Intelligent voice question-answering system and knowledge reasoning method for operation and maintenance of power equipment

Abstract

The invention discloses an intelligent voice question-answering system and a knowledge reasoning method for operation and maintenance of power equipment. The system comprises a knowledge base construction module and a core question and answer engine module, the knowledge base construction module is used for constructing a power equipment knowledge graph and a fault case library, and the core question and answer engine module comprises a natural language understanding unit, a multistage semantic retrieval unit and a voice interaction ambiguity resolution unit. A natural language understanding unit identifies intent categories and power entities. The multi-stage semantic retrieval unit executes the following steps: matching in an ontology layer of the knowledge graph; performing extended query to obtain detailed information; and similar case matching is carried out in the fault case library. The voice interaction ambiguity resolution unit is used for executing disambiguation processing; analyzing the confidence and clarifying the dialogue; and triggering knowledge retrieval and reasoning and generating a sorting answer. The operation and maintenance efficiency is improved, the dependence on experts is reduced, the operation and maintenance quality and safety are improved, knowledge precipitation and sharing are promoted, the field work experience is improved, and scientific decision making is assisted.

Description

Technical Field

[0001] This invention relates to an intelligent voice question-and-answer system in the field of power system technology, and more particularly to an intelligent voice question-and-answer system for power equipment operation and maintenance, and also to a knowledge reasoning method. Background Technology

[0002] The following problems are commonly encountered in the operation and maintenance of existing power systems: First, knowledge acquisition is difficult. Equipment manuals, procedures, historical fault cases, and other knowledge are scattered and unstructured, making it difficult for maintenance personnel to quickly and accurately find the information they need. Second, there is a strong reliance on experts. Complex fault diagnosis and handling solutions heavily depend on experienced experts, resulting in slow response times and limited expert resources. Third, information retrieval efficiency is low. Traditional keyword-based query systems (such as document search) struggle to understand complex natural language, leading to low precision and recall rates, especially in mobile or noisy field environments. Fourth, dynamic reasoning capabilities are weak. Existing systems are mostly based on static knowledge bases and lack the ability to dynamically analyze and reason using real-time equipment operating status data (such as SCADA and online monitoring data), failing to provide context-sensitive and accurate suggestions. Fifth, the interaction methods are unnatural. Keyboard and mouse operation is inconvenient in equipment sites (such as substations and distribution rooms), especially when wearing gloves or climbing. Existing voice assistants have insufficient understanding capabilities in specialized fields, especially in the power sector where there are many technical terms and potential ambiguities, resulting in a poor user experience.

[0003] Therefore, there is a lack of intelligent question-and-answer and decision-making support tools that can perform natural voice interaction, deeply understand the semantics of power industry, integrate static knowledge and dynamic operation data, possess intelligent reasoning capabilities, and be optimized for complex on-site environments, in order to solve the problems of low efficiency and strong reliance on experts in existing power equipment operation and maintenance systems. Summary of the Invention

[0004] To address the technical problems of low efficiency and high reliance on experts in existing power equipment operation and maintenance systems, this invention provides an intelligent voice question-and-answer system and knowledge reasoning method for power equipment operation and maintenance.

[0005] This invention is achieved using the following technical solution: an intelligent voice question-and-answer system for power equipment operation and maintenance, comprising:

[0006] The knowledge base construction module is used to construct a knowledge graph of power equipment and a structured fault case library associated with the nodes of the graph.

[0007] The core question-answering engine module includes:

[0008] The natural language understanding unit is used to identify the intent category and power entity in the user's question; the multi-level semantic retrieval unit is used to perform: (1) matching in the ontology layer of the knowledge graph to locate relevant concept nodes based on the identified intent and entity; (2) expanding the query to obtain detailed information according to the node attributes and associations; (3) matching similar cases in the fault case library in combination with the user's historical dialogue context and the current equipment operating status.

[0009] The voice interaction ambiguity resolution unit is used to perform: (1) disambiguation processing based on domain disambiguation dictionary and context-awareness; (2) analysis of confidence and clarification dialogue when confidence is low; and (3) triggering knowledge retrieval and reasoning and generating ranked answers.

[0010] This invention addresses the technical challenges of knowledge acquisition difficulties, strong reliance on experts, low interaction efficiency, and weak dynamic reasoning capabilities in existing power operation and maintenance systems by constructing an association system between a power equipment knowledge graph and a structured fault case library. This is achieved through a multi-level semantic retrieval strategy and a dynamic data fusion reasoning mechanism, resulting in the following technical effects: 1. Significantly improved operation and maintenance efficiency: Maintenance personnel can quickly obtain accurate information and suggestions anytime, anywhere via voice, greatly shortening fault location and handling time; 2. Reduced reliance on experts: Providing powerful intelligent assistance to frontline personnel, alleviating expert pressure, especially suitable for novices or at night / in emergencies; 3. Improved operation and maintenance quality and safety: Suggestions based on structured knowledge and dynamic reasoning are more standardized and accurate, reducing the risk of human error and providing safety precautions; 4. Promoted knowledge accumulation and sharing: The constructed knowledge graph and case library become valuable knowledge assets for enterprises, facilitating retrieval and learning; 5. Improved on-site work experience: Hands-free operation; natural voice interaction is more suitable for on-site work scenarios, improving work convenience and satisfaction; 6. Support for scientific decision-making: Dynamic reasoning combined with real-time data provides a more scientific basis for condition assessment and preventative maintenance.

[0011] As a further improvement to the above solution, the intelligent voice question-and-answer system also includes:

[0012] The voice interaction front-end module is deployed on mobile or fixed terminals. It is used to first collect the user's voice and preprocess it, then recognize the preprocessed voice signal as text, and finally synthesize the text into voice and broadcast it to the user.

[0013] As a further improvement to the above solution, the intelligent voice question-and-answer system also includes:

[0014] The data interface module is used to access real-time or historical operating data of the power production system; the power production system includes SCADA, PMS and online monitoring system, and the operating data includes current, voltage, temperature, oil chromatography data, partial discharge data and equipment status.

[0015] As a further improvement to the above solution, the core question-answering engine module also includes a dynamic reasoning engine; the dynamic reasoning engine includes:

[0016] A rule reasoning unit, which executes a predefined expert rule base;

[0017] The case reasoning unit is used to calculate the vector similarity between the current fault features and historical cases;

[0018] The dynamic knowledge graph completion unit is used to create temporary nodes from real-time data outliers and associate them with the knowledge graph.

[0019] Furthermore, when calculating similarity, the case reasoning unit assigns higher weights to real-time running data features than to text features, and the similar case matching results include a priority ranking of recommended treatment measures.

[0020] As a further improvement to the above solution, the voice interaction ambiguity resolution unit integrates a domain disambiguation mechanism and includes a mapping dictionary of synonyms / polysems of power industry terms, a context-aware disambiguation subunit based on dialogue history and real-time device status, and an interactive closed-loop subunit that initiates a clarification dialogue when the confidence level is below a threshold.

[0021] As a further improvement to the above scheme, the method for constructing the knowledge graph includes the following steps:

[0022] Establish a data source that includes equipment ledgers, technical manuals, maintenance procedures, industry standards, and expert experience;

[0023] Perform ontology definition: Define the concepts and relationships of power equipment, equipment components, attributes, failure modes, failure phenomena, failure causes, handling measures, and related relationships;

[0024] Knowledge Extraction: Automatically extract <entity, relation, entity> triples from structured / unstructured data;

[0025] Knowledge fusion and storage: Integrate knowledge from multiple sources, resolve conflicts, construct a unified and standardized knowledge graph, and store it in a graph database or relational database.

[0026] As a further improvement to the above solution, the method for constructing the fault case library includes the following steps:

[0027] Collect historical fault work orders, incident reports, and expert handling records;

[0028] Structured representation: The case is broken down into multiple elements: faulty equipment, fault phenomenon, time of occurrence, environmental conditions, relevant operational data, diagnostic process, root cause, treatment measures, and effect evaluation;

[0029] Association with the knowledge graph: Link multiple elements to the corresponding nodes of the knowledge graph to form instantiated knowledge.

[0030] This invention also provides a knowledge reasoning method, which is applied to any of the above-described intelligent voice question-and-answer systems for power equipment operation and maintenance. The knowledge reasoning method includes the following steps:

[0031] When a user's question involves fault diagnosis or status assessment, the knowledge graph and the fault case library are queried to obtain key real-time / historical operating data of the target device.

[0032] Reasoning is based on rules defined by expert experience;

[0033] The current device's phenomena and operational data are combined with cases in the fault case library to calculate similarity.

[0034] Abnormal states reflected in real-time operational data or calculated features are dynamically associated with the knowledge graph as temporary nodes or attributes.

[0035] By combining the results of rule-based reasoning, the handling measures for similar cases, and the procedures and steps associated with the knowledge graph, a priority list of handling suggestions for the current problem is generated.

[0036] This invention also provides a knowledge reasoning method, which is applied to any of the above-described intelligent voice question-and-answer systems for power equipment operation and maintenance. The knowledge reasoning method includes the following steps:

[0037] Receive user queries processed by the natural language understanding unit and identify the target device and fault symptoms;

[0038] Obtain the real-time operating data of the device;

[0039] Perform dynamic reasoning: call the rule base to match the cause of the fault, and at the same time search similar cases in the fault case library based on the device model, phenomenon characteristics and real-time data;

[0040] Generating disposal suggestions: Integrating rule-based reasoning results, similar case measures, and knowledge graph-related procedures, and outputting them in order of priority.

[0041] Compared to existing power equipment operation and maintenance systems, the intelligent voice question-and-answer system and knowledge reasoning method for power equipment operation and maintenance of the present invention have the following advantages:

[0042] 1. This intelligent voice question-and-answer system, by constructing an association system between a knowledge graph of power equipment and a structured fault case database, and combining a multi-level semantic retrieval strategy and a dynamic data fusion reasoning mechanism, solves the technical problems of difficult knowledge acquisition, strong reliance on experts, low interaction efficiency, and weak dynamic reasoning ability in existing power operation and maintenance systems, achieving the following technical effects:

[0043] (1) Significantly improves operation and maintenance efficiency. Operation and maintenance personnel can quickly obtain accurate information and suggestions through voice anytime and anywhere, greatly shortening the time for fault finding and handling.

[0044] (2) Reduce reliance on experts, provide powerful intelligent assistance to frontline staff, alleviate the pressure on experts, and are especially suitable for novices or at night / in emergency situations;

[0045] (3) Improve the quality and security of operation and maintenance. Suggestions based on structured knowledge and dynamic reasoning are more standardized and accurate, reducing the risk of human error and reminding you of safety precautions.

[0046] (4) Promote knowledge accumulation and sharing. The knowledge graph and case library built become valuable knowledge assets of enterprises, which are convenient for querying and learning.

[0047] (5) Improve the on-site work experience, free up your hands, and natural voice interaction is more in line with the on-site work scenario, improving work convenience and satisfaction;

[0048] (6) Assist scientific decision-making, dynamic reasoning combined with real-time data provides a more scientific basis for condition assessment and preventive maintenance.

[0049] 2. This knowledge reasoning method, by integrating static knowledge graphs and dynamic operational data, and combining rule-based reasoning with case-based reasoning, achieves intelligent diagnosis and decision support for power equipment faults. This method constructs dynamic knowledge nodes by accessing real-time equipment operational data, uses a weighted similarity algorithm for case matching, and establishes a feedback optimization mechanism. This effectively solves the problems of traditional systems having limited reasoning capabilities and poor real-time performance, significantly improving fault diagnosis accuracy and enhancing the intelligence level and decision-making efficiency of power operation and maintenance. Attached Figure Description

[0050] Figure 1 This is an architecture diagram of the intelligent voice question-and-answer system for power equipment operation and maintenance according to Embodiment 1 of the present invention.

[0051] Figure 2 This is a business process diagram of the intelligent voice question-and-answer system for power equipment operation and maintenance according to Embodiment 1 of the present invention. Detailed Implementation

[0052] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0053] Example 1

[0054] Please see Figure 1 and Figure 2 This embodiment provides an intelligent voice question-and-answer system for power equipment operation and maintenance. The system mainly includes a knowledge base construction module and a core question-and-answer engine module, and may also include a voice interaction front-end module, a data interface module, and other modules. This system is an intelligent question-and-answer and decision-making support tool capable of natural voice interaction, deep understanding of power industry semantics, integration of static knowledge and dynamic operational data, intelligent reasoning capabilities, and optimization for complex on-site environments. It can be comprehensively applied in the operation and maintenance process of power systems.

[0055] The knowledge base construction module is used to build a knowledge graph of power equipment and a structured fault case library associated with the graph nodes. The knowledge base construction module can be deployed on a backend server. The knowledge graph construction method includes the following steps: (1) Establishing data sources including equipment ledgers, technical manuals, maintenance procedures, industry standards, and expert experience; (2) Defining ontology: defining the concepts and relationships of power equipment (such as transformers, circuit breakers, GIS, cables, etc.), equipment components, attributes (such as parameters, models, manufacturers), fault modes, fault phenomena, fault causes, handling measures, and associations (such as "belongs to", "causes", "needs inspection"); (3) Extracting knowledge: automatically extracting <entity, relation, entity> triples from structured / unstructured data, where NLP technology (entity recognition, relation extraction) can be used; (4) Knowledge fusion and storage: fusing multi-source knowledge, resolving conflicts, building a unified and standardized knowledge graph, and storing it in a graph database (such as Neo4j) or a relational database.

[0056] The method for constructing a fault case library includes the following steps: (1) collecting historical fault work orders, accident reports, expert handling records, etc.; (2) structured representation: decomposing the case into multiple elements: faulty equipment, fault phenomenon, occurrence time, environmental conditions, relevant operating data, diagnosis process, root cause, handling measures, effect evaluation (other elements can be added in other embodiments); (3) association with knowledge graph: linking multiple elements to the corresponding nodes of the knowledge graph to form instantiated knowledge.

[0057] The voice interaction front-end module is deployed on mobile terminals (such as tablets, smart helmets, and mobile phones) or fixed terminals. It first collects the user's voice and preprocesses it (such as noise reduction and echo cancellation), then recognizes the preprocessed voice signal as text, and finally synthesizes the text into voice for broadcast to the user. The voice interaction front-end module can be installed as an app on the mobile terminal of maintenance personnel (explosion-proof tablets, smart helmet accessories) or deployed on a fixed industrial control computer in the field. In this embodiment, Automatic Speech Recognition (ASR) converts the voice signal into text. The recognition accuracy for power industry-specific terms (such as "partial discharge," "oil chromatography," and "tap switch") is specifically optimized, and a domain-customized ASR model can be trained or a professional dictionary can be loaded. Text-to-Speech (TTS) converts the system-generated text answer into voice for broadcast to the user.

[0058] The core question-answering engine module includes a Natural Language Understanding (NLU) unit, a multi-level semantic retrieval unit, and a voice interaction ambiguity resolution unit. This core question-answering engine module can be deployed on an application server. The NLU unit is used to identify the intent category (such as query parameters, fault diagnosis, inquiry procedures, and requests for handling suggestions) and electrical entities in user questions. These electrical entities can include electrical equipment entities (such as "#1 main transformer"), components (such as "cooler"), fault phenomena (such as "overheating"), parameters (such as "oil temperature"), etc.

[0059] The multi-level semantic retrieval unit is used to perform: (1) matching in the ontology layer of the knowledge graph to locate relevant concept nodes based on the identified intent and entity; (2) expanding the query to obtain detailed information based on node attributes and relationships (such as “cause”, “need to be checked”, “processing method”); (3) combining the user’s historical dialogue context (such as the currently concerned device) and the current device operating status (from the data interface) to perform similar case matching in the fault case library (using text similarity, case feature vector matching and other technologies).

[0060] The voice interaction ambiguity resolution unit is used to perform: (1) disambiguation processing based on the domain disambiguation dictionary and context-awareness; (2) analysis of confidence and clarification dialogue when confidence is low; and (3) triggering knowledge retrieval and reasoning and generating ranked answers. Among them, the domain disambiguation dictionary may include: establishing a synonym and polysemous word mapping table for professional terms in the power field (e.g., "switch" may refer to a circuit breaker or disconnector, and "protection action" may refer to a specific protection device). Context-aware disambiguation mainly uses dialogue history (e.g., the equipment confirmed in the previous round) and the current operating scenario (e.g., equipment status) to disambiguate the identified ambiguous words. Confidence analysis and clarification dialogue can evaluate the confidence of ASR results and NLU understanding results. When the confidence is low or there is key ambiguity, a clarification dialogue is initiated (e.g., "Are you referring to the cooler fan of the #1 main transformer?"). The knowledge retrieval and reasoning interface can send structured query requests processed by NLU to the knowledge reasoning module. Answer generation and sorting mainly involves receiving the results returned by the knowledge reasoning module, combining the user's intent and context, generating answer text in natural language form (or a structured suggestion list), and sorting them according to relevance, confidence, etc.

[0061] In some embodiments, the voice interaction ambiguity resolution unit integrates a domain disambiguation mechanism and includes a mapping dictionary of synonyms / polysems of power industry terms, a context-aware disambiguation subunit based on dialogue history and real-time device status, and an interactive closed-loop subunit that initiates a clarification dialogue when the confidence level is below a threshold.

[0062] The core question-answering engine module also includes a dynamic reasoning engine, which comprises a rule-based reasoning unit, a case-based reasoning unit, and a dynamic knowledge graph completion unit. The rule-based reasoning unit executes a predefined expert rule base. The case-based reasoning unit calculates the vector similarity between current fault features and historical cases. When calculating similarity, the case-based reasoning unit assigns higher weights to real-time operational data features than to text features, and the similar case matching results include a priority ranking of recommended handling measures. The dynamic knowledge graph completion unit creates temporary nodes from real-time data outliers and associates them with the knowledge graph.

[0063] The data interface module is used to access real-time or historical operational data from the power production system. This system includes SCADA, PMS, and online monitoring systems, and the operational data includes current, voltage, temperature, oil chromatography data, partial discharge data, and equipment status. The data interface module provides standardized interfaces (such as APIs) to facilitate data communication. It is important to note that the data interface module must establish secure and reliable data communication with the existing power monitoring system (e.g., a firewall-isolated private network, MQTT / HTTPS API).

[0064] In this embodiment, the intelligent voice question-and-answer system achieves the following objectives: 1. Constructing a structured knowledge graph of power equipment and a rich fault case database to provide core knowledge support for intelligent question answering and reasoning. 2. Developing a voice-driven intelligent question-and-answer engine to support on-site maintenance personnel in conveniently consulting about problems using natural speech. 3. Designing a multi-level semantic retrieval strategy to improve the accuracy of understanding and retrieval efficiency for complex, professional, and colloquial questions. 4. Innovatively integrating real-time / historical operating data of equipment to achieve dynamic knowledge reasoning and provide more timely and targeted solutions. 5. Applying advanced voice interaction ambiguity resolution technology to effectively handle ambiguities caused by professional terminology, synonyms, polysemous words, and speech recognition errors in the power field, improving the robustness and accuracy of interaction. 6. Providing maintenance personnel with accurate fault diagnosis assistance, recommended handling solutions, and procedure queries, improving maintenance efficiency and safety.

[0065] In practical implementation, this embodiment provides an example, where the business process is as follows:

[0066] 1. Voice input stage: On-site personnel initiate voice queries via mobile terminals; the system performs noise reduction and echo cancellation processing; domain-optimized ASR completes the speech-to-text conversion.

[0067] 2. Ambiguity handling closed loop: The confidence level of the identification result is evaluated; when the confidence level is low, a clarification dialogue is initiated (e.g., "Please confirm if you mean the QF1 circuit breaker?"); thus forming a human-computer interaction verification closed loop.

[0068] 3. Three-level retrieval mechanism: Level 1: Ontology layer matching device type; Level 2: Attribute relationship expansion (e.g., "transformer → oil temperature → threshold"); Level 3: Context case matching (integrating real-time data).

[0069] 4. Dynamic Reasoning Stage: Rule-based reasoning: based on expert rule base (IF-THEN); Case-based reasoning: similarity calculation (fault features + operational data); Dynamic knowledge graph completion (temporary node creation).

[0070] 5. Output and Evolution: Generate structured suggestions (operation steps / safety precautions); multimodal output: voice broadcast + graphic display; execution result feedback updates the knowledge base.

[0071] In addition, the system architecture is explained in detail below:

[0072] 1. Terminal layer (mobile devices): smart safety helmet / explosion-proof tablet; local lightweight model supporting offline mode; multimodal interactive interface (voice + text).

[0073] 2. Core Engine Layer: Natural Language Understanding (NLU): Domain Entity Recognition (Power Equipment / Fault Type); Intent Classification (Query / Diagnosis / Procedure).

[0074] 3. Multi-level retrieval system: Dynamic reasoning engine: Rule-based reasoning: based on the Drools rule engine; Case-based reasoning: cosine similarity algorithm; Real-time knowledge graph completion.

[0075] 4. Knowledge Base Layer: Knowledge Graph Construction: Ontology Definition: OWL language describes device relationships; Storage: Neo4j graph database.

[0076] 5. Fault Case Library: Structured fields: Equipment / Phenomenon / Cause / Measure; Vectorized storage (FAISS index).

[0077] 6. Data Interface Layer: Real-time data access: SCADA: Equipment operating status; PMS: Maintenance history record; Online monitoring: Oil chromatography / partial discharge data.

[0078] 7. Secure communication: HTTPS+MQTT protocol; dedicated network channel isolated by firewall.

[0079] Here's an example to illustrate, such as a question-answering engine workflow:

[0080] User voice input: "The chromatographic analysis of the main transformer oil shows that the acetylene level is too high. What could be the reason?"

[0081] Voice front-end: Acquisition and noise reduction -> ASR -> Text: "The main transformer oil chromatography analysis shows that acetylene levels are too high. What could be the reason?"

[0082] NLU:

[0083] Intent recognition: Fault diagnosis / cause lookup.

[0084] Entity identification: Equipment = Main transformer (disambiguation required, assuming context or location information confirms it as #1 main transformer), Phenomenon / parameter = acetylene exceeds standard in oil chromatography.

[0085] Multi-level search:

[0086] Body layer: Locating transformer -> Oil chromatography analysis -> Acetylene (C2H2) -> Exceeding standard -> Associated fault mode node (such as discharge fault).

[0087] Attribute Relationship Layer: Query the possible causes associated with nodes exceeding acetylene limits (e.g., high-temperature arc discharge, partial discharge, low-energy discharge, overheating involving solid insulation, etc.). Query attributes such as the model and historical data of main transformer #1.

[0088] Context Case Layer: Obtain operational data such as current load, temperature, and other gas content of the #1 main transformer through the data interface. Search the case library for cases containing "acetylene exceeding the standard" and with similar equipment models and operational data characteristics.

[0089] In summary, compared with existing power equipment operation and maintenance systems, the intelligent voice question-and-answer system for power equipment operation and maintenance in this embodiment has the following advantages:

[0090] This intelligent voice question-and-answer system, by constructing a link between a knowledge graph of power equipment and a structured fault case database, and combining a multi-level semantic retrieval strategy and a dynamic data fusion reasoning mechanism, solves the technical problems of difficult knowledge acquisition, strong reliance on experts, low interaction efficiency, and weak dynamic reasoning ability in existing power operation and maintenance systems, achieving the following technical effects:

[0091] (1) Significantly improves operation and maintenance efficiency. Operation and maintenance personnel can quickly obtain accurate information and suggestions through voice anytime and anywhere, greatly shortening the time for fault finding and handling.

[0092] (2) Reduce reliance on experts, provide powerful intelligent assistance to frontline staff, alleviate the pressure on experts, and are especially suitable for novices or at night / in emergency situations;

[0093] (3) Improve the quality and security of operation and maintenance. Suggestions based on structured knowledge and dynamic reasoning are more standardized and accurate, reducing the risk of human error and reminding you of safety precautions.

[0094] (4) Promote knowledge accumulation and sharing. The knowledge graph and case library built become valuable knowledge assets of enterprises, which are convenient for querying and learning.

[0095] (5) Improve the on-site work experience, free up your hands, and natural voice interaction is more in line with the on-site work scenario, improving work convenience and satisfaction;

[0096] (6) Assist scientific decision-making, dynamic reasoning combined with real-time data provides a more scientific basis for condition assessment and preventive maintenance.

[0097] Example 2

[0098] This embodiment provides a knowledge reasoning method, which is applied to the intelligent voice question-and-answer system for power equipment operation and maintenance in Embodiment 1. Specifically, it is embedded in the question-and-answer engine and is responsible for deep analysis using knowledge base and real-time data. The method mainly includes the following steps.

[0099] (1) When the user’s problem involves fault diagnosis or status assessment, the inference engine not only queries the knowledge graph and fault case library, but also obtains the key real-time / historical operating data of the target device through the data interface module.

[0100] (2) Reasoning based on rules defined by expert experience. For example: IF transformer top oil temperature > threshold AND load current > rated value THEN Possible causes: insufficient cooler efficiency OR overload; Recommendation: Check the cooler's operating status and verify the load. Rules can utilize causal chains in knowledge graphs.

[0101] (3) Case analogy reasoning: Combine the current equipment's phenomena and operating data with cases in the fault case library to calculate similarity; find the K most similar historical cases and refer to their diagnostic conclusions and handling measures. The similarity calculation will give higher weight to operating data characteristics (such as specific temperature trends, specific gas contents).

[0102] (4) Dynamic knowledge graph completion: Abnormal states reflected in real-time running data (such as over-limit alarms) or calculated features (such as trend analysis results) are dynamically associated with the knowledge graph as temporary nodes or attributes to enrich the reasoning basis.

[0103] (5) Based on the comprehensive rule reasoning results, the handling measures for similar cases, and the procedures and steps associated in the knowledge graph, generate a priority list of handling suggestions for the current problem.

[0104] This embodiment is illustrated by example. The example below only represents one implementation scheme, and other schemes may be used in other embodiments.

[0105] Based on the retrieved possible causes (rules), the real-time data of the #1 main transformer (e.g., normal load, no abnormalities in other gases), and similar cases (e.g., a case showing that the isolated rise in acetylene was caused by poor internal contact leading to low-energy discharge), a comprehensive inference was made.

[0106] If the operational data shows a recent operational shock, it may indicate an arc discharge; if the data is stable and there is only a slight increase in acetylene, it may indicate a low-energy discharge or an early failure.

[0107] Answer Generation: Based on the comprehensive reasoning results, the following answer is generated: "#1 The acetylene level in the main transformer oil chromatography exceeds the standard. Possible causes include low-energy discharge (such as poor internal contact), partial discharge, or early overheating. Recommendations: 1. Further analysis based on online partial discharge monitoring data (if any); 2. Check for any recent operational or abnormal operating conditions; 3. Shorten the monitoring cycle and pay attention to the gas production rate. Historical cases indicate that similar situations at the XX substation were caused by poor contact at the bushing end screen. Please refer to Chapter XXX of the regulations for detailed inspection."

[0108] TTS: Converts the answer text into speech and plays it back, while displaying detailed text and images on the screen.

[0109] Example of ambiguity resolution:

[0110] User's voice: "The switch tripped."

[0111] ASR text: "The switch tripped." (Possibly high confidence level, but semantically ambiguous)

[0112] NLU: Intent = Report an event / Query the cause. Entity = Switch (ambiguity: circuit breaker? disconnector? protection device output?).

[0113] Ambiguity resolution:

[0114] According to the domain dictionary, "switch" is a broad term.

[0115] Contextual awareness: If the previous conversation involved "110kV line protection action", it is more likely to refer to "circuit breaker tripping"; if it is mentioned during an inspection of GIS equipment, it may refer to an abnormal change in the status of a disconnecting switch.

[0116] Confidence and Clarification: If the key entity cannot be identified, initiate clarification: "Which switch are you referring to? Has the circuit breaker tripped? Or has the status of the disconnector changed?"

[0117] User clarifies: "Oh, it was circuit breaker 101 that tripped." -> Disambiguation successful, proceed with follow-up processing.

[0118] This knowledge reasoning method integrates static knowledge graphs with dynamic operational data, combining rule-based reasoning and case-based reasoning mechanisms to achieve intelligent diagnosis and decision support for power equipment faults. By constructing dynamic knowledge nodes through real-time access to equipment operational data, employing a weighted similarity algorithm for case matching, and establishing a feedback optimization mechanism, this method effectively solves the problems of limited reasoning capabilities and poor real-time performance in traditional systems. This significantly improves the accuracy of fault diagnosis and substantially enhances the intelligence level and decision-making efficiency of power operation and maintenance.

[0119] Example 3

[0120] This embodiment provides a knowledge reasoning method, which is applied to the intelligent voice question-and-answer system for power equipment operation and maintenance in Embodiment 1. The method mainly includes the following steps: (i) receiving user queries processed by the natural language understanding unit and identifying the target equipment and fault phenomena; (ii) obtaining the real-time operating data of the equipment; (iii) performing dynamic reasoning: calling the rule base to match the cause of the fault, and simultaneously searching for similar cases in the fault case library based on the equipment model, phenomenon characteristics and real-time data; (iv) generating disposal suggestions: integrating the rule reasoning results, similar case measures and knowledge graph association procedures, and outputting them in priority order.

[0121] Example 4

[0122] This embodiment provides a computer terminal, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps of the knowledge reasoning method of Embodiment 1.

[0123] When applying the methods of Embodiment 2 or 3, they can be implemented in software form, such as by designing them as stand-alone programs and installing them on a computer terminal, which can be a computer, smartphone, control system, or other IoT devices. Alternatively, the methods of Embodiment 2 or 3 can be designed as embedded programs and installed on a computer terminal, such as on a microcontroller.

[0124] Example 5

[0125] This embodiment provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, it implements the steps of the knowledge reasoning method of Embodiment 2 or the present embodiment.

[0126] When applying the method of Embodiment 2 or 3, it can be applied in the form of software, such as by designing it as a program that can run independently on a computer-readable storage medium. The computer-readable storage medium can be a USB flash drive, designed as a USB security token, and the program is designed to start the entire method through an external trigger.

[0127] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An intelligent voice question-and-answer system for power equipment operation and maintenance, characterized in that, It includes: The knowledge base construction module is used to construct a knowledge graph of power equipment and a structured fault case library associated with the nodes of the graph. The core question-answering engine module includes: The Natural Language Understanding Unit is used to identify the intent category and electrical entity in the user's question; A multi-level semantic retrieval unit is used to perform: (1) matching in the ontology layer of the knowledge graph to locate relevant concept nodes based on the identified intent and entity; (2) expanding the query to obtain detailed information according to node attributes and relationships; and (3) matching similar cases in the fault case library in combination with the user's historical dialogue context and the current device operating status. The voice interaction ambiguity resolution unit is used to perform: (1) disambiguation processing based on domain disambiguation dictionary and context-awareness; (2) analysis of confidence and clarification dialogue when confidence is low; and (3) triggering knowledge retrieval and reasoning and generating ranked answers.

2. The intelligent voice question-and-answer system for power equipment operation and maintenance as described in claim 1, characterized in that, The intelligent voice question-and-answer system also includes: The voice interaction front-end module is deployed on mobile or fixed terminals. It is used to first collect the user's voice and preprocess it, then recognize the preprocessed voice signal as text, and finally synthesize the text into voice and broadcast it to the user.

3. The intelligent voice question-and-answer system for power equipment operation and maintenance as described in claim 1, characterized in that, The intelligent voice question-and-answer system also includes: The data interface module is used to access real-time or historical operating data of the power production system; the power production system includes SCADA, PMS and online monitoring system, and the operating data includes current, voltage, temperature, oil chromatography data, partial discharge data and equipment status.

4. The intelligent voice question-and-answer system for power equipment operation and maintenance as described in claim 1, characterized in that, The core question-answering engine module also includes a dynamic reasoning engine; the dynamic reasoning engine includes: A rule reasoning unit, which executes a predefined expert rule base; The case reasoning unit is used to calculate the vector similarity between the current fault features and historical cases; The dynamic knowledge graph completion unit is used to create temporary nodes from real-time data outliers and associate them with the knowledge graph.

5. The intelligent voice question-and-answer system for power equipment operation and maintenance as described in claim 4, characterized in that, When calculating similarity, the case reasoning unit assigns higher weights to real-time running data features than to text features, and the similar case matching results include a priority ranking of recommended treatment measures.

6. The intelligent voice question-and-answer system for power equipment operation and maintenance as described in claim 1, characterized in that, The voice interaction ambiguity resolution unit integrates a domain disambiguation mechanism and includes a mapping dictionary of synonyms / polysems of power industry terms, a context-aware disambiguation subunit based on dialogue history and real-time device status, and an interactive closed-loop subunit that initiates a clarification dialogue when the confidence level is below a threshold.

7. The intelligent voice question-and-answer system for power equipment operation and maintenance as described in claim 1, characterized in that, The method for constructing the knowledge graph includes the following steps: Establish a data source that includes equipment ledgers, technical manuals, maintenance procedures, industry standards, and expert experience; Perform ontology definition: Define the concepts and relationships of power equipment, equipment components, attributes, failure modes, failure phenomena, failure causes, handling measures, and related relationships; Knowledge Extraction: Automatically extract <entity, relation, entity> triples from structured / unstructured data; Knowledge fusion and storage: Integrate knowledge from multiple sources, resolve conflicts, construct a unified and standardized knowledge graph, and store it in a graph database or relational database.

8. The intelligent voice question-and-answer system for power equipment operation and maintenance as described in claim 1, characterized in that, The method for constructing the fault case library includes the following steps: Collect historical fault work orders, incident reports, and expert handling records; Structured representation: The case is broken down into multiple elements: faulty equipment, fault phenomenon, time of occurrence, environmental conditions, relevant operating data, diagnostic process, root cause, treatment measures, and effect evaluation; Association with the knowledge graph: Link multiple elements to the corresponding nodes of the knowledge graph to form instantiated knowledge.

9. A knowledge reasoning method, applied to an intelligent voice question-and-answer system for power equipment operation and maintenance as described in any one of claims 1-8, characterized in that, The knowledge reasoning method includes the following steps: When a user's question involves fault diagnosis or status assessment, the knowledge graph and the fault case library are queried to obtain key real-time / historical operating data of the target device; Reasoning is based on rules defined by expert experience; The current device's phenomena and operational data are combined with cases in the fault case library to calculate similarity. Abnormal states reflected in real-time operational data or calculated features are dynamically associated with the knowledge graph as temporary nodes or attributes. By combining the results of rule-based reasoning, the handling measures for similar cases, and the procedures and steps associated with the knowledge graph, a priority list of handling suggestions for the current problem is generated.

10. A knowledge reasoning method, applied to an intelligent voice question-and-answer system for power equipment operation and maintenance as described in any one of claims 1-8, characterized in that, The knowledge reasoning method includes the following steps: Receive user queries processed by the natural language understanding unit and identify the target device and fault symptoms; Obtain the real-time operating data of the device; Perform dynamic reasoning: call the rule base to match the cause of the fault, and at the same time search similar cases in the fault case library based on the device model, phenomenon characteristics and real-time data; Generating disposal suggestions: Integrating rule-based reasoning results, similar case measures, and knowledge graph-related procedures, and outputting them in order of priority.

Images