Intelligent abnormal sound sensing power fault diagnosis question-answering system and method thereof
Through the intelligent abnormal noise sensing power fault diagnosis question and answer system, deep learning and large language models are used to identify and analyze power faults, solving the problem that power abnormal noise faults are difficult to describe in language, and achieving efficient and accurate power fault diagnosis and solution provision.
Patent Information
- Application Number
- CN202510325311.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-19
- Publication Date
- 2025-06-20
AI Technical Summary
In existing power systems, it is difficult to accurately diagnose abnormal noise faults through language description, resulting in low fault diagnosis efficiency and accuracy.
The intelligent abnormal noise sensing power fault diagnosis question and answer system is adopted to train the power fault sound samples through deep learning algorithms, identify multiple fault types, and analyze the cause of the fault in combination with large language models to provide users with detailed explanations and solutions.
It realizes efficient and accurate power fault diagnosis, significantly improves the efficiency and accuracy of power system fault diagnosis, and provides an integrated solution for fault diagnosis and solutions.
Smart Images
Figure CN120179877A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of artificial intelligence technology, and specifically relates to an intelligent abnormal sound perception power failure diagnosis Q&A system and method thereof. Background Art
[0002] The intelligent abnormal sound perception power failure diagnosis Q&A system and method proposed by the present invention aim at the problem that it is difficult to describe the common power abnormal sound faults in the operation of the existing power system in language. Combining sound recognition technology and large language models, an efficient and accurate power failure diagnosis method is realized. The system can identify various fault types through training on a large number of power failure sound samples using advanced deep learning algorithms, and analyze the fault causes in combination with large language models, providing users with detailed explanations and corresponding solutions. In addition, when the user asks supplementary questions, the system can call the online search function of the large model to perform real-time search to provide more comprehensive information, so as to realize the integration of fault diagnosis and solutions, and significantly improve the efficiency and accuracy of power system fault diagnosis. Summary of the Invention
[0003] In view of the above problems, the present invention is proposed to provide an intelligent abnormal sound perception power failure diagnosis Q&A system and method thereof that overcome the above problems.
[0004] In one aspect of the present invention, there is provided an intelligent abnormal sound perception power failure diagnosis Q&A system and method thereof, and the method includes:
[0005] Obtain a power abnormal sound file, and identify the sound file through an abnormal sound recognition module; call a preset fast Q&A language model to retrieve answers from a preset vector database, where the fast Q&A language model is implemented using a large language model with a small number of parameters, and the vector data in the vector database is a real number vector formed according to text segments related to power failure diagnosis knowledge; if an answer is retrieved, the fast Q&A language model fills the question text and the retrieved answer into a preset first guiding word template to generate a first standardized answer text and output it to the user; if there is an additional question, call a preset deep reasoning language model to perform an online answer search by calling an execution tool related to the reasoning task corresponding to the current question according to the additional question of the user, and fill the search result into a preset second guiding word template to generate a second standardized answer text and output it to the user, where the deep reasoning language model is implemented using a large language model with a large number of parameters, and a preset fine-tuning data set is used to perform fine-tuning training on the selected large language model with a large number of parameters, so that the large model has the ability of cross-modal power failure diagnosis knowledge Q&A.
[0006] Further, the generation method of the fine-tuning dataset includes: searching for key text information from a pre-constructed power fault diagnosis knowledge text library, constructing a power fault diagnosis Q&A text pair dataset according to the key text information, and forming the power fault diagnosis Q&A text pair dataset into a fine-tuning dataset in a specified format, where the power fault diagnosis Q&A text pair includes a power fault diagnosis question text and a corresponding answer text.
[0007] Further, the creation method of the vector database includes: performing text slicing on the text data in a pre-constructed power fault diagnosis knowledge text library to form a text segment dataset suitable for large model processing and maintaining the integrity of context semantic information; converting each text segment in the text segment dataset into vector data and storing it in a preset vector database.
[0008] Further, the method further includes: forming a historical Q&A record from the question text of each user and the first standardized answer text generated by the fast Q&A language model or the second standardized answer text generated by the deep reasoning language model; performing text slicing on the historical Q&A record to split it into multiple text segments; converting the sliced text segments into vector data and storing the obtained vector data in a preset memory database with the identification information of the user as the index.
[0009] Further, before calling a preset fast Q&A agent to retrieve an answer from a preset vector database according to the abnormal sound classification text, the method further includes: retrieving an answer from the memory database according to the abnormal sound classification text, and filling the retrieved historical Q&A record into a specified position in the first guiding word template p1 as part of the final answer; if the fast Q&A language model does not retrieve a historical Q&A record, the specified position in the first guiding word template p1 is empty.
[0010] Further, retrieving an answer from the memory database according to the abnormal sound classification text includes: obtaining the current abnormal sound classification information, and preferentially retrieving an answer from the historical records of the current abnormal sound classification in the memory database; if no answer is retrieved from the historical Q&A records, retrieving an answer from other historical Q&A records in the memory database.
[0011] Further, calling a preset deep reasoning agent to perform an online search by invoking an execution tool related to the current question's corresponding reasoning task according to the user's additional question includes: performing syntactic and semantic analysis on the question text input by the user to obtain at least one reasoning task to be executed corresponding to the question raised by the user; sequentially invoking the execution tool corresponding to each reasoning task to be executed according to the logical relationship of each reasoning task to be executed for an online search.
[0012] Another aspect of the present invention provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor; when the computer program is executed by the processor, the steps of the power failure diagnosis knowledge Q&A method based on a large language model described in any one of the above are implemented.
[0013] The intelligent abnormal sound perception power failure diagnosis Q&A system and method provided by the embodiments of the present invention identify the collected power abnormal sounds through an abnormal sound recognition module, and for the recognized results, according to whether additional questions are needed, two intelligent models work together, taking into account the accuracy and response speed of Q&A in the field of power failure diagnosis. Among them, the fast Q&A language model can quickly answer knowledge questions in the field of power failure diagnosis by retrieving the vector database and the memory, and the deep reasoning language model can effectively expand the real-time information of the large language model by calling the execution tools related to the reasoning tasks corresponding to the questions currently proposed by the user to search for answers online, and then obtain the reasoning and analysis results. The present invention helps to improve the intelligence, accuracy and speed in the field of power failure diagnosis, and has high application value for the maintenance and fault diagnosis of power systems.
[0014] The above description is only an overview of the technical solution of the present invention. In order to be able to understand the technical means of the present invention more clearly, it can be implemented according to the content of the specification. And in order to make the above and other purposes, features and advantages of the present invention more obvious and understandable, the following specifically illustrates the specific embodiments of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS
[0015] Figure 1 It is a flowchart of an intelligent abnormal sound perception power failure diagnosis Q&A method according to an embodiment of the present invention;
[0016] Figure 2 It is a schematic structural diagram of an intelligent abnormal sound perception power failure diagnosis Q&A system according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0017] Hereinafter, exemplary embodiments of the present disclosure will be described in more detail with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be fully conveyed to those skilled in the art.
[0018] Those skilled in the art can understand that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as the general understanding of those of ordinary skill in the art to which this invention belongs. It should also be understood that terms such as those defined in a general dictionary should be understood to have a meaning consistent with their meaning in the context of the prior art, and will not be interpreted with an idealized or overly formal meaning unless specifically defined.
[0019] Embodiment 1. An embodiment of the present invention provides an intelligent abnormal sound perception power failure diagnosis Q&A system and its method. As Figure 1 shown, the power failure diagnosis knowledge Q&A method based on a large language model proposed by the present invention includes the following steps:
[0020] S11. Obtain a power abnormal sound audio file.
[0021] In this embodiment, the user collects a power abnormal sound audio file through a sound collection device, and this file is used to reflect power failures through the sound modality.
[0022] S12. Identify the audio file.
[0023] The audio file collected in S11 is identified through a preset power abnormal sound recognition model to obtain a text expression of the abnormal sound classification.
[0024] S13. According to the recognized power abnormal sound classification text, retrieve answers in the vector database preset by the fast Q&A AI agent. The fast Q&A language model is implemented using a large language model with a small number of parameters. The vector data in the vector database is a real number vector formed according to text fragments related to power failure diagnosis knowledge. Determine whether there is an additional question. If there is an additional question, execute step S14; if there is no additional question, execute step S15;
[0025] S14. Call the preset deep reasoning AI agent to search for online answers by calling the execution tools related to the reasoning task corresponding to the current question according to the additional question raised by the user. Fill the question text and the search results into the preset prompt template p2 to generate a second standardized answer text and output it to the user. The deep reasoning language model is implemented using a large language model with a large number of parameters, and a preset fine-tuning dataset is used to fine-tune the selected large language model with a large number of parameters, so that the large model has the ability to answer power failure diagnosis knowledge questions.
[0026] S15. The fast Q&A AI agent fills the answer into the preset prompt template p1 to generate a first standardized answer and output it to the user.
[0027] In this embodiment, the fast Q&A AI agent uses an open-source model with a small number of parameters, and the deep reasoning AI agent uses a fine-tuned large-scale parameter model and an online connection function. Specifically, fine-tuning training can be performed by constructing a fine-tuning dataset. After fine-tuning training, the deep reasoning AI agent will have good knowledge Q&A and logical reasoning abilities in the field of power fault diagnosis. Among them, methods such as full-parameter fine-tuning can be selected for fine-tuning training, and the present invention does not make specific limitations on this.
[0028] The cross-modal power fault diagnosis method based on the large language model AI agent provided by the embodiment of the present invention realizes the detection of power faults through sound modality while taking into account the accuracy and response speed of Q&A in the field of power fault diagnosis by identifying power abnormal sounds and the collaborative work of two AI agents. Among them, the fast Q&A AI agent can quickly answer knowledge questions in the field of power fault diagnosis by retrieving the vector database and the memory. The deep reasoning AI agent can effectively expand the real-time information of the AI agent by calling the execution tool related to the inference task corresponding to the question newly proposed by the user and searching for answers online, and then obtain the inference and analysis results. The present invention helps to improve the intelligence, accuracy and speed in the field of power fault diagnosis, and has high application value for the maintenance and fault diagnosis of power systems.
[0029] In the embodiment of the present invention, the generation method of the fine-tuning dataset includes: searching for key text information from a pre-constructed power fault diagnosis knowledge text library, constructing a power fault diagnosis Q&A text pair dataset according to the key text information, and forming a fine-tuning dataset in a specified format from the power fault diagnosis Q&A text pair dataset. Among them, the power fault diagnosis Q&A text pair includes a power fault diagnosis question text and a corresponding answer text.
[0030] In the embodiment of the present invention, the creation method of the vector database includes: slicing the text data in a pre-constructed power fault diagnosis knowledge text library to form a text fragment dataset suitable for large model processing and maintaining the integrity of context semantic information; converting each text fragment in the text fragment dataset into vector data and storing it in a preset vector database.
[0031] In the embodiment of the present invention, before calling the preset fast Q&A AI agent to retrieve an answer from the preset vector database according to the question proposed by the user, the method further includes: retrieving an answer from the memory database according to the question proposed by the user, and filling the retrieved historical Q&A record into the specified position in the first guiding word template p1 as part of the final answer; if the fast Q&A AI agent does not retrieve the historical Q&A record, the specified position in the first guiding word template p1 is empty.
[0032] The modeling of prompt engineering plays an important role in the development of AI agent conversations. A good prompt template can assist the AI agent in making more accurate predictions and obtaining more accurate answers. The prompt template can be modified according to the specific Q&A scenario requirements. In this embodiment, examples of the prompt template are as follows:
[0033] "You are a senior power fault diagnosis engineer who can accurately judge power faults based on power abnormal sounds and give specific solutions and methods to solve the faults. The abnormal sound classification is the classified text information obtained through the abnormal sound recognition module. For the sake of the rigor of the answer, please remind the user at the end that this answer is the generated data of AI for reference only, and accurate answers need to consult experts in the field of power fault diagnosis." Quick Q&A agent generated answer: "Type of abnormal sound: Discharge sound. Description of abnormal sound: There may be "crackling" or "hissing" discharge sounds inside the transformer, similar to the sound of fine electric sparks. Causes: 1. Bubbles or impurities in the oil cause partial discharge. 2. Insulation aging or damage results in a decrease in insulation performance. 3. The distance between the high-voltage lead and the transformer shell or other components is too close, causing arc discharge. 4. Poor contact of the tap changer causes partial discharge. Solutions: 1. Check the quality of the transformer oil and, if necessary, perform oil treatment or replace with new oil. 2. Conduct an insulation test on the transformer to find the aged insulation part and repair or replace it. 3. Check the relative positions of the high-voltage lead and other components to ensure sufficient insulation distance and adjust if necessary. 4. Check the tap changer, clean the contact surface to ensure good contact, and replace the tap changer if necessary. When dealing with any transformer abnormal sound problem, safety should be ensured first, the power supply should be cut off, and the correct operating procedures should be followed. If you are unsure about the severity of the problem or the solution, you should consult a professional transformer maintenance personnel or the manufacturer's technical support. A low insulation resistance value may be due to moisture or contamination of the insulation material, resulting in a decrease in insulation performance. In addition, long-term operation of the equipment and aging of the insulation material will also cause the insulation resistance value to decrease. It is recommended that you conduct a detailed inspection of the equipment to determine the specific cause and take corresponding measures."
[0034] In addition, a logical judgment on whether to call the deep inference AI agent needs to be made. Specifically, if there are follow-up questions, the deep inference AI agent needs to be directly called.
[0035] The deep inference agent is a large language model fine-tuned to have stronger language understanding, logical reasoning, and online retrieval capabilities.
[0036] Specific additional Q&A examples are as follows: 1) User's additional question: "In addition to abnormal noises, why does the transformer get severely heated during operation and how to handle it?" 2) Reply from the in-depth reasoning intelligent agent: "The severe heating of the transformer during operation may be caused by the following reasons: 1. Overload: The transformer operates for a long time under a load exceeding the rated load, resulting in heat generation. 2. Poor heat dissipation: The heat dissipation conditions around the transformer are not good, such as blocked radiators, fan failures, or too high ambient temperature, etc. 3. Insulation aging: The insulation material of the transformer ages, resulting in a decline in insulation performance, an increase in resistance, and thus more heat is generated. 4. Internal faults: Such as winding short circuits, poor contact of joints, etc., which will cause local overheating. Handling methods: 1. Check the load condition to ensure that the transformer operates within the rated load range. If necessary, reduce the load or upgrade the transformer capacity. 2. Check and improve the heat dissipation conditions of the transformer, clean the radiator, check whether the fan is working properly, and ensure good ventilation in the surrounding environment. Please conduct a thorough overhaul if necessary."
[0037] In summary, the intelligent abnormal noise perception power fault diagnosis Q&A system and its method provided by the embodiments of the present invention mainly consist of three parts: a power abnormal noise retrieval module, a quick Q&A AI intelligent agent, and an in-depth reasoning AI intelligent agent. During the Q&A process, the quick Q&A AI intelligent agent is preferentially called. If no suitable answer is found, the in-depth reasoning AI intelligent agent will be called.
[0038] Embodiment 2 The embodiments of the present invention provide a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, it implements the steps in the above-mentioned embodiments of each power fault diagnosis knowledge Q&A method based on the large language model, such as Figure 1 the steps S11 - S15 shown. Alternatively, when the processor executes the computer program, it implements the functions of each module in the above-mentioned embodiments of each power fault diagnosis knowledge Q&A system based on the large language model, such as Figure 2 the abnormal noise perception module 201, abnormal noise recognition module 202, quick Q&A AI intelligent agent 203, and in-depth reasoning AI intelligent agent 204 shown.
[0039] In the specific implementation process of Embodiment 2, reference can be made to Embodiment 1, and it has corresponding technical effects.
[0040] In addition, those skilled in the art can understand that although some embodiments herein include certain features included in other embodiments rather than other features, the combination of features of different embodiments means that it is within the scope of the present invention and forms different embodiments. For example, any one of the claimed embodiments can be used in any combination.
[0041] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. Claim 1: An intelligent abnormal sound perception power fault diagnosis question and answer system, characterized in that: include: An abnormal sound recognition module is used to obtain and recognize the sound file of abnormal power sound; a fast question and answer language model is used to retrieve answers from a preset vector database, and the fast question and answer language model is implemented using a large language model with a small-scale parameter amount; a deep reasoning language model is used to call the execution tool related to the reasoning task corresponding to the current question according to the question added by the user to search for answers online, and the deep reasoning language model is implemented using a large language model with a large-scale parameter amount and has been fine-tuned and trained; a vector database is used to store real number vectors formed according to text fragments of power fault diagnosis knowledge; a memory database is used to store vector data of historical question and answer records; wherein the system identifies abnormal power sound through the abnormal sound recognition module, and uses the fast question and answer language model and the deep reasoning language model to work together according to the recognition result to provide question and answer services for power fault diagnosis knowledge.
2. Claim 2: The intelligent abnormal sound perception power fault diagnosis question and answer system according to claim 1 is characterized in that: The method also includes: generating a fine-tuning dataset for fine-tuning the deep reasoning language model, wherein the fine-tuning dataset includes power fault diagnosis question text and corresponding answer text.
3. Claim 3: The intelligent abnormal sound perception power fault diagnosis question and answer system according to claim 1, characterized in that: The method further includes: creating a vector database, converting the electric power fault diagnosis knowledge text fragments into vector data and storing the vector data.
4. Claim 4: The intelligent abnormal sound perception power fault diagnosis question and answer system according to claim 1, characterized in that: The method further includes: forming a historical question and answer record, and storing the historical question and answer record in a memory database.
5. Claim 5: The intelligent abnormal sound perception power fault diagnosis question and answer system according to claim 1 is characterized in that: The method also includes: before calling the fast question-answering language model, retrieving historical question-answering records from a memory database, and using the retrieved historical question-answering records as part of the final answer.
6. Claim 6: The intelligent abnormal sound perception power fault diagnosis question and answer system according to claim 1, characterized in that: The method further includes: performing grammatical and semantic analysis on the user's additional question, and calling an execution tool corresponding to the reasoning task to be executed to perform an online search.
7. Claim 7: A computer device, characterized in that The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor; when the computer program is executed by the processor, the steps of the method according to any one of claims 1 to 6 are implemented.