Method and device for constructing fault diagnosis model, fault diagnosis method and device

Abstract

The application relates to a fault diagnosis model construction method and device and a fault diagnosis method and device, and is applied to the technical field of steam turbines. The model construction method comprises the following steps: acquiring fault diagnosis model construction instructions of single components respectively; recognizing the instructions by using a steam turbine professional large language model, extracting candidate monitoring parameters, calling a steam turbine knowledge base to determine first importance degree values of the candidate monitoring parameters; calculating correlation coefficients between each candidate monitoring parameter and other candidate monitoring parameters and determining second importance degree values; weighting and averaging the first importance degree values and the second importance degree values to obtain comprehensive importance degree values, and taking candidate monitoring parameters with comprehensive importance degree values greater than a preset threshold as target monitoring parameters; and constructing fault diagnosis models corresponding to single components according to time series corresponding to the target monitoring parameters. The fault diagnosis models corresponding to the components are encapsulated to obtain a steam turbine fault diagnosis model. The efficiency and accuracy of model construction can be improved.

Description

Technical Field

[0001] This application relates to the field of steam turbine technology, and in particular to a method and apparatus for constructing a fault diagnosis model, and a fault diagnosis method and apparatus. Background Technology

[0002] Steam turbines, as key equipment in thermal power generation, play a vital role in my country's energy and power sectors. During operation, various factors can lead to performance degradation and frequent malfunctions. Therefore, to improve the operating efficiency and safety of steam turbines, there is an urgent need for a highly efficient, accurate, and intelligent fault diagnosis system.

[0003] Existing turbine fault diagnosis primarily relies on data-driven deep learning models. These models use a deep learning architecture to process complex time-series data based on key monitoring data of turbine operation, capturing crucial time dependencies and correlations between variables. By mining data patterns, they differentiate between normal and abnormal data, thereby enabling anomaly detection and fault diagnosis.

[0004] To improve the accuracy of turbine fault diagnosis, a corresponding fault diagnosis model needs to be built for each component. However, turbines have numerous monitoring parameters, and manually selecting these parameters is usually required when building a fault diagnosis model for each component. This manual selection process demands high levels of expertise and is time-consuming and labor-intensive, resulting in low efficiency and accuracy in building fault diagnosis models. Summary of the Invention

[0005] To address the aforementioned technical issues, this application provides a method and apparatus for constructing a turbine fault diagnosis model based on an AI large-scale model, a turbine fault diagnosis method and apparatus based on an AI large-scale model, electronic equipment, and a storage medium.

[0006] According to the first aspect of this application, a method for constructing a turbine fault diagnosis model based on an AI large model is provided, including:

[0007] Obtain the instructions for constructing fault diagnosis models for individual components in the steam turbine;

[0008] Using a pre-built large language model for steam turbines, the instructions for building a fault diagnosis model for a single component are identified, the candidate monitoring parameters corresponding to the single component are extracted, and the first importance value of the candidate monitoring parameters is determined by calling a pre-built steam turbine knowledge base.

[0009] For each candidate monitoring parameter, the Pearson correlation coefficient between the candidate monitoring parameter and other candidate monitoring parameters is calculated, and the second importance value of the candidate monitoring parameter is determined based on the first importance value of other candidate monitoring parameters and the Pearson correlation coefficient between other candidate monitoring parameters and the candidate monitoring parameter.

[0010] The first importance value and the second importance value of the candidate monitoring parameter are weighted and averaged to obtain the comprehensive importance value of the candidate monitoring parameter. The candidate monitoring parameter with a comprehensive importance value greater than a preset threshold is taken as the target monitoring parameter.

[0011] Based on the time series corresponding to the target monitoring parameters, a fault diagnosis model for a single component is constructed.

[0012] After constructing the fault diagnosis models corresponding to each component, the fault diagnosis models corresponding to each component are encapsulated to obtain the turbine fault diagnosis model.

[0013] Optionally, the method for constructing the turbine fault diagnosis model based on the AI ​​large model further includes:

[0014] Obtain instructions for constructing a fault diagnosis model for the entire steam turbine.

[0015] Using the turbine professional language model, the instructions for constructing a fault diagnosis model for the entire turbine are identified, candidate monitoring parameters corresponding to the entire turbine are extracted, and the turbine knowledge base is called to determine the first importance value of the candidate monitoring parameters.

[0016] For each candidate monitoring parameter, the Pearson correlation coefficient between the candidate monitoring parameter and other candidate monitoring parameters is calculated, and the second importance value of the candidate monitoring parameter is determined based on the first importance value of other candidate monitoring parameters and the Pearson correlation coefficient between other candidate monitoring parameters and the candidate monitoring parameter.

[0017] The first importance value and the second importance value of the candidate monitoring parameter are weighted and averaged to obtain the comprehensive importance value of the candidate monitoring parameter. The candidate monitoring parameter with a comprehensive importance value greater than a preset threshold is taken as the target monitoring parameter.

[0018] Based on the time series corresponding to the target monitoring parameters, a fault diagnosis model for the whole machine is constructed.

[0019] The process of encapsulating the fault diagnosis models corresponding to each component to obtain the turbine fault diagnosis model includes:

[0020] The fault diagnosis models corresponding to each component and the fault diagnosis model corresponding to the whole machine are encapsulated to obtain the turbine fault diagnosis model.

[0021] Optionally, determining the second importance value of the candidate monitoring parameter based on the first importance value of other candidate monitoring parameters and the Pearson correlation coefficient between the other candidate monitoring parameters and the candidate monitoring parameter includes:

[0022] Assumption Indicates components C The corresponding number i The first importance value of each candidate monitoring parameter Indicates components C The corresponding number j The first importance value of each candidate monitoring parameter Indicates the first i The candidate monitoring parameter and the first j The Pearson correlation coefficient between the candidate monitoring parameters is calculated using the following formula: and Determine the first i The second importance value of each candidate monitoring parameter ,in, For the first i The correlation importance value of each candidate monitoring parameter.

[0023] Optionally, the method for constructing the large language model for steam turbines includes:

[0024] Obtain steam turbine-related corpus and convert it into question-answer pairs.

[0025] A general large language model is obtained, and the general large language model is fine-tuned using the question-and-answer pair data to obtain the steam turbine professional large language model.

[0026] Optionally, the method for constructing the steam turbine knowledge base includes:

[0027] Obtain steam turbine-related documents, perform text recognition on the steam turbine-related documents, and obtain knowledge text;

[0028] The knowledge text is segmented to obtain knowledge text blocks;

[0029] The knowledge text blocks are converted into vectors by an embedding model to obtain a vector database, which is then used as a steam turbine knowledge base.

[0030] Optionally, the step of constructing a fault diagnosis model for a single component based on the time series corresponding to the target monitoring parameters includes:

[0031] Training samples are constructed based on the time series corresponding to the target monitoring parameters, and a monitoring data prediction model for a single component is constructed using LSTM (Long Short-Term Memory Network).

[0032] For each target monitoring parameter, the deviation between the actual value and the predicted value of the target monitoring parameter on each training sample is calculated, and the average value and standard deviation are obtained by statistical analysis based on the deviations of all training samples.

[0033] Based on the monitoring data prediction model, the average value, and the standard deviation, a fault diagnosis model corresponding to a single component is constructed.

[0034] According to a second aspect of this application, a turbine fault diagnosis method based on an AI large model is provided, comprising:

[0035] The user command is obtained, and the user command is identified using a pre-built large language model for steam turbines to determine whether the user command belongs to a steam turbine fault diagnosis command.

[0036] If the user command is not a turbine fault diagnosis command, output the corresponding query result;

[0037] If the user instruction is a turbine fault diagnosis instruction, the monitoring parameters corresponding to the component corresponding to the turbine fault diagnosis instruction are extracted, and the fault diagnosis model corresponding to the component is called from the pre-built turbine fault diagnosis model to process the monitoring parameters corresponding to the component and obtain the fault diagnosis result. The turbine fault diagnosis model is constructed based on the method described in the first aspect.

[0038] According to a third aspect of this application, an apparatus for constructing a turbine fault diagnosis model based on an AI large model is provided, comprising:

[0039] The instruction acquisition module is used to acquire instructions for constructing fault diagnosis models for individual components in a steam turbine.

[0040] The first importance value determination module is used to identify the fault diagnosis model construction instructions for a single component using a pre-built steam turbine professional large language model, extract the candidate monitoring parameters corresponding to the single component, and call the pre-built steam turbine knowledge base to determine the first importance value of the candidate monitoring parameters.

[0041] The second importance value determination module is used to calculate the Pearson correlation coefficient between the candidate monitoring parameter and other candidate monitoring parameters for each candidate monitoring parameter, and determine the second importance value of the candidate monitoring parameter based on the first importance value of other candidate monitoring parameters and the Pearson correlation coefficient between other candidate monitoring parameters and the candidate monitoring parameter.

[0042] The comprehensive importance value determination module is used to perform a weighted average of the first importance value and the second importance value of the candidate monitoring parameter to obtain the comprehensive importance value of the candidate monitoring parameter;

[0043] The target monitoring parameter determination module is used to select candidate monitoring parameters whose comprehensive importance value is greater than a preset threshold as target monitoring parameters.

[0044] The sub-fault diagnosis model construction module is used to construct a fault diagnosis model for a single component based on the time series corresponding to the target monitoring parameters.

[0045] The turbine fault diagnosis model generation module is used to encapsulate the fault diagnosis models corresponding to each component after the fault diagnosis models corresponding to each component are constructed, so as to obtain the turbine fault diagnosis model.

[0046] Optionally, the instruction acquisition module is further configured to acquire instructions for constructing a fault diagnosis model for the entire steam turbine.

[0047] The first importance value determination module is also used to use the steam turbine professional big language model to identify the fault diagnosis model construction instructions for the whole steam turbine, extract the candidate monitoring parameters corresponding to the whole steam turbine, and call the steam turbine knowledge base to determine the first importance value of the candidate monitoring parameters.

[0048] The second importance value determination module is further configured to calculate the Pearson correlation coefficient between the candidate monitoring parameter and other candidate monitoring parameters for each candidate monitoring parameter, and determine the second importance value of the candidate monitoring parameter based on the first importance value of the other candidate monitoring parameters and the Pearson correlation coefficient between the other candidate monitoring parameters and the candidate monitoring parameter;

[0049] The comprehensive importance value determination module is further configured to perform a weighted average of the first importance value and the second importance value of the candidate monitoring parameter to obtain the comprehensive importance value of the candidate monitoring parameter.

[0050] The target monitoring parameter determination module is also used to select candidate monitoring parameters whose comprehensive importance value is greater than a preset threshold as target monitoring parameters;

[0051] The sub-fault diagnosis model construction module is also used to construct a fault diagnosis model for the whole machine based on the time series corresponding to the target monitoring parameters.

[0052] The turbine fault diagnosis model generation module is specifically used to encapsulate the fault diagnosis models corresponding to each component and the fault diagnosis model corresponding to the whole machine to obtain the turbine fault diagnosis model.

[0053] Optionally, the second importance value determination module is specifically used to assume... Indicates the first part corresponding to component C. i The first importance value of each candidate monitoring parameter Indicates components C The corresponding number j The first importance value of each candidate monitoring parameter Indicates the first i The candidate monitoring parameter and the first j The Pearson correlation coefficient between the candidate monitoring parameters is calculated using the following formula: and Determine the first i The second importance value of each candidate monitoring parameter ,in, For the first i The correlation importance value of each candidate monitoring parameter.

[0054] Optionally, the device for constructing the turbine fault diagnosis model based on the AI ​​large model further includes:

[0055] The question-and-answer pair data generation module is used to acquire steam turbine-related corpus and convert the steam turbine-related corpus into question-and-answer pair data.

[0056] The general large language model fine-tuning module is used to obtain the general large language model and fine-tune the general large language model using the question-and-answer pair data to obtain the turbine-specific large language model.

[0057] Optionally, the device for constructing the turbine fault diagnosis model based on the AI ​​large model further includes:

[0058] The text recognition module is used to acquire steam turbine professional documents, perform text recognition on the steam turbine professional documents, and obtain knowledge text;

[0059] The text segmentation module is used to segment the knowledge text to obtain knowledge text blocks;

[0060] The vector conversion module is used to convert the knowledge text blocks into vectors through an embedding model to obtain a vector database.

[0061] The turbine knowledge base generation module is used to use the vector database as a turbine knowledge base.

[0062] Optionally, the sub-fault diagnosis model construction module is specifically used to construct training samples based on the time series corresponding to the target monitoring parameters, and use LSTM to construct a monitoring data prediction model corresponding to a single component; for each target monitoring parameter, calculate the deviation between the actual value and the predicted value of the target monitoring parameter on each training sample, and perform statistics on the deviations corresponding to all training samples to obtain the mean and standard deviation; and construct a fault diagnosis model corresponding to a single component based on the monitoring data prediction model, the mean and standard deviation.

[0063] According to a fourth aspect of this application, a turbine fault diagnosis device based on an AI large model is provided, the device comprising:

[0064] The user instruction acquisition module is used to acquire user instructions.

[0065] The user instruction recognition module is used to identify the user instruction using a pre-built steam turbine professional language model and determine whether the user instruction belongs to a steam turbine fault diagnosis instruction.

[0066] The query result output module is used to output the corresponding query result if the user instruction recognition module determines that the user instruction does not belong to the turbine fault diagnosis instruction.

[0067] The fault diagnosis result output module is used to, if the user instruction recognition module determines that the user instruction belongs to the turbine fault diagnosis instruction, extract the monitoring parameters corresponding to the component according to the component corresponding to the turbine fault diagnosis instruction, and call the fault diagnosis model corresponding to the component from the pre-built turbine fault diagnosis model to process the monitoring parameters corresponding to the component and obtain the fault diagnosis result, wherein the turbine fault diagnosis model is constructed based on the method described in the first aspect.

[0068] According to a fifth aspect of this application, an electronic device is provided, comprising: a processor configured to execute a computer program stored in a memory, wherein the computer program, when executed by the processor, implements the method described in the first or second aspect.

[0069] According to a sixth aspect of this application, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the method described in the first or second aspect.

[0070] According to a seventh aspect of this application, a computer program product is provided that, when the computer program product is run on a computer, causes the computer to perform the method described in the first or second aspect.

[0071] The technical solution provided in this application has the following advantages compared with the prior art:

[0072] By utilizing a pre-constructed comprehensive language model for steam turbines, a strong understanding of professional knowledge can be achieved. This allows for in-depth understanding and professional answers to issues related to the steam turbine field, in conjunction with professional data and documentation. Therefore, when constructing a steam turbine fault diagnosis model, the comprehensive language model's expertise in the steam turbine domain is leveraged to identify commands, output candidate monitoring parameters, and simultaneously call upon the steam turbine knowledge base to output the first importance value corresponding to each candidate monitoring parameter—a mechanism-based importance value. Simultaneously, the correlation between candidate monitoring parameters is used to calculate the second importance value—a statistically based importance value. These two values ​​are then fused to calculate a comprehensive importance value, which is then used to select monitoring parameters based on a threshold. It is evident that using the comprehensive language model and steam turbine knowledge base to assist in the selection of monitoring parameters effectively integrates the mechanistic knowledge from the comprehensive language model and the steam turbine knowledge base with data-based statistical information, improving the efficiency and accuracy of monitoring parameter selection. This, in turn, enhances the efficiency and accuracy of fault diagnosis model construction. Attached Figure Description

[0073] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0074] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0075] Figure 1 This is a system architecture diagram of the method and apparatus for constructing a turbine fault diagnosis model and the turbine fault diagnosis method and apparatus applicable to the embodiments of this application;

[0076] Figure 2 This is a flowchart of a method for constructing a turbine fault diagnosis model based on an AI large model in the embodiments of this application;

[0077] Figure 3 This is a schematic diagram illustrating the construction and retrieval of a steam turbine knowledge base in an embodiment of this application;

[0078] Figure 4 This is another flowchart illustrating the method for constructing a turbine fault diagnosis model based on an AI large model in the embodiments of this application;

[0079] Figure 5 This is a flowchart of a turbine fault diagnosis method based on an AI large model in an embodiment of this application;

[0080] Figure 6 This is a schematic diagram of a device for constructing a turbine fault diagnosis model based on an AI large model, as described in this application.

[0081] Figure 7 This is a schematic diagram of a turbine fault diagnosis device based on an AI large model in an embodiment of this application.

[0082] Figure 8 This is a schematic diagram of the structure of an electronic device in an embodiment of this application. Detailed Implementation

[0083] To better understand the above-mentioned objectives, features, and advantages of this application, the solution of this application will be further described below. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0084] Many specific details are set forth in the following description in order to provide a full understanding of this application, but this application may also be implemented in other ways different from those described herein; obviously, the embodiments in the specification are only some embodiments of this application, and not all embodiments.

[0085] With the continuous advancement of technology, artificial intelligence has witnessed significant development, especially in recent years, with major breakthroughs achieved in large language models centered on the Transformer model architecture. Representative large language models such as BERT (Bidirectional Encoder Representations from Transformers), GPT (Generative Pre-trained Transformer), and GLM (General Language Model) have emerged. As large language models continue to improve, their applications in various fields are gradually expanding. From text generation, machine translation, and sentiment analysis to information extraction, large language models have provided powerful support for solving complex problems and have already triggered revolutionary changes in many industries.

[0086] However, the current capabilities of large language models are mainly focused on solving general language problems, lacking industry-specific knowledge accumulation and exhibiting weak understanding of specialized scenarios and question-answering abilities. Furthermore, large language models can only process text and lack the ability to process thermodynamic models, operational data, and images widely present in the steam turbine field. Therefore, their ability to handle industry-specific problems within the steam turbine industry context needs further improvement.

[0087] Based on this, this application constructs a comprehensive language model for steam turbines, applicable to the specialized field. This model possesses a strong understanding of professional knowledge and can combine professional data and documents to deeply understand and provide professional answers to issues related to steam turbines. This specifically enhances the professional capabilities of the general language model in steam turbine operation and maintenance, enabling faster identification and response to professional issues, reducing error rates, and improving work efficiency. Simultaneously, a steam turbine knowledge base is constructed. Its main function is to store and manage textual materials such as professional books and documents related to steam turbines, and to provide basic or timely professional knowledge references. When constructing a steam turbine fault diagnosis model, the comprehensive language model and the steam turbine knowledge base are used to assist in the selection of monitoring parameters, thereby improving the efficiency and accuracy of fault diagnosis model construction.

[0088] Figure 1 This is a system architecture diagram of the method and apparatus for constructing a steam turbine fault diagnosis model, applicable to embodiments of this application, and the method and apparatus for steam turbine fault diagnosis. The system includes a steam turbine knowledge base and a steam turbine fault diagnosis model, which are integrated using intelligent proxy technology based on a steam turbine professional large language model to form a steam turbine professional large language model platform responsible for interacting with users based on natural language.

[0089] See Figure 2 , Figure 2 This is a flowchart illustrating a method for constructing a turbine fault diagnosis model based on an AI large model, as described in this application. The method may include the following steps:

[0090] Step S202: Obtain the fault diagnosis model construction instructions for each individual component in the steam turbine.

[0091] The turbine fault diagnosis model includes fault diagnosis models corresponding to each component of the turbine. The construction method of the fault diagnosis model corresponding to each component is the same. Here, we will take the fault diagnosis model corresponding to a single component as an example for explanation.

[0092] When building a fault diagnosis model for a specific component, the user can output a command to build the fault diagnosis model for that component. For example, if the user wants to build a fault diagnosis model for the high-pressure cylinder, the user can output the command "Build a fault diagnosis model for the high-pressure cylinder of the steam turbine".

[0093] Step S204: Using the pre-built turbine professional large language model, identify the fault diagnosis model construction instructions for a single component, extract the candidate monitoring parameters corresponding to the single component, and call the pre-built turbine knowledge base to determine the first importance value of the candidate monitoring parameters.

[0094] By constructing a large language model for steam turbines, its powerful general understanding and reasoning capabilities can provide the ability to understand complex relationships and patterns within the energy and power industry, as well as the ability to process in-depth business knowledge.

[0095] Optionally, the method for constructing a large language model for steam turbines may include the following steps:

[0096] Acquire turbine-related corpora, such as books and technical documents. Preprocess the collected corpora, performing data cleaning and quality assessment to ensure accuracy, comprehensiveness, and representativeness. Convert the turbine-related corpora into question-answer pairs, which are suitable for fine-tuning a general large language model.

[0097] A general-purpose large language model is obtained. Based on the general-purpose large language model, and fine-tuned using question-and-answer pair data, a steam turbine-specific large language model is obtained. This steam turbine-specific large language model has the ability to understand and reason about the steam turbine-specific knowledge involved in the corpus.

[0098] For example, in response to the instruction "Construct a fault diagnosis model for the high-pressure cylinder of a steam turbine," the steam turbine professional language model can recognize the user's intent and select monitoring parameters related to the high-pressure cylinder from the steam turbine's monitoring parameters as candidate monitoring parameters. The steam turbine professional language model can also access a pre-built steam turbine knowledge base. After understanding its own professional knowledge and accessing the knowledge in the steam turbine knowledge base, it can gain a certain understanding of the importance of the candidate monitoring parameters, thereby determining the first importance value of the candidate monitoring parameters. The first importance value can be a value between 0 and 1.

[0099] Figure 3 This is a schematic diagram illustrating the construction and retrieval of a steam turbine knowledge base in an embodiment of this application. The method for constructing the steam turbine knowledge base can be as follows: Obtain steam turbine-related documents; perform text recognition on the steam turbine-related documents to obtain knowledge text, meeting the input requirements of a large-scale steam turbine language model; segment the knowledge text to obtain knowledge text blocks; convert the knowledge text blocks into vectors using an embedding model to obtain a vector database; and use the vector database as the steam turbine knowledge base.

[0100] The steam turbine knowledge base has management functions, allowing users to upload new professional documents. After uploading a professional document, repeat the above steps to update the steam turbine knowledge base.

[0101] When the turbine-specific large-scale language model interacts with users based on natural language, it can leverage its own professional knowledge and access a knowledge base to output content related to user commands. Specifically, for any user command, the steps for accessing the knowledge base include: converting the user command into a vector using an embedding model; calculating the similarity between the user command and the vectors of various text blocks in the turbine knowledge base; retrieving the text block with the highest similarity to the user command vector; and constructing prompt words for the turbine-specific large-scale language model. These prompt words are relatively standardized words that can be used as input to the turbine-specific large-scale language model, and can include both the user command and the text block with the highest similarity. After inputting the prompt words into the turbine-specific large-scale language model, the model outputs knowledge base retrieval content related to the user command.

[0102] Step S206: For each candidate monitoring parameter, calculate the Pearson correlation coefficient between the candidate monitoring parameter and other candidate monitoring parameters, and determine the second importance value of the candidate monitoring parameter based on the first importance value of other candidate monitoring parameters and the Pearson correlation coefficient between other candidate monitoring parameters and the candidate monitoring parameter.

[0103] For any two candidate monitoring parameters and The Pearson correlation coefficients between them can be expressed as follows:

[0104] ;

[0105] in, express and Covariance between express variance express The variance.

[0106] For each candidate monitoring parameter, the Pearson correlation coefficient between that candidate monitoring parameter and all other candidate monitoring parameters can be calculated. Then, by combining the first importance values ​​of the other candidate monitoring parameters, the second importance value of the candidate monitoring parameter can be statistically determined.

[0107] Alternatively, assume Indicates components C The corresponding number i The first importance value of each candidate monitoring parameter Indicates components C The corresponding number j The first importance value of each candidate monitoring parameter Indicates the first i The candidate monitoring parameter and the firstj The Pearson correlation coefficient between the candidate monitoring parameters can be calculated using the following formula: and Determine the first i The second importance value of each candidate monitoring parameter ,in, For the first i The correlation importance value of each candidate monitoring parameter.

[0108] It can be seen that the second importance value of the candidate monitoring parameter is positively correlated with the first importance value of other candidate monitoring parameters and with the Pearson correlation coefficient between the candidate monitoring parameter and other candidate monitoring parameters.

[0109] Step S208: Take a weighted average of the first importance value and the second importance value of the candidate monitoring parameter to obtain the comprehensive importance value of the candidate monitoring parameter, and take the candidate monitoring parameter whose comprehensive importance value is greater than a preset threshold as the target monitoring parameter.

[0110] It can be based on the following formula: ;

[0111] Calculate the overall importance value of the candidate monitoring parameter. ,in, This represents the weighting coefficient, which takes values ​​of [0,1]. Different candidate monitoring parameters can correspond to different weighting coefficients.

[0112] A higher overall importance value indicates a greater impact of the candidate monitoring parameter on the fault diagnosis model corresponding to that component. Therefore, candidate monitoring parameters with an overall importance value greater than a preset threshold can be used as target monitoring parameters for subsequent fault diagnosis model construction. Different preset thresholds can be set for different components.

[0113] Step S210: Construct a fault diagnosis model for a single component based on the time series corresponding to the target monitoring parameters.

[0114] After selecting the target monitoring parameters, the target monitoring parameter data can be cleaned to remove outliers caused by sensor errors, communication anomalies, etc., and to exclude monitoring values ​​generated during fault periods, including only normal data. Standardization processing can also be performed on the target monitoring parameter data. Then, the processed target monitoring parameter data can be used to construct a fault diagnosis model for each individual component.

[0115] The input to the fault diagnosis model includes the time series of target monitoring parameters. X = ,in, T The length of the time series. DNumber of target monitoring parameters ，d The number of training samples. For time steps.

[0116] Optionally, training samples can be constructed based on the time series corresponding to the target monitoring parameters, and an LSTM (Long Short-Term Memory) network can be used to build a monitoring data prediction model for a single component. The input to the monitoring data prediction model is the time series of the target monitoring parameters for one period, and the label data can be the time series of the target monitoring parameters that differ from this time series by one sampling time. For example, the input is {X1,X2,…,XT}, and the label data is {X2,X3,…,X(T+1)}.

[0117] After training and generating the monitoring data prediction model, for each target monitoring parameter, the deviation between the actual value and the predicted value of the target monitoring parameter on each training sample is calculated, and the average value and standard deviation are obtained by statistical analysis based on the deviations of all training samples.

[0118] For example, regarding the first i There are target monitoring parameters, assuming the number of training samples is [number]. d It can be calculated using the following formula:

[0119] ;

[0120] Calculate the actual value and predicted value Deviation between Then, based on the deviations of all training samples, a statistical average is obtained. and standard deviation .

[0121] Subsequently, based on the monitoring data prediction model, average value, and standard deviation, a fault diagnosis model corresponding to a single component is constructed.

[0122] Upon receiving monitoring parameter sample input, the deviation can be calculated using the method described above, followed by anomaly detection, outputting a normal or abnormal label (1 for abnormal, 0 for normal), which serves as the output of the fault diagnosis model. This can be expressed as the following formula:

[0123] ;

[0124] When all monitoring parameter samples are labeled The fault diagnosis model outputs 0 (normal) only when both values ​​are 0; otherwise, it outputs 1 (abnormal).

[0125] Step S212: After constructing the fault diagnosis model corresponding to each component, the fault diagnosis model corresponding to each component is encapsulated to obtain the turbine fault diagnosis model.

[0126] By encapsulating the fault diagnosis models corresponding to each component, a turbine fault diagnosis model can be formed. For each component's fault diagnosis model, a tool declaration can be provided, including the data types and physical meanings of its inputs and outputs, as well as an overall functional description of the model.

[0127] The method for constructing a steam turbine fault diagnosis model based on an AI-powered large-scale model in this application utilizes a pre-constructed steam turbine professional language model. This model possesses a strong understanding of professional knowledge and can combine professional data and documents to deeply understand and provide professional answers to issues related to the steam turbine field. Therefore, when constructing the steam turbine fault diagnosis model, the professional capabilities of the steam turbine professional language model in the steam turbine field are leveraged to identify instructions, output candidate monitoring parameters, and simultaneously call upon the steam turbine knowledge base to output the first importance value corresponding to the candidate monitoring parameters, i.e., the importance value based on mechanism. Simultaneously, the correlation between the candidate monitoring parameters is used to calculate the second importance value corresponding to the candidate monitoring parameters, i.e., the importance value based on statistics. The two are then fused to calculate the comprehensive importance value, and monitoring parameters are selected based on a threshold. It is evident that using the steam turbine professional language model and the steam turbine knowledge base to assist in the selection of monitoring parameters can effectively integrate the mechanistic knowledge and data-based statistical information in the steam turbine professional language model and the steam turbine knowledge base, improving the efficiency and accuracy of monitoring parameter selection. Furthermore, this can improve the efficiency and accuracy of fault diagnosis model construction. Compared to traditional algorithms, it has higher integration and user-friendliness.

[0128] See Figure 4 , Figure 4 This is a flowchart illustrating a method for constructing a turbine fault diagnosis model based on an AI large-scale model, as described in this application. Figure 2 Based on this, the following steps are included:

[0129] Step S402: Obtain the fault diagnosis model construction instruction for the entire steam turbine.

[0130] Step S404: Using the large language model of steam turbine, identify the instructions for constructing the fault diagnosis model for the whole steam turbine, extract the candidate monitoring parameters corresponding to the whole steam turbine, and call the steam turbine knowledge base to determine the first importance value of the candidate monitoring parameters.

[0131] Step S406: For each candidate monitoring parameter, calculate the Pearson correlation coefficient between the candidate monitoring parameter and other candidate monitoring parameters, and determine the second importance value of the candidate monitoring parameter based on the first importance value of other candidate monitoring parameters and the Pearson correlation coefficient between other candidate monitoring parameters and the candidate monitoring parameter.

[0132] Step S408: The first importance value and the second importance value of the candidate monitoring parameter are weighted and averaged to obtain the comprehensive importance value of the candidate monitoring parameter. The candidate monitoring parameter with the comprehensive importance value greater than the preset threshold is taken as the target monitoring parameter.

[0133] Step S410: Construct a fault diagnosis model for the whole machine based on the time series corresponding to the target monitoring parameters.

[0134] The method for constructing the fault diagnosis model for the entire machine in steps S402 to S410 above is the same as the method for constructing the fault diagnosis model for a single component. See details... Figure 1 The descriptions in the embodiments are sufficient and will not be repeated here.

[0135] Step S412: After constructing the fault diagnosis models corresponding to each component and the whole machine, the fault diagnosis models corresponding to each component and the whole machine are encapsulated to obtain the turbine fault diagnosis model.

[0136] In this embodiment, the fault diagnosis models for the whole system and for each component can also be encapsulated. This allows for the invocation of the fault diagnosis model for the entire system or the fault diagnosis model for a single component, as needed.

[0137] See Figure 5 , Figure 5 This is a flowchart of a turbine fault diagnosis method based on an AI large model in this application, including the following steps:

[0138] Step S502: Obtain user instructions.

[0139] Step S504: Using a pre-built large language model for steam turbines, the user commands are identified to determine whether they belong to steam turbine fault diagnosis commands.

[0140] The turbine-specific large language model can recognize user commands based on its own understanding capabilities, determining whether the user command belongs to a turbine fault diagnosis command, that is, whether to call the turbine fault diagnosis model. If the user command does not belong to a turbine fault diagnosis command, step S506 is executed; if the user command does belong to a turbine fault diagnosis command, step S508 is executed.

[0141] For example, if the user command is "What is a high-pressure cylinder?", the turbine-specific large language model can recognize this as a general knowledge answer command and directly output the corresponding query result. If the user command is "Perform fault diagnosis on the high-pressure cylinder", the turbine-specific large language model can recognize this as a turbine fault diagnosis command, extract the corresponding monitoring parameters, call the turbine fault diagnosis model to perform fault diagnosis, and finally output the fault diagnosis result.

[0142] Step S506: Output the corresponding query results.

[0143] Step S508: Based on the component corresponding to the turbine fault diagnosis command, extract the monitoring parameters corresponding to the component, and call the fault diagnosis model corresponding to the component from the pre-built turbine fault diagnosis model to process the monitoring parameters corresponding to the component and obtain the fault diagnosis result. The turbine fault diagnosis model is constructed based on the above-mentioned construction method of the turbine fault diagnosis model based on the AI ​​large model.

[0144] This application embodiment intelligently integrates a data-driven turbine fault diagnosis algorithm, which can understand and autonomously interpret user instructions in language form, automatically make decisions and plans, call fault diagnosis models according to needs, and complete complex tasks contained in user instructions.

[0145] Corresponding to the above method embodiments, this application also provides a device for constructing a turbine fault diagnosis model based on an AI large model, see [link to relevant documentation]. Figure 6 The device 600 for constructing a turbine fault diagnosis model based on an AI large model includes:

[0146] The instruction acquisition module 602 is used to acquire instructions for constructing fault diagnosis models for individual components in a steam turbine.

[0147] The first importance value determination module 604 is used to identify the fault diagnosis model construction instructions for a single component by using a pre-built steam turbine professional large language model, extract the candidate monitoring parameters corresponding to the single component, and call the pre-built steam turbine knowledge base to determine the first importance value of the candidate monitoring parameters.

[0148] The second importance value determination module 606 is used to calculate the Pearson correlation coefficient between the candidate monitoring parameter and other candidate monitoring parameters for each candidate monitoring parameter, and determine the second importance value of the candidate monitoring parameter based on the first importance value of other candidate monitoring parameters and the Pearson correlation coefficient between other candidate monitoring parameters and the candidate monitoring parameter.

[0149] The comprehensive importance value determination module 608 is used to perform a weighted average of the first importance value and the second importance value of the candidate monitoring parameter to obtain the comprehensive importance value of the candidate monitoring parameter.

[0150] The target monitoring parameter determination module 610 is used to select candidate monitoring parameters whose comprehensive importance value is greater than a preset threshold as target monitoring parameters.

[0151] The sub-fault diagnosis model construction module 612 is used to construct a fault diagnosis model for a single component based on the time series corresponding to the target monitoring parameters.

[0152] The turbine fault diagnosis model generation module 614 is used to encapsulate the fault diagnosis models corresponding to each component after the fault diagnosis models corresponding to each component are constructed, so as to obtain the turbine fault diagnosis model.

[0153] Optionally, the instruction acquisition module 602 is also used to acquire instructions for constructing a fault diagnosis model for the entire steam turbine.

[0154] The first importance value determination module 604 is also used to identify the fault diagnosis model construction instructions for the whole steam turbine by using the steam turbine professional big language model, extract the candidate monitoring parameters corresponding to the whole steam turbine, and call the steam turbine knowledge base to determine the first importance value of the candidate monitoring parameters.

[0155] The second importance value determination module 606 is also used to calculate the Pearson correlation coefficient between the candidate monitoring parameter and other candidate monitoring parameters for each candidate monitoring parameter, and determine the second importance value of the candidate monitoring parameter based on the first importance value of other candidate monitoring parameters and the Pearson correlation coefficient between other candidate monitoring parameters and the candidate monitoring parameter.

[0156] The comprehensive importance value determination module 608 is also used to perform a weighted average of the first importance value and the second importance value of the candidate monitoring parameter to obtain the comprehensive importance value of the candidate monitoring parameter;

[0157] The target monitoring parameter determination module 610 is also used to select candidate monitoring parameters whose comprehensive importance value is greater than a preset threshold as target monitoring parameters;

[0158] The sub-fault diagnosis model construction module 612 is also used to construct the fault diagnosis model corresponding to the whole machine based on the time series corresponding to the target monitoring parameters.

[0159] The turbine fault diagnosis model generation module 614 is specifically used to encapsulate the fault diagnosis models corresponding to each component and the fault diagnosis model corresponding to the whole machine to obtain the turbine fault diagnosis model.

[0160] Optionally, the second importance value determination module 606 is specifically used for assuming... Indicates components C The corresponding number i The first importance value of each candidate monitoring parameter Indicates components C The corresponding number j The first importance value of each candidate monitoring parameter Indicates the first i The candidate monitoring parameter and the first j The Pearson correlation coefficient between the candidate monitoring parameters is calculated using the following formula: and Determine the first i The second importance value of each candidate monitoring parameter ,in, For the first i The correlation importance value of each candidate monitoring parameter.

[0161] Optionally, the device 600 for constructing a turbine fault diagnosis model based on an AI large model also includes:

[0162] The question-and-answer pair data generation module is used to acquire steam turbine-related corpus and convert it into question-and-answer pair data.

[0163] The general large language model fine-tuning module is used to obtain a general large language model and fine-tune it using question-and-answer pair data to obtain a steam turbine-specific large language model.

[0164] Optionally, the device 600 for constructing a turbine fault diagnosis model based on an AI large model also includes:

[0165] The text recognition module is used to acquire steam turbine-related documents, perform text recognition on the steam turbine-related documents, and obtain knowledge text.

[0166] The text segmentation module is used to segment knowledge text into knowledge text blocks.

[0167] The vector transformation module is used to convert knowledge text blocks into vectors through an embedding model, resulting in a vector database.

[0168] The turbine knowledge base generation module is used to generate a vector database as a turbine knowledge base.

[0169] Optionally, the sub-fault diagnosis model construction module 612 is specifically used to construct training samples based on the time series corresponding to the target monitoring parameters, and to construct a monitoring data prediction model corresponding to a single component using LSTM; for each target monitoring parameter, calculate the deviation between the actual value and the predicted value of the target monitoring parameter on each training sample, and statistically analyze the deviations corresponding to all training samples to obtain the mean and standard deviation; and construct a fault diagnosis model corresponding to a single component based on the monitoring data prediction model, the mean, and the standard deviation.

[0170] This application also provides a turbine fault diagnosis device based on an AI large model, see [link to relevant documentation]. Figure 7 The turbine fault diagnosis device 700 based on an AI large model includes:

[0171] User instruction acquisition module 702 is used to acquire user instructions;

[0172] The user command recognition module 704 is used to identify user commands using a pre-built steam turbine professional language model and determine whether the user command belongs to a steam turbine fault diagnosis command.

[0173] The query result output module 706 is used to output the corresponding query result if the user instruction recognition module 704 determines that the user instruction does not belong to the turbine fault diagnosis instruction.

[0174] The fault diagnosis result output module 708 is used to extract the monitoring parameters corresponding to the component based on the component corresponding to the turbine fault diagnosis instruction if the user instruction recognition module 704 determines that the user instruction belongs to the turbine fault diagnosis instruction. Then, it calls the fault diagnosis model corresponding to the component from the pre-built turbine fault diagnosis model to process the monitoring parameters corresponding to the component and obtain the fault diagnosis result. The turbine fault diagnosis model is constructed based on the above-mentioned construction method of the turbine fault diagnosis model based on the AI ​​large model.

[0175] The specific details of each module or unit in the above-mentioned device have been described in detail in the corresponding methods, so they will not be repeated here.

[0176] It should be noted that although several modules or units for the device used to perform actions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to the embodiments of this application, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0177] This application also provides an electronic device, including: a processor; a memory for storing processor-executable instructions; wherein the processor is configured to execute the above-described AI-based large model-based turbine fault diagnosis method in this example embodiment.

[0178] Reference Figure 8 , Figure 8 This is a schematic diagram of the structure of an electronic device in an embodiment of this application. The specific embodiments of this application do not limit the specific implementation of the electronic device.

[0179] like Figure 8 As shown, the electronic device may include: a processor 802, a communication interface 804, a memory 806, and a communication bus 808.

[0180] The processor 802, communication interface 804, and memory 806 communicate with each other via communication bus 808.

[0181] Communication interface 804 is used to communicate with other electronic devices or servers.

[0182] The processor 802 is used to execute program 810, specifically the relevant steps in the above method embodiments.

[0183] Specifically, program 810 may include program code that includes computer operation instructions.

[0184] The processor 802 may be a central processing unit, a specific integrated circuit, or one or more integrated circuits configured to implement the embodiments of this application. The one or more processors included in the smart device may be processors of the same type, such as one or more CPUs; or they may be processors of different types, such as one or more CPUs and one or more ASICs.

[0185] Memory 806 is used to store program 810. Memory 806 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0186] Specifically, program 810 can be used to cause processor 802 to execute the steps in the above embodiment of the turbine fault diagnosis method based on AI large model.

[0187] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the devices and modules described above can be referred to the corresponding process descriptions in the foregoing method embodiments, and will not be repeated here.

[0188] This application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the above-described turbine fault diagnosis method based on an AI large model.

[0189] It should be noted that the computer-readable storage medium shown in this application can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory, read-only memory, erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this application, the computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. The program code contained on the computer-readable storage medium can be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, radio frequency, etc., or any suitable combination thereof.

[0190] In this embodiment of the application, a computer program product is also provided, which, when run on a computer, causes the computer to execute the above-mentioned turbine fault diagnosis method based on an AI large model.

[0191] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0192] The above description is merely a specific embodiment of this application, enabling those skilled in the art to understand or implement this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments described herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for constructing a turbine fault diagnosis model based on an AI large-scale model, characterized in that, include: Obtain the instructions for constructing fault diagnosis models for individual components in the steam turbine; Using a pre-built turbine professional language model, the instructions for constructing a fault diagnosis model for a single component are identified, and candidate monitoring parameters corresponding to the single component are extracted. The turbine professional language model is then used to call a pre-built turbine knowledge base, so that the turbine professional language model determines the first importance value of the candidate monitoring parameter based on the knowledge in the turbine professional language model and the knowledge in the turbine knowledge base. The steam turbine knowledge base is used to store and manage text materials related to steam turbines; For each candidate monitoring parameter, the Pearson correlation coefficient between the candidate monitoring parameter and other candidate monitoring parameters is calculated, and the second importance value of the candidate monitoring parameter is determined based on the first importance value of other candidate monitoring parameters and the Pearson correlation coefficient between other candidate monitoring parameters and the candidate monitoring parameter. The first importance value and the second importance value of the candidate monitoring parameter are weighted and averaged to obtain the comprehensive importance value of the candidate monitoring parameter. The candidate monitoring parameter with a comprehensive importance value greater than a preset threshold is taken as the target monitoring parameter. Based on the time series corresponding to the target monitoring parameters, a fault diagnosis model for a single component is constructed. After constructing the fault diagnosis models corresponding to each component, the fault diagnosis models corresponding to each component are encapsulated to obtain the turbine fault diagnosis model. The step of determining the second importance value of the candidate monitoring parameter based on the first importance value of other candidate monitoring parameters and the Pearson correlation coefficient between the other candidate monitoring parameters and the candidate monitoring parameter includes: Assumption This represents the first importance value of the i-th candidate monitoring parameter corresponding to component C. corr represents the first importance value of the j-th candidate monitoring parameter corresponding to component C. ij Let represent the Pearson correlation coefficient between the i-th candidate monitoring parameter and the j-th candidate monitoring parameter; according to the following formula: and Determine the second importance value of the i-th candidate monitoring parameter. in, denoted as the relevance importance value of the i-th candidate monitoring parameter.

2. The method for constructing a turbine fault diagnosis model based on an AI large model according to claim 1, characterized in that, The method further includes: Obtain instructions for constructing a fault diagnosis model for the entire steam turbine. Using the turbine professional language model, the instructions for constructing a fault diagnosis model for the entire turbine are identified, candidate monitoring parameters corresponding to the entire turbine are extracted, and the turbine knowledge base is called to determine the first importance value of the candidate monitoring parameters. For each candidate monitoring parameter, the Pearson correlation coefficient between the candidate monitoring parameter and other candidate monitoring parameters is calculated, and the second importance value of the candidate monitoring parameter is determined based on the first importance value of other candidate monitoring parameters and the Pearson correlation coefficient between other candidate monitoring parameters and the candidate monitoring parameter. The first importance value and the second importance value of the candidate monitoring parameter are weighted and averaged to obtain the comprehensive importance value of the candidate monitoring parameter. The candidate monitoring parameter with a comprehensive importance value greater than a preset threshold is taken as the target monitoring parameter. Based on the time series corresponding to the target monitoring parameters, a fault diagnosis model for the whole machine is constructed. The process of encapsulating the fault diagnosis models corresponding to each component to obtain the turbine fault diagnosis model includes: The fault diagnosis models corresponding to each component and the fault diagnosis model corresponding to the whole machine are encapsulated to obtain the turbine fault diagnosis model.

3. The method for constructing a turbine fault diagnosis model based on an AI large model according to claim 1, characterized in that, The method for constructing the large language model for steam turbines includes: Obtain steam turbine-related corpus and convert it into question-answer pairs. A general large language model is obtained, and the general large language model is fine-tuned using the question-and-answer pair data to obtain the steam turbine professional large language model.

4. The method for constructing a turbine fault diagnosis model based on an AI large model according to claim 1, characterized in that, The method for constructing the steam turbine knowledge base includes: Obtain steam turbine-related documents, perform text recognition on the steam turbine-related documents, and obtain knowledge text; The knowledge text is segmented to obtain knowledge text blocks; The knowledge text blocks are converted into vectors by an embedding model to obtain a vector database, which is then used as a steam turbine knowledge base.

5. The method for constructing a turbine fault diagnosis model based on an AI large model according to claim 1, characterized in that, The step of constructing a fault diagnosis model for a single component based on the time series corresponding to the target monitoring parameters includes: Training samples are constructed based on the time series corresponding to the target monitoring parameters, and a monitoring data prediction model for a single component is constructed using a Long Short-Term Memory (LSTM) network. For each target monitoring parameter, the deviation between the actual value and the predicted value of the target monitoring parameter on each training sample is calculated, and the average value and standard deviation are obtained by statistical analysis based on the deviations of all training samples. Based on the monitoring data prediction model, the average value, and the standard deviation, a fault diagnosis model corresponding to a single component is constructed.

6. A turbine fault diagnosis method based on an AI large model, characterized in that, include: The user command is obtained, and the user command is identified using a pre-built large language model for steam turbines to determine whether the user command belongs to a steam turbine fault diagnosis command. If the user command is not a turbine fault diagnosis command, output the corresponding query result; If the user instruction is a turbine fault diagnosis instruction, the monitoring parameters corresponding to the component corresponding to the turbine fault diagnosis instruction are extracted, and the fault diagnosis model corresponding to the component is called from the pre-built turbine fault diagnosis model to process the monitoring parameters corresponding to the component and obtain the fault diagnosis result. The turbine fault diagnosis model is constructed based on the construction method of the turbine fault diagnosis model based on the AI ​​large model as described in any one of claims 1 to 5.

7. A device for constructing a turbine fault diagnosis model based on an AI large-scale model, characterized in that, include: The instruction acquisition module is used to acquire instructions for constructing fault diagnosis models for individual components in a steam turbine. The first importance value determination module is used to identify the fault diagnosis model construction instructions for a single component using a pre-built turbine professional language model, extract the candidate monitoring parameters corresponding to the single component, and call the pre-built turbine knowledge base using the turbine professional language model, so that the turbine professional language model determines the first importance value of the candidate monitoring parameter based on the knowledge in the turbine professional language model and the knowledge in the turbine knowledge base; The steam turbine knowledge base is used to store and manage text materials related to steam turbines; The second importance value determination module is used to calculate the Pearson correlation coefficient between the candidate monitoring parameter and other candidate monitoring parameters for each candidate monitoring parameter, and determine the second importance value of the candidate monitoring parameter based on the first importance value of other candidate monitoring parameters and the Pearson correlation coefficient between other candidate monitoring parameters and the candidate monitoring parameter. The comprehensive importance value determination module is used to perform a weighted average of the first importance value and the second importance value of the candidate monitoring parameter to obtain the comprehensive importance value of the candidate monitoring parameter; The target monitoring parameter determination module is used to select candidate monitoring parameters whose comprehensive importance value is greater than a preset threshold as target monitoring parameters. The sub-fault diagnosis model construction module is used to construct a fault diagnosis model for a single component based on the time series corresponding to the target monitoring parameters. The turbine fault diagnosis model generation module is used to encapsulate the fault diagnosis models corresponding to each component after the fault diagnosis models corresponding to each component are constructed, so as to obtain the turbine fault diagnosis model. Specifically, the second importance value determination module is used to assume... This represents the first importance value of the i-th candidate monitoring parameter corresponding to component C. corr represents the first importance value of the j-th candidate monitoring parameter corresponding to component C. ij Let represent the Pearson correlation coefficient between the i-th candidate monitoring parameter and the j-th candidate monitoring parameter; according to the following formula: and Determine the second importance value of the i-th candidate monitoring parameter. in, denoted as the relevance importance value of the i-th candidate monitoring parameter.

8. The apparatus for constructing a turbine fault diagnosis model based on an AI large model according to claim 7, characterized in that, The instruction acquisition module is also used to acquire instructions for constructing a fault diagnosis model for the entire steam turbine. The first importance value determination module is also used to use the steam turbine professional big language model to identify the fault diagnosis model construction instructions for the whole steam turbine, extract the candidate monitoring parameters corresponding to the whole steam turbine, and call the steam turbine knowledge base to determine the first importance value of the candidate monitoring parameters. The second importance value determination module is further configured to calculate the Pearson correlation coefficient between the candidate monitoring parameter and other candidate monitoring parameters for each candidate monitoring parameter, and determine the second importance value of the candidate monitoring parameter based on the first importance value of the other candidate monitoring parameters and the Pearson correlation coefficient between the other candidate monitoring parameters and the candidate monitoring parameter; The comprehensive importance value determination module is further configured to perform a weighted average of the first importance value and the second importance value of the candidate monitoring parameter to obtain the comprehensive importance value of the candidate monitoring parameter. The target monitoring parameter determination module is also used to select candidate monitoring parameters whose comprehensive importance value is greater than a preset threshold as target monitoring parameters; The sub-fault diagnosis model construction module is also used to construct a fault diagnosis model for the whole machine based on the time series corresponding to the target monitoring parameters. The turbine fault diagnosis model generation module is specifically used to encapsulate the fault diagnosis models corresponding to each component and the fault diagnosis model corresponding to the whole machine to obtain the turbine fault diagnosis model.

9. A turbine fault diagnosis device based on an AI large model, characterized in that, The device includes: The user instruction acquisition module is used to acquire user instructions. The user instruction recognition module is used to identify the user instruction using a pre-built steam turbine professional language model and determine whether the user instruction belongs to a steam turbine fault diagnosis instruction. The query result output module is used to output the corresponding query result if the user instruction recognition module determines that the user instruction does not belong to the turbine fault diagnosis instruction. The fault diagnosis result output module is used to, if the user instruction recognition module determines that the user instruction belongs to the turbine fault diagnosis instruction, extract the monitoring parameters corresponding to the component according to the component corresponding to the turbine fault diagnosis instruction, and call the fault diagnosis model corresponding to the component from the pre-built turbine fault diagnosis model to process the monitoring parameters corresponding to the component to obtain the fault diagnosis result. The turbine fault diagnosis model is constructed based on the construction method of the turbine fault diagnosis model based on the AI ​​large model as described in any one of claims 1 to 5.

Images