Intelligent question answering method and device based on large model, electronic equipment and program product

Abstract

The invention discloses an intelligent question and answer method and device based on a large model, electronic equipment and a program product, and relates to the field of artificial intelligence, the method comprises the steps of collecting original data from a plurality of data sources, preprocessing the original data to obtain a training data set, responding to a standby power generation question request output by a user side through an interactive interface, after the standby power generation questioning request is analyzed, a knowledge base corresponding to the standby power generation system is called, and the knowledge base is constructed by extracting data of specified standby power generation associated information based on a standby power generation system demand scheme and a standby power generation database; searching reply information corresponding to the standby power generation question request from a knowledge base through an intelligent question and answer model, and generating a standby power generation answer; and returning the standby power generation answer to a front-end display interface of the user side. According to the method and the device, the technical problem of high cost caused by high resource consumption when an intelligent question and answer application based on a large model is applied to a specific field in related technologies is solved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology or other related fields. Specifically, it relates to an intelligent question-answering method and device, an electronic device, and a program product based on a large model. Background Art

[0002] Intelligent question-answering systems are an important application in the field of artificial intelligence. They use technologies such as natural language processing, machine learning, and information retrieval to understand user questions and return accurate answers. In recent years, with the development of deep learning technology, especially the emergence of large models, the performance of intelligent question-answering systems has been significantly improved, and they can handle more complex and abstract user questions.

[0003] However, in related technologies, intelligent question-answering applications based on large models still have the following disadvantages when dealing with specific fields, such as power generation and standby system operation and maintenance, equipment fault diagnosis, etc.: 1. Resource consumption problem: Traditional intelligent question-answering applications based on large models consume a large amount of computing resources and energy during training and inference, which leads to high operating costs. For systems that need to operate with limited resources, efficiency and cost control become difficult problems. 2. High data dependence: The performance of large models highly depends on high-quality training data. However, in professional fields such as power generation and standby system operation and maintenance, it is often difficult to obtain and integrate high-quality data, and incomplete or biased data will affect the accuracy and reliability of the model. 3. Privacy and security risks: Sensitive information may be involved when processing user questions. Therefore, how to ensure the security of user data and the protection of user privacy while ensuring service quality has become an urgent problem to be solved.

[0004] For the above problems, no effective solutions have been proposed yet. Summary of the Invention

[0005] Embodiments of the present invention provide an intelligent question-answering method and device, an electronic device, and a program product based on a large model, so as to at least solve the technical problem that intelligent question-answering applications based on large models consume a large amount of resources and result in high costs when dealing with specific fields in related technologies.

[0006] To achieve the above object, according to one aspect of the present application, an intelligent question-answering method based on a large model is provided, including: collecting raw data from multiple data sources and preprocessing the raw data to obtain a training data set, where the training data set is used to train a pre-constructed initial deep learning framework to obtain an intelligent question-answering model; in response to a power generation question request output by a client through an interaction interface, after parsing the power generation question request, calling a knowledge base corresponding to the power generation system, where the knowledge base is constructed by extracting specified power generation-related information data based on the power generation system requirement scheme and the power generation database; retrieving, through the intelligent question-answering model, a reply message corresponding to the power generation question request from the knowledge base to generate a power generation answer; and returning the power generation answer to the front-end display interface of the client.

[0007] Optionally, the step of preprocessing the raw data to obtain a training data set includes: classifying the raw data to obtain multi-modal data, where the types of the multi-modal data at least include: text, image, audio, video; performing data cleaning, data denoising, normalization processing, and data format conversion on the multi-modal data to complete the preprocessing process of the raw data and obtain the training data set.

[0008] Optionally, the knowledge base corresponding to the power generation system is constructed in the following manner: extracting power generation equipment structure information, power generation signal quantities, and power outage signal quantities from the power generation database using a preset compilation language to obtain first extraction data; generating an unstructured knowledge document based on the first extraction data; extracting power generation system site information and path information from the power generation system requirement scheme and design scheme using a preset compilation language to obtain second extraction data; extracting functional operation steps from the operation manual and test cases associated with the power generation system using a preset compilation language to obtain third extraction data; comprehensively generating a structured knowledge graph based on the second extraction data and the third extraction data; and constructing the knowledge base corresponding to the power generation system based on the structured knowledge graph and the unstructured knowledge document.

[0009] Optionally, the step of retrieving, by the intelligent question-answering model, reply information corresponding to the standby power generation question request from the knowledge base and generating a standby power generation answer includes: when the type of the standby power generation question request is power generation equipment positioning, parsing the standby power generation question request to obtain identification information of a target power generation equipment to be positioned; based on the identification information, invoking the knowledge base corresponding to the standby power generation system, and outputting, by the intelligent question-answering model, function positioning information and navigation path information associated with the target power generation equipment to obtain the standby power generation answer; or, when the type of the standby power generation question request is power generation data analysis, parsing the standby power generation question request to obtain identification information of a target power generation equipment to be positioned; based on the identification information, invoking the knowledge base corresponding to the standby power generation system, and outputting, by the intelligent question-answering model, power generation data associated with the target power generation equipment to obtain the standby power generation answer, where the power generation data at least includes: date and corresponding power generation, period and corresponding power generation, power generation peak and valley.

[0010] Optionally, the step of retrieving, by the intelligent question-answering model, reply information corresponding to the standby power generation question request from the knowledge base and generating a standby power generation answer includes: when the type of the standby power generation question request is substation site power outage analysis, parsing the standby power generation question request to obtain identification information of a target power generation equipment; based on the identification information of the target power generation equipment, invoking the knowledge base corresponding to the standby power generation system, analyzing whether the power generation of the target power generation equipment meets the power consumption requirements of a predetermined area, and when the analysis result indicates that the power generation of the target power generation equipment does not meet the power consumption requirements of the predetermined area, outputting, by the intelligent question-answering model, power outage prompt information associated with the predetermined area and equipment information of standby power generation equipment that can be invoked to obtain the standby power generation answer.

[0011] Optionally, the step of retrieving, by the intelligent question-answering model, reply information corresponding to the standby power generation question request from the knowledge base and generating a standby power generation answer includes: when the type of the standby power generation question request is faulty equipment analysis, parsing the standby power generation question request to obtain identification information of all standby power generation equipment to be analyzed, and the operating status and equipment detection pictures of each standby power generation equipment; based on the identification information of each standby power generation equipment and the operating status and equipment detection pictures of the standby power generation equipment, invoking the knowledge base corresponding to the standby power generation system to perform a fault analysis on each standby power generation equipment, and outputting, by the intelligent question-answering model, the standby power generation answer including faulty equipment information and a fault repair suggestion report.

[0012] Optionally, the step of returning the standby power generation answer to the front-end display interface of the client includes: encrypting the standby power generation answer using a homomorphic encryption strategy; returning the encryption result to the front-end display interface of the client through an end-to-end encryption transmission strategy; collecting the answer feedback evaluation of the client, and iteratively updating the intelligent question-answering model and the knowledge base based on the answer feedback evaluation.

[0013] According to another aspect of the embodiments of the present invention, there is also provided an intelligent question-answering device based on a large model, including: a data preprocessing unit for collecting raw data from multiple data sources and preprocessing the raw data to obtain a training data set, where the training data set is used to perform model training on a pre-constructed initial deep learning framework to obtain an intelligent question-answering model; a knowledge base calling unit for responding to a standby power generation question request output by a client through an interaction interface, and after parsing the standby power generation question request, calling the knowledge base corresponding to the standby power generation system, where the knowledge base is constructed by extracting specified standby power generation association information data based on the standby power generation system requirement scheme and the standby power generation database; an answer retrieval unit for retrieving, through the intelligent question-answering model, reply information corresponding to the standby power generation question request from the knowledge base to generate a standby power generation answer; and an answer sending unit for returning the standby power generation answer to the front-end display interface of the client.

[0014] Optionally, the data preprocessing unit includes: a data classification module for classifying the raw data to obtain multi-modal data, where the types of the multi-modal data at least include: text, image, audio, video; and a preprocessing module for performing data cleaning, data denoising, normalization processing, and data format conversion on the multi-modal data to complete the preprocessing process of the raw data and obtain the training data set.

[0015] Optionally, the knowledge base corresponding to the standby power generation system is constructed in the following manner: extracting standby power generation equipment structure information, power generation signal quantities, and power outage signal quantities from the standby power generation database using a preset compilation language to obtain first extraction data; generating an unstructured knowledge document based on the first extraction data; extracting standby power generation system site information and path information from the standby power generation system requirement scheme and design scheme using a preset compilation language to obtain second extraction data; extracting functional operation steps from the operation manual and test cases associated with the standby power generation system using a preset compilation language to obtain third extraction data; synthesizing the second extraction data and the third extraction data to generate a structured knowledge graph; and constructing the knowledge base corresponding to the standby power generation system based on the structured knowledge graph and the unstructured knowledge document.

[0016] Optionally, the answer retrieval unit includes: a first parsing module, configured to parse the standby power generation question request to obtain the identification information of the target power generation device to be located when the type of the standby power generation question request is power generation device positioning; a first knowledge base calling module, configured to call the knowledge base corresponding to the standby power generation system based on the identification information, and output, through the intelligent question and answer model, the function positioning information and navigation path information associated with the target power generation device to obtain the standby power generation answer; or, a second parsing module, configured to parse the standby power generation question request to obtain the identification information of the target power generation device to be located when the type of the standby power generation question request is power generation quantity data analysis; a second knowledge base calling module, configured to call the knowledge base corresponding to the standby power generation system based on the identification information, and output, through the intelligent question and answer model, the power generation quantity data associated with the target power generation device to obtain the standby power generation answer, where the power generation quantity data at least includes: the date and the corresponding power generation quantity, the period and the corresponding power generation quantity, and the peak and valley of the power generation quantity.

[0017] Optionally, the answer retrieval unit includes: a third parsing module, configured to parse the standby power generation question request to obtain the identification information of the target power generation device when the type of the standby power generation question request is site power outage analysis; a third knowledge base calling module, configured to call the knowledge base corresponding to the standby power generation system based on the identification information of the target power generation device, analyze whether the power generation quantity of the target power generation device meets the power consumption demand of a predetermined area, and when the analysis result indicates that the power generation quantity of the target power generation device does not meet the power consumption demand of the predetermined area, output, through the intelligent question and answer model, the power outage prompt information associated with the predetermined area and the device information of the standby power generation device that can be called to obtain the standby power generation answer.

[0018] Optionally, the answer retrieval unit includes: a fourth parsing module, configured to parse the standby power generation question request to obtain the identification information of all standby power generation devices to be analyzed, the operating status of each standby power generation device, and the device detection pictures of each standby power generation device when the type of the standby power generation question request is faulty device analysis; a fourth knowledge base calling module, configured to perform fault analysis on each standby power generation device by calling the knowledge base corresponding to the standby power generation system based on the identification information of each standby power generation device, the operating status of the standby power generation device, and the device detection pictures of the standby power generation device, and output, through the intelligent question and answer model, the standby power generation answer including the faulty device information and the fault repair suggestion report.

[0019] Optionally, the answer sending unit includes: an answer encryption module for encrypting the backup power generation answer using a homomorphic encryption strategy; an answer returning module for returning the encryption result to the front-end display interface of the user terminal through an end-to-end encryption transmission strategy; and an updating module for collecting the answer feedback evaluation of the user terminal and iteratively updating the intelligent question-answering model and the knowledge base based on the answer feedback evaluation.

[0020] According to another aspect of the embodiments of the present invention, there is also provided a computer-readable storage medium, which includes a stored computer program. When the computer program runs, it controls the device where the computer-readable storage medium is located to execute the intelligent question-answering method based on a large model as described in any one of the above.

[0021] According to another aspect of the embodiments of the present invention, there is also provided an electronic device, including one or more processors and a memory. The memory is used to store one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors are caused to implement the intelligent question-answering method based on a large model as described in any one of the above.

[0022] According to another aspect of the embodiments of the present invention, there is also provided a computer program product, including a computer program. When the computer program is executed by a processor, it implements the steps of the intelligent question-answering method based on a large model as described in any one of the above.

[0023] In the present disclosure, raw data is collected from multiple data sources and preprocessed to obtain a training data set. The training data set is used to train a pre-constructed initial deep learning framework to obtain an intelligent question-answering model. In response to a backup power generation question request output by the user terminal through an interaction interface, after parsing the backup power generation question request, the knowledge base corresponding to the backup power generation system is called. The knowledge base is constructed by extracting data of specified backup power generation association information based on the backup power generation system requirement scheme and the backup power generation database. The intelligent question-answering model retrieves the reply information corresponding to the backup power generation question request from the knowledge base to generate a backup power generation answer, and the backup power generation answer is returned to the front-end display interface of the user terminal.

[0024] From the above disclosure, in the question-answering process in the backup power generation field, the task of parsing and initially answering questions can be assigned to a lightweight model, and the pre-constructed knowledge base is called to search for reply answers that meet the specific backup power generation system field, reducing the system retrieval time, reducing the resource consumption during reasoning, and reducing the server retrieval cost, thereby solving the technical problem in the related art that the intelligent question-answering application based on a large model consumes a large amount of resources and results in high costs when dealing with a specific field. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The accompanying drawings described herein are used to provide a further understanding of the present invention and form a part of this application. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0026] Figure 1 A hardware structure block diagram of a computer terminal for implementing an intelligent question-answering method based on a large model is shown;

[0027] Figure 2 is a flowchart of an optional intelligent question-answering method based on a large model according to an embodiment of the present invention;

[0028] Figure 3 is a system structure diagram of an optional intelligent question-answering application based on a large model according to an embodiment of the present invention;

[0029] Figure 4 is a schematic diagram of an optional intelligent question-answering device based on a large model according to an embodiment of the present invention;

[0030] Figure 5 is a structure block diagram of an electronic device according to an embodiment of the present application. Detailed implementation manners

[0031] In order to enable those skilled in the art of this technology to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0032] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and do not necessarily need to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present invention described here can be implemented in an order other than those illustrated or described here. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or are inherent to these processes, methods, products, or devices.

[0033] To facilitate the understanding of the present invention by those skilled in the art, the following explanations are made for some terms or nouns involved in the embodiments of the present invention:

[0034] Transformer, a deep learning model architecture for processing sequence data, does not rely on the sequential dependencies in the sequence for information transfer. Instead, it uses the self-attention mechanism to process all elements in the sequence in parallel. In the present invention, it is applied as a core component in the large model intelligent question-answering system to handle natural language understanding tasks for complex and long texts.

[0035] BERT, Bidirectional Encoder Representations from Transformers, is a pre-trained model based on the Transformer architecture that can be fine-tuned in various downstream tasks, such as text classification, sentiment analysis, question-answering systems, etc., to achieve specific natural language processing tasks. In the standby power generation intelligent question-answering application, using the pre-trained BERT model can greatly improve the question-answering system's ability to understand user questions and generate accurate answers.

[0036] Hbase is a distributed, multi-dimensional, sorted mapping table data storage system based on the Hadoop framework. It can store a large amount of data (PB level) and provides real-time read and write capabilities. HBase stores data using column families and can perform dynamic column expansion, making it very suitable for storing structured and semi-structured massive data. In the standby power generation system of the present invention, the HBase database may be used to store real-time or historical data related to power generation, equipment status, signal volume, etc. These data are crucial for question-answering scenarios such as predicting power outages and analyzing equipment failures.

[0037] The Mongodb database is an open-source NoSQL database suitable for processing unstructured and semi-structured data, such as multimedia data like text, images, audio, etc., as well as data files in various formats, such as Word documents and CSV tables. In the standby power generation intelligent question-answering application of the present invention, MongoDB may be used to store document data such as the requirement plans and operation manuals of the standby power generation system, as well as the structured knowledge information extracted from these documents. These data are of great value for building a knowledge base, understanding user questions, and generating accurate answers.

[0038] It should be noted that the large model-based intelligent question-answering method and its device in the present disclosure can be used in the field of artificial intelligence technology. In the case of realizing intelligent question-answering in a specific field based on artificial intelligence, it can also be used in any field other than the artificial intelligence technology field. In the case of realizing intelligent question-answering in a specific field based on artificial intelligence, the application field of the large model-based intelligent question-answering method and its device in the present disclosure is not limited.

[0039] It should be noted that the information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) collected in this disclosure are information and data that have been authorized by the user or fully authorized by all parties. Moreover, for the processing of relevant data such as collection, storage, use, processing, transmission, provision, disclosure, and application, all comply with the relevant laws, regulations, and standards of the relevant regions, adopt necessary confidentiality measures, do not violate public order and good customs, and provide corresponding operation entrances for users to choose to authorize or refuse. For example, there is an interface between this system and relevant users or institutions. Before obtaining relevant information, a request for acquisition needs to be sent to the aforementioned users or institutions through the interface, and after receiving the consent information feedback from the aforementioned users or institutions, the relevant information can be obtained.

[0040] It should be noted that in this disclosure, when collecting customer information, analyzing customer information, an operation entrance is provided for users to choose to agree or refuse the automated decision-making result; if the user chooses to refuse, the expert decision-making process will be entered.

[0041] The following embodiments of the present invention can be applied to various systems / applications / devices of intelligent question answering based on large models. The present invention adopts advanced deep learning technologies and natural language processing technologies, as well as optimized designs for database management systems and front-end interactions. The combination of these technologies enables the system to process massive data more quickly, reducing the time for data processing and model inference. At the same time, the optimized front-end design reduces the waiting time of users and improves the overall response speed of the system.

[0042] In addition, the present invention supports multiple input methods (text, voice) and personalized recommendations and interaction optimizations based on the user's historical behavior and preferences, enabling the system to be closer to the actual needs and usage habits of users. This personalized service method enhances the user's sense of participation and satisfaction.

[0043] At the same time, advanced encryption technologies and anonymization processing means are integrated in the present invention to ensure the security of user data during transmission and processing, preventing data leakage and malicious attacks.

[0044] The present invention will be described in detail below in conjunction with each embodiment.

[0045] Embodiment 1

[0046] According to an embodiment of the present invention, an embodiment of a method for intelligent question answering based on a large model is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. And although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0047] The embodiment of the intelligent question-answering method based on a large model provided by Embodiment 1 of the present application can be executed in a computer terminal or a similar computing device. Figure 1 A hardware structure block diagram of a computer terminal for implementing an intelligent question-answering method based on a large model is shown. As Figure 1 shown, the computer terminal 10 may include one or more ( Figure 1 shown as 102a, 102b,..., 102n in the figure) processors 102 (the processor 102 may include, but is not limited to, processing devices such as a microprocessor MCU (Microcontroller Unit) or a field programmable gate array FPGA (Field Programmable Gate Array)), a memory 104 for storing data, and a transmission device 106 for communication functions. In addition, it may further include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the BUS bus), a network interface, a power supply, and / or a camera. Those of ordinary skill in the art can understand that Figure 1 the structure shown is only illustrative and does not limit the structure of the above-mentioned electronic device. For example, the computer terminal 10 may further include more or fewer components than Figure 1 shown in the figure, or have a different configuration from Figure 1 shown in the figure.

[0048] It should be noted that the above one or more processors 102 and / or other data processing circuits are generally referred to as "data processing circuits" in this article. The data processing circuit may be fully or partially embodied as software, hardware, firmware, or any combination thereof. In addition, the data processing circuit may be a single independent processing module, or fully or partially incorporated into any one of the other elements in the computer terminal 10 (or mobile device). As involved in the embodiments of the present application, the data processing circuit is a processor control (such as the selection of a variable resistance terminal path connected to an interface).

[0049] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the large model-based intelligent question-answering method in the embodiments of the present application. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, that is, implements the above-mentioned large model-based intelligent question-answering method. The memory 104 may include high-speed random access memory and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memories. In some instances, the memory 104 may further include a memory remotely disposed relative to the processor 102, and these remote memories can be connected to the computer terminal 10 through a network. Examples of the above-mentioned network include but are not limited to the Internet, enterprise intranet, local area network, mobile communication network, and combinations thereof.

[0050] The transmission device 106 is used to receive or send data via a network. Specific examples of the above-mentioned network may include the wireless network provided by the communication provider of the computer terminal 10. In one instance, the transmission device 106 includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices through a base station and thus can communicate with the Internet. In one instance, the transmission device 106 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0051] The display can be, for example, a touch-screen liquid crystal display (Liquid Crystal Display, abbreviated as LCD), and this liquid crystal display enables users to interact with the user interface of the computer terminal 10 (or mobile device).

[0052] The large model-based intelligent question-answering application in this embodiment is a question-answering system for web environments based on artificial intelligence technology. Through a large language model, it learns language patterns on a huge dataset, so as to be able to understand and answer natural language queries of users in specific fields. Users only need to input questions on the web page to obtain detailed and personalized answers without manual intervention, greatly improving the convenience and efficiency of information acquisition.

[0053] In the above operating environment, the present application provides a large model-based intelligent question-answering method as Figure 2 shown. Figure 2 is a flowchart of an optional large model-based intelligent question-answering method according to an embodiment of the present invention, as Figure 2 shown, and the method includes the following steps:

[0054] Step S201: Collect raw data from multiple data sources and preprocess the raw data to obtain a training dataset, where the training dataset is used to train a pre-constructed initial deep learning framework to obtain an intelligent question-answering model.

[0055] In the embodiments of the present invention, the intelligent question-answering model and the subsequent constructed knowledge base are applied to a specific technical field - the intelligent question-answering field of power generation and standby power, to realize the reply to the question related to the power generation and standby power system.

[0056] In the initial stage of the intelligent question-answering application for power generation and standby power, the system needs to collect raw data from different data sources, including but not limited to the power generation and standby power system requirement documents, design documents, operation manuals, and HBase and MongoDB databases. The data collection process is automated using programming languages such as Python to improve efficiency and reduce human errors.

[0057] Once the raw data is collected, the system will perform a series of preprocessing operations. Optionally, the steps of preprocessing the raw data to obtain a training dataset include: classifying the raw data to obtain multimodal data, where the types of multimodal data at least include: text, image, audio, video; performing data cleaning, data denoising, normalization processing, and data format conversion on the multimodal data to complete the preprocessing process of the raw data and obtain a training dataset.

[0058] Among them, classifying the raw data to obtain multimodal data means that the raw data often comes from multiple data sources such as the power generation and standby power system requirement documents, design documents, operation manuals, and HBase and MongoDB databases. These data sources may contain different types of multimodal data, including: Text: Requirement documents, design documents, operation manuals, etc. of the power generation and standby power system, providing text descriptions of system functions, operation processes, and operation and maintenance knowledge. Image: Structure diagrams of power generation and standby power equipment, system layout diagrams, image materials of fault cases, etc., providing visual information support for the intelligent question-answering system. Audio: Audio guides for power generation and standby power system operations, voice records of fault diagnosis, etc., providing operation guidance or fault descriptions in voice mode. Video: Operation tutorial videos of the power generation and standby power system, videos of the fault diagnosis process, etc., integrating dynamic visual and auditory information, providing a more intuitive operation and maintenance scenario for the intelligent question-answering system. Through programming languages such as Python, the system can automatically extract the specified power generation and standby power-related information from the above data sources to form a multimodal dataset. The construction of the multimodal dataset takes into account the characteristics and application scenarios of different modal data, providing rich and comprehensive data input for the intelligent question-answering model.

[0059] The preprocessing operations include data cleaning, denoising, standardization, and format conversion. The purpose of data cleaning is to remove errors, redundancy, and inconsistent information from the data. For example, for text data, the system needs to detect and correct spelling mistakes, punctuation errors, grammar errors, and delete duplicate or irrelevant information. For image or audio data, it may be necessary to filter out noise or identify the main features. Data denoising is a process to further improve the data quality, especially for image and audio data. This may involve applying filtering techniques to remove background noise, or using image processing algorithms to eliminate irrelevant or blurred details in the image to ensure that the model can more accurately identify and understand the key information in the data. Standardization processing is to ensure the comparability and consistency of the data numerically. For example, converting numerical data to the same scale, such as using z-score standardization, can avoid certain features having a bias on model training due to a large numerical range. For text data, it may include operations such as converting text to lowercase and removing stop words to reduce data complexity. Data format conversion is to uniformly convert the original data obtained from different data sources into a format suitable for model training. For example, converting an image to a pixel value vector, converting audio to a spectrogram or Mel-frequency cepstral coefficients (MFCC), and converting text to a word embedding vector, etc. This process ensures the consistency of the data, enabling the model to correctly parse and utilize data of each modality.

[0060] After the preprocessing process is completed, the system will obtain a high-quality training dataset, which contains multi-modal data that is clearly classified, clean, standardized, and in a unified format. This dataset will be used to train deep learning models for intelligent question answering, such as models based on the Transformer architecture or BERT models. The models during the training process can learn rich features and patterns from the multi-modal data, thus providing more accurate and comprehensive response information in the question answering scenario.

[0061] The preprocessed data is organized into a training dataset, which is then used to train an initial deep learning framework, such as a model based on the Transformer architecture. The construction of the training dataset needs to consider the diversity and balance of the data to ensure that the model can generalize to unseen data, thereby improving the accuracy and robustness of the intelligent question answering model.

[0062] Step S202, in response to the power generation standby question request output by the client through the interaction interface, after parsing the power generation standby question request, call the knowledge base corresponding to the power generation standby system, where the knowledge base is constructed by extracting data of specified power generation standby associated information based on the power generation standby system requirement plan and the power generation standby database.

[0063] It should be noted that this embodiment does not limit the types of power generation backup question requests output by the user terminal through the interaction interface, realizing the intelligent integration of various input modalities such as text and voice, and being able to automatically analyze and convert data of different modalities (i.e., realizing multi-modal input and intelligent integration), providing a richer and more diverse input source for the question-and-answer system.

[0064] When the user inputs a question related to the power generation backup system through the Web front-end interface, the system first analyzes the user's question request. During the analysis process, natural language processing technology is involved to understand the intention and semantics of the user's input question. The analyzed question will be further converted into a query format that the system can recognize and process.

[0065] The system then calls the corresponding knowledge base of the power generation backup system. This knowledge base is constructed by deeply analyzing the requirements scheme of the power generation backup system and the information in the database, and extracting relevant associated information related to the power generation backup system from text and tabular data using tools such as Python. The knowledge base structure includes a structured knowledge graph and an unstructured document library to support the answering of different types of questions.

[0066] Step S203, retrieve the reply information corresponding to the power generation backup question request from the knowledge base through the intelligent question-and-answer model to generate a power generation backup answer.

[0067] After calling the knowledge base, the intelligent question-and-answer model starts to work. The model first searches for the most relevant reply information from the knowledge base based on the analyzed question request. During the search process, deep semantic matching technology is involved to ensure that the retrieved information best meets the user's question intention.

[0068] The retrieved reply information is then used by the intelligent question-and-answer model to generate the final answer. The model synthesizes various information sources, including structured knowledge graphs, unstructured document information, and external data sources, to generate a power generation backup answer that is as comprehensive and accurate as possible.

[0069] Step S204, return the power generation backup answer to the front-end display interface of the user terminal.

[0070] Finally, the generated power generation backup answer is displayed to the user through the Web front-end interface. The front-end display interface is designed to be responsive and user-friendly, can adapt to the displays of different devices (such as desktop computers, tablets, mobile phones), and can present the answer in multiple formats (such as text, images, charts) to meet the needs and preferences of different users.

[0071] This interface also allows users to provide feedback on the answer, whether it is a positive confirmation or pointing out errors. These feedback messages will be captured by the system and used for subsequent model training and knowledge base update, forming a continuously optimized closed loop.

[0072] Through the above steps, raw data can be collected from multiple data sources, and the raw data can be preprocessed to obtain a training dataset. The training dataset is used to train a model for a pre-constructed initial deep learning framework to obtain an intelligent question-answering model. In response to a power generation standby question request output by the client through the interaction interface, after parsing the power generation standby question request, the knowledge base corresponding to the power generation standby system is called. The knowledge base is constructed by extracting data of specified power generation standby related information based on the power generation standby system requirement scheme and the power generation standby database. The intelligent question-answering model is used to retrieve reply information corresponding to the power generation standby question request from the knowledge base and generate a power generation standby answer. The power generation standby answer is returned to the front-end display interface of the client. In this embodiment, the parsing and preliminary answering tasks of the question can be assigned to a lightweight model, and the pre-constructed knowledge base is called to search for reply answers that conform to a specific power generation standby system domain, reducing the system retrieval time, reducing the resource consumption during inference, and reducing the server retrieval cost, thereby solving the technical problem in the related art that the intelligent question-answering application based on a large model has a large resource consumption and high cost when dealing with a specific domain.

[0073] Optionally, the knowledge base corresponding to the power generation standby system is constructed in the following manner: The power generation standby equipment structure information, power generation signal quantity, and power outage signal quantity are extracted from the power generation standby database using a preset compilation language to obtain first extraction data; Based on the first extraction data, an unstructured knowledge document is generated; The power generation standby system site information and path information are extracted from the power generation standby system requirement scheme and design scheme using a preset compilation language to obtain second extraction data; The functional operation steps are extracted from the operation manual and test cases related to the power generation standby system using a preset compilation language to obtain third extraction data; The second extraction data and the third extraction data are combined to generate a structured knowledge graph; Based on the structured knowledge graph and the unstructured knowledge document, the knowledge base corresponding to the power generation standby system is constructed.

[0074] Among them, the first extraction data includes: extracting equipment structure information, power generation signal quantity, and power outage signal quantity from the power generation standby database. Here, the power generation standby database, such as HBase and MongoDB, stores the operation data of the power generation standby system, including information such as the real-time status of equipment, historical records, fault signals, and power generation quantity. During the data extraction process, a preset programming language, such as Python, can be used to write scripts or programs to extract data related to the power generation standby equipment structure, power generation signal quantity, and power outage signal quantity from the database. The obtained first extraction data will be sorted and analyzed. For example, the equipment structure information will be converted into an easy-to-understand format, and statistical and trend analyses will be performed on the power generation quantity and power outage signal quantity. These data will be used to construct an unstructured knowledge document to provide background information on the operation status of the power generation standby system.

[0075] Based on the first extracted data retrieved from the database, the system generates unstructured knowledge documents. These documents may include equipment descriptions, analysis of historical power generation records, outage case descriptions, etc., which are used to provide the basic knowledge and operation background of the backup power generation system.

[0076] For the second extracted data, the data sources include: the requirement plan and design plan of the backup power generation system. These documents detail information such as the construction objectives, technical specifications, site distribution, and functional paths of the backup power generation system. During the data extraction process, a preset programming language, such as Python, is also used to read and parse these documents, and extract data related to site information and functional paths. Site information and path information are crucial for understanding and locating the operation and maintenance issues of the backup power generation system and can be used to construct the key nodes and association relationships in the knowledge graph.

[0077] For the third extracted data, the data sources include: operation manuals and test case documents related to the backup power generation system. These materials detail how to operate the backup power generation system and the specific steps for testing. During the data extraction process, through automated scripts or programs, the steps and processes of functional operations are extracted from the operation manuals and test cases. This may include using natural language processing techniques to understand the operation instructions in the text and using pattern matching techniques to identify the key operations and results in the test cases.

[0078] Integrate the site information, path information, and functional operation steps to construct a knowledge graph. A knowledge graph is a graphical data structure with nodes representing entities and edges representing relationships between entities, which is used to represent the relationships between the components, locations, functions, and operation processes of the backup power generation system. When constructing the knowledge graph, graph database technologies (such as Neo4j) or entity recognition and relationship extraction technologies based on deep learning can be used. Each node in the graph represents an entity in the backup power generation system (such as equipment, site, operation step), and the edges represent the associations between entities (such as the ownership relationship between equipment and site, and the execution order between operation steps and functional paths).

[0079] Combine the unstructured knowledge documents with the structured knowledge graph to construct a comprehensive knowledge base. The knowledge base not only contains the operation data of the backup power generation system but also includes the design, operation, and operation and maintenance knowledge of the system, forming a comprehensive information resource library.

[0080] Optionally, the step of retrieving reply information corresponding to the standby power generation question request from the knowledge base through the intelligent Q&A model and generating the standby power generation answer includes: when the type of the standby power generation question request is power generation equipment location, parsing the standby power generation question request to obtain the identification information of the target power generation equipment to be located; based on the identification information, calling the knowledge base corresponding to the standby power generation system, and outputting the function location information and navigation path information associated with the target power generation equipment through the intelligent Q&A model to obtain the standby power generation answer; or, when the type of the standby power generation question request is power generation quantity data analysis, parsing the standby power generation question request to obtain the identification information of the target power generation equipment to be located; based on the identification information, calling the knowledge base corresponding to the standby power generation system, and outputting the power generation quantity data associated with the target power generation equipment through the intelligent Q&A model to obtain the standby power generation answer, where the power generation quantity data at least includes: date and corresponding power generation quantity, period and corresponding power generation quantity, peak and valley of power generation quantity.

[0081] When a user submits a question request regarding power generation equipment location through the interaction interface, the system first parses this request to understand the intent and key information of the user's question. For example, if the user asks "How to find the generator numbered 12345?", the system will identify "the generator numbered 12345" as the target power generation equipment to be located, and "How to find" indicates that the user needs function location information and navigation path information. Then, from the parsed standby power generation question request, the system extracts the identification information of the target power generation equipment, such as the equipment number, name, or location code. The identification information is crucial for accurately locating and obtaining the function information of the specific equipment. Then, the knowledge base corresponding to the standby power generation system is called. This knowledge base contains a structured knowledge graph and unstructured knowledge documents. In the knowledge base, the system uses the intelligent Q&A model to retrieve based on the identification information of the target power generation equipment.

[0082] For the retrieval of equipment location information, the intelligent Q&A model may query the nodes in the knowledge graph that match the equipment identification, and obtain the location of the equipment, the affiliated site, and the related function description. At the same time, the model may also retrieve unstructured knowledge documents to provide more detailed equipment information and operation guidance.

[0083] For the retrieval of navigation path information, the model may need further processing, such as calculating the optimal path between the equipment and the user's current location, which involves the site network structure in the knowledge graph, as well as actual geographical information and road data. According to the retrieved function location information and navigation path information, the intelligent Q&A model generates the standby power generation answer. The answer may be returned in the form of structured data, such as the specific location of the equipment, the detailed information of the affiliated site, the function description, and the navigation path from the user's current location to the equipment location. In addition, this embodiment may also provide an intuitive map or roadmap for the user to more clearly understand how to reach the target equipment.

[0084] When the user asks questions about power generation data analysis, the system also needs to parse the request and understand the user's query requirements for power generation data. For example, when the user asks "What is the power generation of the generator numbered 12345 in the past week?", the system will identify "the generator numbered 12345" and "the power generation in the past week" as the key query conditions. The identification information of the target power generation equipment to be located is obtained from the parsing, such as the equipment number, and then the knowledge base is called. The intelligent question-answering model retrieves the relevant power generation data based on the equipment identification information. The data may be stored in databases such as HBase or MongoDB, including information such as the date and corresponding power generation, the cycle and corresponding power generation, and the peak and valley of power generation. The model needs to understand the time range (such as "the past week") and extract the corresponding data records from the database.

[0085] The model further processes the retrieved power generation data, which may include statistical analysis, trend analysis, and periodic pattern recognition. For example, calculate the average power generation per day in a week, identify the peak and trough periods of power generation, and analyze the change trend of power generation over time.

[0086] Based on the analysis results of the power generation data, the intelligent question-answering model generates standby power answers. The answers may include specific numbers, such as the daily power generation, or may provide charts or graphs, such as the power generation trend chart or periodic pattern chart. In addition, the model may also predict the power generation for a period of time in the future based on historical data, providing decision support for the operation and management of the standby power system.

[0087] Optionally, the steps of generating the standby power answer by retrieving the reply information corresponding to the standby power question request from the knowledge base through the intelligent question-answering model include: in the case where the type of the standby power question request is site power outage analysis, parsing the standby power question request to obtain the identification information of the target power generation equipment; based on the identification information of the target power generation equipment, calling the corresponding knowledge base of the standby power system, analyzing whether the power generation of the target power generation equipment meets the electricity demand of a predetermined area, and when the analysis result indicates that the power generation of the target power generation equipment does not meet the electricity demand of the predetermined area, outputting, through the intelligent question-answering model, the power outage prompt information associated with the predetermined area and the equipment information of the standby power generation equipment that can be called to obtain the standby power answer.

[0088] When a user asks a question about substation power outage analysis through the interaction interface, the system first performs parsing. This process involves natural language understanding technology to ensure accurate understanding of the user's intention and the specific content of the request. For example, the user may ask "Is there a possibility of power outage at XX substation and the situation of backup power generation equipment?" After parsing, the system can extract key information from the question, including the identification information of the target power generation equipment. Then, according to the target power generation equipment identification obtained from the parsing, the system accesses the knowledge base corresponding to the backup power generation system, which contains detailed information about backup power generation equipment, historical operation data, equipment failure records, and system design requirements, etc. Querying the knowledge base may involve structured data query (such as SQL query) and unstructured data retrieval (such as keyword-based text search) to obtain the current status and historical power generation of the target power generation equipment.

[0089] It should be noted that the system in this embodiment can call the HBase and MongoDB database information stored in the knowledge base to analyze the real-time power generation of the target power generation equipment and the power generation trend in a recent period. By comparing with the power consumption demand in a predetermined area, it is evaluated whether the current power generation is sufficient to meet the power demand of this area. If the analysis result shows that the power generation of the target power generation equipment is sufficient to meet the power consumption demand of the predetermined area, the system may not need to perform further operations and directly feedback the analysis result to the user. However, if the power generation does not meet the demand, the system will enter the next step to provide power outage warning and backup power generation equipment scheduling suggestions.

[0090] In the case where the power generation does not meet the demand, the intelligent question-and-answer model will generate power outage warning information based on the information in the knowledge base, including the possibility of power outage, the estimated power outage time, the possible affected range, etc. At the same time, the model will retrieve the information about backup power generation equipment in the knowledge base, including the status, location, power generation capacity of the equipment, etc., to determine which backup equipment can be called to make up for the power gap. The system will output scheduling suggestions according to the availability of the equipment and the distance from the predetermined area, that is, which backup power generation equipment can be quickly started and meet the power demand of this area.

[0091] Based on the above analysis and retrieval results, the system will generate a detailed backup power generation answer, the content of which may include power outage warning, power demand analysis, a recommended list of backup power generation equipment, as well as specific operation steps and precautions for equipment scheduling. The backup power generation answer will be presented to the user in a clear and easy-to-understand format, including text description, chart display, equipment status snapshot, etc., to ensure that the user can quickly grasp the situation and make a response.

[0092] It should be noted that during the power outage analysis and spare equipment scheduling process, the intelligent question-answering model can utilize the structured knowledge graph to identify the relevance between devices, such as the affiliated station address of the device, the relationship between devices (such as the primary and backup relationship), and the matching degree between the device and the regional demand. This helps to more accurately analyze the power supply and demand situation and formulate scheduling suggestions. The system may integrate a prediction model based on machine learning to predict device power generation, regional power demand, and future power outage risks. The model can provide an estimate of future situations based on historical data and real-time information, so as to make responses in advance and reduce the impact of power outages on users.

[0093] In addition, when generating the backup power generation answer, the intelligent question-answering application may take into account the user's usage habits and preferences, such as providing interactive charts through which users can view the power generation at specific time points, or providing real-time updates on the device status, allowing users to immediately understand whether the device has been scheduled and the power generation situation.

[0094] Optionally, the steps of generating the backup power generation answer by retrieving the reply information corresponding to the backup power generation question request from the knowledge base through the intelligent question-answering model include: when the type of the backup power generation question request is the analysis of faulty devices, parsing the backup power generation question request to obtain the identification information of all backup power generation devices to be analyzed, as well as the operating status and device detection pictures of each backup power generation device; based on the identification information of each backup power generation device, the operating status and device detection pictures of the backup power generation device, calling the corresponding knowledge base of the backup power generation system to conduct a fault analysis on each backup power generation device, and outputting a backup power generation answer including faulty device information and a fault repair suggestion report through the intelligent question-answering model.

[0095] When the user submits a question request regarding the analysis of faulty devices through the Web front-end interface, the system first needs to parse this request. After parsing, it obtains the identification information of the device, the current operating status of the device, and possibly attached device detection pictures, and extracts the semantics of the device name, fault phenomenon, and detection pictures described in the request, so as to accurately locate the device and fault type to be analyzed. After parsing the question request, the system will call the corresponding knowledge base of the backup power generation system to conduct a fault analysis based on the identification information, operating status, and device detection pictures of each backup power generation device.

[0096] It should be noted that each standby power generation device has its unique identification information, including device ID, model, manufacturer, etc. These information are used to locate specific devices and access relevant materials. Operating status: The operating status data of the device, such as real-time or historical monitoring data of voltage, current, temperature, pressure, etc., are important bases for judging whether the device is faulty, the degree of fault and the type of fault. Device detection pictures: For some faults, users may upload detection pictures of the device. These pictures may contain information such as damage to the device appearance and abnormal instrument readings, which are intuitive evidences for fault analysis.

[0097] Using the information in the knowledge base, the intelligent Q&A model begins to analyze the fault conditions of each standby power generation device. The model may use image recognition technology to detect abnormalities in the device detection pictures, and at the same time combine the operating status data of the device to conduct in-depth fault cause analysis and identify possible fault types and locations. After the fault analysis is completed, the intelligent Q&A model will generate a report containing the information of the faulty device and maintenance suggestions as the standby power generation answer to be returned. Faulty device information: Accurately indicate which devices have faults, the types of faults and possible causes. Maintenance suggestion report: According to the results of the fault analysis, generate specific maintenance steps, required tools and spare parts, safety precautions and possible preventive measures. The information in the report will be as detailed and specific as possible so that the operation and maintenance personnel can perform accurate fault location and effective maintenance operations based on the report.

[0098] The intelligent Q&A model can use natural language generation (NLG) technology to generate a clear and detailed maintenance suggestion report according to the results of the fault analysis. The report will use easy-to-understand language, combine the specific information of the faulty device and the maintenance knowledge in the knowledge base to provide personalized maintenance guidance. The finally generated standby power generation answer will be displayed to the user through the Web front-end interface. The interface will present the detailed information of the faulty device and the maintenance suggestion report, which may be displayed in the form of text, table and pictures to ensure the intuitiveness and readability of the information. At the same time, the system allows users to give feedback on the answer, so as to iteratively update the intelligent Q&A model and the knowledge base in the future to further improve the accuracy and efficiency of fault analysis.

[0099] Through the above embodiments, the standby power generation intelligent Q&A application can effectively process the question requests for faulty device analysis, provide professional and personalized maintenance suggestions, reduce the time and energy of the operation and maintenance personnel in fault troubleshooting and maintenance decision-making, and improve the operation and maintenance efficiency and safety of the standby power generation system.

[0100] Optionally, the step of returning the standby power generation answer to the front-end display interface of the user terminal includes: encrypting the standby power generation answer using a homomorphic encryption strategy; returning the encrypted result to the front-end display interface of the user terminal through an end-to-end encryption transmission strategy; collecting the answer feedback evaluation of the user terminal, and iteratively updating the intelligent Q&A model and the knowledge base based on the answer feedback evaluation.

[0101] When the intelligent Q&A system generates backup power generation answers, in order to protect user privacy and the corresponding transmitted data, the system uses homomorphic encryption technology to encrypt the answers. Homomorphic encryption is a cryptographic technique that allows computations to be performed on encrypted data without first decrypting the data. This means that even if the answer is intercepted during transmission, the attacker cannot understand its true content, thus protecting the security of the answer and user privacy. Among them, according to application requirements, the encryption methods that the system can choose include: fully homomorphic encryption, partially homomorphic encryption, or homomorphic encryption for multi-keyword search. Fully homomorphic encryption allows arbitrary computations to be performed on encrypted data, while partially homomorphic encryption supports specific types of computations, such as addition or multiplication.

[0102] The encrypted backup power generation answers are sent to the user side through an end-to-end encryption transmission strategy, ensuring the security of the data during transmission. End-to-end encryption ensures that only the sender and receiver of the message can access the plaintext data. Even if the data is intercepted by a third party during transmission, it cannot be decrypted to read the content. During transmission, in order to ensure that the answer can only be accessed by legitimate users, this embodiment can implement a user authentication mechanism. After receiving the encrypted backup power generation answers, the user side decrypts and views them, and then can provide feedback on the accuracy and satisfaction of the answers. The system collects these feedbacks for iterative updating of the intelligent Q&A model and knowledge base to continuously improve the system performance and user experience.

[0103] Optionally, the intelligent Q&A model of this embodiment can design a feedback collection interface that allows users to express their satisfaction with the answers by rating, commenting, or selecting whether the answer is correct. At the same time, the system can automatically record the user's feedback behavior, such as query time, device information, etc., to assist subsequent data analysis. The collected feedback data will be analyzed to identify errors, deficiencies, and changes in user needs in the system. According to the feedback analysis results, the intelligent Q&A model and knowledge base will be updated and optimized. For example, if the feedback shows that a certain type of answer is inaccurate, the system can retrain the model, adjust the algorithm parameters, or update the information in the knowledge base to improve the processing ability of such problems. The update process may be an ongoing cycle, and as more user feedback accumulates, the system performance will be continuously improved.

[0104] Through the above steps, the backup power generation intelligent Q&A system can not only safely return answers to users, but also continuously improve itself through user feedback, forming a dynamic and adaptive learning system, ensuring the security, accuracy, and user satisfaction of the system, and is an essential part of building modern intelligent Q&A applications.

[0105] The following will be described in detail in combination with another optional specific implementation manner.

[0106] Figure 3 It is a system structure diagram of an optional large model-based intelligent Q&A application according to an embodiment of the present invention. As Figure 3 shown, the system includes: a data preprocessing module, a model training module, a knowledge base construction module, a Web front-end interface module, and a Q&A service module. The following will explain these modules separately.

[0107] 1. Data preprocessing module: Responsible for collecting, cleaning, and organizing raw data from various channels, including multi-modal data such as text, pictures, and audio. By data cleaning and standardization processing, improve data quality and consistency, and provide a high-quality data set for subsequent model training.

[0108] 2. Model training module: Adopt deep learning frameworks and algorithms, such as Transformer, BERT, etc., to train large-scale data sets to obtain a large model with deep semantic understanding and multi-modal processing capabilities. Through continuous iteration and optimization, improve the accuracy and generalization ability of the model.

[0109] 3. Knowledge base construction module: Based on multi-data sources such as the demand plan, design plan, operation manual, Hbase, and Mongodb database of the power generation and standby system, use forms such as word, csv, and direct database connection as input, and extract data such as demand information, functional operation steps, equipment fault signal quantities, and power outage signal quantities through python to construct a structured and unstructured knowledge base.

[0110] Through the knowledge base of the power generation and standby system, it can provide intelligent Q&A covering aspects such as function path positioning of the power generation and standby system, power generation data analysis, site power outage prediction, and faulty equipment prediction, providing rich knowledge base support and analysis and prediction.

[0111] Document data: Document data such as demand plans, design plans, operation manuals, and test cases are extracted through the python language and transferred to a vector database.

[0112] Semi-structured data: Collect basic data such as site addresses and equipment in Mongodb, and performance data such as power generation and signal quantities in Hbase, and convert them into structured data.

[0113] Illustrate with examples, user input questions and analysis:

[0114] The operator asks: Output the power generation trend of XX equipment in the past 7 days?

[0115] The system receives the question, understands the operator's needs, and wants to view the power generation trend chart of XX equipment in the past 7 days.

[0116] Retrieve XX equipment and power generation from the knowledge base, and match the power generation in the past 7 days.

[0117] Use a multi-modal large model to generate answers. The power generation of the XX device in the past 7 days is N degrees, M degrees, ... respectively, and generate a line chart and output it to the Q&A box for presentation to the user.

[0118] 4. Web Front-end Interface Module: Design an intuitive and user-friendly Web front-end interface that supports users to ask questions in various ways such as text, pictures, and voices. The interface adopts a responsive design, is compatible with various devices and browsers, and provides a good user experience.

[0119] Optionally, the interface in this embodiment adopts a responsive design, is compatible with various devices and browsers, and at the same time develops back-end interfaces to support data interaction and function implementation between the front end and the back end.

[0120] 5. Q&A Service Module: After receiving a user's question, use natural language processing technology to analyze and semantically understand the management questions of the power generation system, then retrieve relevant information from the knowledge base and generate answers. The answers are presented to the user side in various forms such as text, pictures, and voices, and support users to give feedback and make corrections.

[0121] After the user asks a question through the Web front-end interface, the back end receives the request and performs parsing and semantic understanding. Then, based on the knowledge base, information retrieval and answer generation are carried out, and the results are returned to the front end for display. At the same time, it supports users to give feedback and make corrections to improve the accuracy and personalization of the Q&A service.

[0122] Through the above implementation methods, the processing efficiency and response speed are significantly improved: Due to the adoption of advanced deep learning technology and natural language processing technology, as well as the optimized design of the database management system and front-end interaction, the combination of these technologies enables the system to process massive data more quickly, reducing the time for data processing and model inference. At the same time, the optimized front-end design reduces the user's waiting time and improves the overall response speed of the system.

[0123] Through the above implementation methods, the accuracy and reliability of the Q&A service can be enhanced, that is, by continuously iterating and optimizing model parameters and training strategies, and combining a rich and timely knowledge base, the system can more accurately understand user questions and retrieve the most relevant answers from massive information. This dual guarantee mechanism significantly improves the accuracy and reliability of the Q&A.

[0124] In addition, this implementation method can also improve the personalization and satisfaction of the user experience: Support multiple input methods (text, voice) and personalized recommendations and interaction optimizations based on the user's historical behavior and preferences, enabling the system to be closer to the actual needs and usage habits of users. This personalized service method enhances the user's sense of participation and satisfaction.

[0125] An intelligent Q&A application based on a large model adopting this embodiment shows significant advantages in terms of efficiency, accuracy, personalization, and scalability, effectively overcoming the shortcomings of the prior art and providing users with a better-quality, more efficient, and personalized Q&A service experience.

[0126] The following will be described in detail in conjunction with another embodiment.

[0127] Embodiment 2

[0128] An intelligent Q&A device based on a large model provided in this embodiment includes multiple implementation units, and each implementation unit corresponds to each implementation step in the above Embodiment 1. Its specific implementation manner and beneficial effects can be referred to the foregoing method embodiment and will not be elaborated here.

[0129] Figure 4 It is a schematic diagram of an optional intelligent Q&A device based on a large model according to an embodiment of the present invention. As Figure 4 shown, the intelligent Q&A device based on a large model may include: a data preprocessing unit 41, a knowledge base calling unit 42, an answer retrieval unit 43, and an answer sending unit 44.

[0130] Among them, the data preprocessing unit 41 is used to collect raw data from multiple data sources and preprocess the raw data to obtain a training data set, where the training data set is used to train a pre-constructed initial deep learning framework to obtain an intelligent Q&A model.

[0131] The knowledge base calling unit 42 is used to respond to the standby power generation question request output by the client through the interaction interface. After parsing the standby power generation question request, it calls the knowledge base corresponding to the standby power generation system, where the knowledge base is constructed by extracting data of specified standby power generation associated information based on the standby power generation system requirement plan and the standby power generation database.

[0132] The answer retrieval unit 43 is used to retrieve reply information corresponding to the standby power generation question request from the knowledge base through the intelligent Q&A model and generate a standby power generation answer.

[0133] The answer sending unit 44 is used to return the standby power generation answer to the front-end display interface of the client.

[0134] The above intelligent Q&A device based on large models can collect raw data from multiple data sources through the data preprocessing unit 41, and preprocess the raw data to obtain a training dataset. Among them, the training dataset is used to train the pre-constructed initial deep learning framework to obtain an intelligent Q&A model. The knowledge base calling unit 42 responds to the standby power generation question request output by the client through the interaction interface. After parsing the standby power generation question request, it calls the knowledge base corresponding to the standby power generation system. The knowledge base is constructed by extracting data of specified standby power generation related information based on the standby power generation system requirement plan and the standby power generation database. The answer retrieval unit 43 retrieves the reply information corresponding to the standby power generation question request from the knowledge base through the intelligent Q&A model to generate a standby power generation answer. The answer sending unit 44 returns the standby power generation answer to the front-end display interface of the client. In this embodiment, the parsing and preliminary answering tasks of the question can be assigned to a lightweight model, and the pre-constructed knowledge base is called to search for reply answers that meet the specific standby power generation system field, reducing the system retrieval time, reducing the resource consumption during inference, and reducing the server retrieval cost, thereby solving the technical problem in the related art that the intelligent Q&A application based on large models has a large resource consumption and high cost when dealing with specific fields.

[0135] Optionally, the data preprocessing unit includes: a data classification module for classifying the raw data to obtain multi-modal data, where the types of the multi-modal data at least include: text, image, audio, video; a preprocessing module for performing data cleaning, data denoising, normalization processing, and data format conversion on the multi-modal data to complete the preprocessing process of the raw data and obtain a training dataset.

[0136] Optionally, the knowledge base corresponding to the standby power generation system is constructed in the following manner: extracting standby power generation equipment structure information, power generation signal quantities, and power outage signal quantities from the standby power generation database using a preset compilation language to obtain first extraction data; generating an unstructured knowledge document based on the first extraction data; extracting standby power generation system site information and path information from the standby power generation system requirement plan and design plan using a preset compilation language to obtain second extraction data; extracting functional operation steps from the operation manual and test cases related to the standby power generation system using a preset compilation language to obtain third extraction data; synthesizing the second extraction data and the third extraction data to generate a structured knowledge graph; constructing the knowledge base corresponding to the standby power generation system based on the structured knowledge graph and the unstructured knowledge document.

[0137] Optionally, the answer retrieval unit includes: a first parsing module, configured to parse the standby power generation question request to obtain the identification information of the target power generation equipment to be located when the type of the standby power generation question request is power generation equipment location; a first knowledge base calling module, configured to call the knowledge base corresponding to the standby power generation system based on the identification information, and output the function location information and navigation path information associated with the target power generation equipment through an intelligent question-answering model to obtain the standby power generation answer; or, a second parsing module, configured to parse the standby power generation question request to obtain the identification information of the target power generation equipment to be located when the type of the standby power generation question request is power generation quantity data analysis; a second knowledge base calling module, configured to call the knowledge base corresponding to the standby power generation system based on the identification information, and output the power generation quantity data associated with the target power generation equipment through an intelligent question-answering model to obtain the standby power generation answer, where the power generation quantity data at least includes: date and corresponding power generation quantity, period and corresponding power generation quantity, power generation peak and valley.

[0138] Optionally, the answer retrieval unit includes: a third parsing module, configured to parse the standby power generation question request to obtain the identification information of the target power generation equipment when the type of the standby power generation question request is substation site power outage analysis; a third knowledge base calling module, configured to call the knowledge base corresponding to the standby power generation system based on the identification information of the target power generation equipment, analyze whether the power generation quantity of the target power generation equipment meets the power consumption demand of a predetermined area, and when the analysis result indicates that the power generation quantity of the target power generation equipment does not meet the power consumption demand of the predetermined area, output the power outage prompt information associated with the predetermined area and the equipment information of the standby power generation equipment that can be called through an intelligent question-answering model to obtain the standby power generation answer.

[0139] Optionally, the answer retrieval unit includes: a fourth parsing module, configured to parse the standby power generation question request to obtain the identification information of all standby power generation equipment to be analyzed, the operating status of each standby power generation equipment, and the equipment detection pictures of each standby power generation equipment when the type of the standby power generation question request is faulty equipment analysis; a fourth knowledge base calling module, configured to perform fault analysis on each standby power generation equipment by calling the knowledge base corresponding to the standby power generation system based on the identification information of each standby power generation equipment, the operating status of the standby power generation equipment, and the equipment detection pictures, and output the standby power generation answer including the faulty equipment information and the fault repair suggestion report through an intelligent question-answering model.

[0140] Optionally, the answer sending unit includes: an answer encryption module, configured to encrypt the standby power generation answer by adopting a homomorphic encryption strategy; an answer return module, configured to return the encryption result to the front-end display interface of the user terminal through an end-to-end encryption transmission strategy; an update module, configured to collect the answer feedback evaluation of the user terminal, and iteratively update the intelligent question-answering model and the knowledge base based on the answer feedback evaluation.

[0141] The above intelligent question-answering device based on a large model may further include a processor and a memory. The above data preprocessing unit 41, knowledge base calling unit 42, answer retrieval unit 43, answer sending unit 44, etc. are all stored in the memory as program units, and the processor executes the above program units stored in the memory to implement corresponding functions.

[0142] The above processor contains a kernel, and the kernel retrieves the corresponding program units from the memory. One or more kernels can be set, and by adjusting the kernel parameters, answer responses in a specific field based on the intelligent question-answering model can be achieved.

[0143] The above memory may include non-permanent memory in a computer-readable medium, random access memory (RAM), and / or non-volatile memory in the form of, for example, read-only memory (ROM) or flash RAM. The memory includes at least one storage chip.

[0144] Embodiment III

[0145] An embodiment of the present application can provide an electronic device. Figure 5 It is a structural block diagram of an electronic device according to an embodiment of the present application. As Figure 5 shown, the electronic device may include: one or more ( Figure 5 only one is shown in the figure) processors 502, a memory 504, a storage controller, and a peripheral interface. Among them, the peripheral interface is connected to a radio frequency module, an audio module, and a display.

[0146] Among them, the memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the intelligent question-answering method and device based on a large model in the embodiment of the present application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, implements the above intelligent question-answering method based on a large model. The memory may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memories. In some instances, the memory may further include a memory remotely set relative to the processor, and these remote memories can be connected to the terminal through a network. Examples of the above network include but are not limited to the Internet, an enterprise intranet, a local area network, a mobile communication network, and combinations thereof.

[0147] The processor can call the information and application programs stored in the memory through the transmission device to execute the following steps: collect raw data from multiple data sources, and preprocess the raw data to obtain a training dataset, where the training dataset is used to train the pre-constructed initial deep learning framework to obtain an intelligent question-answering model; in response to the standby power generation question request output by the client through the interaction interface, after parsing the standby power generation question request, call the knowledge base corresponding to the standby power generation system, where the knowledge base is constructed by extracting data of specified standby power generation related information based on the standby power generation system requirement plan and the standby power generation database; retrieve the reply information corresponding to the standby power generation question request from the knowledge base through the intelligent question-answering model to generate a standby power generation answer; return the standby power generation answer to the front-end display interface of the client.

[0148] The processor can also call the information and application programs stored in the memory through the transmission device to execute the following steps: classify the raw data to obtain multimodal data, where the types of the multimodal data at least include: text, image, audio, video; perform data cleaning, data denoising, normalization processing, and data format conversion on the multimodal data to complete the preprocessing process of the raw data and obtain a training dataset.

[0149] The processor can also call the information and application programs stored in the memory through the transmission device to execute the following steps: The knowledge base corresponding to the standby power generation system is constructed in the following way: use a preset compilation language to extract standby power generation equipment structure information, power generation signal quantity, and power outage signal quantity from the standby power generation database to obtain the first extraction data; generate an unstructured knowledge document based on the first extraction data; use a preset compilation language to extract the standby power generation system site information and path information from the standby power generation system requirement plan and design plan to obtain the second extraction data; use a preset compilation language to extract functional operation steps from the operation manual and test cases related to the standby power generation system to obtain the third extraction data; synthesize the second extraction data and the third extraction data to generate a structured knowledge graph; construct the knowledge base corresponding to the standby power generation system based on the structured knowledge graph and the unstructured knowledge document.

[0150] Those of ordinary skill in the art can understand that Figure 5 the structure shown is only for illustration, and the electronic device can also be a terminal device such as a smart phone, a tablet computer, a palm computer, and a Mobile Internet Device (MID), a PAD, etc. Figure 5 It does not limit the structure of the above electronic device. For example, the electronic device may also include more or fewer components (such as a network interface, a display device, etc.) than those shown Figure 5 in, or have a different configuration from that shown Figure 5 in.

[0151] Those of ordinary skill in the art can understand that all or part of the steps in the various large model-based intelligent question-answering methods of the above embodiments can be completed by a program instructing the relevant hardware of the terminal device. This program can be stored in a computer-readable storage medium, and the storage medium can include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, an optical disc, etc.

[0152] Embodiment 4

[0153] An embodiment of the present application also provides a storage medium. Optionally, in this embodiment, the above storage medium can be used to store the program code executed by the large model-based intelligent question-answering method provided in Embodiment 1 above.

[0154] On the other hand, according to an embodiment of the present invention, there is also provided a computer-readable storage medium. The computer-readable storage medium includes a stored computer program, wherein when the computer program runs, it controls the device where the computer-readable storage medium is located to execute the large model-based intelligent question-answering method of any one of the above Embodiment 1.

[0155] Optionally, in this embodiment, the above storage medium can be located in any one of the computer terminals in the computer terminal group in the computer network, or in any one of the mobile terminals in the mobile terminal group.

[0156] The present application also provides a computer program product, including a computer program, and the steps of the large model-based intelligent question-answering method described in each embodiment of the present application are implemented when the computer program is executed by a processor.

[0157] The present application also provides a computer program product, including a non-volatile computer-readable storage medium. The non-volatile computer-readable storage medium stores a computer program, and the steps of the large model-based intelligent question-answering method described in each embodiment of the present application are implemented when the computer program is executed by a processor.

[0158] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages and disadvantages of the embodiments.

[0159] In the above embodiments of the present invention, the descriptions of each embodiment have their own emphases. For the parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0160] In several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are merely illustrative. For example, the division of the units can be a logical function division. In actual implementation, there can be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the couplings or direct couplings or communication connections shown or discussed with each other can be through some interfaces. The indirect couplings or communication connections of units or modules can be in electrical or other forms.

[0161] The units described as separate components may or may not be physically separated. The components shown as units may or may not be physical units, that is, they can be located in one place or distributed to multiple units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0162] In addition, in each embodiment of the present invention, the functional units can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above-mentioned integrated units can be implemented in the form of hardware or in the form of software functional units.

[0163] If the above-mentioned integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of the present invention. The foregoing storage medium includes: various media such as USB flash drives, read-only memories (ROMs), random access memories (RAMs), mobile hard disks, magnetic disks or optical discs that can store program codes.

[0164] The above is only the preferred embodiment of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. An intelligent question-answering method based on a large model, characterized in that Including: Collecting raw data from multiple data sources, and preprocessing the raw data to obtain a training data set, where the training data set is used to train a pre-constructed initial deep learning framework to obtain an intelligent question-answering model; Responding to a power generation backup question request output by a client through an interaction interface, after parsing the power generation backup question request, calling a knowledge base corresponding to the power generation backup system, where the knowledge base is constructed by extracting data of specified power generation backup association information based on the power generation backup system requirement scheme and the power generation backup database; Retrieving, from the knowledge base, a reply information corresponding to the power generation backup question request through the intelligent question-answering model to generate a power generation backup answer; Returning the power generation backup answer to the front-end display interface of the client.

2. The method according to claim 1, wherein The steps of preprocessing the raw data to obtain a training data set include: Classifying the raw data to obtain multi-modal data, where the types of the multi-modal data at least include: text, image, audio, video; Performing data cleaning, data denoising, standardization processing, and data format conversion on the multi-modal data to complete the preprocessing process of the raw data and obtain the training data set.

3. The method according to claim 1, wherein The knowledge base corresponding to the power generation backup system is constructed in the following manner: Extracting power generation backup equipment structure information, power generation signal quantities, and power outage signal quantities from the power generation backup database using a preset compilation language to obtain first extraction data; Generating an unstructured knowledge document based on the first extraction data; Extracting power generation backup system site information and path information from the power generation backup system requirement scheme and design scheme using a preset compilation language to obtain second extraction data; Extracting functional operation steps from the operation manual and test cases associated with the power generation backup system using a preset compilation language to obtain third extraction data; Integrating the second extraction data and the third extraction data to generate a structured knowledge graph; Constructing the knowledge base corresponding to the power generation backup system based on the structured knowledge graph and the unstructured knowledge document.

4. The method according to claim 3, characterized in that, The steps of retrieving, from the knowledge base, a reply information corresponding to the power generation backup question request through the intelligent question-answering model to generate a power generation backup answer include: In the case where the type of the power generation backup question request is power generation equipment positioning, parsing the power generation backup question request to obtain identification information of a target power generation equipment to be positioned; Based on the identification information, calling the knowledge base corresponding to the power generation backup system, and outputting, through the intelligent question-answering model, functional positioning information and navigation path information associated with the target power generation equipment to obtain the power generation backup answer; or, In the case where the type of the power generation backup question request is power generation quantity data analysis, parsing the power generation backup question request to obtain identification information of a target power generation equipment to be positioned; Based on the identification information, calling the knowledge base corresponding to the power generation backup system, and outputting, through the intelligent question-answering model, power generation quantity data associated with the target power generation equipment to obtain the power generation backup answer, where the power generation quantity data at least includes: date and corresponding power generation quantity, period and corresponding power generation quantity, power generation peak and valley.

5. The method according to claim 3, characterized in that, The steps of retrieving response information corresponding to the power generation and power supply question request from the knowledge base through the intelligent question and answer model and generating the power generation and power supply answer include: When the type of the power generation and power supply question request is site power outage analysis, parsing the power generation and power supply question request to obtain the identification information of the target power generation equipment; Based on the identification information of the target power generation equipment, calling the knowledge base corresponding to the power generation and power supply system, analyzing whether the power generation amount of the target power generation equipment meets the power consumption demand of a predetermined area. When the analysis result indicates that the power generation amount of the target power generation equipment does not meet the power consumption demand of the predetermined area, outputting, through the intelligent question and answer model, a power outage prompt information associated with the predetermined area and the equipment information of the standby power generation equipment that can be called to obtain the power generation and power supply answer.

6. The method according to claim 3, wherein The steps of retrieving response information corresponding to the power generation and power supply question request from the knowledge base through the intelligent question and answer model and generating the power generation and power supply answer include: When the type of the power generation and power supply question request is faulty equipment analysis, parsing the power generation and power supply question request to obtain the identification information of all power generation and power supply equipment to be analyzed, as well as the operating status and equipment detection pictures of each power generation and power supply equipment; Based on the identification information of each power generation and power supply equipment, as well as the operating status and equipment detection pictures of the power generation and power supply equipment, calling the knowledge base corresponding to the power generation and power supply system to perform fault analysis on each power generation and power supply equipment, and outputting, through the intelligent question and answer model, the power generation and power supply answer including the faulty equipment information and the fault repair suggestion report.

7. The method according to claim 1, wherein The steps of returning the power generation and power supply answer to the front-end display interface of the user terminal include: Performing encryption processing on the power generation and power supply answer by adopting a homomorphic encryption strategy; Returning the encryption result to the front-end display interface of the user terminal through an end-to-end encryption transmission strategy; Collecting the answer feedback evaluation of the user terminal, and iteratively updating the intelligent question and answer model and the knowledge base based on the answer feedback evaluation.

8. An intelligent question-answering device based on a large model, characterized in that, Include: A data preprocessing unit, configured to collect raw data from multiple data sources and preprocess the raw data to obtain a training data set, where the training data set is used to perform model training on a pre-constructed initial deep learning framework to obtain an intelligent question and answer model; A knowledge base calling unit, configured to respond to a power generation and power supply question request output by the user terminal through an interaction interface, and call the knowledge base corresponding to the power generation and power supply system after parsing the power generation and power supply question request, where the knowledge base is constructed by extracting specified power generation and power supply associated information data based on the power generation and power supply system requirement scheme and the power generation and power supply database; An answer retrieval unit, configured to retrieve response information corresponding to the power generation and power supply question request from the knowledge base through the intelligent question and answer model and generate a power generation and power supply answer; An answer sending unit, configured to return the power generation and power supply answer to the front-end display interface of the user terminal.

9. An electronic device, characterized in that, Comprising one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors are caused to implement the large model-based intelligent question-answering method according to any one of claims 1 to 7.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, the steps of the large model-based intelligent question-answering method according to any one of claims 1 to 7 are implemented.

Images