Intelligent agent assisted operation and maintenance technology service remote guidance interaction method and system

Abstract

This application provides a remote guidance and interaction method and system for operation and maintenance (O&M) technical services that combines intelligent agent assistance. It relates to the field of O&M technical services. First, it receives a remote guidance interaction request initiated by O&M personnel, containing information about the O&M object, a description of the O&M problem, and the desired guidance direction. Then, it uses a virtual artificial intelligence agent to identify O&M scenario features, extracts key features, and generates an O&M guidance scheme containing a scenario-adaptive operation guidance sequence and real-time semantic interaction nodes, which is then pushed to the O&M personnel's terminal. It receives operation feedback descriptions in real time, adjusts the guidance scheme through a feedback semantic parsing module, obtains an updated scheme, and pushes it back. Simultaneously, it stores the evolution trajectory of the guidance scheme and operation feedback descriptions during the interaction process, achieving efficient, accurate, and dynamic remote guidance interaction.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and more specifically, to a remote guidance and interaction method and system for operation and maintenance technical services that combines intelligent agent assistance. Background Technology

[0002] In the traditional field of operations and maintenance (O&M) technical services, when O&M personnel encounter complex technical problems and require remote guidance, they typically rely on human experts for communication. However, human expert resources are limited, and they may not be able to respond to O&M personnel's requests in a timely manner due to factors such as location and time constraints. Furthermore, during human guidance, there may be discrepancies in the transmission and understanding of information, leading to low guidance efficiency.

[0003] While some existing remote guidance systems offer some guidance capabilities, they often lack the ability to accurately identify operational scenarios. They cannot automatically extract key features based on the specific information of the object being maintained, the detailed description of the problem, and the desired guidance direction. These features include the types of associated operating parameters, the scope of the fault's impact, and the level of operational complexity matching the desired guidance direction. This results in generated guidance plans lacking specificity and adaptability, failing to adequately meet the needs of actual operational scenarios. Furthermore, during the execution of operations by maintenance personnel, existing systems struggle to dynamically adjust guidance plans based on real-time feedback, hindering efficient and accurate remote guidance interaction. Summary of the Invention

[0004] In view of this, the purpose of this application is to provide a remote guidance and interaction method and system for operation and maintenance technical services that combines intelligent agent assistance.

[0005] According to a first aspect of this application, a remote guidance and interaction method for operation and maintenance technical services combined with intelligent agent assistance is provided, the method comprising:

[0006] Receive remote guidance interaction requests for operation and maintenance technical services initiated by operation and maintenance personnel. The remote guidance interaction requests include operation and maintenance object information, operation and maintenance problem description and expected guidance direction.

[0007] The virtual artificial intelligence agent identifies the operation and maintenance scenario features of the remote guidance interaction request, and extracts the types of operating parameters associated with the operation and maintenance object, the scope of fault impact corresponding to the operation and maintenance problem, and the level of operational complexity matching the expected guidance direction.

[0008] Based on the operation and maintenance scenario feature recognition results, the dynamic guidance strategy generation module of the virtual artificial intelligence agent is invoked to generate an operation and maintenance guidance scheme that includes scenario-adaptive operation guidance sequences and real-time semantic interaction nodes.

[0009] The operation and maintenance guidance plan is pushed to the operation and maintenance personnel's terminal, and the operation feedback description returned by the operation and maintenance personnel after performing the operation based on the operation and maintenance guidance plan is received in real time. The operation feedback description is then input into the feedback semantic parsing module of the virtual artificial intelligence agent.

[0010] By associating the real-time semantic interaction nodes in the operation and maintenance guidance scheme with the feedback semantic parsing module, adjusting the step content and presentation order of the scenario-adaptive operation guidance sequence, an updated operation and maintenance guidance scheme is obtained. The updated operation and maintenance guidance scheme is then pushed to the operation and maintenance personnel's terminal, and the evolution trajectory of the guidance scheme and operation feedback description during this interaction process are stored.

[0011] According to a second aspect of this application, a remote guidance and interaction system for operation and maintenance technical services combined with intelligent agent assistance is provided. The remote guidance and interaction system for operation and maintenance technical services combined with intelligent agent assistance 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 remote guidance and interaction system for operation and maintenance technical services combined with intelligent agent assistance implements the aforementioned remote guidance and interaction method for operation and maintenance technical services combined with intelligent agent assistance.

[0012] 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 remote guidance and interaction method for operation and maintenance technical services combined with intelligent agent assistance is implemented.

[0013] Based on any of the above aspects, the technical effect of this application is as follows:

[0014] By receiving remote guidance interaction requests initiated by operations and maintenance (O&M) personnel, a virtual AI agent comprehensively identifies the O&M scenario features of the requests. This allows for the accurate extraction of key information, such as the types of operating parameters associated with the O&M object, the scope of the fault impact corresponding to the O&M problem, and the operational complexity level matching the desired guidance direction. Based on the scenario feature identification results, an O&M guidance scheme is generated, containing a scenario-adaptive operation guidance sequence and real-time semantic interaction nodes. This effectively guides O&M personnel in solving practical problems. During the operation execution process, real-time operation feedback descriptions are received and linked to real-time semantic interaction nodes through a feedback semantic parsing module. The steps and presentation order of the scenario-adaptive operation guidance sequence are dynamically adjusted to obtain an updated O&M guidance scheme. This achieves real-time optimization of the guidance scheme, improving the efficiency and accuracy of remote guidance. Simultaneously, storing the evolution trajectory of the guidance scheme and operation feedback descriptions during this interaction helps improve the overall O&M technical service level, realizing efficient, accurate, and dynamic remote guidance interaction for O&M technical services. Attached Figure Description

[0015] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 A flowchart illustrating the remote guidance and interaction method for operation and maintenance technical services combined with intelligent agent assistance provided in an embodiment of this application is shown.

[0017] Figure 2 This illustration shows a schematic diagram of the component structure of a remote guidance and interaction system for operation and maintenance technical services that combines intelligent agents to implement the above-described remote guidance and interaction method for operation and maintenance technical services that combines intelligent agents. Detailed Implementation

[0018] The embodiments of this application are described below with reference to the accompanying drawings. It should be understood that the embodiments described below with reference to the accompanying drawings are exemplary descriptions for explaining the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions of the embodiments of this application.

[0019] Those skilled in the art will understand that, unless otherwise stated, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the terms “comprising” and “including” as used in embodiments of this application mean that the corresponding feature can be implemented as the presented feature, information, data, step, operation, element, and / or component, but do not exclude implementation as other features, information, data, step, operation, element, component, and / or combinations thereof supported by the art. It should be understood that when an element is said to be “connected” or “coupled” to another element, the element may be directly connected or coupled to the other element, or it may mean that the element and the other element are connected through an intermediate element. Furthermore, “connected” or “coupled” as used herein may include wireless connection or wireless coupling, and the term “and / or” as used herein indicates at least one of the items defined by the term; for example, “A and / or B” may be implemented as “A,” or as “B,” or as “A and B.”

[0020] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings. The technical solutions of the embodiments of this application and the technical effects produced by the technical solutions of this application will be explained below through the description of several exemplary embodiments. It should be noted that the following embodiments can be referenced, borrowed from, or combined with each other, and the same terms, similar features, and similar implementation steps in different embodiments will not be described again.

[0021] Figure 1 This document illustrates a flowchart of a remote guidance and interaction method and system for operation and maintenance technical services combined with intelligent agent assistance, provided in an embodiment of this application. It should be understood that in other embodiments, the order of some steps in the remote guidance and interaction method for operation and maintenance technical services combined with intelligent agent assistance in this embodiment can be shared according to actual needs, or some steps can be omitted or maintained. The detailed steps of this remote guidance and interaction method for operation and maintenance technical services combined with intelligent agent assistance include:

[0022] Step S110: Receive a remote guidance interaction request for operation and maintenance technical services initiated by operation and maintenance personnel. The remote guidance interaction request includes operation and maintenance object information, operation and maintenance problem description and expected guidance direction.

[0023] This example focuses on the operation and maintenance scenario of smart air conditioners in chain stores. When an on-site engineer performs maintenance on a smart air conditioner at a store, they initiate a remote guidance interaction request through a mobile terminal maintenance application. The maintenance object information includes the air conditioner brand and model, installation date, store location identifier, installation area, and unique hardware code, which are automatically retrieved by the application from the store's equipment database.

[0024] The maintenance issue description should be entered by the engineer in text, including the time of occurrence (e.g., at startup, during operation), specific phenomena (e.g., sudden drop in cooling, error codes, operating noise), preliminary checks already performed (e.g., filter cleanliness, power connection), and environmental changes (e.g., voltage fluctuations, sudden increase in ambient temperature). The requested guidance should clearly specify the type of technical support, such as fault code interpretation, component replacement guidelines, noise troubleshooting steps, or parameter adjustment solutions. The request should be transmitted in encrypted form to the technical center's virtual AI agent server.

[0025] Step S120: The virtual artificial intelligence agent identifies the operation and maintenance scenario features of the remote guidance interaction request, and extracts the operation parameter type associated with the operation and maintenance object, the fault impact range corresponding to the operation and maintenance problem, and the operation complexity level matching the expected guidance direction.

[0026] Upon receiving a request, the virtual AI agent initiates a feature recognition process, first converting unstructured text descriptions into analyzable feature data. It then accesses the equipment feature database based on the maintenance object information, matches it with the model's technical manual to determine key operating parameter types, and filters core parameters based on problem symptoms. It analyzes the related components and store requirements, assessing the impact of the fault on functional modules and operations, such as the risk of cooling issues affecting store environmental comfort and customer experience. Finally, based on an operational complexity assessment model, it determines the complexity level by combining the operation type and technical difficulty.

[0027] Step S121: Input the remote guidance interaction request into the scene feature extraction unit of the virtual artificial intelligence agent, and extract the domain to which the operation and maintenance object belongs, the type of the core component of the operation and maintenance object, and the problem triggering conditions and problem-related components in the operation and maintenance problem description from the operation and maintenance object information.

[0028] The scene feature extraction unit performs multi-level segmentation of the request. It parses the maintenance object information to determine the domain as commercial air conditioning maintenance, further subdividing it into the intelligent inverter air conditioning sub-domain, and then invokes a dedicated feature recognition model. Based on the equipment model and problem symptoms, it extracts the core component types; for example, refrigeration problems are associated with the compressor, condenser, and refrigerant circulation system; abnormal codes are associated with the control board and sensor components; and each component is associated with technical parameters and a fault mode library.

[0029] The problem description is broken down into triggering conditions, including equipment operation phases (start-up, stable operation, shutdown), environmental conditions (temperature, humidity, voltage range), and operational behaviors (parameter adjustment, mode switching). A structured list is formed through keyword matching and semantic understanding. Causal relationships are used to identify associated components, such as associating abnormal noise with fan components and compressor mounting structures, and display abnormalities with display screen components and drive circuits. The breakdown results are stored in a feature vector database.

[0030] Step S122: Call the operation and maintenance domain parameter library associated with the virtual artificial intelligence agent, and match the corresponding operation parameter types that need to be monitored by the operation and maintenance object under normal operation state according to the domain to which the operation and maintenance object belongs and the core component type of the operation and maintenance object.

[0031] The operation and maintenance parameter library stores key parameter standards hierarchically by equipment domain and component type. The intelligent agent performs multi-level searches based on the commercial intelligent air conditioning domain and core component types (compressor, condenser, control system). For compressor components, parameters such as operating current, operating pressure, exhaust temperature, and operating frequency are matched; for condenser components, parameters such as fin temperature, fan speed, and inlet / outlet temperature difference are matched; and for control systems, parameters such as deviation between set and actual temperature, mode switching response time, and sensor feedback values ​​are matched.

[0032] Each parameter type is associated with a monitoring location, normal range, measurement tool requirements, and allowable fluctuation range of operating conditions. Extracted parameters are then matched a second time with the equipment model characteristics, eliminating generic parameters and retaining specific parameters, such as the unique pressure monitoring parameters for a particular brand of compressor, ensuring that the parameter types are practical and effective.

[0033] Step S123: Based on the problem triggering conditions and problem-related components in the operation and maintenance problem description, and combined with the fault impact analysis logic of the virtual artificial intelligence agent, determine the range of adjacent components that may be affected by the operation and maintenance problem and the service interruption risk level, and form the fault impact range corresponding to the operation and maintenance problem.

[0034] The fault impact analysis logic is based on a fault propagation model, which includes component connectivity, functional dependencies, and fault propagation paths. Inputting the problem triggering conditions and associated components determines the initial fault point. Based on physical connections and functional collaboration, the scope of fault propagation to adjacent components is deduced; for example, a compressor fault may affect the condenser and evaporator, while a control board fault may affect sensors and actuators.

[0035] Service interruption risk levels are assessed by combining store operating hours (peak / off-peak), equipment importance level (core area / auxiliary area), and repair difficulty. During peak hours, the risk of air conditioning failure is high in core areas, and low in auxiliary areas during off-peak hours. The scope of impact of a failure is described by combining the range of adjacent components and their risk levels.

[0036] Step S1231: Input the problem triggering conditions and problem-associated components in the operation and maintenance problem description into the fault impact analysis unit of the virtual artificial intelligence agent.

[0037] After receiving the problem triggering conditions and related components, the fault impact analysis unit performs time series analysis on the triggering conditions to determine the duration, frequency characteristics, and periodicity of the problem. Simultaneously, it creates a 3D model of the related components, reconstructing their installation location, connection ports, and physical distance from other components within the overall equipment structure.

[0038] Step S1232: Call the component association graph of the virtual artificial intelligence agent, locate the connection relationship of the problem-related components in the operation and maintenance system according to the problem-related components, and identify the directly connected first-level adjacent components and the indirectly related second-level adjacent components.

[0039] The component association graph stores the connection relationships between components of the device using a node-edge structure. Nodes represent components, and edges represent connection types (e.g., mechanical, electrical, signal connections). The graph is traversed by associating component nodes to extract directly connected first-level adjacent components (e.g., the compressor and refrigerant piping are directly connected). Further traversal through these first-level adjacent component nodes identifies indirectly connected second-level adjacent components (e.g., the condenser connected to the refrigerant piping). Each adjacent component is marked with connection strength and signal transmission direction; higher connection strength indicates a greater probability of fault propagation.

[0040] Step S1233: Analyze the impact of the problem triggering conditions on the functions of the problem-related components, determine whether the functional impact will be transmitted to the first-level adjacent components and the second-level adjacent components through the connection relationship, and determine the range of adjacent components that may be affected.

[0041] Analyze the impact of problem triggering conditions (such as voltage anomalies) on related components (such as the control board) (e.g., signal processing anomalies, unstable power supply). Based on the connection types and strengths in the component association diagram, simulate the transmission paths of functional impacts: electrical connections may transmit voltage anomalies, and signal connections may transmit data errors. Assess the transmission probability of each primary and secondary adjacent component, including components with transmission risks in the potentially affected adjacent component range, and label the degree of impact (e.g., severe impact, minor impact).

[0042] Step S1234: Based on the range of potentially affected adjacent components and the service weight of each component in the operation and maintenance system, calculate the probability of service interruption and the number of affected users, and determine the service interruption risk level.

[0043] Service weights are assigned to potentially affected adjacent components, with weights determined based on the component's functional importance within the system (e.g., core cooling components have a higher weight than auxiliary display components) and the store's operational dependence (e.g., lobby air conditioning components have a higher weight than warehouse air conditioning components). Historical failure data is used to statistically analyze the probability of service interruption caused by such component failures, and this is combined with real-time customer flow data to estimate the number of affected users. The probability of service interruption (high / medium / low) is combined with the number of affected users (many / medium / few) to map to the corresponding service interruption risk level (e.g., extremely high, high, medium, low).

[0044] Step S1235: Combine the range of potentially affected adjacent components with the service interruption risk level to form the fault impact range corresponding to the operation and maintenance problem.

[0045] The potential impact on adjacent components is presented in a structured table format, including component name, impact level, and transmission path description. The service interruption risk level is accompanied by a summary of the calculation basis, explaining the real-time passenger flow data period and historical interruption probability referenced during the assessment. These two sets of data are combined with time-related information (such as the expected duration of impact) to form a complete description of the fault's impact range.

[0046] Step S124: Analyze the operational target description in the desired guidance direction, combine it with the operational complexity evaluation model of the virtual artificial intelligence agent, and determine the operational complexity level matching the desired guidance direction based on the number of components involved in the operation, the correlation of operation steps, and the operation fault tolerance requirements.

[0047] Semantic parsing aims to guide the expression of operational goals, extracting core operational verbs (such as replace, adjust, and detect) and operational objects to clarify the operation type. The operation complexity assessment model evaluates complexity from three dimensions: Component quantity dimension: counting the total number of components touched or adjusted (more components result in a higher base value); Step correlation dimension: analyzing step dependencies and sequence constraints (stronger correlations increase decomposition difficulty); Operation fault tolerance requirement dimension: determining fault tolerance thresholds based on the severity of consequences of errors (equipment damage, safety risks) (higher requirements result in higher complexity).

[0048] The model calculates a weighted overall score, which is then mapped to three complexity levels: Basic, Intermediate, and Expert. The Basic level corresponds to single-step, low-risk operations; the Intermediate level corresponds to multi-step, interconnected operations; and the Expert level corresponds to multi-component collaboration and high-fault-tolerance operations. Each level is associated with different levels of guidance depth and presentation methods.

[0049] Step S125: Summarize the types of operating parameters, the scope of fault impact, and the level of operational complexity to form the operation and maintenance scenario feature identification results.

[0050] The system summarizes the types of operating parameters, the scope of impact of faults, and the level of operational complexity. Operating parameter types are sorted by their importance and their correlation weight with the fault is indicated. The scope of impact of faults lists affected components and their risk levels. The level of operational complexity includes a summary of the assessment criteria. Results are stored in a standardized format, including feature identifiers, feature values, confidence scores (based on data matching and model accuracy), and data source identifiers. These serve as the core input for generating dynamic guidance strategies, ensuring that the guidance plan accurately adapts to the scenario.

[0051] Step S130: Based on the operation and maintenance scenario feature recognition results, call the dynamic guidance strategy generation module of the virtual artificial intelligence agent to generate an operation and maintenance guidance scheme that includes scenario-adaptive operation guidance sequence and real-time semantic interaction nodes.

[0052] After receiving the feature recognition results, the dynamic guidance strategy generation module initiates a multi-dimensional generation process. Operational priorities are determined based on the scope of the fault's impact, with high-risk faults prioritized for emergency handling. The granularity and depth of step decomposition are matched according to the level of operational complexity. Monitoring logic is embedded into the operational steps based on the type of operating parameters, and the step sequence is optimized by referencing historical cases to reduce redundant steps.

[0053] Real-time semantic interaction nodes are set up before and after key steps, including parameter monitoring and confirmation, operation effect feedback, and risk warning nodes, clearly defining the types and formats of feedback information. The final guidance plan includes a list of operation steps, interaction node configurations, supporting documents, and risk warnings, forming a complete guidance system.

[0054] Step S131: Input the operation and maintenance scenario feature recognition result into the dynamic guidance strategy generation module of the virtual artificial intelligence agent, trigger the strategy generation rule engine in the dynamic guidance strategy generation module, and extract the basic guidance framework that matches the type of running parameters, the scope of fault impact, and the level of operation complexity.

[0055] The strategy generation rule engine contains multiple preset rule sets, each corresponding to a specific scenario feature combination. After the feature recognition result is input, the engine uses a feature matching algorithm to select the most matching rule combination and extracts the corresponding basic guidance framework from the guidance framework library. The framework includes the overall structure of the operation process, the logical sequence of core steps, and safety specification modules. For example, the high-complexity operation of a refrigeration system includes a structure of steps for safety depressurization, component inspection, parameter calibration, and functional verification.

[0056] Step S132: Adjust the operation priority order in the basic guidance framework according to the scope of the fault impact, and put the operation steps corresponding to the associated components of multiple core service components whose scope of the fault impact covers the front, and at the same time supplement the risk avoidance descriptions corresponding to the operation of the associated components.

[0057] Analyze the number and importance of core components affected by the fault, and prioritize the steps in the basic guidance framework. Prioritize operations on related components affecting multiple core components; for example, if the fault affects the compressor, condenser, and temperature control system, prioritize refrigerant pressure detection and power safety isolation steps to control the fault's spread. Supplement these prioritization steps with risk avoidance statements, including potential risks (refrigerant leaks, high-voltage electric shock), preventative measures (wearing protective equipment, using specialized tools), and emergency plans (leakage ventilation, electric shock first aid). Transform these statements into concise operational prompts embedded in the step descriptions.

[0058] Step S133: Based on the operation complexity level, match the detail description depth for each operation step in the basic guidance framework. For operation complexity level steps with detailed requirements, add component collaborative operation instructions. For operation complexity level steps with non-detail requirements, simplify redundant descriptions to form a scenario-adaptive operation guidance sequence.

[0059] A pre-defined mapping relationship exists between operational complexity levels and description depth, with higher complexity corresponding to finer-grained descriptions. Description depth parameters are determined for each step to control information density and level of detail. Steps requiring high detail include additional component collaboration specifications, such as operational sequence constraints (open valve A before starting pump B), parameter linkage relationships (adjust parameter X while simultaneously monitoring Y), and collaboration precautions (avoiding operation D while component C is running). Supplementary tool specifications and environmental control standards are also provided.

[0060] The steps requiring low detail have been streamlined, removing duplicate prompts, redundant background information, and irrelevant extended explanations, while retaining the target operation, core actions, and key result metrics. The adjusted steps are arranged sequentially to form a scenario-adaptive operation guidance sequence.

[0061] Step S1331: Extract all operation steps in the basic guidance framework. Each operation step includes the operation object, operation action, and operation goal.

[0062] The process iterates through the step list of the basic guidance framework, analyzing the three key elements of each operation step: The object of operation clearly defines the specific component or parameter (e.g., "compressor power switch," "refrigerant pressure value"); the operation action defines the specific execution behavior (e.g., "shut down," "adjust," "detect"); and the operation goal describes the desired result (e.g., "cut off power," "pressure stabilizes within normal range," "confirm no leaks"). Steps lacking these three elements are supplemented and improved to ensure the operational intent of each step is clear and unambiguous.

[0063] Step S1332: Associate the operation complexity level with the description depth matching rule of the virtual artificial intelligence agent to determine the detailed description dimension corresponding to different operation complexity levels. When the operation complexity level corresponds to detailed requirements, it includes component collaboration logic, operation timing requirements, and parameter adjustment gradient. When the operation complexity level corresponds to non-detail requirements, it only retains the core operation actions and operation objectives.

[0064] The description of the deep matching rules defines the mapping relationship between complexity levels and detail dimensions: Expert level (high detail requirements) corresponds to component collaboration logic (cooperation relationships between multiple components), operation timing requirements (time intervals and order of step execution), and parameter adjustment gradients (the magnitude and range of each adjustment); Advanced level corresponds to some detail dimensions; Basic level (non-detail requirements) only retains the core operation actions and operation objectives. Based on the current operation complexity level, determine the detail description dimensions that each operation step should include.

[0065] Step S1333: For the operation steps corresponding to the detailed requirements of the operation complexity level, supplement the collaborative triggering conditions between multiple components involved in the operation step, the sequential order of operation execution, and the gradual adjustment gradient description of key parameters, and improve the details of the operation steps.

[0066] For expert-level complex operating procedures, supplementary component coordination triggering conditions are provided (e.g., "The compressor can only be turned on after the condenser fan has started and been running stably for 30 seconds"); the sequence of operation execution is clearly defined (e.g., "First step: close the high-pressure valve; second step: close the low-pressure valve; third step: disconnect the main power supply"); and the gradual adjustment gradient of key parameters is refined (e.g., "Adjust the temperature setpoint in 1-degree Celsius increments, observe for 5 minutes after adjustment before proceeding with the next adjustment"). Supplementary content references the equipment technical manual and expert operating experience to ensure the accuracy and operability of the detailed descriptions.

[0067] Step S1334: For operation steps that do not require detailed information, delete duplicate operation target descriptions, redundant safety prompts, and background descriptions that are irrelevant to the operation actions, and simplify the description of operation steps for operation steps corresponding to the operation complexity level.

[0068] For basic-level complex operation steps, check and remove recurring operation objectives (such as repeatedly mentioning "ensure the equipment is powered off" in the same step); streamline redundant safety prompts (e.g., retain only the core prompt "confirm power off before operation," and remove repeated warnings about the risk of electric shock); remove background explanations unrelated to the operation (such as introductions to the equipment's working principle). The simplified step descriptions should maintain concise language, logical coherence, and highlight core operational information.

[0069] Step S1335: Integrate, supplement, improve and simplify all operation steps according to the initial order of the basic guidance framework to form a scenario-adaptive operation guidance sequence.

[0070] The simplified and detailed operation steps are arranged in the original logical order of the basic guidance framework. The connection between steps is checked to ensure that the operation objectives of the preceding steps provide the necessary conditions for the subsequent steps. The process sequence is simulated to verify the rationality of the operation order. If logical conflicts exist, local adjustments are made. The final scenario-adaptive operation guidance sequence includes step numbers, operation objects, operation actions, operation objectives, and corresponding detailed descriptions, fully presenting the operation process.

[0071] Step S134: In the scenario-adaptive operation guidance sequence, select key steps involving monitoring operating parameters, confirming component status, and verifying operation results, and set real-time semantic interaction nodes. Each real-time semantic interaction node contains the core information type that needs to be fed back by the operation and maintenance personnel and the timing of the feedback trigger.

[0072] Traverse the scenario-adaptive operation guidance sequence and identify the key step types: operation parameter monitoring steps (such as detecting refrigerant pressure and measuring current values), component status confirmation steps (such as checking component appearance and verifying connection tightness), and operation result verification steps (such as evaluating trial operation effects and observing stability after parameter adjustments). These steps directly affect the correctness of subsequent operations and require the establishment of real-time semantic interaction nodes.

[0073] Each real-time semantic interaction node has clearly defined core information type requirements. For example, the operation parameter monitoring node needs to report specific parameter measurement values ​​and measurement tool models; the component status confirmation node needs to report the component's appearance characteristics, connection status, and functional indicators; and the operation result verification node needs to report changes in phenomena, abnormal situations, and suggestions for the next operation after the operation.

[0074] Feedback triggering timing is set according to the nature of the step, including pre-step preparation confirmation (e.g., "Please report tool calibration status before starting measurement"), key nodes during step execution (e.g., "Please report current reading after pressure stabilizes"), and post-step result feedback (e.g., "Please report equipment response status after operation is completed"). Real-time semantic interaction nodes are associated with corresponding operation steps through unique identifiers to ensure that interaction information accurately corresponds to the operation process.

[0075] Step S135: Integrate the scenario-adaptive operation guidance sequence with the real-time semantic interaction nodes, add association identifiers between interaction nodes and operation steps, and generate an operation and maintenance guidance scheme.

[0076] The scenario-adaptive operation guidance sequence and real-time semantic interaction nodes are integrated according to time sequence and logical relationship to form a unified operation and maintenance guidance scheme structure. Each real-time semantic interaction node is bound to the corresponding operation step through an association identifier, which includes the step number, node type, and interaction sequence number, ensuring accurate location of the node during guidance execution.

[0077] The operation and maintenance guidance plan also includes solution metadata, such as the solution generation time, applicable equipment models, solution version number, associated fault type tags, and referenced technical document numbers. This metadata facilitates solution management and traceability. Additionally, the plan incorporates a format conversion module that automatically adjusts the content layout based on the display characteristics of the operation and maintenance personnel's terminals, ensuring clear presentation across different devices.

[0078] The generated operation and maintenance guidance plan undergoes a compliance check to verify that its operational steps comply with security specifications and technical standards, and that the settings of interaction nodes cover all critical aspects. Once the check is passed, the plan is stored in encrypted format and a unique plan identifier is generated for subsequent updates and tracking.

[0079] Step S140: Push the operation and maintenance guidance plan to the operation and maintenance personnel's terminal, receive the operation feedback description returned by the operation and maintenance personnel after performing the operation based on the operation and maintenance guidance plan in real time, and input the operation feedback description into the feedback semantic parsing module of the virtual artificial intelligence agent.

[0080] The virtual AI agent's solution delivery unit sends the operation and maintenance guidance solution to the operation and maintenance personnel's mobile terminal via an encrypted transmission channel. During transmission, data fragmentation and verification mechanisms are employed to ensure the integrity and accuracy of the solution content. Upon receiving the operation and maintenance guidance solution, the terminal application automatically parses the solution structure and displays a sequence of scenario-adaptive operation instructions on the interface, step by step. It also marks the location of real-time semantic interaction nodes next to each step, such as displaying a "Feedback Required" label after the "Check Refrigerant Pressure" step.

[0081] When maintenance personnel perform each operation step, the terminal application provides prompts based on the trigger conditions of real-time semantic interaction nodes. For example, when the step reaches the node "Please provide feedback on the current reading after the pressure stabilizes," a feedback input box pops up on the interface, supporting multiple feedback methods such as text input, voice input, and image upload. The operation feedback description entered by the maintenance personnel includes the actual situation of the operation execution (e.g., "Pressure gauge has been connected according to the steps"), the observed component status (e.g., "Pressure pointer is stable in a certain range"), the measured parameter values ​​(e.g., "Current pressure value is in a certain state"), and any abnormal phenomena encountered (e.g., "Pressure gauge reading continues to fluctuate").

[0082] The terminal application performs real-time preprocessing on the input operation feedback description, including format standardization (unifying technical terminology) and completeness verification (checking whether it contains the core information required by the node). If any information is missing, it prompts the maintenance personnel to supplement it. The verified operation feedback description is sent to the feedback semantic parsing module of the virtual AI agent via an instant messaging protocol. During transmission, the feedback timestamp and corresponding node identifier are recorded to ensure that the feedback information is accurately associated with the operation steps. After receiving the operation feedback description, the feedback semantic parsing module stores it in a temporary data buffer, awaiting further semantic analysis processing.

[0083] Step S141: Through the interactive push unit of the virtual artificial intelligence agent, the operation step description and the feedback prompts of the corresponding real-time semantic interaction node are synchronously pushed to the operation and maintenance personnel terminal according to the step sequence of the scenario-adaptive operation guidance sequence in the operation and maintenance guidance scheme.

[0084] The interactive push unit generates a push queue based on the step sequence of the operation and maintenance guidance plan, and pushes content sequentially according to the operation progress of the operation and maintenance personnel. When pushing a description of each operation step, it includes the operation object, core actions, precautions, and reference diagrams. For example, when pushing the step of "turn off the compressor power", it includes a diagram of the power switch location and an emphasis on the core action of "confirming that the switch is off".

[0085] When a step involving real-time semantic interaction is pushed to a specific point, the interactive push unit simultaneously sends a feedback prompt, clearly informing the maintenance personnel of the type and specific requirements of the information to be reported, such as "Please report the current refrigerant pressure measurement value and the model of the measuring tool." The feedback prompt is displayed in a prominent pop-up window, including a feedback deadline (set according to the complexity of the operation, such as within 5 minutes) and an example format (such as "Pressure value: certain status, tool: digital pressure gauge"), guiding the maintenance personnel to provide standardized feedback.

[0086] The push unit monitors the terminal's reception status in real time. If no confirmation is received from the terminal, it re-pushes the message after 30 seconds to ensure that maintenance personnel can obtain the operation instructions in a timely manner. Simultaneously, the push unit records the push time, terminal reception time, and maintenance personnel's viewing status for each step, forming a push trajectory log.

[0087] Step S142: When each real-time semantic interaction node is triggered, the feedback receiving channel of the virtual artificial intelligence agent is started to receive the operation execution status, component response phenomena and operation questions input by the operation and maintenance personnel through the terminal, and form an operation feedback description.

[0088] Once the real-time semantic interaction node is triggered, the feedback receiving channel is automatically activated, establishing a two-way communication link between the terminal and the virtual AI agent. Maintenance personnel input the operation execution status through the terminal, including the completion status of each step (e.g., "Completed," "In Progress," "Not Executed"), the model of the tool used during execution (e.g., "Multimeter model is [model number]," "Pressure gauge range is [range]"), and the operation time (e.g., "3 minutes elapsed from connecting the tool to obtaining the reading").

[0089] The feedback on component response phenomena includes detailed descriptions of changes in the component's physical state (e.g., "no leakage sound after valve closure" and "smooth operation after fan startup"), indicator light display status (e.g., "power indicator light is constantly on" and "fault indicator light is flashing"), and abnormal behaviors (e.g., "slight frost at pipe connections" and "abnormal noise from the motor"). Operational questions are entered by maintenance personnel expressing any confusion encountered during the execution of procedures, such as "whether it is necessary to immediately release pressure if the pressure value is below the normal range" or "how to handle situations where the tool's displayed value does not match the standard unit."

[0090] The feedback receiving channel supports multi-round input, allowing maintenance personnel to supplement feedback multiple times at the same node. The terminal application automatically integrates these multiple inputs into a complete operation feedback description and marks each supplement with a timestamp. After the feedback content is entered, the maintenance personnel click the "Submit" button. The terminal application then encrypts the operation feedback description and sends it to the feedback receiving channel. Once the channel confirms receipt, it returns a successful receipt receipt to the terminal.

[0091] Step S143: Transmit the operation feedback description in real time so that the operation feedback description is sent to the feedback semantic parsing module of the virtual artificial intelligence agent immediately after the operation and maintenance personnel complete the input.

[0092] After the operation feedback description is submitted, the terminal application encapsulates it, adding metadata such as the terminal device identifier, the maintenance personnel's identity identifier, the corresponding real-time semantic interaction node identifier, and the feedback timestamp, forming a complete feedback data packet. The data packet is encrypted using a symmetric encryption algorithm and sent to the feedback receiving server of the virtual artificial intelligence agent through a dedicated communication port.

[0093] A breakpoint resumption mechanism is employed during transmission. If a network interruption causes transmission failure, the terminal application will automatically resend the unsuccessfully transmitted data packets once the network is restored. Upon receiving the data packet, the feedback receiving server first decrypts and verifies its integrity. After successful verification, it forwards the data packet to the feedback semantic parsing module and returns confirmation information to the terminal, including the data packet reception time and verification result.

[0094] After receiving the data packet, the feedback semantic parsing module extracts the operation feedback description and associated metadata, stores them in the parsing buffer, and records the data reception log, including reception time, data size, and source identifier, to ensure the traceability of the feedback data. The parsing module queues the operation feedback descriptions in the order they are received, awaiting semantic word segmentation processing.

[0095] Step S144: After receiving the operation feedback description, the feedback semantic parsing module starts semantic word segmentation processing to separate the factual and interrogative statements in the operation feedback description and mark them as factual feedback fragments and interrogative feedback fragments, respectively.

[0096] Upon receiving the operation feedback description, the feedback semantic parsing module immediately activates its built-in semantic processing engine. This engine comprises multiple functional units, including word segmentation, matching, and verification, which work together to complete the semantic parsing task. First, the original text is standardized to remove interfering information. Then, a professional thesaurus is used to distinguish between factual and interrogative statements. Finally, the completeness of the segmentation results is verified to ensure that subsequent analysis is based on accurate feedback information.

[0097] Step S1441: Call the semantic word segmentation model of the virtual artificial intelligence agent to perform word segmentation on the operation feedback description to obtain multiple semantic word units.

[0098] The semantic word segmentation model loads a dedicated word segmentation dictionary for the operations and maintenance (O&M) domain. This dictionary covers core domain vocabulary such as equipment component terms, operational action vocabulary, parameter type vocabulary, and fault phenomenon vocabulary. The operational feedback description is scanned sentence by sentence, and a bidirectional maximum matching algorithm is used for word segmentation, matching the longest word sequence starting from both the beginning and end of the text. For example, "compressor pressure continuously drops and is accompanied by abnormal noise" is segmented into multiple semantic word units: "compressor / running / pressure / continuously dropping / and / accompanied / abnormal noise". Each word unit is assigned a part-of-speech tag (such as "noun", "verb", "adjective", "fault word"), and its start and end positions in the original text are recorded for easy tracing of the word's origin.

[0099] Step S1442: Match the semantic vocabulary unit with the factual statement lexicon of the virtual artificial intelligence agent, filter out vocabulary units containing factual information such as operation results, component status, and parameter values, and combine them to form factual feedback fragments.

[0100] The factual statement lexicon is stored in a hierarchical structure. The base layer contains parameter value vocabulary and unit vocabulary, the middle layer contains state description vocabulary and component state vocabulary, and the top layer covers operation result vocabulary. The matching engine calculates the similarity between each semantic word unit and the factual statement lexicon, uses the edit distance algorithm to quantify the word differences, and selects word units with a matching degree higher than a preset threshold as candidate factual words. Dependency parsing is used to identify semantic relationships between words, such as subject-verb and verb-object relationships, combining "pressure / continuously decreasing" into the factual feedback fragment "pressure continuously decreasing", and combining "compressor / operating normally" into the factual feedback fragment "compressor operating normally". Each factual feedback fragment is labeled with its corresponding information type, such as parameter change, component state, and operation result.

[0101] Step S1443: Match the remaining semantic vocabulary units with the question expression vocabulary of the virtual artificial intelligence agent, filter out vocabulary units containing question particles, uncertain expressions, and descriptions of operational confusion, and combine them to form question feedback fragments.

[0102] The remaining semantic lexical units from the unmatched factual statement lexicon enter the question statement lexicon matching process. The question statement lexicon is stored categorized by question type, including subsets of question auxiliary words, subsets of uncertain expressions, and subsets of requests for help. The relevance between words and the question statement lexicon is determined through keyword matching and semantic vector calculation. Lexical units containing question auxiliary words such as "how," "whether," and "why," uncertain expressions such as "maybe" and "suspected," and descriptions of operational confusion such as "don't know how to operate" and "request guidance" are selected. Semantic role labeling technology is used to extract the core elements of the question, combining "How to handle a continuous decrease in stress" into the question feedback fragment "Consultation on methods for handling a continuous decrease in stress," clarifying the operational scenario and core demand of the question.

[0103] Step S1444: Perform a completeness check on the fact feedback segment, check whether it contains the core result information corresponding to the operation steps. If it is missing, generate a supplementary inquiry prompt and push it to the operation and maintenance personnel's terminal. After receiving the supplementary feedback, complete the fact feedback segment.

[0104] The integrity assessment unit checks whether the factual feedback fragment contains necessary key elements based on the core information requirements corresponding to the real-time semantic interaction nodes. For example, the factual feedback fragment of the operating parameter monitoring node must include information such as parameter name, measured value, and measuring tool; the component status confirmation node must include information such as component name, appearance characteristics, and function indicators. If the check finds that the factual feedback fragment is missing core information, such as only reporting "pressure abnormality" without specifying the value, a standardized supplementary inquiry prompt is generated, such as "Please supplement the specific measured value of the current refrigerant pressure," and sent to the maintenance personnel's terminal through the interactive push unit. After receiving the supplementary feedback from the maintenance personnel, the added information is integrated into the original factual feedback fragment to ensure the completeness of the fragment information.

[0105] Step S1445: Extract the core question from the question feedback segment, determine the specific points of confusion for the operation and maintenance personnel regarding the operation steps, form the core identifier of the question feedback segment, and associate it with the corresponding real-time semantic interaction node identifier.

[0106] The core question extraction unit uses semantic understanding technology to analyze question feedback fragments, identifying the type of question and key needs. Through keyword extraction and syntactic analysis, the core point of confusion in the question "The pressure is still unstable after adjusting the expansion valve; I don't know whether to continue adjusting it" is extracted as "the subsequent operation judgment for unstable pressure after adjusting the expansion valve". A core identifier is generated for each question feedback fragment, including the question type (such as operation method type, anomaly judgment type, parameter standard type), the involved operation object, and a description of the core confusion. The core identifier is bound to the corresponding real-time semantic interaction node identifier to ensure that the question feedback can be accurately associated with specific operation steps and interaction nodes.

[0107] Step S145: Associate the factual feedback fragment with the question feedback fragment and the corresponding real-time semantic interaction node identifier to form the initial input data for feedback semantic parsing.

[0108] The feedback semantic parsing module extracts the real-time semantic interaction node identifiers associated with the operation feedback descriptions and binds them to factual feedback fragments and question feedback fragments respectively, ensuring that each feedback fragment accurately corresponds to a specific node location in the operation and maintenance guidance plan. During the binding process, information such as fragment number, fragment type, and step number of the associated node are added to form structured associated data.

[0109] The initial input data is organized in the form of data frames. Each data frame contains a list of factual feedback snippets, a list of question feedback snippets, a node identifier, a feedback timestamp, and an operations and maintenance personnel identifier. The module performs an integrity check on the data frames to ensure that all necessary field information is included. If any is missing, it automatically adds default values ​​or marks it as abnormal data.

[0110] The initial input data that passes the inspection is stored in the parsing results database, and a unique parsing task identifier is generated to track subsequent parsing processes. The module also pushes the initial input data to the feedback analysis unit, triggering a comparative analysis of the factual feedback segments and the expected results, as well as the extraction of the core questions from the questioning feedback segments.

[0111] Step S150: By associating the real-time semantic interaction nodes in the operation and maintenance guidance scheme through the feedback semantic parsing module, the step content and presentation order of the scenario-adaptive operation guidance sequence are adjusted to obtain the updated operation and maintenance guidance scheme. The updated operation and maintenance guidance scheme is then pushed to the operation and maintenance personnel's terminal, and the evolution trajectory of the guidance scheme and operation feedback description during this interaction process are stored.

[0112] The feedback semantic parsing module associates and matches factual feedback fragments and question feedback fragments in the initial input data with the corresponding real-time semantic interaction nodes, locating the corresponding step of that node in the scenario-adaptive operation guidance sequence. The analysis unit calls the expected result database, extracts the expected result description of this step, compares it with the factual feedback fragment, and determines whether the operation execution result meets expectations.

[0113] If the actual feedback segment matches the expected result, the subsequent steps remain unchanged; if there is a discrepancy (such as abnormal parameters or failure to achieve the target), the description of the abnormal result is extracted, and the anomaly handling strategy library is used to generate operation adjustment suggestions. For the core points of confusion in the question feedback segment, the question answering knowledge base is used to match the corresponding answers, which are then incorporated into the adjusted step description.

[0114] Based on the operational adjustment suggestions and answers, the module modifies the corresponding steps in the scenario-adaptive operational guidance sequence, such as adding anomaly handling sub-steps and adjusting parameter monitoring standards, and reorders subsequent steps according to the scope of the anomaly's impact. After the correction is completed, an updated operation and maintenance guidance plan is generated and sent to the operation and maintenance personnel's terminals via the push unit, and the updated plan description is displayed on the terminal interface.

[0115] Simultaneously, the module records the initial guidance plan, the updated guidance plan, the basis for each adjustment, and the corresponding operation feedback descriptions during this interaction process, forming a guidance plan evolution trajectory in chronological order. This evolution trajectory, along with all operation feedback descriptions, is stored in the interaction archive, along with metadata such as the operation and maintenance object identifier and the start and end times of the interaction.

[0116] Step S151: Extract the factual feedback fragments, question feedback fragments, and associated real-time semantic interaction node identifiers corresponding to the operation feedback description through the feedback semantic parsing module.

[0117] The extraction unit of the feedback semantic parsing module reads factual feedback fragments, question feedback fragments, and associated real-time semantic interaction node identifiers from the initial input data. It performs secondary verification on the fragment content to ensure the completeness and semantic coherence of the fragments. If a fragment is found to have semantic breaks or missing key information during the verification process, it is automatically marked as a low-confidence fragment, and the missing type is recorded (such as missing parameter values ​​or incomplete state descriptions).

[0118] The extraction unit categorizes high-confidence factual feedback fragments by information type, such as parameter measurement, status observation, and operation execution, and associates each type of fragment with corresponding feature tags (such as "pressure parameter," "leakage status," and "switch operation"). Question feedback fragments are categorized by question type, such as operation method, anomaly judgment, and tool usage, to facilitate subsequent question answer matching.

[0119] The extraction unit performs a secondary binding between the categorized segments and real-time semantic interaction node identifiers, generating a segment-node association table containing information such as segment number, node identifier, segment type, feature label, and confidence score. This segment-node association table serves as the foundational data for subsequent steps of localization and content adjustment and is stored in a temporary analysis database.

[0120] Step S152: Match the real-time semantic interaction node identifier with the real-time semantic interaction node in the operation and maintenance guidance scheme, and locate the target operation step in the scenario-adaptive operation guidance sequence corresponding to the real-time semantic interaction node.

[0121] The matching unit receives the real-time semantic interaction node identifiers from the fragment-node association table, traverses the list of real-time semantic interaction nodes in the operation and maintenance guidance scheme, and finds the corresponding node record by comparing the identifiers. The node record contains information such as node number, step number, node type, and associated operation target. The matching unit locates the specific step in the scenario-adaptive operation guidance sequence, i.e., the target operation step, based on the step number to which the node belongs.

[0122] After location is established, the matching unit extracts detailed information about the target operation step, including the step number, the object of the operation, the operation action, the expected result description, and the associated running parameter types, forming a target step information table. If node identification fails to match (e.g., incorrect identification or the node has been deleted), a matching exception log is generated, recording the error identification and the reason for the matching failure, and triggering manual intervention to ensure the accuracy of step location.

[0123] After the target step information table is generated, it is sent to the analysis unit as benchmark data for comparing the factual feedback segments with the expected results. At the same time, it is linked to the answer matching process of the question feedback segments to ensure that the adjustment and answer content accurately corresponds to the target steps.

[0124] Step S153: Analyze whether the factual feedback segment is consistent with the expected result description of the target operation step. If they are inconsistent, extract the abnormal result description from the factual feedback segment, call the abnormal response strategy library of the virtual artificial intelligence agent, and generate operation adjustment suggestions for the abnormal result description.

[0125] In this step, the first step is to accurately locate the expected result description of the target operation step, and then compare the consistency with the factual feedback segments based on this. The expected result description usually includes multi-dimensional verification indicators, such as the range of component operating status parameters, the physical phenomena characteristics after the operation is executed, and the indicative manifestations of functional recovery. Taking the refrigerant pressure adjustment step of the smart air conditioner in a chain store as an example, the expected result description might be "After adjustment, the low-pressure side pressure stabilizes in the standard range, the pressure gauge pointer fluctuation does not exceed the allowable range, and the air conditioner operating noise is reduced to the normal decibel range."

[0126] The specific statements in the factual feedback segment are compared point by point with the expected result description. The comparison dimensions include the degree of matching of parameter values, the degree of consistency of phenomenon characteristics, and the stability of time series. If the factual feedback segment contains content that is inconsistent with the expectation, such as "after pressure adjustment, it continued to drop below the lower limit of the standard range" or "the pointer fluctuated violently for more than the specified duration", it is judged as an inconsistent result, triggering the anomaly handling process.

[0127] Step S1531: Extract the expected result description of the target operation steps in the operation and maintenance guidance plan, and determine the component status parameters, operating parameter range and operation completion identifier included in the expected result description.

[0128] The core elements of the expected results are extracted from the target operation steps described in the operation and maintenance guidance plan. Component status parameters include the physical state of the components (e.g., "no leakage after valve closure"), indicator light display status (e.g., "operation indicator light is constantly on without flashing"), and mechanical action response (e.g., "fan rotates smoothly without jamming after startup"). The operating parameter range clearly defines the normal range of key indicators, such as "the difference between return air temperature and set temperature does not exceed 2 degrees Celsius" and "compressor operating current remains within the rated range." Operation completion indicators are clear end signals, such as "the buzzer sounds an alarm" and "the display screen returns to the normal operation interface," which are directly observable results. These elements are stored through structured fields to form an expected result verification checklist.

[0129] Step S1532: Compare the operation execution status and component response phenomena in the fact feedback segment with the expected result description point by point, mark the inconsistent differences, and determine the abnormal result description.

[0130] The feedback segments are structured and analyzed to extract the operation execution status (e.g., "three adjustments have been completed according to steps"), component response phenomena (e.g., "pressure continues to drop" "abnormal vibration occurs"), and related parameter data (e.g., "current pressure value, fluctuation frequency"). A point-by-point comparison is performed according to the expected result verification checklist, such as comparing the actual pressure value with the standard range and comparing the noise description with normal characteristics. Items with deviations are marked as discrepancies, such as "pressure value is 20% below the lower limit" or "continuous fluctuation duration exceeds expectations by 3 times." These discrepancies are summarized to form an abnormal result description, ensuring that each anomaly is supported by a specific description.

[0131] Step S1533: Send a query request to the anomaly response strategy library of the virtual artificial intelligence agent, so that the historical response strategies corresponding to the anomaly result description and target operation step identifier are matched based on the query request, and the parameter monitoring requirements are supplemented by the operation parameter type to form an initial response suggestion. The query request includes the anomaly result description, the target operation step identifier, and the operation parameter type in the operation and maintenance scenario feature identification result.

[0132] The anomaly response strategy library employs a three-dimensional index structure of "phenomenon-step-strategy," requiring queries to include complete search elements. The anomaly result description serves as the core search term, such as "refrigerant pressure continuously decreasing"; the target operation step clearly identifies the operation type, such as "expansion valve adjustment step"; and the operating parameter type limits the associated monitoring indicators, such as "low-pressure side pressure, compressor current." The system matches response strategies for similar or identical scenarios from historical cases within the library, such as operational suggestions like "gradually close the shut-off valve and observe pressure changes" and "check valve sealing status." Simultaneously, it supplements targeted monitoring requirements based on the operating parameter type, such as "record the pressure value every 1 / 4 turn of adjustment" and "simultaneously monitor the compressor discharge temperature," forming initial response suggestions that include operational actions, monitoring points, and judgment conditions.

[0133] Step S1534: Perform scenario adaptability analysis on the initial response suggestions to ensure that the operation adjustment content in the initial response suggestions conforms to the fault impact range and operation complexity level of the current operation and maintenance scenario, and generate the final operation adjustment suggestions.

[0134] The scenario adaptability analysis is conducted from two dimensions: the scope of the fault's impact and the complexity of the operation. Regarding the scope of the fault's impact, if the anomaly may affect multiple core components (e.g., a pressure anomaly may affect the compressor and condenser), the initial recommendations emphasize safe isolation procedures, such as "turn off the main power supply before inspecting the components." If the impact is limited to a single component, the focus is on precise adjustment recommendations. Based on the level of operational complexity, detailed tool specifications and coordination requirements are added to expert-level operation steps, such as "use a dedicated torque wrench to tighten to the specified torque." Basic-level steps are simplified, retaining only the core actions. During the analysis, generic strategies that are incompatible with the current scenario are eliminated, and scenario-specific constraints are added, such as "avoid prolonged downtime during store operations and adopt a segmented adjustment method," ultimately forming operational adjustment recommendations adapted to the current scenario.

[0135] Step S154: For the core points of confusion in the question feedback segment, call the question answering knowledge base of the virtual artificial intelligence agent, match the answer statements related to the target operation steps, and integrate them into the adjusted operation step description.

[0136] The question analysis unit extracts the core points of confusion from the question feedback fragments. Using semantic understanding technology, it identifies the core needs of the question. For example, the core point of confusion in the question "Does the pressure value need to be released immediately if it is below the normal range?" is "Determining the appropriate time to handle abnormal pressure." After extracting the core points of confusion, a query request containing the target operation step identifier, keywords related to the points of confusion, and operational scenario characteristics is generated and sent to the question answering knowledge base.

[0137] The Q&A knowledge base employs a multi-level index structure, categorizing and storing answers according to dimensions such as operation step type, equipment component, and fault type. Upon receiving a query request, the knowledge base retrieves answers highly relevant to the core points of confusion through keyword matching and scenario similarity calculation, such as "When the pressure is below a certain lower limit and continues to fluctuate, the intake valve should be closed first, and the pressure should be released after it stabilizes."

[0138] The retrieved solutions undergo scenario adaptability verification to ensure that the solutions match the current maintenance scenario's equipment model, operational complexity level, and fault impact range. Once verified, the solutions are formatted into natural language descriptions and added to the adjusted content of the target operation steps. For example, after the "Check refrigerant pressure" step, a supplementary explanation could be added: "If the pressure is below the normal range, operation suggestion: First close the intake valve..."

[0139] Step S155: Based on the operation adjustment suggestions and answers, revise the content of the target operation steps in the scenario-adaptive operation guidance sequence, and adjust the presentation order of subsequent operation steps based on the impact range corresponding to the abnormal result description, to form an updated operation and maintenance guidance plan.

[0140] The solution correction unit receives operation adjustment suggestions and explanations. First, it corrects the content of the target operation steps in the scenario-adaptive operation guidance sequence, such as updating the operation action description (changing "record pressure value" to "record pressure value, and perform the operation if it is lower than the normal range"), adding abnormal handling sub-steps (such as "close the intake valve" and "wait for the pressure to stabilize"), and adding precautions (such as "open the valve slowly when depressurizing to avoid a sudden drop in pressure").

[0141] Simultaneously, the solution correction unit analyzes the impact range corresponding to the abnormal results and assesses the impact of the abnormality on the execution conditions and risk level of subsequent steps. If the impact range of the abnormality is small (e.g., only the parameters of the current step need to be adjusted), the order of subsequent steps remains unchanged; if the abnormality may affect multiple subsequent steps (e.g., components need to be replaced to continue), the subsequent steps are reordered, and the relevant component inspection and replacement steps are moved to the forefront to ensure the rationality and safety of the operation process.

[0142] After the corrections are completed, the solution correction unit updates the step numbers and associated identifiers of the scenario-adaptive operation guidance sequence, and regenerates the location information of the real-time semantic interaction nodes to ensure that the nodes accurately correspond to the adjusted steps. The final updated operation and maintenance guidance solution undergoes integrity verification and compliance checks. Once confirmed to be error-free, it is marked as the official version and ready to be pushed to the operation and maintenance personnel's terminals.

[0143] Step S156: Through the solution push unit of the virtual artificial intelligence agent, the updated operation and maintenance guidance solution is pushed to the operation and maintenance personnel terminal in the order of the adjusted scenario-adaptive operation guidance sequence. At the same time, the solution update identifier and adjustment instructions are displayed on the terminal interface of the operation and maintenance personnel terminal.

[0144] The solution push unit receives the updated operation and maintenance guidance solution and generates a solution update notification, which includes the update time, the reason for the adjustment (e.g., "adjustment based on pressure anomaly feedback"), and a summary of the main adjustments (e.g., "adding pressure anomaly handling steps"). The notification and the updated guidance solution are pushed to the operation and maintenance personnel's terminals through an encrypted channel. Incremental transmission technology is used during the push process, sending only the content that differs from the previous version to reduce the amount of data transmitted.

[0145] After receiving the information, the maintenance personnel's terminal displays an updated solution identifier (such as a red "Update" label) at the top of the interface, and a pop-up adjustment explanation window shows the reason for the update and the main adjustments. The terminal application automatically replaces the original guidance solution content, displays the scenario-adaptive operation guidance sequence in the adjusted order, highlights newly added or modified steps (such as a yellow background), and marks "Added" or "Modified" next to the steps.

[0146] The terminal application records solution update logs, including update time, version number, summary of adjustments, and receipt status. It also supports operations and maintenance personnel to view historical versions of the solution. The version switching button allows for comparison of the differences in steps between different versions, ensuring that operations and maintenance personnel understand the specific content of the solution adjustments.

[0147] Step S157: Receive confirmation information from the operations and maintenance personnel regarding the updated operations and maintenance guidance plan. If the confirmation is made, record the generation time, adjustment basis, and version identifier of the updated operations and maintenance guidance plan.

[0148] The terminal interface displays the updated maintenance guidance plan and provides two operation buttons: "Confirm Execution" and "Feedback Questions." After reviewing the plan's adjustments, maintenance personnel click the "Confirm Execution" button to acknowledge the changes. The terminal then converts this confirmation into structured confirmation information, including the maintenance personnel's identifier, confirmation timestamp, and plan version identifier, which is transmitted via an encrypted channel to the interaction recording module of the virtual AI agent. Upon receiving the confirmation information, the interaction recording module immediately associates it with the metadata of the updated maintenance guidance plan, recording the current system time in the generation time field, accurate to the millisecond level, to ensure accurate time tracking.

[0149] The adjustment basis field records in detail the operational feedback description fragments, corresponding real-time semantic interaction node identifiers, and abnormal result analysis conclusions upon which this plan adjustment is based. For example, "Based on the 'refrigerant pressure remains low' phenomenon reported in step S143, and combined with the pressure monitoring requirements of node N3, the refrigerant leak response strategy numbered STR-2023-045 in the abnormal response strategy library is invoked for adjustment." The version identifier is generated in the format of "base version number.adjustment number," such as the initial plan being V1.0, the first adjustment being V1.1, and so on. Each version identifier uniquely corresponds to one plan update.

[0150] This information is stored in conjunction with the complete content of the updated operation and maintenance guidance plan, forming a full record of this version of the plan. If the operation and maintenance personnel click the "Feedback Questions" button, the terminal will be redirected to the question input interface, where the personnel can enter specific questions about the updated plan. The virtual artificial intelligence agent will then restart the feedback semantic parsing process to further optimize the plan based on the questions raised.

[0151] Step S158: Summarize the initial operation and maintenance guidance plan and the operation and maintenance guidance plan after each update during this interaction process, arrange them in order of generation time, mark the real-time semantic interaction nodes and operation feedback descriptions corresponding to each adjustment, and form the evolution trajectory of the guidance plan.

[0152] Retrieve all solution data generated during this interaction from the solution storage module of the virtual AI agent, including the initial operation and maintenance guidance solution and the operation and maintenance guidance solutions after each round of updates. Add a timestamp tag to each solution and establish a linear sequence of solutions arranged in chronological order of their generation.

[0153] For each updated operation and maintenance guidance plan in the sequence, its corresponding adjustment trigger information is associated: the real-time semantic interaction node number that triggered the adjustment, the location of the operation step where the node is located, and the full text of the operation feedback received by the node. Adjacent plans are connected by arrows in the plan sequence, with the adjustment type (such as "parameter anomaly adjustment", "Q&A adjustment", "step optimization adjustment") marked next to the arrows, clearly presenting the causal relationship of the plan evolution.

[0154] The evolution trajectory of the guidance scheme also includes a comparison table of the core differences for each scheme. This table lists the specific changes in operational steps, step order, depth of detail, and interaction node settings across different schemes, facilitating the tracking and adjustment of logic. The trajectory data is stored in a visual format, allowing users to view the status of each stage of the scheme by sliding along a timeline.

[0155] Step S159: Associate the evolution trajectory of the guidance scheme with all operation feedback descriptions, add the operation and maintenance object identifier, operation and maintenance personnel identifier, and the start and end time of the interaction for this interaction, and form a complete interaction record dataset.

[0156] Establish an index linking the evolution trajectory of the guidance scheme with the description of operational feedback. By using real-time semantic interaction node identifiers, each operational feedback description is precisely bound to the corresponding scheme adjustment node in the trajectory, ensuring that the correspondence between feedback content and scheme adjustment is traceable.

[0157] Add basic information fields to the interaction log dataset: the maintenance object identifier uses the device's unique hardware code, the maintenance personnel identifier is the engineer's employee number, and the interaction start and end times are accurate to the second (including the request initiation time and the final solution confirmation time). Also, supplement the records of key time nodes during the interaction process, such as the time of the first solution push, the time of the first feedback received, and the times of each solution update, forming a complete time chain.

[0158] The dataset also includes terminal device information (such as device model and operating system version) and network status records (such as network latency and signal strength at each stage of data transmission). All data is integrated in a structured format to ensure complete fields and clear logical relationships.

[0159] Step S1510: Store the interaction record dataset in the interaction archive of the virtual artificial intelligence agent.

[0160] After the interaction log dataset is generated, it undergoes a data validation process to check field completeness, accuracy of relationships, and data format compliance. If any missing or incorrect data is found, it is returned for correction. Once validation is successful, the dataset is encrypted using an encryption algorithm to ensure security during data transmission and storage.

[0161] The virtual AI agent's interaction archive stores data in a two-level directory structure of "Operation Object - Interaction Time". Multiple interaction records of the same operation object are grouped into the same first-level directory, and subdirectories are created under the directory according to the interaction timestamp to store the corresponding datasets. The archive supports multi-dimensional retrieval by operation object identifier, operation personnel identifier, interaction time range, and fault type tag, which facilitates subsequent data statistics and case analysis.

[0162] After storage is complete, the archive automatically generates a data backup, which is stored off-site to prevent data loss. Simultaneously, lifecycle management tags are added to datasets, and retention periods are set based on data importance. Non-core data exceeding its retention period is automatically archived to the historical database, ensuring efficient operation of the archive.

[0163] Figure 2This application illustrates a remote guidance and interaction system 100 for operation and maintenance technical services combined with intelligent agent assistance, comprising a processor 1001 and a memory 1003. The processor 1001 and the memory 1003 are connected, for example, via a bus 1002. Optionally, the remote guidance and interaction system 100 for operation and maintenance technical services combined with intelligent agent assistance may further include a transceiver 1004. The transceiver 1004 can be used for data interaction between this remote guidance and interaction system for operation and maintenance technical services combined with intelligent agent assistance and other remote guidance and interaction systems for operation and maintenance technical services combined with intelligent agent assistance, 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 remote guidance and interaction system for operation and maintenance technical services combined with intelligent agent assistance does not constitute a limitation on the embodiments of this application.

[0164] 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.

[0165] This application provides a computer-readable storage medium storing program code, which, when executed by a processor, can implement the steps and corresponding content of the aforementioned method embodiments.

[0166] It should be understood that although arrows indicate various operation steps in the flowcharts of the embodiments of this application, the order in which these steps are implemented is not limited to the order indicated by the arrows. Unless explicitly stated herein, in some implementation scenarios of the embodiments of this application, the implementation steps in each flowchart may be executed in other orders based on requirements. Furthermore, some or all steps in each flowchart may include multiple sub-steps or multiple stages depending on the actual implementation scenario. Some or all of these sub-steps or stages may be executed at the same time, and each sub-step or stage may also be executed at different times. In scenarios where execution times differ, the execution order of these sub-steps or stages can be flexibly configured based on requirements, and the embodiments of this application do not limit this.

[0167] 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 remote guidance and interaction method for operation and maintenance technical services combined with intelligent agent assistance, characterized in that, The method includes: Receive remote guidance interaction requests for operation and maintenance technical services initiated by operation and maintenance personnel. The remote guidance interaction requests include operation and maintenance object information, operation and maintenance problem description and expected guidance direction. The virtual artificial intelligence agent identifies the operation and maintenance scenario features of the remote guidance interaction request, and extracts the types of operating parameters associated with the operation and maintenance object, the scope of fault impact corresponding to the operation and maintenance problem, and the level of operational complexity matching the expected guidance direction. Based on the operation and maintenance scenario feature recognition results, the dynamic guidance strategy generation module of the virtual artificial intelligence agent is invoked to generate an operation and maintenance guidance scheme that includes scenario-adaptive operation guidance sequences and real-time semantic interaction nodes. The operation and maintenance guidance plan is pushed to the operation and maintenance personnel's terminal, and the operation feedback description returned by the operation and maintenance personnel after performing the operation based on the operation and maintenance guidance plan is received in real time. The operation feedback description is then input into the feedback semantic parsing module of the virtual artificial intelligence agent. By associating the real-time semantic interaction nodes in the operation and maintenance guidance scheme with the feedback semantic parsing module, adjusting the step content and presentation order of the scenario-adaptive operation guidance sequence, an updated operation and maintenance guidance scheme is obtained. The updated operation and maintenance guidance scheme is then pushed to the operation and maintenance personnel's terminal and the evolution trajectory of the guidance scheme and operation feedback description during this interaction process are stored. Based on the operation and maintenance scenario feature recognition results, the dynamic guidance strategy generation module of the virtual artificial intelligence agent is invoked to generate an operation and maintenance guidance scheme containing scenario-adaptive operation guidance sequences and real-time semantic interaction nodes, including: Input the operation and maintenance scenario feature recognition result into the dynamic guidance strategy generation module of the virtual artificial intelligence agent, trigger the strategy generation rule engine in the dynamic guidance strategy generation module, and extract the basic guidance framework that matches the type of operation parameter, the scope of fault impact and the level of operation complexity; Adjust the operation priority order in the basic guidance framework according to the scope of the fault impact, and put the operation steps corresponding to the related components of multiple core service components whose scope of the fault impact covers the front, while supplementing the risk avoidance descriptions corresponding to the operation of the related components. Based on the operation complexity level, a detailed description depth is matched for each operation step in the basic guidance framework. Steps with detailed requirements corresponding to the operation complexity level are given additional component collaboration operation instructions, while steps with non-detail requirements corresponding to the operation complexity level are given simplified redundant descriptions, forming a scenario-adaptive operation guidance sequence. In the scenario-adaptive operation guidance sequence, key steps involving monitoring of operating parameters, confirmation of component status and verification of operation results are selected, and real-time semantic interaction nodes are set. Each real-time semantic interaction node contains the core information type that needs to be fed back by the operation and maintenance personnel and the timing of the feedback trigger. By integrating the scenario-adaptive operation guidance sequence with real-time semantic interaction nodes, and adding association identifiers between interaction nodes and operation steps, an operation and maintenance guidance scheme is generated.

2. The remote guidance and interaction method for operation and maintenance technical services combined with intelligent agent assistance as described in claim 1, characterized in that, The process involves using a virtual AI agent to identify operational scenario features in the remote guidance interaction request, extracting the types of operating parameters associated with the operational object, the scope of fault impact corresponding to the operational problem, and the operational complexity level matching the desired guidance direction. This includes: The remote guidance interaction request is input into the scene feature extraction unit of the virtual artificial intelligence agent to extract the domain to which the operation and maintenance object belongs, the type of the core component of the operation and maintenance object, and the problem triggering conditions and problem-related components in the operation and maintenance problem description; Call the operation and maintenance domain parameter library associated with the virtual artificial intelligence agent, and match the corresponding operation parameter types that need to be monitored by the operation and maintenance object under normal operation state according to the domain to which the operation and maintenance object belongs and the core component type of the operation and maintenance object; Based on the problem triggering conditions and associated components in the operation and maintenance problem description, and combined with the fault impact analysis logic of the virtual artificial intelligence agent, the range of adjacent components that may be affected by the operation and maintenance problem and the service interruption risk level are determined, thus forming the fault impact range corresponding to the operation and maintenance problem. The operational objective description in the desired guidance direction is analyzed, and combined with the operational complexity assessment model of the virtual artificial intelligence agent, the operational complexity level matching the desired guidance direction is determined based on the number of components involved in the operation, the correlation of operation steps, and the operation fault tolerance requirements. The operational parameter types, fault impact range, and operational complexity levels are summarized to form the operation and maintenance scenario feature identification results.

3. The remote guidance and interaction method for operation and maintenance technical services combined with intelligent agent assistance as described in claim 2, characterized in that, Based on the problem triggering conditions and associated components in the operation and maintenance problem description, and combined with the fault impact analysis logic of the virtual artificial intelligence agent, the scope of adjacent components that may be affected by the operation and maintenance problem and the level of service interruption risk are determined, forming the fault impact range corresponding to the operation and maintenance problem, including: Input the problem triggering conditions and problem-associating components in the operation and maintenance problem description into the fault impact analysis unit of the virtual artificial intelligence agent; The component association graph of the virtual artificial intelligence agent is invoked to locate the connection relationship of the problem-related components in the operation and maintenance system based on the problem-related components, and to identify the directly connected first-level adjacent components and the indirectly related second-level adjacent components. Analyze the impact of the problem triggering conditions on the functionality of the problem-related components, determine whether the functional impact will be transmitted to first-level and second-level adjacent components through the connection relationship, and determine the range of adjacent components that may be affected; Based on the potential range of adjacent components, and combined with the service weight of each component in the operation and maintenance system, the probability of service interruption and the number of affected users are calculated to determine the service interruption risk level. The scope of potentially affected adjacent components is combined with the service interruption risk level to form the fault impact range corresponding to the operation and maintenance problem.

4. The remote guidance and interaction method for operation and maintenance technical services combined with intelligent agent assistance as described in claim 1, characterized in that, The method combines the operational complexity level to match the level of detailed description for each operational step in the basic guidance framework. Steps with detailed requirements corresponding to the operational complexity level are given additional component collaboration instructions, while steps with non-detailed requirements are simplified with redundant descriptions, forming a scenario-adaptive operational guidance sequence, including: Extract all operation steps from the basic guidance framework. Each operation step includes the operation object, operation action, and operation goal. The operation complexity level is associated with the description depth matching rule of the virtual artificial intelligence agent to determine the detailed description dimension corresponding to different operation complexity levels. When the operation complexity level corresponds to detailed requirements, it includes component collaboration logic, operation timing requirements, and parameter adjustment gradient. When the operation complexity level corresponds to non-detail requirements, it only retains the core operation actions and operation objectives. For the operation steps corresponding to the level of operation complexity, supplement the details of the operation steps by describing the collaborative triggering conditions between multiple components involved in the operation steps, the sequential order of operation execution, and the gradual adjustment gradient of key parameters. For operation steps that correspond to non-detailed requirements based on the operation complexity level, remove duplicate operation objective descriptions, redundant safety prompts, and background descriptions unrelated to the operation actions, and simplify the description of operation steps. Following the initial order of the basic guidance framework, all operational steps are integrated, supplemented, improved, and simplified to form a scenario-adaptive operational guidance sequence.

5. The remote guidance and interaction method for operation and maintenance technical services combined with intelligent agent assistance as described in claim 1, characterized in that, The process of pushing the operation and maintenance guidance plan to the operation and maintenance personnel's terminal, receiving the operation feedback description returned by the operation and maintenance personnel after performing operations based on the operation and maintenance guidance plan in real time, and inputting the operation feedback description into the feedback semantic parsing module of the virtual artificial intelligence agent includes: Through the interactive push unit of the virtual artificial intelligence agent, the operation step description and the corresponding real-time semantic interaction node feedback prompts are synchronously pushed to the operation and maintenance personnel's terminal according to the step sequence of the scenario-adaptive operation guidance sequence in the operation and maintenance guidance scheme. When each real-time semantic interaction node is triggered, the feedback receiving channel of the virtual artificial intelligence agent is activated to receive the operation execution status, component response phenomena and operation questions input by the operation and maintenance personnel through the terminal, and form an operation feedback description. The operation feedback description is transmitted in real time so that it is sent to the feedback semantic parsing module of the virtual artificial intelligence agent immediately after the operation and maintenance personnel complete the input. After receiving the operation feedback description, the feedback semantic parsing module starts semantic word segmentation processing to separate the factual and interrogative statements in the operation feedback description and mark them as factual feedback fragments and interrogative feedback fragments, respectively. Associating the factual feedback fragments with the question feedback fragments with corresponding real-time semantic interaction node identifiers forms the initial input data for feedback semantic parsing.

6. The remote guidance and interaction method for operation and maintenance technical services combined with intelligent agent assistance as described in claim 5, characterized in that, After receiving the operation feedback description, the feedback semantic parsing module initiates semantic word segmentation processing to separate the factual and interrogative statements in the operation feedback description, marking them as factual feedback fragments and interrogative feedback fragments, respectively, including: The semantic word segmentation model of the virtual artificial intelligence agent is invoked to perform word segmentation on the operation feedback description, resulting in multiple semantic word units; The semantic vocabulary units are matched with the factual statement vocabulary of the virtual artificial intelligence agent, and vocabulary units containing factual information such as operation results, component status, and parameter values ​​are selected and combined to form factual feedback fragments. The remaining semantic vocabulary units are matched with the question expression vocabulary of the virtual artificial intelligence agent, and vocabulary units containing question particles, uncertain expressions, and descriptions of operational confusion are selected and combined to form question feedback fragments. The fact feedback segment is judged for completeness. It is checked whether it contains the core result information corresponding to the operation steps. If it is missing, a supplementary inquiry prompt is generated and pushed to the operation and maintenance personnel's terminal. After receiving the supplementary feedback, the fact feedback segment is improved. The core questions of the question feedback segments are extracted to identify the specific points of confusion for the operation and maintenance personnel regarding the operation steps, forming the core identifier of the question feedback segments and associating them with the corresponding real-time semantic interaction node identifiers.

7. The remote guidance and interaction method for operation and maintenance technical services combined with intelligent agent assistance as described in claim 1, characterized in that, The step of associating the real-time semantic interaction nodes in the operation and maintenance guidance scheme through the feedback semantic parsing module, adjusting the step content and presentation order of the scenario-adaptive operation guidance sequence, and obtaining the updated operation and maintenance guidance scheme includes: The feedback semantic parsing module extracts the factual feedback fragments, question feedback fragments, and associated real-time semantic interaction node identifiers corresponding to the operation feedback description. The real-time semantic interaction node identifier is matched with the real-time semantic interaction node in the operation and maintenance guidance scheme, and the target operation step in the scenario-adaptive operation guidance sequence corresponding to the real-time semantic interaction node is located. Analyze whether the factual feedback segment is consistent with the expected result description of the target operation step. If they are inconsistent, extract the abnormal result description from the factual feedback segment, call the abnormal response strategy library of the virtual artificial intelligence agent, and generate operation adjustment suggestions for the abnormal result description. For the core points of confusion in the question feedback segment, the question answering knowledge base of the virtual artificial intelligence agent is invoked to match the answer statements related to the target operation steps and integrate them into the adjusted operation step description; Based on the operation adjustment suggestions and answers, the content of the target operation steps in the scenario-adaptive operation guidance sequence is revised, and the presentation order of subsequent operation steps is adjusted based on the impact range corresponding to the abnormal results, thus forming an updated operation and maintenance guidance scheme.

8. The remote guidance and interaction method for operation and maintenance technical services combined with intelligent agent assistance as described in claim 7, characterized in that, The analysis determines whether the factual feedback segment matches the expected result description of the target operation step. If they do not match, the abnormal result description in the factual feedback segment is extracted, and the abnormal response strategy library of the virtual artificial intelligence agent is invoked to generate operation adjustment suggestions for the abnormal result description, including: Extract the expected result description of the target operation steps in the operation and maintenance guidance plan, and determine the component status parameters, operating parameter ranges and operation completion identifiers included in the expected result description; The operation execution status and component response phenomena in the factual feedback segment are compared point by point with the expected result description, and inconsistent differences are marked to determine the abnormal result description. A query request is sent to the anomaly response strategy library of the virtual artificial intelligence agent so that the historical response strategies corresponding to the anomaly result description and target operation step identifier are matched based on the query request, and the parameter monitoring requirements are supplemented by the operation parameter type to form an initial response suggestion. The query request includes the anomaly result description, target operation step identifier and operation parameter type in the operation and maintenance scenario feature identification result. A scenario adaptability analysis is performed on the initial response suggestions to ensure that the operational adjustments in the initial response suggestions are consistent with the fault impact range and operational complexity level of the current operation and maintenance scenario, thereby generating the final operational adjustment suggestions.

9. A remote guidance and interactive system for operation and maintenance technical services combined with intelligent agent assistance, 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 remote guidance and interaction method for operation and maintenance technical services combined with intelligent agent assistance as described in any one of claims 1-8.

Images