly shows the entire process of knowledge generation from initialization to completion.

[0025] Step S113: Collect iterative text of equipment maintenance knowledge from the equipment manufacturer's technical upgrade document library, compare the differences in knowledge content before and after the iteration, determine the range of equipment models affected by the iterative knowledge, maintenance process links and related knowledge modules, and form a list of the scope of impact of the iteration.

[0026] The iterative text processing in the equipment manufacturer's technical upgrade document library, taking an air conditioner manufacturer's technical upgrade document as an example, involves collecting iterative text related to air conditioner maintenance knowledge from the document and comparing the differences between the before and after iterations. For example, in the previous knowledge, the filter replacement cycle for a certain model of air conditioner was one month; after the iteration, it was updated to two months, and a new method for detecting the filter's condition was added to the replacement steps. Through comparison, it was determined that the equipment models affected by this iterative knowledge are specific series of air conditioners manufactured by this manufacturer within the last two years. In terms of maintenance processes, it affects the filter replacement step and related equipment performance testing. Related knowledge modules may involve the air conditioner operating efficiency assessment module, as the change in the filter replacement cycle affects the air conditioner's operating efficiency. The above information is then compiled into a list of the scope of impact of the iteration, clearly defining the various aspects affected by the knowledge iteration.

[0027] Step S114: Obtain the revised text of standard knowledge from the industry operation and maintenance standard revision library, extract the reference relationship between the revised clause and the original clause, the association relationship between the revised clause and other standard clauses, calculate the content similarity between the revised clause and the related clause, set a similarity threshold, and filter clauses with similarity values ​​exceeding the similarity threshold as strongly related clauses to form a standard knowledge association list.

[0028] The processing of the industry operation and maintenance standard revision library takes the revision of the energy efficiency standard for air conditioning operation in commercial premises as an example. After obtaining the revised text of the standard knowledge, the citation relationship between the revised clauses and the original clauses is extracted. For example, the revised clauses may cite clauses in the original standard regarding the classification of air conditioning energy efficiency levels. Simultaneously, the relationships between the revised clauses and other standard clauses are extracted. For example, the revised clauses may be related to clauses in fire safety standards regarding the wiring of air conditioning equipment. The content similarity between the revised clauses and these related clauses is calculated. The content similarity calculation is performed by segmenting the clause text, extracting keywords, and then based on the overlap and semantic relevance of keywords. A similarity threshold is set, and clauses with similarity values ​​exceeding this threshold are filtered out as strongly related clauses, thus forming a standard knowledge association list. This list helps to consider other related standard clauses when applying the revised standard knowledge.

[0029] Step S115: Extract the processing scheme of real-time events from the real-time processing library of operation and maintenance events, associate it with existing similar event processing knowledge, calculate the content similarity between the new knowledge and the existing knowledge, and use the content similarity as the association strength between different knowledge.

[0030] In the real-time event handling library, when a chain store reports a black screen event in the information publishing system, the handling solution for that event is extracted, such as checking the power connection, restarting the system, or reinstalling the driver. Existing similar event handling knowledge is then linked, such as methods used by other stores to handle similar black screen events in the past. The content similarity between the newly added black screen event handling solution and existing similar knowledge is calculated, using methods based on keyword matching and semantic understanding of the text content. The calculated content similarity is directly used as the association strength between these two pieces of different knowledge; the higher the association strength value, the stronger the correlation between the two pieces of knowledge.

[0031] Step S116: Collect user feedback information on the effectiveness of knowledge application from the operation and maintenance knowledge user feedback library, including knowledge practicality score, usability score and supplementary suggestions. Associate the feedback information with the corresponding knowledge module. Adjust the association strength between different knowledge based on the practicality score in the feedback information. Feedback with a practicality score higher than the preset score corresponds to an increase in the association strength value, and feedback with a practicality score lower than the preset score corresponds to a decrease in the association strength value.

[0032] The processing of the operation and maintenance knowledge user feedback database involves collecting feedback from on-site engineers regarding knowledge on handling response delays in IoT control systems. Engineers gave the knowledge a high practicality rating and offered supplementary suggestions regarding usability. This feedback is then correlated with the knowledge module on handling response delays in IoT control systems. A preset practicality rating threshold is used. If an engineer's practicality rating is higher than this threshold, it indicates that the knowledge is effective in practical applications, and the correlation strength between this knowledge and other related knowledge needs to be increased, such as the correlation strength with IoT control module hardware detection knowledge. If other users give a certain knowledge a practicality rating lower than the preset rating, the correlation strength between this knowledge and related knowledge is correspondingly decreased.

[0033] Step S117: Integrate the knowledge generation timeline of the single task, the iterative impact range list, the standard knowledge association list, and the association strength data between different knowledge, construct knowledge generation trigger records using a time-series modeling method, integrate the association relationship data using a graph construction method to form a cross-knowledge association network, classify and organize them according to knowledge type, and form a knowledge full-link evolution set.

[0034] This process integrates the knowledge generation timeline, iteration impact scope list, standard knowledge association list, and association strength data between different knowledge sets obtained from the above steps for each individual task. Using a temporal modeling approach, combining the chronological order of the knowledge generation timeline with knowledge generation triggering factors, a knowledge generation trigger record is constructed. This record clearly reflects when and why knowledge was generated. A graph construction method is used to integrate the association data, such as the clause relationships in the standard knowledge association list and the association strength data between different knowledge sets, to form a cross-knowledge association network. This network uses nodes to represent knowledge and edges to represent the association relationships and strengths between knowledge sets. Knowledge is then categorized and organized by type, such as air conditioning operation and maintenance knowledge, information publishing system operation and maintenance knowledge, and IoT control system operation and maintenance knowledge, ultimately forming a complete knowledge evolution set. This set contains comprehensive information on knowledge generation, iteration, and association.

[0035] Step S120: Based on the real-time scenario features and knowledge full-link evolution set of the operation and maintenance task, generate dynamic coupling rules between scenarios and knowledge. The dynamic coupling rules include scenario knowledge matching conditions, knowledge call priority sequence and related knowledge linkage logic. The real-time scenario features of the operation and maintenance task include task target type, equipment operating status parameters and business impact range.

[0036] In the smart operation and maintenance scenario of chain stores, when a new operation and maintenance task is received, dynamic coupling rules need to be generated based on its real-time scenario characteristics and the evolution of the entire knowledge chain to ensure that the large model can accurately call the appropriate knowledge. For example, if a store reports that its information publishing system cannot display advertising content normally, the real-time scenario characteristics at this time include the task objective type being information publishing system fault repair, equipment operating status parameters such as system memory usage and network connection speed, and the business impact scope involving the store's advertising effectiveness, which may affect customer traffic and sales.

[0037] Step S121: Collect real-time scene characteristics of operation and maintenance tasks, and extract task target type information from the operation and maintenance task management platform.

[0038] Extract the task target type information of the information release system fault repair task from the operation and maintenance task management platform to clarify that this task is to repair the fault of the information release system, rather than other types of tasks such as routine maintenance or system upgrade.

[0039] Step S122: Obtain equipment operating status parameters from the equipment status monitoring system, including equipment CPU utilization, memory usage, disk read / write speed, network transmission rate, and equipment temperature data.

[0040] The equipment status monitoring system obtains the operating status parameters of the store's information publishing system equipment. The CPU utilization rate reflects the equipment's computing load; excessive utilization may lead to slow system response. The memory usage rate shows the memory space occupied by the currently running programs; excessive memory usage may cause system lag. Disk read / write speed affects data loading and storage efficiency. Network transmission rate is related to the download and updating of advertising content. Equipment temperature data reflects whether the equipment is at risk of overheating. These parameters together constitute a comprehensive description of the equipment's operating status.

[0041] Step S123: Extract business impact scope information from the business impact analysis module, including the number of affected business systems, the scale of affected users, and the estimated duration of business interruption.

[0042] The business impact analysis module will analyze the business impact caused by the failure of the information publishing system. The number of affected business systems may be limited to the information publishing system of this store, or it may be related to the advertising content management system of the headquarters. The affected user scale includes the store's customers and potential customers who may see the store's advertisements through online channels. The estimated duration of business interruption is based on past experience in handling similar failures, and this duration is directly related to the degree of impact on the store's business.

[0043] Step S124: Perform feature encoding on the task target type, the device operating status parameters, and the business impact range to generate an encoding vector of real-time scene features. Each dimension of the encoding vector corresponds to a specific value of a scene feature.

[0044] Feature encoding is performed on the task objective type, equipment operating status parameters, and business impact scope. The task objective type is information publishing system fault repair, which can be encoded as a specific category value. Numerical parameters in the equipment operating status parameters, such as CPU utilization and memory usage, are normalized and used as corresponding dimension values ​​in the encoding vector. The number of affected business systems, the scale of affected users, and the estimated duration of business interruption in the business impact scope are also converted into corresponding dimension values ​​in the encoding vector. Finally, a multi-dimensional real-time scene feature encoding vector is generated, which can be effectively identified and processed by large models.

[0045] Step S125: Extract the application scenario features of each piece of knowledge from the knowledge end-to-end evolution set, including the task target type of the knowledge application, the range of adapted device operating status parameters and the scope of business impact to be dealt with, and generate the encoding vector of the knowledge application scenario.

[0046] The application scenario features of each piece of knowledge are extracted from the knowledge end-to-end evolution set. Taking a piece of knowledge about handling a black screen failure in an information publishing system as an example, the task target type it was applied to was information publishing system fault repair. The applicable device operating status parameters might be that the CPU utilization rate is within a certain range, the memory usage rate is below a certain threshold, etc., and the business impact it dealt with was the interruption of the information publishing system in a single store. The above application scenario features are also encoded to generate an encoding vector for the knowledge application scenario. The structure of this vector is consistent with the real-time scenario feature encoding vector to facilitate similarity calculation.

[0047] Step S126: Calculate the vector similarity between the encoding vector of the real-time scene features and the encoding vector of the knowledge application scene, set a similarity threshold, and filter out knowledge with similarity values ​​exceeding the threshold as candidate scene knowledge.

[0048] The similarity between the real-time scene feature encoding vector and the knowledge application scene encoding vector is calculated, and their similarity is measured by methods such as calculating the cosine similarity between the two vectors. A similarity threshold is set, for example, 0.7, and knowledge with similarity values ​​exceeding this threshold is selected as candidate scene knowledge. For the above-mentioned information publishing system fault repair task, knowledge whose application scenarios have high similarity to the current real-time scene features, such as knowledge from the past handling of faults in information publishing systems for similar stores or similar equipment models, will be selected into the candidate scene knowledge set.

[0049] Step S127: Extract the generation triggering factors, iterative impact assessment results, and correlation strength between different knowledge from the knowledge end-to-end evolution set of candidate scenario knowledge, and analyze the applicability of candidate scenario knowledge in the current real-time scenario, including whether the matching degree value between the generation triggering factor and the current scenario triggering event exceeds the preset matching degree threshold, whether the ratio of the iterative impact scope covering the current device and business exceeds the preset coverage ratio threshold, and whether the correlation strength value between different knowledge meets the preset strength requirements.

[0050] The generation triggering factors, iterative impact assessment results, and correlation strengths between different knowledge types are extracted from the knowledge end-to-end evolution set. The applicability of these candidate knowledge types in the current real-time scenario is analyzed, and the matching degree between the generation triggering factors and the current scenario triggering events is determined. For example, if the current fault triggering event is a black screen in the information publishing system, and the generation triggering factor of candidate knowledge is also a similar black screen event, the matching degree is high, and it needs to be determined whether it exceeds a preset matching degree threshold. The proportion of the iterative impact scope covering the current device and business is also considered; that is, the proportion of the current faulty device model and business scope included in the iterative impact scope of the candidate knowledge. If it exceeds a preset coverage ratio threshold, it indicates that the knowledge is highly applicable to the current scenario. The correlation strength between different knowledge types must also meet preset strength requirements to ensure that the correlation between knowledge types is reasonable and effective.

[0051] Step S128: Based on the applicability analysis results, set scenario knowledge matching conditions. The matching conditions include task target type matching requirements, equipment operating status parameter adaptation range, and business impact range response requirements. The task target type matching requirements ensure that the current task target type is consistent with the task target type to which the knowledge has been applied. The equipment operating status parameter adaptation range ensures that the current equipment operating status parameters are within the parameter range of knowledge adaptation. The business impact range response requirements ensure that the current business impact range is within the scope of knowledge response.

[0052] Based on the applicability analysis results, scenario knowledge matching conditions are set. The task target type matching requirement clearly states that the current task target type must be consistent with the task target type to which the knowledge was previously applied. That is, if the current task is a fault repair task for an information publishing system, the candidate knowledge must also be applied to this type of task. The device operating status parameter adaptation range determines that the current device operating status parameters must be within the parameter range of the knowledge adaptation. For example, if the CPU utilization range adapted to a candidate knowledge is a specific interval, the CPU utilization of the current device must be within this interval. The business impact scope response requirement determines that the current business impact scope must be within the scope of the knowledge response. If the current business impact scope is a single store, but the candidate knowledge addresses a large-scale impact on multiple stores, then the knowledge does not meet the matching conditions.

[0053] Step S129: Statistically analyze the application effect data of candidate scenario knowledge in similar historical scenarios, including task completion time, task success rate, equipment recovery time and business recovery efficiency after knowledge application. Sort the candidate scenario knowledge in the order of task completion time from shortest to longest, task success rate from highest to lowest, equipment recovery time from shortest to longest, and business recovery efficiency from fastest to slowest.

[0054] The application effects of candidate scenario knowledge in similar historical scenarios are statistically analyzed. Taking two candidate knowledge sets as examples, knowledge A, when applied in similar historical scenarios, resulted in shorter task completion times, higher task success rates, faster equipment recovery times, and faster business recovery efficiency. Knowledge B, on the other hand, performed relatively worse in all aspects. The candidate scenario knowledge sets are then sorted in the following order: task completion time (shortest to longest), task success rate (highest to lowest), equipment recovery time (shortest to longest), and business recovery efficiency (fastest to slowest). Knowledge A will rank higher than Knowledge B.

[0055] Step S1210: Generate a knowledge call priority sequence based on the sorting results. Based on the cross-knowledge association network of the knowledge ranked first in the priority sequence, extract the association strength value of the associated knowledge, and arrange them in descending order of association strength value to determine the call order of the associated knowledge, thus forming the linked logic of associated knowledge.

[0056] A knowledge retrieval priority sequence is generated based on the ranking results, with higher-ranked knowledge having higher retrieval priority. Based on the cross-knowledge association network of the higher-ranked knowledge in the priority sequence, the association strength values ​​of the associated knowledge are extracted. For example, knowledge A is associated with the information publishing system power supply check knowledge and the system software version knowledge, with association strength values ​​of 0.8 and 0.6 respectively. The knowledge is then sorted from highest to lowest association strength value, determining that the power supply check knowledge is invoked first, followed by the system software version knowledge, forming a linked logic for associated knowledge.

[0057] Step S1211: Integrate the scene knowledge matching conditions, the knowledge call priority sequence, and the associated knowledge linkage logic, and generate dynamic coupling rules between scenes and knowledge according to task target types. Each task target type corresponds to an independent set of coupling rules, and the content of the coupling rules is used to define the specific values ​​and association logic of each element.

[0058] By integrating scenario knowledge matching conditions, knowledge invocation priority sequences, and related knowledge linkage logic, dynamic coupling rules are generated according to task objective type. For the task objective type of information publishing system fault repair, an independent set of coupling rules is formed. This set of rules clarifies the specific values ​​and association logic of each element, such as scenario knowledge matching conditions, the order of knowledge invocation, and the linkage method of related knowledge, in this type of task, ensuring that the large model can accurately apply relevant knowledge when processing this type of task.

[0059] Step S130: Use the dynamic coupling rules to couple and filter the knowledge content in the knowledge full-link evolution set, extract the knowledge content that meets the matching conditions, and integrate them to obtain the scene-coupled knowledge set.

[0060] In the smart operation and maintenance scenario of chain stores, after the dynamic coupling rules are generated, they need to be used to process the knowledge content in the knowledge end-to-end evolution set to obtain a scenario-coupled knowledge set applicable to the current operation and maintenance task. For example, if the current task is the response delay problem of an IoT control system in a store, the scenario knowledge matching conditions in the dynamic coupling rules will filter out knowledge related to this task.

[0061] Step S131: Traverse all knowledge content in the knowledge full-link evolution set, and extract the application scenario characteristics, core knowledge content, and correlation information between different knowledge for each knowledge.

[0062] This process iterates through all knowledge content in the entire knowledge evolution set. For each piece of knowledge, such as the knowledge about troubleshooting response delays in IoT control systems, it extracts its application scenario characteristics. This includes the type of task objectives previously applied to addressing IoT control system response issues, the range of applicable device operating parameters (e.g., network latency within a specific range, processor load within a certain level), and the scope of business impact (e.g., the IoT control function of a single store is affected). The core knowledge content includes operational steps such as troubleshooting network connectivity, checking server load, and optimizing control command transmission protocols; the network testing tools and server monitoring software used; the parameter settings required for each step (e.g., network timeout settings, server thread pool size adjustments); precautions (e.g., avoiding impact on other control functions); and expected results (e.g., shortening system response time to a normal range). The correlation information between different pieces of knowledge includes the relationship and strength of this troubleshooting knowledge with IoT control module hardware testing knowledge, network device maintenance knowledge, etc.

[0063] Step S132: Compare the application scenario features of each piece of knowledge with the scenario knowledge matching conditions in the dynamic coupling rules between scenarios and knowledge one by one. Check whether the task target type of the knowledge is consistent with the matching requirements, whether the range of adapted device operating status parameters includes the current device operating status parameters, and whether the scope of business impact to be addressed covers the current business impact. Retain all knowledge that passes the comparison to form an initial set of filtered knowledge.

[0064] Each piece of knowledge is compared one by one with the application scenario characteristics and scenario knowledge matching conditions in the dynamic coupling rules. For knowledge about troubleshooting response delays in IoT control systems, it is checked whether its task objective type is IoT control system fault repair, which is consistent with the current task objective type. The range of adapted device operating status parameters is checked to see if it includes parameters such as network latency and processor load of the current device. For example, if the current device's network latency is a certain value, and this value is within the network latency range of the knowledge adaptation, then it meets the condition. The scope of the business impact it addresses covers the business impact scope of a single store. If the scope of the knowledge addresses includes a single store, then it passes the comparison. All knowledge that passes the comparison is retained to form an initial set of filtered knowledge, in which the knowledge initially meets the matching conditions of the current scenario.

[0065] Step S133: Extract the association information between different knowledge in each knowledge from the initial screening knowledge set. Based on the association knowledge linkage logic in the dynamic coupling rules between scenarios and knowledge, find other knowledge that is associated with each knowledge. If the associated knowledge is not included in the initial screening knowledge set, check whether the application scenario characteristics of the associated knowledge meet the scenario knowledge matching conditions. If they do, add it to the initial screening knowledge set.

[0066] The system extracts the association information for each piece of knowledge from the initial knowledge set and searches for related knowledge based on the association logic. For example, knowledge about troubleshooting response delays in an IoT control system is associated with knowledge about network device maintenance. If this network device maintenance knowledge is not included in the initial knowledge set, the system checks whether its application scenario characteristics meet the scenario knowledge matching conditions, namely, whether the task target type is relevant, whether the range of applicable device operating status parameters includes the parameters of the current network device, and whether the scope of the business impact it addresses covers the current scope. If these conditions are met, the network device maintenance knowledge is added to the initial knowledge set to enrich the knowledge content.

[0067] Step S134: Prioritize the knowledge in the supplemented initial filtered knowledge set, and arrange the knowledge from high to low priority according to the knowledge call priority sequence in the dynamic coupling rules between scenarios and knowledge to form an ordered knowledge list.

[0068] The knowledge in the supplemented initial knowledge set is prioritized according to the knowledge call priority sequence in the dynamic coupling rules. Based on the application effect data of knowledge in similar historical scenarios, such as task completion time and task success rate, the knowledge is arranged from high to low priority to form an ordered knowledge list. The knowledge at the top of the list has a better application effect in the current scenario and has a higher call priority.

[0069] Step S135: Extract the core content of each knowledge item in the ordered knowledge list, including the corresponding operation steps, tools used, parameter setting requirements, precautions and expected results.

[0070] Extracting the core content of each knowledge item from the ordered knowledge list, taking the highest-priority knowledge item—troubleshooting response delays in IoT control systems—as an example, the operational steps include logging into the system backend to check network connection status, using network testing tools to test network latency, and checking server CPU and memory usage. Tools used include specific network analysis software and server monitoring tools. Parameter settings requirements include setting the sampling frequency of the network testing tool and setting thresholds for server monitoring indicators. Precautions include backing up system configurations during operation and avoiding interruptions to executing control commands. The expected outcome is that the system response time returns to normal levels, and the control command execution delay is below the preset threshold.

[0071] Step S136: Associate and integrate the core content, take the core content of the knowledge ranked first in the priority sequence as the main knowledge content, take the core content of the related knowledge as the supplementary knowledge content, and embed the supplementary knowledge content into the relevant operation steps of the main knowledge content according to the calling order determined by the linkage logic of the related knowledge.

[0072] The core content is integrated and linked, with the core content of IoT control system response delay troubleshooting, which ranks high in the priority sequence, serving as the main knowledge content. The core content of the related network device maintenance knowledge serves as supplementary knowledge content. Following the calling order determined by the linkage logic of the related knowledge, the supplementary operation steps of network device maintenance knowledge, such as checking the status of network switch ports and cleaning port dust, are embedded after the network connection check steps in the main knowledge content, making the knowledge content more complete and coherent.

[0073] Step S137: The integrated knowledge content is structured and organized in the following order: knowledge identifier, task target adaptation type, equipment status adaptation range, business impact response range, core operation process, supplementary operation details, and expected effect. Each structured knowledge unit corresponds to a complete integrated knowledge.

[0074] The integrated knowledge content is structured, with each piece of integrated knowledge assigned a unique knowledge identifier, clearly defining its task objective as IoT control system fault repair. The device status adaptation range lists the applicable network latency range, server load range, etc. The business impact response scope is clearly defined as the IoT control function of a single store being affected. The core operation process lists the operation steps of the main knowledge content in sequence. Supplementary operation details include supplementary steps embedded with related knowledge. Expected results clearly define the expected values ​​for indicators such as system response time; each structured knowledge unit contains all of the above complete information.

[0075] Step S138: Combine all structured knowledge units in priority order to form a scenario-coupled knowledge set, which contains complete association information and structured core content of each knowledge.

[0076] All structured knowledge units are combined according to priority to form a scenario-coupled knowledge set. The knowledge in this set is arranged from high to low priority of invocation. Each piece of knowledge contains complete association information, such as the relationship and strength of association with other knowledge, as well as structured core content, which prepares for subsequent embedding into a large model knowledge architecture.

[0077] Step S140: Embed the scenario-coupled knowledge set into the hierarchical knowledge architecture of the large-scale operation and maintenance technical service model, and simultaneously adjust the storage node parameters and retrieval logic of the hierarchical knowledge architecture based on the knowledge full-link evolution set to obtain the large-scale operation and maintenance technical service model after basic update.

[0078] In the smart operation and maintenance scenario of chain stores, the knowledge set coupled with the scenario is embedded into the knowledge architecture of the operation and maintenance technical service big model, and the relevant parameters and logic are adjusted to achieve effective storage and retrieval of knowledge.

[0079] Step S141: Analyze the hierarchical knowledge architecture of the large model of operation and maintenance technical services, and determine the layering logic of the knowledge architecture. The knowledge architecture includes a basic operation and maintenance knowledge layer, a scenario-specific knowledge layer, an emergency response knowledge layer, and a related knowledge layer.

[0080] This paper analyzes the hierarchical knowledge architecture of the large-scale operation and maintenance (O&M) technical service model, clarifying its layered logic. The basic O&M knowledge layer stores fundamental knowledge applicable to various routine O&M scenarios, such as general maintenance methods for various brands of air conditioners, daily maintenance knowledge for common information publishing systems, and basic operating procedures for IoT control systems. This knowledge is universal and applicable to the daily O&M tasks of most chain stores. The scenario-specific knowledge layer stores O&M knowledge for specific scenarios, such as special fault handling knowledge for different brands and models of air conditioners, installation and debugging knowledge for information publishing systems in specific store layouts, and O&M knowledge for customized IoT control functions for different chain brands. The emergency response knowledge layer stores knowledge for dealing with emergency O&M events, such as rapid handling plans for sudden large-scale air conditioner shutdowns in stores, emergency response knowledge for information publishing systems suffering cyberattacks, and emergency recovery knowledge for store operations interrupted due to IoT control system paralysis. The related knowledge layer records the relationships between various knowledge items, such as the relationship between air conditioner fault handling knowledge and air conditioner component replacement knowledge, and the relationship between information publishing system fault knowledge and cybersecurity knowledge.

[0081] Step S142: Determine the hierarchical affiliation of each structured knowledge unit in the scenario-coupled knowledge set. Based on the scope and urgency of the adapted scenarios, allocate knowledge that covers all regular scenarios to the basic operation and maintenance knowledge layer; allocate knowledge that is limited to specific scenarios to the scenario-specific knowledge layer; allocate knowledge that is adapted to tasks with an urgency level of immediate response to the emergency response knowledge layer; and allocate data that records the relationships between knowledge to the related knowledge layer.

[0082] For each structured knowledge unit in the scenario-coupled knowledge set, its hierarchical affiliation is determined based on the scope of its applicable scenarios and its urgency. A piece of knowledge applicable to the daily cleaning and maintenance of air conditioners in all chain stores, due to its coverage of all common scenarios, is assigned to the basic maintenance knowledge layer. A piece of advanced troubleshooting knowledge applicable only to a specific model of IoT control system, because its applicable scenarios are limited to that specific model of equipment, is assigned to the scenario-specific knowledge layer. An emergency handling piece of knowledge regarding a hacker attack on a store information publishing system, with an urgency level of immediate response, is assigned to the emergency response knowledge layer. Data recording the relationships between the above knowledge items, such as the relationship between air conditioner maintenance knowledge and component knowledge, is assigned to the association knowledge layer.

[0083] Step S143: Set initial parameters for each storage node in the hierarchical knowledge layer, including the upper limit of knowledge capacity, the knowledge update frequency threshold, and the knowledge retrieval weight of the storage node. The initial parameters are set based on the historical operation data of the hierarchical knowledge architecture.

[0084] Initial parameters are set for storage nodes in each hierarchical knowledge layer. The upper limit of the storage node knowledge capacity for the basic operations and maintenance knowledge layer is set based on the average quantity and growth trend of knowledge in this layer during historical operation, ensuring sufficient general knowledge can be accommodated. The knowledge update frequency threshold is set to a relatively low value because basic operations and maintenance knowledge is relatively stable and updates are infrequent. The knowledge retrieval weight is set to a medium level to ensure reasonable retrieval in general searches. The upper limit of the storage node knowledge capacity for the scenario-specific knowledge layer is set based on the amount of knowledge for a specific scenario, and the knowledge update frequency threshold is slightly higher than that of the basic layer because specific scenario knowledge may be updated with device model updates. The knowledge retrieval weight is set based on the commonness of the scenario, with higher weights for more common scenarios. The upper limit of the storage node knowledge capacity for the emergency response knowledge layer is relatively small, and the knowledge update frequency threshold is higher to accommodate the rapid updates of emergency event handling knowledge. The knowledge retrieval weight is set to the highest level to ensure priority retrieval of relevant knowledge in emergency scenarios. The storage node parameters for the associated knowledge layer are set based on the amount of associated data and the update frequency.

[0085] Step S144: Extract the generation triggering factors and iteration impact assessment results of each knowledge in the scenario-coupled knowledge set from the knowledge full-link evolution set, count the occurrence frequency of knowledge generation triggering factors and the number of devices covered by the iteration impact range, and determine the knowledge update frequency requirements and retrieval priority requirements based on the statistical results.

[0086] The generation triggering factors and iterative impact assessment results of each piece of knowledge in the scenario-coupled knowledge set are extracted from the knowledge end-to-end evolution set. Taking an emergency response knowledge as an example, its generation triggering factor may be multiple hacker attacks on the store information publishing system that occurred in the past year, and the frequency of occurrence of this triggering factor is high. The iterative impact assessment results show that the number of devices affected by this knowledge covers the information publishing devices of most chain stores. Based on these statistical results, it is determined that the emergency response knowledge needs to be updated frequently, and new attack methods and protection methods need to be followed up in a timely manner; the retrieval priority also needs to be high to deal with possible recurring emergencies.

[0087] Step S145: Based on the determined needs, adjust the parameters of the corresponding storage nodes. For storage nodes containing knowledge where the frequency of generating trigger factors is higher than a preset frequency threshold and the number of devices covered by the iteration influence range exceeds a preset number threshold, increase the knowledge update frequency threshold of the storage node. For storage nodes containing knowledge where the correlation value between the generated trigger factors and the current scene is higher than a preset correlation threshold and the iteration influence range evaluation result includes the currently running device, increase the knowledge retrieval weight of the storage node.

[0088] Based on the determined needs, adjust the parameters of the corresponding storage nodes. For storage nodes containing knowledge where the frequency of generated trigger factors exceeds a preset frequency threshold and the number of devices covered by the iteration impact range exceeds a preset device number threshold, such as the emergency response knowledge layer storage node containing the aforementioned emergency response knowledge, increase its knowledge update frequency threshold so that the node can receive and update knowledge more frequently. For storage nodes containing knowledge where the correlation between generated trigger factors and the current scenario exceeds a preset correlation threshold and the iteration impact range assessment results include currently operating devices, for example, where the current scenario is a specific model of air conditioner malfunction in a store, and a storage node containing knowledge where the generated trigger factor is highly correlated with the malfunction of that model of air conditioner and the iteration impact range includes the air conditioners in that store, increase its knowledge retrieval weight to increase the probability that the knowledge will be preferentially called during retrieval.

[0089] Step S146: Determine the initial retrieval logic of the hierarchical knowledge architecture. The initial retrieval logic is set to retrieve knowledge in the order of emergency response knowledge layer, scenario-specific knowledge layer, and basic operation and maintenance knowledge layer, with related knowledge layers being called synchronously with the main knowledge retrieval.

[0090] The initial retrieval logic of the knowledge architecture is determined, with the retrieval order set as: emergency response knowledge layer, scenario-specific knowledge layer, and basic operations and maintenance knowledge layer. When an operations and maintenance task is received, the emergency response knowledge layer is searched first to check for matching emergency handling knowledge; if not, the scenario-specific knowledge layer is searched to find operations and maintenance knowledge for a specific scenario; finally, the basic operations and maintenance knowledge layer is searched. The associated knowledge layer is invoked synchronously with the main knowledge retrieval; when the main knowledge is retrieved, its associated knowledge is automatically invoked.

[0091] Step S147: Extract the cross-knowledge association network of each knowledge in the scenario-coupled knowledge set from the knowledge full-link evolution set, analyze the association strength value and calling order between knowledge, and adjust the association knowledge calling weight in the retrieval logic based on the association strength value.

[0092] The cross-knowledge association network for each piece of knowledge is extracted from the knowledge end-to-end evolution set, and the association strength and invocation order between knowledge are analyzed. For example, a scenario-specific knowledge is associated with three related knowledge, with association strength values ​​of 0.9, 0.7, and 0.5, respectively. Based on these association strength values, the invocation weight of related knowledge in the retrieval logic is adjusted. The knowledge with an association strength of 0.9 has the highest invocation weight and is prioritized when the main knowledge is invoked; the knowledge with an association strength of 0.7 has the second highest weight, and the knowledge with an association strength of 0.5 has the lowest weight, so that the invocation order of related knowledge matches the association strength.

[0093] Step S148: Write the structured knowledge units in the scenario-coupled knowledge set into the corresponding storage nodes according to their hierarchical affiliation. After writing, test the retrieval function of the hierarchical knowledge architecture, simulate knowledge retrieval requests under different operation and maintenance task scenarios, and record whether the retrieved knowledge belongs to the scenario-coupled knowledge set, the retrieval response time, and whether the related knowledge calls are accurate.

[0094] Structured knowledge units from the scenario-coupled knowledge set are written to corresponding storage nodes according to their hierarchical affiliation. For example, basic operation and maintenance knowledge is written to the basic operation and maintenance knowledge layer storage node, and scenario-specific knowledge is written to the scenario-specific knowledge layer storage node. After writing, the knowledge architecture's retrieval function is tested by simulating different operation and maintenance task scenarios, such as routine air conditioning maintenance, specific model fault scenarios in information publishing systems, and emergency paralysis scenarios in IoT control systems. Whether the retrieved knowledge belongs to the scenario-coupled knowledge set is recorded; if irrelevant knowledge is retrieved, the retrieval is inaccurate. The retrieval response time is recorded to determine if it is within an acceptable range. The accuracy of related knowledge invocation is checked; that is, when the main knowledge is retrieved, whether the related knowledge is invoked as expected.

[0095] Step S149: If the retrieved knowledge does not belong to the scene-coupled knowledge set, the response time exceeds the preset response time threshold, or the associated knowledge call is inaccurate, then readjust the storage node parameters and retrieval logic, repeat the test process, until the retrieval function meets the usage requirements.

[0096] If the retrieved knowledge is found to be outside the context-coupled knowledge set during testing, the knowledge retrieval weights of the storage nodes may be improperly set, requiring readjustment of the relevant storage nodes' retrieval weights. If the response time exceeds the preset response time threshold, the knowledge capacity of the storage nodes may be too large or the retrieval logic too complex; the knowledge organization method of the storage nodes can be optimized or the retrieval logic simplified. If the related knowledge calls are inaccurate, the strength of the association between knowledge points needs to be re-analyzed, and the weights of related knowledge calls adjusted. After readjustment, test again, repeating this process until the retrieval function meets the usage requirements, i.e., accurate retrieval, timely response, and correct related knowledge calls.

[0097] Step S1410: After completing the adjustment of storage node parameters and retrieval logic, the updated operation and maintenance technical service big model with the knowledge architecture divided by hierarchy is used as the basis for the updated operation and maintenance technical service big model.

[0098] After the storage node parameters and retrieval logic were adjusted and the retrieval function was tested and passed, the large-scale operation and maintenance technical service model equipped with the updated knowledge architecture became the basic updated operation and maintenance technical service model. This model has updated knowledge content and optimized knowledge storage and retrieval mechanisms.

[0099] Step S150: Perform multi-dimensional dynamic adaptation on the basic updated operation and maintenance technical service big model, adjust the knowledge call priority sequence and storage node parameters in combination with the real-time scenario feature changes of operation and maintenance tasks, and generate the adapted updated operation and maintenance technical service big model.

[0100] In the process of smart operation and maintenance of chain stores, the real-time scenario characteristics of operation and maintenance tasks may change. After the basic update, the large model of operation and maintenance technical services needs to be dynamically adapted in multiple dimensions to adapt to the new scenario requirements. For example, during the handling of the response delay problem of the IoT control system of a certain store, the system suddenly became completely paralyzed, and the real-time scenario characteristics changed. At this time, it is necessary to adjust the knowledge call priority sequence and storage node parameters.

[0101] Step S151: Collect real-time change data of operation and maintenance task scenario characteristics, and obtain change information of task target type from the operation and maintenance task management platform.

[0102] The system collects real-time data on changes in the characteristics of operation and maintenance tasks, and retrieves information on changes in task objective types from the operation and maintenance task management platform. The original task objective type was "Repairing the Response Delay Fault of the IoT Control System," but due to a complete system failure, the task objective type has been changed to "Emergency Recovery from the Paralysis of the IoT Control System."

[0103] Step S152: Capture the changing trends of equipment operating status parameters from the equipment status monitoring system, including the magnitude of increase or decrease in parameter values, whether the parameters exceed the warning threshold, and the duration of parameter changes.

[0104] By capturing the changing trends of equipment operating status parameters in the equipment status monitoring system, before the IoT control system fails, there may be a sharp increase in server CPU utilization, exceeding the normal fluctuation range and surpassing the warning threshold. This parameter change is short-lived, rising from normal levels to the pre-failure peak within a few minutes. Memory usage also shows a rapid upward trend, similarly exceeding the warning threshold, with the change duration synchronized with CPU utilization. Network transmission rate, on the other hand, rapidly drops from normal values ​​to near zero, with a significant decrease exceeding the warning threshold, and the change is short-lived.

[0105] Step S153: Receive adjustment information on the scope of business impact, including an increase in the number of affected business systems, an expansion in the scale of affected users, or an extension of the expected duration of business interruption.

[0106] The system received information regarding adjustments to the scope of business disruption. Originally, only the store's IoT control system was affected; now, the disruption may extend to the store's sales system, as IoT-controlled door locks, POS devices, and other devices are linked to the sales system, increasing the number of affected systems. The affected user base has expanded from store customers to online order users, as the IoT system outage prevents online orders from being synchronized to the store. The expected duration of business disruption has also increased from a few hours to a longer period.

[0107] Step S154: Encode the real-time scene features of the changed operation and maintenance tasks, generate the encoded vector of the updated scene features, compare the encoded vectors of the scene features before and after the update, and find the dimensions that differ.

[0108] The real-time scenario characteristics of the changed operation and maintenance task are encoded. The task objective type changes to "emergency recovery from IoT control system paralysis," and the corresponding encoded values ​​change. The encoded values ​​of device operating status parameters such as CPU utilization, memory usage, and network transmission rate also change due to the parameter changes. The encoded values ​​of the number of affected business systems, the scale of affected users, and the estimated duration of business interruption in the business impact scope also change, generating an updated encoded vector of scenario characteristics. By comparing the encoded vectors before and after the update, the dimensions with differences, such as the task objective type dimension, the dimensions of each device operating status parameter, and the dimensions of the business impact scope, are identified.

[0109] Step S155: Extract the knowledge call priority adjustment strategy corresponding to the difference dimension from the dynamic coupling rule, update the knowledge call priority sequence in the basic updated operation and maintenance technical service big model according to the adjustment strategy, and modify the priority parameters of the corresponding knowledge so that the adjusted priority sequence is consistent with the changed scenario features.

[0110] The knowledge call priority adjustment strategy corresponding to the difference dimension is extracted from the dynamic coupling rules. For the dimension where the task target type changes to emergency recovery, the adjustment strategy may be to significantly increase the priority of the IoT system paralysis emergency recovery knowledge in the emergency response knowledge layer. According to this adjustment strategy, the knowledge call priority sequence in the basic update operation and maintenance technical service big model is updated, and the originally low-priority emergency recovery knowledge is promoted to the first place. The priority parameters of the corresponding knowledge are modified so that the adjusted priority sequence matches the changed scenario characteristics.

[0111] Step S156: Extract the iterative impact evaluation results of knowledge whose application frequency exceeds the preset application frequency threshold under the changed scenario features from the knowledge full-link evolution set, as well as the correlation strength between different knowledge, and analyze the requirements of the knowledge on the storage node parameters.

[0112] This study extracts knowledge that has been applied more frequently than a preset application frequency threshold under changed scenario characteristics from the knowledge end-to-end evolution set. For example, knowledge about emergency recovery from IoT system failures is frequently used in similar past scenarios. The iterative impact assessment results are extracted, showing a wide range of affected devices and business scopes, with frequent iterations. The correlation strength between different knowledge sets is also analyzed; for instance, this emergency recovery knowledge has a high correlation with data backup and recovery knowledge and hardware replacement knowledge. The study also examines the storage node parameter requirements of this knowledge, emphasizing the need for a high knowledge update frequency threshold and a high retrieval weight.

[0113] Step S157: Based on the analysis results, adjust the storage node parameters of the hierarchical knowledge architecture, and at the same time adjust the knowledge eviction policy parameters of the storage nodes.

[0114] Based on the analysis results, the storage node parameters were adjusted. The knowledge update frequency threshold for the storage nodes in the emergency response knowledge layer that store knowledge related to the emergency recovery of IoT system failures was increased to accommodate their frequent updates. Simultaneously, the knowledge retrieval weight of these storage nodes was increased to ensure priority retrieval of this knowledge in emergency scenarios. The knowledge eviction policy parameters for the storage nodes were adjusted; a lower knowledge eviction probability was set for these emergency response knowledge layer storage nodes to retain more emergency recovery-related knowledge, even if some knowledge has been stored for a long time and still has potential value.

[0115] Step S158: Run the model adaptation test to simulate the operation and maintenance task processing flow under the changed scenario, and record the priority order of model knowledge calls, knowledge retrieval response time, accuracy of associated knowledge calls, and task processing efficiency after knowledge application.

[0116] Run model adaptation tests to simulate the operation and maintenance task processing flow under an emergency recovery scenario of an IoT control system failure. Record the priority order of knowledge called by the model, and check whether the emergency recovery knowledge for the IoT system failure is listed first, followed by data backup and recovery knowledge and hardware replacement knowledge in order of their relevance. Check whether the knowledge retrieval response time is within the preset emergency scenario response time threshold. Check the accuracy of associated knowledge calls, and verify whether the model accurately calls the backup and recovery and hardware replacement knowledge associated with the emergency recovery knowledge. Evaluate the task processing efficiency after knowledge application, such as the time from model knowledge call to system recovery start and the total time for the system to return to normal operation.

[0117] Step S159: If the knowledge call priority is inconsistent with the adjustment strategy, the response time exceeds the preset response time threshold, the associated call is inaccurate, or the task processing efficiency does not meet the preset efficiency standard, then the knowledge call priority sequence and storage node parameters are readjusted, and the test process is repeated.

[0118] If the knowledge retrieval priority in the test results is inconsistent with the adjustment strategy, and emergency recovery knowledge for IoT system paralysis is not prioritized, the knowledge retrieval priority sequence needs to be readjusted, and the priority parameter settings need to be checked for correctness. If the response time exceeds the preset response time threshold, the retrieval logic and hardware performance of the storage nodes need to be checked, and storage node parameters need to be readjusted, such as increasing the number of retrieval threads. If the associated retrieval is inaccurate, the correlation strength between knowledge points needs to be re-analyzed, and the weight of associated knowledge retrieval needs to be adjusted. If the task processing efficiency does not meet the preset efficiency standard, the knowledge content or retrieval process needs to be optimized, and the relevant parameters need to be readjusted before repeating the test process.

[0119] Step S1510: After completing all adjustments and tests, save the updated knowledge call priority sequence and storage node parameters, and output the large model of the updated operation and maintenance technical service.

[0120] After completing all adjustments and tests, ensuring that the knowledge call priority conforms to the adjustment strategy, the response time is within the threshold, the associated calls are accurate, and the task processing efficiency meets the standards, the updated knowledge call priority sequence and storage node parameters are saved, and a large model of the updated operation and maintenance technical service is output. This model can adapt to the real-time scenario characteristics of the changed operation and maintenance tasks.

[0121] Furthermore, the method may also include:

[0122] Step S210: Run the adapted and updated operation and maintenance technical service big model to process actual operation and maintenance tasks, collect knowledge application data and task execution result data during the task processing, and generate a knowledge application effect analysis report.

[0123] Step S211: Select different types of actual operation and maintenance tasks, covering equipment fault repair tasks, regular maintenance tasks, system optimization tasks and business recovery tasks. The task types cover the main scenarios in the dynamic coupling rules between scenarios and knowledge.

[0124] In the smart operation and maintenance scenario of chain stores, different types of actual operation and maintenance tasks are selected. Equipment fault repair tasks include repairing a damaged air conditioner compressor in a store, repairing a motherboard fault in the information publishing system, and repairing a short circuit in the IoT control module. Regular maintenance tasks include quarterly maintenance of air conditioners in each store, monthly software updates for the information publishing system, and semi-annual performance testing of the IoT control system. System optimization tasks include improving the running speed of the store's information publishing system, optimizing the transmission efficiency of control commands in the IoT control system, and optimizing air conditioning energy consumption. Business recovery tasks include restoring business after sales are interrupted due to information system failures and restoring business after the membership system becomes unusable due to IoT control system failures. These task types cover the main scenarios in the dynamic coupling rules.

[0125] Step S212: Assign the selected actual operation and maintenance tasks to the adapted and updated operation and maintenance technical service big model, trigger the model's knowledge call and task processing flow, and record the complete process of the model processing tasks.

[0126] The selected actual operation and maintenance tasks are assigned to the updated operation and maintenance technical service model. Taking the task of repairing a damaged air conditioner compressor in a store as an example, after task assignment, the model's knowledge invocation and task processing flow are triggered. The model first receives task information, analyzes real-time scenario characteristics, retrieves relevant knowledge according to dynamic coupling rules, invokes the main knowledge of air conditioner compressor damage repair and related knowledge of component replacement, refrigerant addition, etc., to guide engineers in performing fault diagnosis, component replacement, system debugging, and other operations. The model records the complete processing process from receiving the task to task completion, including the knowledge invocation order, parameter adjustment suggestions, and operation step prompts.

[0127] Step S213: Collect knowledge application data, including the knowledge identifier called by the model, the order of knowledge calls, the call duration of each knowledge, the parameter adjustment records during the knowledge application process, the number of times related knowledge is called, and the feedback information after the knowledge application; at the same time, collect task execution result data, and extract the task start time, task completion time, task completion duration, operation steps records during task execution, the time point when the equipment resumes normal operation, and the time point when the business resumes normal operation from the task execution record system.

[0128] The system collects knowledge application data, including knowledge identifiers such as the knowledge ID for air conditioner compressor damage repair and the knowledge ID for parts replacement. The knowledge call order is as follows: compressor fault diagnosis knowledge, compressor replacement knowledge, and refrigerant addition knowledge. The call duration for each knowledge item is recorded, with fault diagnosis knowledge taking shorter times and replacement knowledge taking longer. Parameter adjustment records are also included during knowledge application, such as refrigerant addition parameter adjustments. The number of times related knowledge is called is also recorded, for example, the parts replacement knowledge was associated with three calls to tool usage knowledge. Feedback information after knowledge application is also included, such as engineers' ratings of the clarity of knowledge steps and feedback on parts model matching. Simultaneously, task execution result data is collected. The task execution record system extracts data showing a task start time of 9:00 AM, a completion time of 11:30 AM, and a completion time of two and a half hours. The operation steps during task execution are recorded, including power outage, disassembly of the old compressor, installation of the new compressor, connection of pipelines, refrigerant addition, and power-on testing. The equipment returned to normal operation at 11:15 AM, coinciding with the business recovery time, indicating that the air conditioner malfunction at this store did not affect the operation of the business system.

[0129] Step S214: Obtain task execution quality data, including equipment fault repair success rate, stable equipment operation time after maintenance, performance improvement after system optimization, and user satisfaction score after business recovery.

[0130] Acquire task execution quality data. Equipment fault repair success rate: The air conditioner compressor repair task was successfully completed with a success rate of 100%. Stable equipment runtime after maintenance: The air conditioner's operation was tracked for one week after repair; stable runtime was measured for one week. For system optimization tasks, such as optimizing the information publishing system's running speed, the performance improvement was calculated by comparing system page loading times before and after optimization. User satisfaction rating after business recovery: The rating was obtained through a questionnaire survey of store customers and management regarding their satisfaction with the business recovery.

[0131] Step S215: Associate the knowledge application data with the task execution result data to establish a mapping relationship between each knowledge application record and the corresponding task execution result, with each application record corresponding to one task execution result.

[0132] By linking knowledge application data with task execution result data, a mapping relationship is established between each knowledge application record and its corresponding task execution result. For example, the application record of knowledge about replacing an air conditioner compressor corresponds to the execution result of the compressor repair task, including task completion time and equipment recovery time. Similarly, the application record of knowledge about replacing parts is also linked to the task execution result; each application record can find corresponding task execution result data.

[0133] Step S216: Analyze the correlation between knowledge application data and task execution result data.

[0134] This analysis examines the correlation between knowledge application data and task execution results, and the impact of knowledge retrieval order on task completion time. A reasonable knowledge retrieval order, such as diagnosing first, then replacing, and finally debugging, can shorten task completion time; conversely, a disordered retrieval order may lead to repetitive operations, extending the time. The relationship between knowledge retrieval time and task execution efficiency is also analyzed. Excessive knowledge retrieval time may indicate complex knowledge content or difficulty in engineer understanding, reducing task execution efficiency. Finally, the impact of the number of knowledge retrievals on task success rate is explored. Appropriately linking knowledge retrieval frequencies, such as retrieving tool usage knowledge or safety operation knowledge, can improve task success rate; too few retrieval frequencies may miss critical operations, leading to a lower success rate.

[0135] Step S217: Based on the association analysis results, identify problems where the knowledge application effect has not met the preset standards.

[0136] Based on the association analysis results, issues were identified where the knowledge application effect did not meet preset standards. For example, the task completion time after a certain knowledge application exceeded a preset time threshold; for instance, after applying knowledge about troubleshooting an information publishing system, the task completion time was one hour longer than the preset time. The number of associated knowledge applications was lower than a preset application number threshold; for example, during the application of knowledge about air conditioner maintenance, the associated safety operating procedures were applied zero times, lower than the preset threshold of at least one application. The stable operating time of the equipment after knowledge application was shorter than a preset stable operating time threshold; for example, after applying knowledge about repairing an IoT control module, the stable operating time of the equipment was only two days, shorter than the preset threshold of one week.

[0137] Step S218: Statistically analyze the application effectiveness indicators of different types of knowledge, including knowledge application success rate, knowledge application efficiency, and knowledge association effectiveness; the knowledge application success rate is calculated as the proportion of the number of tasks that successfully apply knowledge to the total number of application tasks; the knowledge application efficiency is calculated as the average task completion time after knowledge application; and the knowledge association effectiveness is calculated as the numerical improvement in task quality after the application of associated knowledge.

[0138] The application effectiveness metrics for different types of knowledge were analyzed. For example, in the basic operations and maintenance knowledge layer, knowledge about routine air conditioning maintenance was applied to 100 tasks, with 95 successful applications, resulting in a success rate of 95%. In the scenario-specific knowledge layer, knowledge about a specific model's information publishing system was applied to 50 tasks, with an average task completion time of two hours. Finally, in the emergency response knowledge layer, the success rate of tasks improved by 15% after linking emergency recovery knowledge with backup knowledge; this improvement is the knowledge association effectiveness metric.

[0139] Step S219: Categorize and organize application effect indicators according to knowledge type and task type to form a knowledge application effect statistical table. The knowledge application effect statistical table includes knowledge identifier, knowledge type, application task type, application success rate, application efficiency, association effectiveness and problem description.

[0140] Application effectiveness metrics are categorized and organized according to knowledge type and task type. Taking the basic operation and maintenance knowledge layer of air conditioning daily maintenance knowledge with knowledge identifier K001 as an example, the knowledge type is basic operation and maintenance knowledge, the application task type is regular maintenance task, the application success rate is 95%, the application efficiency is an average task duration of one hour, the association effectiveness is a 20% increase in stable equipment runtime after the associated knowledge is called, and the problem description is no obvious problems. All knowledge is organized in this format to form a knowledge application effectiveness statistical table.

[0141] Step S220: Based on the knowledge application effect statistics table and problem analysis results, generate a knowledge application effect analysis report. The knowledge application effect analysis report includes an overview of the overall knowledge application effect, a comparison of the application effects of different types of knowledge, existing problems and their causes.

[0142] Based on the statistical table of knowledge application effectiveness and the results of problem analysis, a knowledge application effectiveness analysis report is generated. The overall knowledge application effectiveness overview summarizes the average success rate, average efficiency, and average association effectiveness of all knowledge applications. A comparison of the application effectiveness of different types of knowledge is presented, comparing the differences in application success rate and efficiency among basic operation and maintenance knowledge, scenario-specific knowledge, and emergency response knowledge. Emergency response knowledge may have the highest success rate, but its efficiency may be slightly lower due to the emergency situation. Problems identified include low efficiency in the application of some scenario-specific knowledge and insufficient association calls for certain knowledge. Problem attribution analysis suggests that low efficiency may be due to cumbersome knowledge steps, and insufficient association calls may be due to unreasonable association strength settings.

[0143] Step S230: Based on the knowledge application effect analysis report, optimize the matching conditions and linkage logic of the dynamic coupling rules between the scenario and knowledge, and generate rule optimization parameters.

[0144] Step S231: Extract the knowledge application problems and their causes recorded in the knowledge application effect analysis report. Classify the knowledge application problems according to their causes to form a list of problem types. Problem types include: knowledge adaptation deviation caused by mismatch between matching conditions and actual scenario characteristics; omission of related knowledge calls caused by misalignment between linkage logic and knowledge association; and inefficient knowledge priority calls caused by mismatch between priority sequence and knowledge application effect.

[0145] Extract knowledge application issues and their causes from the knowledge application effect analysis report, categorizing them by their causes. Mismatches between matching conditions and actual scenario characteristics lead to knowledge adaptation biases. For example, if the device model range in the matching conditions for scenario-specific knowledge is set too narrowly, newly purchased devices of the same series may not be compatible with the knowledge. Misalignment between linkage logic and knowledge relationships leads to omissions in calling related knowledge. For instance, the correlation strength between a main piece of knowledge and an important related piece of knowledge may be underestimated, causing the linkage logic to omit the related knowledge and resulting in omissions during calling. Mismatches between priority sequences and knowledge application effects lead to the priority calling of inefficient knowledge. Knowledge with low application efficiency may have a high priority due to numerous historical calls, resulting in priority calling and impacting task processing efficiency, forming a list of problem types.

[0146] Step S232: For the problem type of knowledge adaptation deviation caused by the mismatch between matching conditions and actual scenario characteristics, analyze the differences between the application scenario characteristics of the problem knowledge and the actual task scenario characteristics, and determine the optimization direction of the matching conditions. If the difference is concentrated in the adaptation range of equipment operating status parameters, the optimization direction is to adjust the parameter range; if the difference is concentrated in the response requirements of the business impact range, the optimization direction is to expand or narrow the response range.

[0147] For problem types where matching conditions do not match the characteristics of the actual scenario, the differences between the application scenario characteristics of the problem knowledge and the actual task scenario characteristics are analyzed. For example, in the application scenario characteristics of air conditioning maintenance knowledge, the applicable range of equipment operating status parameters is the parameter range of older air conditioning models. However, the parameters of newer air conditioning models in the actual task scenario exceed this range. The difference lies in the applicable range of equipment operating status parameters. The optimization direction for the matching conditions is determined to be adjusting the parameter range, expanding the applicable parameter interval to include the parameters of the newer equipment models. Another difference lies in the requirement to address the scope of business impact. Originally intended to address the business impact of a single store, this knowledge was applied to a chain reaction of failures across multiple stores in the actual task, resulting in insufficient response. The optimization direction is to expand the response scope to cover the business impact of multiple stores.

[0148] Step S233: For the problem type of missing related knowledge calls caused by misalignment between linkage logic and knowledge association relationship, analyze the cross-knowledge association network of problem knowledge, find related knowledge that should be called but was not called, determine the optimization direction of the linkage logic of related knowledge, if the related knowledge was not called because the association strength value was lower than the preset association strength threshold, the optimization direction is to adjust the association strength value threshold; if the related knowledge was not called because it was not included in the calling order, the optimization direction is to supplement the calling order.

[0149] To address the issue of misalignment between linkage logic and knowledge relationships, the cross-knowledge association network of the problem knowledge was analyzed. In one instance, within the cross-knowledge association network of a main knowledge for IoT control fault repair, a related knowledge about control protocol compatibility, which should have been invoked, was not. Inspection revealed that the association strength value of this knowledge was 0.6, while the preset association strength threshold was 0.7, resulting in its non-invocation. The optimization direction is to adjust the association strength threshold, lowering it to 0.6, so that this related knowledge can be included in the linkage logic. Another instance of related knowledge not being invoked was because it was not included in the invocation order. The optimization direction is to supplement the invocation order by adding an invocation step for this related knowledge after the main knowledge invocation.

[0150] Step S234: For the problem type of inefficient knowledge priority calling caused by mismatch between priority sequence and knowledge application effect, analyze the reasons for inefficient knowledge priority calling, and determine the optimization direction of priority sequence. If the reason is that the current scene characteristics are not considered, the optimization direction is to add scene feature matching degree weight; if the reason is that the sorting basis does not cover key application effect indicators, the optimization direction is to increase the dimensions of sorting basis.

[0151] To address the issue of mismatch between priority sequences and knowledge application effectiveness, this paper analyzes the reasons for inefficient knowledge prioritization. One example is a high priority given to knowledge about an information publishing system malfunction, despite poor application effectiveness. This is because the priority sequence was generated without considering the specific device model in the current store. This knowledge is more applicable to other device models. The optimization strategy is to add a scenario feature matching weight, considering the matching degree between knowledge and the current scenario features during sorting, thus reducing the priority of mismatched knowledge. Another example of inefficient knowledge prioritization is that the sorting criteria only consider task success rate, neglecting the crucial application effectiveness metric of task completion time. The optimization strategy is to increase the dimension of the sorting criteria, including task completion time, thus reducing the priority of knowledge with longer completion times.

[0152] Step S235: Based on the optimization direction, formulate specific optimization measures, including resetting the upper and lower limits of the device operation status parameters in the matching conditions, supplementing the knowledge calling link of the related knowledge linkage logic with the knowledge whose association strength value exceeds the preset threshold, and adding scene feature matching degree weight to the priority sequence.

[0153] Based on the optimization direction, specific optimization measures were formulated. For optimizing the equipment operating status parameter adaptation range in the matching conditions, the specific measure was to reset the upper and lower limits of the parameter adaptation range for air conditioning maintenance knowledge. The original upper limit of temperature parameters was increased from 50 degrees Celsius to 60 degrees Celsius, and the lower limit was decreased from 10 degrees Celsius to 5 degrees Celsius to include the parameter range of new air conditioning models. For optimizing the linkage logic of related knowledge, the specific measure was to supplement the knowledge invocation process for knowledge whose association strength value exceeds the adjusted preset threshold of 0.6. After the main knowledge invocation for IoT control fault repair, a step for invoking control protocol compatibility related knowledge was added. For optimizing the priority sequence, the specific measure was to increase the scene feature matching degree weight. During sorting, the priority score of knowledge was calculated by multiplying the original task success rate score by 70% and adding the scene feature matching degree score multiplied by 30%.

[0154] Step S236: Convert the optimization measures into rule optimization parameters. Each optimization measure corresponds to one or more parameters. Adjust the equipment operating status parameter adaptation range to generate upper and lower limit adjustment values ​​of the parameters. Supplement the associated knowledge call link to generate associated knowledge identifiers and call sequence numbers. Add scene feature matching degree weight to generate weight coefficients.

[0155] Optimization measures are transformed into rule-based optimization parameters. Optimization measures that adjust the adaptation range of equipment operating status parameters generate corresponding upper and lower limit adjustment values ​​for parameters. For example, the upper limit adjustment value for temperature parameters is +10 degrees Celsius, and the lower limit adjustment value is -5 degrees Celsius. Optimization measures that supplement the knowledge invocation process generate corresponding knowledge identifiers. For example, the identifier ID for control protocol compatibility knowledge is K056, and the invocation sequence number is 2, meaning that after the main knowledge invocation, this related knowledge is invoked in sequence 2. Optimization measures that increase the weight of scene feature matching degree generate corresponding weight coefficients; that is, the weight coefficient for scene feature matching degree is 0.3.

[0156] Step S237: Set the value range and application scenario for the optimization parameters of each rule. The value range of the upper and lower limit adjustment values ​​of the parameters is set based on the statistical data of the device parameters adapted to the problem knowledge. The application scenario of the associated knowledge identifier and the calling sequence number is set based on the application scenario of the problem knowledge. The value range of the weight coefficient is set to be between 0 and 1.

[0157] For each rule, the parameter optimization parameters are set with a value range and application scenario. The upper and lower limit adjustment values ​​are set based on the statistical data of device parameters adapted to the problem knowledge. For example, the temperature parameter range of a new air conditioner model is statistically analyzed, and the upper limit adjustment value is set to 5 to 15 degrees Celsius, while the lower limit adjustment value is set to -3 to -8 degrees Celsius. The application scenario of the associated knowledge identifier and call sequence number is set based on the application scenario of the problem knowledge. For instance, the application scenario of K056 associated knowledge identifier and call sequence number 2 is set to a scenario involving control protocol issues in the fault repair task of an IoT control system. The weight coefficient is set to a value range of 0 to 1, and the scenario feature matching degree weight coefficient of 0.3 is within this range.

[0158] Step S238: Verify the effectiveness of the rule optimization parameters, select a test task similar to the application scenario of the problem knowledge, apply the optimization parameters to the dynamic coupling rule between the scenario and the knowledge, run the rule to filter knowledge and process the test task, and record the knowledge adaptation accuracy, the completeness of associated knowledge retrieval, and the task processing efficiency of the test task.

[0159] To verify the effectiveness of the optimized rules, select test tasks with similar application scenarios to the problem knowledge, such as maintenance tasks for other air conditioner models with similar parameter ranges to the new model. Apply the upper and lower limit adjustment values ​​of the parameters to the dynamic coupling rules, run the rules to filter knowledge, record the knowledge adaptation accuracy, and check whether the knowledge for the new model air conditioner is accurately filtered out. For the optimized parameters related to knowledge identifiers and call sequence numbers, select test tasks related to IoT control protocol issues, apply the optimized parameters, record the completeness rate of related knowledge calls, and check whether the K056 knowledge is accurately called and whether the completeness rate meets the preset standard. For the optimized parameters of weight coefficients, select test tasks with different scenario characteristics, apply the parameters, and record the task processing efficiency, such as whether the task completion time is shortened.

[0160] Step S239: Compare the test results before and after optimization. If the improvement rate of knowledge adaptation accuracy exceeds the preset improvement rate threshold, the completeness rate of associated knowledge retrieval reaches the preset completeness rate threshold, and the improvement of task processing efficiency exceeds the preset improvement rate threshold, then the optimization parameters are deemed effective. If the optimization effect does not reach the corresponding threshold standard, then the values ​​of the optimization parameters are readjusted and the verification process is repeated.

[0161] Comparing the test results before and after optimization, after adjusting the upper and lower limits of the application parameters, the knowledge matching accuracy improved from 70% to 90%, an increase of 20%, exceeding the preset improvement threshold of 15%. The completeness of associated knowledge retrieval improved from 60% to 95%, reaching the preset completeness threshold of 90%. Task processing efficiency improved, with task completion time reduced from 3 hours to 2 hours, an improvement of 33%, exceeding the preset improvement threshold of 20%, thus determining the optimization parameters were effective. If any optimization parameter fails to meet the threshold standard, such as the completeness of associated knowledge retrieval being only 85%, below the 90% threshold, the association strength threshold needs to be readjusted, decreasing it from 0.6 to 0.55, and the verification process repeated.

[0162] Step S240: Collect all valid rule optimization parameters, classify and organize them according to the optimized rule elements, and form a set of rule optimization parameters. Each parameter includes parameter name, optimization direction, value, application scenario and verification result.

[0163] Collect all valid rule optimization parameters and categorize them according to the elements of the optimized rule, such as matching condition optimization parameters, linkage logic optimization parameters, priority sequence optimization parameters, etc. Each parameter includes a parameter name, such as the upper limit adjustment value of the temperature parameter; the optimization direction, such as expanding the parameter range; the value, such as +10 degrees Celsius; the application scenario, such as maintenance tasks for new air conditioner models; and the verification result, such as a 20% improvement in adaptation accuracy, forming a set of rule optimization parameters.

[0164] Step S250: Optimize the modeling dimensions of the knowledge end-to-end evolution model using the rules to optimize the parameters, and supplement the evaluation dimensions of the identification type and iterative influence range of the knowledge generation triggering factors to obtain the optimized knowledge end-to-end evolution model.

[0165] For example, step S251: parse the set of optimization parameters for the rules and extract parameters related to the modeling of the knowledge end-to-end evolution model, including parameters related to knowledge adaptation scenarios, parameters related to knowledge association relationships, and parameters related to the impact of knowledge iteration.

[0166] The parsing rules optimize the parameter set, extracting parameters relevant to the modeling of the knowledge end-to-end evolution model. Parameters related to knowledge adaptation scenarios, such as the upper and lower limits of the adaptation range for device operating status parameters, reflect the changing needs of knowledge adaptation scenario characteristics and can be used to adjust the identification dimension of knowledge generation trigger factors in the model. Parameters related to knowledge relationships, such as the threshold adjustment value for the association strength of associated knowledge and the calling order number, affect the construction of the association network between knowledge and can be used to optimize the modeling dimension of association relationships. Parameters related to the impact of knowledge iteration, such as the adjustment value of the evaluation index of the impact on business systems within the iteration impact range, can be used to supplement the evaluation dimension of the iteration impact range in the model.

[0167] Step S252: For the identification type of knowledge generation trigger factors, analyze the parameters related to scenario matching in the rule optimization parameters, determine the types of knowledge generation trigger factors that need to be added. If the parameters involve specific device states, add device state trigger factor types; if the parameters involve specific business time periods, add business time period trigger factor types.

[0168] To identify the types of knowledge generation trigger factors, we analyzed the scenario-matching parameters in the rule optimization parameters. Among these parameters are those related to adjusting the adaptation range of equipment operating status parameters. These parameters relate to the knowledge requirements of equipment in specific states, such as the special maintenance knowledge requirements of a new air conditioner model under high-temperature conditions. Therefore, we determined that a new equipment status trigger factor type needs to be added. This type is used to identify situations where knowledge generation is triggered due to the equipment being in a specific operating state. Additionally, the parameters involve knowledge adaptation adjustments for business operations during specific time periods, such as the high-load maintenance knowledge requirements of a store information publishing system during holidays. Therefore, we determined that a new business time period trigger factor type needs to be added to identify situations where knowledge generation is triggered due to business operations during specific time periods.

[0169] Step S253: Set identification features for the newly added trigger factor types. The identification features for device status trigger factor types include device operating parameter thresholds and parameter durations. The identification features for business period trigger factor types include the definition of peak business periods and the business volume standard within the period.

[0170] Define identification characteristics for newly added device status trigger factor types. These include device operating parameter thresholds, such as an air conditioner operating temperature exceeding 60 degrees Celsius, and parameter durations, such as the high-temperature state lasting more than 30 minutes. When a device meets both of these characteristics, the device status trigger factor is triggered. Define identification characteristics for business period trigger factor types. Peak business periods are defined as 10:00 AM to 4:00 PM on weekends. The business volume standard during this period is defined as the information publishing system publishing more than 100 advertisements per hour. When business is in this period and the business volume reaches the standard, the business period trigger factor is triggered.

[0171] Step S254: Integrate the newly added trigger factor types and identification features into the trigger factor identification module of the knowledge end-to-end evolution model, update the identification logic in the module, and enable the module to identify the newly added trigger factors.

[0172] The newly added device status trigger factor types and identification features are integrated into the trigger factor identification module of the knowledge end-to-end evolution model. Monitoring logic for device operating parameter thresholds and durations is added to this module. When a device operating parameter reaches a set threshold and persists for the corresponding time, it is identified as a device status trigger factor. Similarly, business period trigger factor types and identification features are integrated into this module. Judgment logic for business periods and business volume is added. When the definition of a peak business period is met and the business volume reaches the standard, it is identified as a business period trigger factor. The identification logic in the module is then updated to ensure that the module can accurately identify the newly added trigger factor types.

[0173] Step S255: For the evaluation dimensions of the scope of impact of the iteration, analyze the parameters related to the scope of knowledge impact in the rule optimization parameters, determine the evaluation dimensions of the iteration impact that need to be supplemented, and supplement the dimensions of impact of business systems if the parameters involve the impact of business systems; supplement the dimensions of impact of user groups if the parameters involve the impact of user groups.

[0174] For the evaluation dimensions of the impact scope of the iteration, the parameters related to the knowledge impact scope in the rule optimization parameters are analyzed. Among the rule optimization parameters are adjustment parameters concerning the number of business systems affected after knowledge iteration. These parameters relate to the impact of knowledge on business systems, identifying the dimensions of business system impact that need to be supplemented, used to assess the degree of impact of knowledge iteration on business systems. There are also adjustment parameters concerning the number of users affected after knowledge iteration, identifying the dimensions of user group impact that need to be supplemented, used to assess the impact of knowledge iteration on the user group.

[0175] Step S256: Set evaluation indicators for the newly added evaluation dimensions. The evaluation indicators for the business system impact dimension include the number of affected business systems and the system recovery time. The evaluation indicators for the user group impact dimension include the number of affected users and changes in user satisfaction.

[0176] Evaluation metrics are set for the newly added business system impact dimension. The number of affected business systems is counted, representing the number of business systems potentially affected after knowledge iteration. For example, knowledge iteration in air conditioning maintenance might affect the store's air conditioning control system and energy management system. System recovery time is evaluated as the time required for a system to return to normal operation when knowledge iteration causes an anomaly. Evaluation metrics are also set for the user group impact dimension. The number of affected users is counted, representing the number of users whose service is affected by knowledge iteration. For example, a knowledge iteration error in the information publishing system might affect all users viewing advertisements for that store. Changes in user satisfaction are collected through questionnaires to gather information on changes in user satisfaction ratings after knowledge iteration.

[0177] Step S257: Integrate the newly added evaluation dimensions and evaluation indicators into the iterative impact evaluation module of the knowledge end-to-end evolution model, update the evaluation logic in the module, and enable the module to calculate the evaluation results of the newly added dimensions.

[0178] The newly added business system impact dimensions and evaluation metrics are integrated into the iterative impact assessment module of the knowledge end-to-end evolution model. Logic for counting the number of affected business systems is added to this module, determined by analyzing the correlation between knowledge content and each business system. Logic for calculating system recovery time is also added, estimating recovery time based on historical data from similar system anomalies and the current system state. Similarly, the user group impact dimension and evaluation metrics are integrated into this module, with logic for counting the number of affected users calculated using a user database and the scope of business impact. Logic for evaluating changes in user satisfaction is added, analyzing user feedback data and iterated knowledge content to update the module's evaluation logic, enabling the module to calculate evaluation results for the newly added dimensions.

[0179] Step S258: Simultaneously, adjust the calculation module for the correlation strength between different knowledge in the knowledge full-link evolution model. Based on the optimization parameters of the correlation knowledge linkage logic in the rule optimization parameters, supplement the calculation dimension of correlation strength. If the parameters involve scenario synergy, supplement the scenario synergy dimension to make the correlation strength calculation more in line with the optimization goal set in the rule optimization parameters.

[0180] The module for calculating the correlation strength between different knowledge points in the knowledge end-to-end evolution model has been adjusted. This adjustment is based on optimization parameters related to the linkage logic of correlated knowledge in the rule optimization parameters, such as the scene matching weight coefficient of correlated knowledge. A new dimension for calculating correlation strength has been added, incorporating parameters related to scene synergy, i.e., the frequency and effectiveness of the joint application of two knowledge points in the same or similar scenarios. When calculating correlation strength, in addition to the original content similarity dimension, a weight for the scene synergy dimension has been added, such as multiplying the scene synergy score by 0.2 and adding the content similarity score by 0.8. This makes the correlation strength calculation more aligned with the optimization goal set in the rule optimization parameters to improve the accuracy of correlated knowledge retrieval.

[0181] Step S259: Run the optimized knowledge end-to-end evolution model, access the latest data from the multi-source operation and maintenance knowledge update source, regenerate the knowledge end-to-end evolution set, compare the differences between the set before and after optimization, and check whether the newly added trigger factor type has been identified, whether the newly added evaluation dimension has calculation results, and whether the correlation strength value is more in line with actual needs.

[0182] The optimized knowledge end-to-end evolution model is run, and the latest data from multiple sources of O&M knowledge updates, such as the latest equipment manufacturer technical upgrade documents and real-time O&M event handling cases, are integrated to regenerate the knowledge end-to-end evolution set. The differences between the sets before and after optimization are compared, checking whether the newly added equipment status trigger factor types are identified, and whether there are knowledge generation records triggered by high-temperature equipment operation in the newly generated set. The calculation results for the newly added business system impact dimension and user group impact dimension are checked, such as whether the impact assessment results of a certain knowledge iteration include data such as the number of affected business systems being 2, the system recovery time being 1 hour, the number of affected users being 500, and the user satisfaction change being -5%. The correlation strength values ​​are also checked to see if they better reflect actual needs. For example, if two pieces of knowledge originally had low correlation strength but high scenario synergy, has the correlation strength value improved after optimization, better reflecting the correlation needs of the two in actual applications?

[0183] Step S260: If the number of newly identified triggering factors in the optimized set is lower than the preset identification number threshold, the newly added evaluation dimension has no calculation results, or the correlation strength value deviates from the actual requirement by more than the preset deviation threshold, then the modeling dimensions and parameters of the model are readjusted, and the running and comparison process is repeated.

[0184] If the optimized set contains only 5 newly identified trigger factors, which is less than the preset threshold of 10, the identification features of the trigger factor identification module in the model need to be readjusted. This could involve lowering the device operating parameter threshold or shortening the duration. The model should then be rerun and the results compared. If no calculation results are obtained for the newly added evaluation dimension, check the logic of the iterative impact evaluation module for errors, correct it, and rerun the model. If the correlation strength value deviates from the actual requirement by more than the preset deviation threshold (e.g., two pieces of knowledge are closely related in actual application, but the correlation strength value calculated by the model is still low), the weights of each dimension in the correlation strength calculation module need to be readjusted. This could involve increasing the weight of the scenario synergy dimension. The process of running and comparing the results should then be repeated.

[0185] Step S261: After completing the adjustment and verification, determine the final modeling logic and parameter settings of the optimized knowledge end-to-end evolution model, and output the optimized knowledge end-to-end evolution model.

[0186] After adjustments and verifications, and after multiple runs and comparisons, the number of newly identified triggering factors reached 15, exceeding the threshold; all newly added evaluation dimensions had accurate calculation results; the deviation between the correlation strength values ​​and actual needs was within the preset deviation threshold. The final modeling logic and parameter settings of the optimized knowledge full-link evolution model were determined, such as the device parameter thresholds of the triggering factor identification module, the weights of each dimension of the iterative impact evaluation module, and the scene synergy weights of the correlation strength calculation module, and the optimized knowledge full-link evolution model was output.

[0187] Step S270: Use the optimized knowledge end-to-end evolution model as the basic modeling framework for the next knowledge update, and repeatedly execute the steps from generating dynamic coupling rules between scenarios and knowledge to generating rule optimization parameters to form a knowledge update closed loop. Based on the rule optimization parameters and model adjustment records of multiple knowledge update closed loops, generate a knowledge update strategy library for the large model of operation and maintenance technical services to guide the knowledge update process under different operation and maintenance scenarios.

[0188] For example, step S271: Collect the set of rule optimization parameters and the record of knowledge end-to-end evolution model adjustment generated during multiple knowledge update loops. Each record includes the execution time of the update loop, the corresponding operation and maintenance scenario type, rule optimization parameter details and model adjustment content.

[0189] We collected records of rule optimization parameter sets and knowledge end-to-end evolution model adjustments generated during multiple knowledge update loops. The first knowledge update loop was executed on March 1st, corresponding to an air conditioning fault repair scenario. Rule optimization parameter details included adjustments to upper and lower limits of temperature parameters and correlation strength thresholds. Model adjustments involved adding new device status trigger factor types. The second loop was executed on April 15th, corresponding to an information publishing system optimization scenario. Rule optimization parameter details included scenario collaboration weight coefficients. Model adjustments involved supplementing the calculation dimension for scenario collaboration correlation strength. Each record contained the aforementioned detailed information.

[0190] Step S272: Classify the collected records and group them according to the operation and maintenance scenario type. For each operation and maintenance scenario type, extract the rule optimization parameters of all knowledge update closed loops under that operation and maintenance scenario type, and count the number of applications and optimization effects of each parameter. The number of applications is calculated according to the number of times the parameter is used in the closed loop, and the optimization effect is calculated according to the numerical improvement of knowledge adaptation accuracy after the parameter is applied.

[0191] The collected records were categorized into groups based on operation and maintenance scenarios, such as air conditioning fault repair, information dissemination system optimization, and IoT control system maintenance. For the air conditioning fault repair scenario, optimization parameters for all knowledge update loops within that scenario were extracted, including temperature and pressure parameter adjustment values. The application frequency of each parameter was counted; the temperature parameter adjustment value was used 3 out of 5 loops. The optimization effect was calculated based on the improvement in knowledge adaptation accuracy after parameter application; the application of this temperature parameter adjustment value resulted in an average improvement of 18% in knowledge adaptation accuracy.

[0192] Step S273: Extract the adjustment content of the knowledge full-link evolution model under this operation and maintenance scenario type, analyze the correspondence between the adjustment direction of the modeling dimension and the scenario requirements, and determine that if the adjustment direction covers the main requirements of the scenario, it is a match; if the adjustment direction does not cover the main requirements of the scenario, it is a mismatch.

[0193] The adjustments to the knowledge chain evolution model under the air conditioning fault repair scenario are extracted, such as adding new equipment status trigger factor types and supplementing the business system impact assessment dimensions. The correspondence between the adjustment direction of the modeling dimensions and the scenario requirements is analyzed. The main requirement of the air conditioning fault repair scenario is to accurately identify the knowledge update requirements triggered by abnormal equipment operation and to assess the impact of knowledge iteration on the air conditioning system and related business systems. The adjustment directions of adding new equipment status trigger factor types and supplementing the business system impact assessment dimensions cover these main requirements and are therefore considered a match. If an adjustment direction is to supplement the user group impact assessment dimension, but the main requirement of the scenario pays little attention to the impact on the user group, the adjustment direction does not cover the main requirements of the scenario and is therefore considered a mismatch.

[0194] Step S274: Based on statistical results and analysis, formulate a unique knowledge update strategy for each operation and maintenance scenario type. The unique knowledge update strategy includes the recommended modeling dimensions of the knowledge end-to-end evolution model under the operation and maintenance scenario type, the recommended matching conditions of the dynamic coupling rules between the scenario and knowledge, the logic for setting the priority of recommended knowledge calls, and the recommended dynamic adaptation adjustment direction. The recommended modeling dimensions are determined based on the matching results of scenario requirements and model adjustment, and the recommended matching conditions are determined based on the parameter optimization effect.

[0195] Based on statistical results and analysis, specific knowledge update strategies are developed for different air conditioner fault repair scenarios. Recommended modeling dimensions are determined based on scenario requirements and model adjustment matching results, including dimensions such as equipment status trigger factor identification and business system impact assessment. The recommended matching conditions for the dynamic coupling rules between scenarios and knowledge are determined based on parameter optimization effects. For example, the recommended temperature upper and lower limits for the adapted range of equipment operating status parameters are +10 degrees Celsius and -5 degrees Celsius, respectively, indicating optimal parameter optimization. The recommended knowledge call priority setting logic combines task success rate and scenario feature matching degree, with weights of 0.7 and 0.3, respectively. The recommended dynamic adaptation adjustment direction is to automatically trigger knowledge call priority adjustment when the change in equipment operating parameters exceeds 20%.

[0196] Step S275: Configure policy application conditions for each exclusive knowledge update policy, determine the applicable operation and maintenance task target type, equipment operating status parameter range and business impact range. The task target type is determined according to the task type that appears most frequently in the historical records under the operation and maintenance scenario type. The equipment operating status parameter range is set according to the common parameter range in the historical statistics under the operation and maintenance scenario type. The business impact range is set according to the typical impact range in the historical evaluation results under the operation and maintenance scenario type.

[0197] The dedicated knowledge update strategy for the air conditioning fault repair scenario is equipped with application conditions. The applicable maintenance task target type is determined. In this scenario, the most frequently occurring task type in historical records is air conditioning cooling failure repair; therefore, the strategy applies to this task type. Equipment operating status parameter ranges are set according to common parameter ranges in historical statistics for this scenario, such as a temperature range of 10-60 degrees Celsius and a pressure range of 0.5-3 MPa. The business impact scope is set according to the typical impact scope in historical assessment results for this scenario. The typical impact scope is the air conditioning system of a single store; therefore, the applicable business impact scope of the strategy is a single store.

[0198] Step S276: At the same time, add strategy effect evaluation criteria for each strategy, including the target value of knowledge adaptation accuracy, the target value of knowledge application efficiency, and the target value of task processing success rate after applying the strategy. The target value is set based on the highest value of the historical optimization effect of this operation and maintenance scenario type.

[0199] To evaluate the effectiveness of the knowledge update strategy specifically designed for air conditioner malfunction repair scenarios, a strategy effectiveness evaluation standard is added. The target value for knowledge adaptation accuracy after applying the strategy is set at 95%, based on the highest historical optimization result for this scenario. This highest value was achieved after applying the optimized parameters in a specific closed-loop iteration. The target value for knowledge application efficiency is set at an average task processing time of no more than 2 hours, based on the highest efficiency result from a closed-loop iteration in the historical optimization process. The target value for task processing success rate is set at 98%, also based on the highest historical success rate.

[0200] Step S277: Integrate the exclusive knowledge update strategies, application conditions, and effect evaluation standards for all operation and maintenance scenario types, classify and store them according to scenario type, and form a knowledge update strategy library for the large model of operation and maintenance technical services.

[0201] The system integrates exclusive knowledge update strategies, application conditions, and effectiveness evaluation standards for all operation and maintenance scenario types. Strategies for scenarios such as air conditioning fault repair, information release system optimization, and IoT control system maintenance are arranged sequentially by scenario type. Each scenario strategy includes recommended modeling dimensions, recommended matching conditions, priority setting logic, and dynamic adaptation adjustment directions. Strategy application conditions include applicable task types, equipment parameter ranges, and business impact scope. Effectiveness evaluation standards include accuracy, efficiency, and success rate target values. These strategies are categorized and stored in the database according to scenario type, forming a knowledge update strategy library for the large-scale operation and maintenance technical service model.

[0202] Step S278: Set up a search function for the knowledge update strategy library. The search dimensions include operation and maintenance scenario type, task target type, and device operating status parameter range. After the user inputs the search dimension information, the corresponding exclusive knowledge update strategy will be output.

[0203] Set up a search function for the knowledge update strategy library. When the user inputs the operation and maintenance scenario type as air conditioner fault repair, the task objective type as air conditioner not cooling fault repair, and the equipment operating status parameter range of temperature 10-60 degrees Celsius, the search function will match the corresponding exclusive knowledge update strategy in the strategy library and output the recommended modeling dimensions, matching conditions, priority setting logic, etc. of the strategy to guide the user to update knowledge in this scenario.

[0204] Figure 2 This application illustrates a dynamic knowledge update system 100 for large-scale modeling of operational and maintenance (O&M) technical services, comprising a processor 1001, a memory 1003, and program code stored in the memory 1003. The processor 1001 executes the program code to implement the steps of the dynamic knowledge update method for O&M technical services. The processor 1001 and the memory 1003 are connected, for example, via a bus 1002. Optionally, the dynamic knowledge update system 100 for O&M technical services may further include a transceiver 1004, which can be used for data interaction between this dynamic knowledge update system and other dynamic knowledge update systems for O&M technical services, such as sending and / or receiving data. It should be noted that in actual scheduling, the transceiver 1004 is not limited to one, and the structure of this dynamic knowledge update system 100 for O&M technical services does not constitute a limitation on the embodiments of this application.

[0205] The memory 1003 is used to store program code for executing the embodiments of this application, and its execution is controlled by the processor 1001. The processor 1001 is used to execute the program code stored in the memory 1003 to implement the steps shown in the foregoing method embodiments.

[0206] The above description is only an optional implementation method for some implementation scenarios of this application. It should be noted that for those skilled in the art, other similar implementation methods based on the technical concept of this application, without departing from the technical concept of this application, also fall within the protection scope of the embodiments of this application.

Claims

1. A method for dynamically updating large-scale model knowledge for operation and maintenance technical services, characterized in that, The method includes: The knowledge generation triggering factors, iteration impact range, and correlation strength between different knowledge in the multi-source operation and maintenance knowledge update source are modeled and processed to obtain the knowledge full-link evolution set, which includes knowledge generation triggering records, iteration impact assessment results, and cross-knowledge association network; Based on the real-time scenario features and knowledge full-link evolution set of operation and maintenance tasks, dynamic coupling rules between scenarios and knowledge are generated. The dynamic coupling rules include scenario knowledge matching conditions, knowledge call priority sequence and related knowledge linkage logic. The real-time scenario features of operation and maintenance tasks include task target type, equipment operating status parameters and business impact range. The dynamic coupling rules are used to couple and filter the knowledge content in the knowledge full-link evolution set, extract the knowledge content that meets the matching conditions and integrate them to obtain the scene-coupled knowledge set. The scenario-coupled knowledge set is embedded into the hierarchical knowledge architecture of the large-scale operation and maintenance technical service model. Simultaneously, the storage node parameters and retrieval logic of the hierarchical knowledge architecture are adjusted based on the knowledge full-link evolution set to obtain the large-scale operation and maintenance technical service model after basic update. The basic updated operation and maintenance technical service big model is dynamically adapted in multiple dimensions. The knowledge call priority sequence and storage node parameters are adjusted in combination with the real-time scenario feature changes of operation and maintenance tasks to generate the adapted updated operation and maintenance technical service big model.

2. The method for dynamic updating of large-scale model knowledge for operation and maintenance technical services according to claim 1, characterized in that, The modeling process of knowledge generation triggering factors, iteration impact range, and correlation strength between different knowledge sources in the multi-source operation and maintenance knowledge update sources yields a knowledge full-link evolution set, including: Access to a multi-source operation and maintenance knowledge update source, which includes an operation and maintenance service historical task case library, an equipment manufacturer technical upgrade document library, an industry operation and maintenance standard revision library, an operation and maintenance event real-time processing library, and an operation and maintenance knowledge user feedback library. Extract the knowledge generation process record of each historical operation and maintenance task from the operation and maintenance service historical task case library, including the initial knowledge when the task starts, the knowledge supplementation during the task execution, and the knowledge summary after the task ends. Arrange them in the order of task execution time to generate a knowledge generation timeline for a single task. Iterative texts of equipment maintenance knowledge are collected from the equipment manufacturer's technical upgrade document library. The differences in knowledge content before and after the iteration are compared to determine the range of equipment models affected by the iterative knowledge, maintenance process steps, and related knowledge modules, thus forming a list of the scope of impact of the iteration. The revised text of standard knowledge is obtained from the industry operation and maintenance standard revision library. The citation relationship between the revised clause and the original clause, the association relationship between the revised clause and other standard clauses are extracted. The content similarity between the revised clause and the related clause is calculated. A similarity threshold is set, and clauses with similarity values ​​exceeding the similarity threshold are selected as strongly related clauses to form a standard knowledge association list. The processing solutions for real-time events are extracted from the real-time processing library of operation and maintenance events, and the existing similar event processing knowledge is associated with them. The content similarity between the new knowledge and the existing knowledge is calculated, and the content similarity is used as the association strength between different knowledge. Feedback information on the effectiveness of knowledge application is collected from the operation and maintenance knowledge user feedback database, including knowledge practicality score, usability score and supplementary suggestions. The feedback information is associated with the corresponding knowledge module. The association strength between different knowledge is adjusted based on the practicality score in the feedback information. Feedback with a practicality score higher than the preset score corresponds to an increase in the association strength value, and feedback with a practicality score lower than the preset score corresponds to a decrease in the association strength value. By integrating the knowledge generation timeline of the individual task, the list of iterative impact ranges, the list of standard knowledge associations, and the association strength data between different knowledge, a knowledge generation trigger record is constructed using a time-series modeling method. A graph construction method is used to integrate the association data to form a cross-knowledge association network. The data is then classified and organized according to knowledge type to form a knowledge full-link evolution set.

3. The method for dynamic updating of large-scale model knowledge for operation and maintenance technical services according to claim 1, characterized in that, The set of real-time scenario features and knowledge evolution based on operation and maintenance tasks generates dynamic coupling rules between scenarios and knowledge, including: Collect real-time scenario characteristics of operation and maintenance tasks, and extract task target type information from the operation and maintenance task management platform; Obtain equipment operating status parameters from the equipment status monitoring system, including equipment CPU utilization, memory usage, disk read / write speed, network transmission rate, and equipment temperature data; Extract business impact scope information from the business impact analysis module, including the number of affected business systems, the scale of affected users, and the estimated duration of business interruption; The task target type, the device operating status parameters, and the business impact range are feature-encoded to generate a coding vector of real-time scene features. Each dimension of the coding vector corresponds to a specific value of a scene feature. The application scenario features of each piece of knowledge are extracted from the knowledge end-to-end evolution set, including the type of task objectives that the knowledge has been applied to, the range of device operating status parameters that it is adapted to, and the scope of business impact it should address, to generate an encoding vector for the knowledge application scenario. Calculate the vector similarity between the encoded vector of real-time scene features and the encoded vector of knowledge application scenarios, set a similarity threshold, and filter knowledge with similarity values ​​exceeding the threshold as candidate scene knowledge; Extract the generation triggering factors, iterative impact assessment results, and correlation strength between different knowledge from the knowledge end-to-end evolution set of candidate scenario knowledge, and analyze the applicability of candidate scenario knowledge in the current real-time scenario, including whether the matching degree between the generation triggering factor and the current scenario triggering event exceeds the preset matching degree threshold, whether the ratio of the iterative impact scope covering the current device and business exceeds the preset coverage ratio threshold, and whether the correlation strength between different knowledge meets the preset strength requirements. Based on the applicability analysis results, scenario knowledge matching conditions are set. The matching conditions include task target type matching requirements, equipment operating status parameter adaptation range, and business impact range response requirements. The task target type matching requirements determine that the current task target type must be consistent with the task target type to which the knowledge has been applied. The equipment operating status parameter adaptation range determines that the current equipment operating status parameters must be within the parameter range of knowledge adaptation. The business impact range response requirements determine that the current business impact range must be within the scope of knowledge response. The application effect data of candidate scenario knowledge in similar historical scenarios are statistically analyzed, including task completion time, task success rate, time for equipment to return to normal operation and business recovery efficiency after knowledge application. The candidate scenario knowledge is sorted in the following order: task completion time from shortest to longest, task success rate from highest to lowest, equipment recovery time from shortest to highest, and business recovery efficiency from fastest to slowest. A knowledge retrieval priority sequence is generated based on the sorting results. Based on the cross-knowledge association network of the knowledge ranked first in the priority sequence, the association strength value of the associated knowledge is extracted. The association strength values ​​are arranged from largest to smallest to determine the retrieval order of the associated knowledge, thus forming the linkage logic of associated knowledge. By integrating the scene knowledge matching conditions, the knowledge call priority sequence, and the associated knowledge linkage logic, dynamic coupling rules between scenes and knowledge are generated according to the task target type. Each task target type corresponds to an independent set of coupling rules, and the content of the coupling rules is used to define the specific values ​​and association logic of each element.

4. The method for dynamic updating of large-scale model knowledge for operation and maintenance technical services according to claim 1, characterized in that, The dynamic coupling rules are used to couple and filter the knowledge content in the knowledge end-to-end evolution set, extract the knowledge content that meets the matching conditions, and integrate them to obtain a scene-coupled knowledge set, including: Traverse all knowledge content in the knowledge full-link evolution set, and extract the application scenario characteristics, core content, and relationship information between different knowledge for each knowledge; Each piece of knowledge is compared one by one with the application scenario characteristics and scenario knowledge matching conditions in the dynamic coupling rules between scenarios and knowledge. The task target type of the knowledge is consistent with the matching requirements, the range of adapted device operating status parameters includes the current device operating status parameters, and the scope of business impact to be addressed covers the current business impact. All knowledge that passes the comparison is retained to form an initial set of filtered knowledge. Extract the association information between different knowledge in each knowledge from the initial screening knowledge set. Based on the association knowledge linkage logic in the dynamic coupling rules between scenarios and knowledge, find other knowledge that is related to each knowledge. If the associated knowledge is not included in the initial screening knowledge set, check whether the application scenario characteristics of the associated knowledge meet the scenario knowledge matching conditions. If they do, add it to the initial screening knowledge set. Prioritize the knowledge in the supplemented initial knowledge set, and arrange the knowledge from high to low priority according to the knowledge call priority sequence in the dynamic coupling rules between scenarios and knowledge to form an ordered knowledge list; Extract the core content of each knowledge item in the ordered knowledge list, including the corresponding operation steps, tools used, parameter setting requirements, precautions and expected results; The core content is linked and integrated. The core content of the knowledge that ranks first in the priority sequence is taken as the main knowledge content, and the core content of the related knowledge is taken as the supplementary knowledge content. According to the calling order determined by the linkage logic of the related knowledge, the supplementary knowledge content is embedded after the relevant operation steps of the main knowledge content. The integrated knowledge content is structured and organized in the following order: knowledge identifier, task target adaptation type, equipment status adaptation range, business impact response range, core operation process, supplementary operation details, and expected effect. Each structured knowledge unit corresponds to a complete integrated knowledge. All structured knowledge units are combined in order of priority to form a scenario-coupled knowledge set, which contains complete association information and structured core content for each piece of knowledge.

5. The method for dynamic updating of large-scale model knowledge for operation and maintenance technical services according to claim 1, characterized in that, The process involves embedding the scenario-coupled knowledge set into a hierarchical knowledge architecture within the large-scale operation and maintenance technical service model. Simultaneously, based on the knowledge chain evolution set, the storage node parameters and retrieval logic of the hierarchical knowledge architecture are adjusted to obtain the basic updated large-scale operation and maintenance technical service model, including: The knowledge architecture of the large model of operation and maintenance technical services is analyzed and divided into layers. The layering logic of the knowledge architecture is determined. The knowledge architecture includes a basic operation and maintenance knowledge layer, a scenario-specific knowledge layer, an emergency response knowledge layer, and a related knowledge layer. Determine the hierarchical affiliation of each structured knowledge unit in the scenario-coupled knowledge set. Based on the scope and urgency of the applicable scenarios, allocate knowledge that covers all regular scenarios to the basic operation and maintenance knowledge layer, allocate knowledge that is limited to specific scenarios to the scenario-specific knowledge layer, allocate knowledge that is adapted to tasks with an urgency level of immediate response to the emergency response knowledge layer, and allocate data that records the relationships between knowledge to the related knowledge layer. Initial parameters are set for each storage node in the hierarchical knowledge layer, including the upper limit of knowledge capacity, the knowledge update frequency threshold, and the knowledge retrieval weight. The initial parameters are set based on the historical operation data of the hierarchical knowledge architecture. Extract the generation triggering factors and iteration impact assessment results of each knowledge in the scenario-coupled knowledge set from the knowledge full-link evolution set, count the occurrence frequency of knowledge generation triggering factors and the number of devices covered by the iteration impact range, and determine the knowledge update frequency requirements and retrieval priority requirements based on the statistical results. Based on the determined needs, the parameters of the corresponding storage nodes are adjusted. For storage nodes containing knowledge where the frequency of generating trigger factors is higher than a preset frequency threshold and the number of devices covered by the iteration impact range exceeds a preset number threshold, the knowledge update frequency threshold of the storage node is increased. For storage nodes containing knowledge where the correlation value between the generated trigger factors and the current scene is higher than a preset correlation threshold and the iteration impact range evaluation result includes the currently running device, the knowledge retrieval weight of the storage node is increased. The initial retrieval logic of the hierarchical knowledge architecture is determined. The initial retrieval logic is set to retrieve knowledge in the order of emergency response knowledge layer, scenario-specific knowledge layer, and basic operation and maintenance knowledge layer, with related knowledge layers being called synchronously with the main knowledge retrieval. Extract the cross-knowledge association network of each knowledge in the scenario-coupled knowledge set from the knowledge full-link evolution set, analyze the association strength value and calling order between knowledge, and adjust the calling weight of associated knowledge in the retrieval logic based on the association strength value; The structured knowledge units in the scenario-coupled knowledge set are written to the corresponding storage nodes according to their hierarchical affiliation. After the writing is completed, the retrieval function of the hierarchical knowledge architecture is tested, and knowledge retrieval requests under different operation and maintenance task scenarios are simulated. The retrieval results are recorded as to whether the retrieved knowledge belongs to the scenario-coupled knowledge set, the retrieval response time, and whether the related knowledge calls are accurate. If the retrieved knowledge does not belong to the scene-coupled knowledge set, the response time exceeds the preset response time threshold, or the associated knowledge is not accurately called, then the storage node parameters and retrieval logic are readjusted, and the testing process is repeated until the retrieval function meets the usage requirements. After adjusting the storage node parameters and retrieval logic, the updated operation and maintenance technical service big model, based on the hierarchical knowledge architecture, is used as the foundation for the updated operation and maintenance technical service big model.

6. The method for dynamic updating of large-scale model knowledge for operation and maintenance technical services according to claim 1, characterized in that, The process of performing multi-dimensional dynamic adaptation on the basic updated operation and maintenance technical service big model, adjusting the knowledge call priority sequence and storage node parameters in conjunction with real-time changes in the characteristics of operation and maintenance tasks, and generating an adapted updated operation and maintenance technical service big model includes: Real-time data collection of changes in real-time scenario characteristics of operation and maintenance tasks; obtaining change information on task target types from the operation and maintenance task management platform. The system captures the changing trends of equipment operating status parameters, including the magnitude of increase or decrease in parameter values, whether the parameters exceed the warning threshold, and the duration of parameter changes. Receive information on adjustments to the scope of business impact, including an increase in the number of affected business systems, an expansion in the number of affected users, or an extension of the expected duration of business interruption; The real-time scene features of the changed operation and maintenance tasks are encoded to generate the encoded vector of the updated scene features. The encoded vectors of the scene features before and after the update are compared to identify the dimensions that differ. Extract the knowledge call priority adjustment strategy corresponding to the difference dimension from the dynamic coupling rules, update the knowledge call priority sequence in the basic updated operation and maintenance technical service big model according to the adjustment strategy, modify the priority parameters of the corresponding knowledge, and make the adjusted priority sequence consistent with the changed scenario features. Extract the iterative impact assessment results of knowledge whose application frequency exceeds a preset application frequency threshold under the changed scenario characteristics from the knowledge full-link evolution set, as well as the correlation strength between different knowledge, and analyze the requirements of the knowledge on the storage node parameters. Based on the analysis results, adjust the storage node parameters of the hierarchical knowledge architecture, and at the same time adjust the knowledge eviction policy parameters of the storage nodes. Run model adaptation tests to simulate the operation and maintenance task processing flow under the changed scenario, and record the priority order of model knowledge calls, knowledge retrieval response time, accuracy of associated knowledge calls, and task processing efficiency after knowledge application. If the test results show that the knowledge call priority is inconsistent with the adjustment strategy, the response time exceeds the preset response time threshold, the associated call is inaccurate, or the task processing efficiency does not meet the preset efficiency standard, then the knowledge call priority sequence and storage node parameters are readjusted, and the test process is repeated. After completing all adjustments and tests, save the updated knowledge call priority sequence and storage node parameters, and output the large model of the updated operation and maintenance technical service.

7. The method for dynamic updating of large-scale model knowledge for operation and maintenance technical services according to claim 1, characterized in that, After the steps of performing multi-dimensional dynamic adaptation on the basic updated operation and maintenance technical service big model, adjusting the knowledge call priority sequence and storage node parameters in combination with the real-time scenario feature changes of operation and maintenance tasks, and generating the adapted updated operation and maintenance technical service big model, the method further includes: After running and adapting to the updated operation and maintenance technical service big model, it processes actual operation and maintenance tasks, collects knowledge application data and task execution result data during the task processing, and generates a knowledge application effect analysis report. Based on the knowledge application effect analysis report, optimize the matching conditions and linkage logic of the dynamic coupling rules between the scenario and knowledge, and generate rule optimization parameters; The modeling dimensions of the knowledge end-to-end evolution model are adjusted by optimizing the parameters according to the rules, and the evaluation dimensions of the identification type and iterative influence range of the knowledge generation triggering factors are added to obtain the optimized knowledge end-to-end evolution model. The optimized knowledge end-to-end evolution model is used as the basic modeling framework for the next knowledge update. The steps of generating dynamic coupling rules between scenarios and knowledge and generating rule optimization parameters are repeatedly executed to form a knowledge update closed loop. Based on the rule optimization parameters and model adjustment records of multiple knowledge update closed loops, a knowledge update strategy library for the large model of operation and maintenance technical services is generated to guide the knowledge update process under different operation and maintenance scenarios.

8. The method for dynamic updating of large-scale model knowledge for operation and maintenance technical services according to claim 7, characterized in that, The updated and adapted operation and maintenance technical service big model processes actual operation and maintenance tasks, collects knowledge application data and task execution result data during task processing, and generates a knowledge application effect analysis report, including: Different types of actual operation and maintenance tasks are selected, covering equipment fault repair tasks, regular maintenance tasks, system optimization tasks and business recovery tasks. The task types cover the main scenarios in the dynamic coupling rules between scenarios and knowledge. The selected actual operation and maintenance tasks are assigned to the adapted and updated operation and maintenance technical service big model, triggering the model's knowledge call and task processing flow, and recording the complete process of the model processing tasks. Collect knowledge application data, including the knowledge identifiers called by the model, the order of knowledge calls, the call duration of each knowledge, the parameter adjustment records during the knowledge application process, the number of calls to related knowledge, and the feedback information after knowledge application; at the same time, collect task execution result data, extracting the task start time, task completion time, task completion duration, operation steps records during task execution, the time point when the equipment resumes normal operation, and the time point when the business resumes normal operation from the task execution record system; Acquire task execution quality data, including equipment failure repair success rate, stable equipment runtime after maintenance, performance improvement after system optimization, and user satisfaction score after business recovery; Link knowledge application data with task execution result data to establish a mapping relationship between each knowledge application record and the corresponding task execution result, with each application record corresponding to one task execution result; Analyze the correlation between knowledge application data and task execution result data, including the impact of knowledge call order on task completion time, the relationship between knowledge call time and task execution efficiency, and the effect of the number of associated knowledge calls on task success rate; Based on the association analysis results, problems that the knowledge application effect did not meet the preset standards were identified, including the task completion time after knowledge call exceeded the preset time threshold, the number of associated knowledge calls was lower than the preset call number threshold, and the stable running time of the device after knowledge application was shorter than the preset stable time threshold. The application effectiveness indicators for different types of knowledge are statistically analyzed, including knowledge application success rate, knowledge application efficiency, and knowledge association effectiveness. The knowledge application success rate is calculated as the proportion of the number of tasks in which knowledge is successfully applied to the total number of application tasks. The knowledge application efficiency is calculated as the average time for task completion after knowledge application. The knowledge association effectiveness is calculated as the numerical improvement in task quality after the application of associated knowledge. Application effectiveness indicators are categorized and organized according to knowledge type and task type to form a knowledge application effectiveness statistical table. The knowledge application effectiveness statistical table includes knowledge identifier, knowledge type, application task type, application success rate, application efficiency, association effectiveness, and problem description. Based on the statistical table of knowledge application effectiveness and the results of problem analysis, a knowledge application effectiveness analysis report is generated. The content of the knowledge application effectiveness analysis report includes an overview of the overall knowledge application effectiveness, a comparison of the application effectiveness of different types of knowledge, existing problems and their causes.

9. The method for dynamic updating of large-scale model knowledge for operation and maintenance technical services according to claim 7, characterized in that, The matching conditions and linkage logic of the dynamic coupling rules between the optimization scenario and knowledge based on the knowledge application effect analysis report are used to generate rule optimization parameters, including: Extract the knowledge application problems and their causes recorded in the knowledge application effect analysis report. Classify the knowledge application problems according to their causes to form a list of problem types. Problem types include: knowledge adaptation deviation caused by mismatch between matching conditions and actual scenario characteristics; omission of related knowledge calls caused by misalignment between linkage logic and knowledge association; and inefficient knowledge priority calls caused by mismatch between priority sequence and knowledge application effect. To address the problem of knowledge adaptation bias caused by the mismatch between matching conditions and actual scenario characteristics, we analyze the differences between the application scenario characteristics of the problem knowledge and the actual task scenario characteristics, and determine the optimization direction of the matching conditions. If the difference is concentrated in the adaptation range of equipment operating status parameters, the optimization direction is to adjust the parameter range; if the difference is concentrated in the response requirements of the business impact range, the optimization direction is to expand or narrow the response range. To address the issue of missing related knowledge calls due to misalignment between linkage logic and knowledge association, we analyze the cross-knowledge association network of the problem knowledge, identify related knowledge that should have been called but was not, and determine the optimization direction for the linkage logic of related knowledge. If the related knowledge was not called because the association strength value was lower than the preset association strength threshold, the optimization direction is to adjust the association strength value threshold; if the related knowledge was not called because it was not included in the calling order, the optimization direction is to supplement the calling order. To address the issue of inefficient knowledge priority retrieval caused by a mismatch between priority sequences and knowledge application effects, this paper analyzes the reasons for inefficient knowledge priority retrieval and determines the optimization direction for priority sequences. If the reason is that the current scenario characteristics are not considered, the optimization direction is to add scenario feature matching weights. If the reason is that the sorting criteria do not cover key application effect indicators, the optimization direction is to increase the dimensions of the sorting criteria. Based on the optimization direction, specific optimization measures were formulated, including resetting the upper and lower limits of the equipment operation status parameters in the matching conditions, supplementing the knowledge calling link in the related knowledge linkage logic with the knowledge whose association strength value exceeds the preset threshold, and adding scene feature matching degree weight to the priority sequence. The optimization measures are transformed into rule optimization parameters. Each optimization measure corresponds to one or more parameters. Adjusting the equipment operating status parameter adaptation range corresponds to generating upper and lower limit adjustment values ​​of parameters. Supplementing the associated knowledge call link corresponds to generating associated knowledge identifiers and call sequence numbers. Adding scene feature matching degree weight corresponds to generating weight coefficients. For each rule, the parameter optimization parameters are set with a range of values ​​and application scenarios. The range of the upper and lower limit adjustment values ​​of the parameters is set based on the statistical data of the device parameters adapted to the problem knowledge. The application scenarios of the associated knowledge identifier and the call sequence number are set based on the application scenarios of the problem knowledge. The range of the weight coefficient is set to 0 to 1. To verify the effectiveness of the rule optimization parameters, select test tasks similar to the application scenarios of the problem knowledge, apply the optimization parameters to the dynamic coupling rules between the scenarios and knowledge, run the rules to filter knowledge and process test tasks, and record the knowledge adaptation accuracy, related knowledge call completeness rate and task processing efficiency of the test tasks. Comparing the test results before and after optimization, if the improvement rate of knowledge adaptation accuracy exceeds the preset improvement rate threshold, the completeness rate of associated knowledge retrieval reaches the preset completeness rate threshold, and the improvement in task processing efficiency exceeds the preset improvement threshold, then the optimization parameters are deemed effective; if the optimization effect does not reach the corresponding threshold standard, the values ​​of the optimization parameters are readjusted and the verification process is repeated. Collect all valid rule optimization parameters, classify and organize them according to the optimized rule elements, and form a set of rule optimization parameters. Each parameter includes a parameter name, optimization direction, value, application scenario and verification results.

10. A large-scale model knowledge dynamic update system for operation and maintenance technical services, characterized in that, The method includes a processor and a computer-readable storage medium storing machine-executable instructions, which, when executed by the processor, implement the method for dynamically updating large model knowledge for operation and maintenance technical services as described in any one of claims 1-9.

Images