A method and apparatus for assisted diagnosis of gas turbine faults

Abstract

This invention discloses a method and device for auxiliary diagnosis of gas turbine faults, belonging to the field of industrial equipment fault diagnosis technology. Utilizing intelligent retrieval algorithms and efficient database support, it can quickly match relevant cases from a massive database of historical cases and provide diagnostic suggestions when a gas turbine experiences a warning or anomaly, shortening the fault diagnosis time. By characterizing historical fault cases of the gas turbine and combining them with current operating background parameters for multi-dimensional, dynamically weighted similarity matching, it can more accurately identify historical cases most similar to the current abnormal operating conditions, thereby improving the accuracy of fault diagnosis. In the specific application scenario of gas turbine power plants, it demonstrates unique technical contributions and significant practical value, which is of great significance for ensuring the safe and stable operation of gas turbines, improving power generation efficiency, and reducing operation and maintenance costs.

Description

Technical Field

[0001] This invention belongs to the field of industrial equipment fault diagnosis technology, specifically relating to a method and equipment for auxiliary fault diagnosis of gas turbines. Background Technology

[0002] As the core power equipment in power plants, gas turbines have complex structures and operate under harsh conditions, requiring extremely high stability and reliability. During operation, due to high temperatures, high pressures, high speeds, and complex combustion and control processes, various faults or abnormal operating conditions inevitably occur. Traditional fault diagnosis methods often rely on the accumulated experience of operators and paper-based procedure manuals, resulting in low diagnostic efficiency and requiring a high level of skill from personnel. Especially for some uncommon or complex fault modes, manual diagnosis is often time-consuming, potentially delaying optimal handling, leading to unplanned downtime or equipment damage, causing serious economic losses and safety risks.

[0003] With the development of industrial internet, big data, and artificial intelligence technologies, intelligent fault diagnosis using historical operating data and fault cases has become a research hotspot. However, existing case-based reasoning (CBR) or similarity matching diagnostic methods still face some challenges when applied to complex dynamic systems such as gas turbines: 1. Diversity and complexity of gas turbine operating conditions: Under the influence of various factors such as start-up and shutdown, load changes, different environmental conditions, and different fuel types, the operating characteristics of gas turbines vary significantly, and the fault manifestations are diverse. Simple parameter matching is insufficient to accurately capture the essential characteristics of the faults.

[0004] 2. Impact of operating background information on fault diagnosis: The operating history of the gas turbine (such as overhaul cycle and component life), the characteristics of the fuel currently used, and environmental conditions have a significant impact on the probability of fault occurrence, type, and diagnostic logic. However, existing methods often ignore or fail to make full use of this contextual information.

[0005] 3. Challenges in Similarity Measurement and Weight Allocation: Under multi-dimensional operating parameters, the key to achieving accurate matching lies in how to scientifically measure the similarity between the current operating condition and historical cases, and dynamically adjust the weights of each dimension according to different features and operating backgrounds.

[0006] 4. Data utilization efficiency and real-time diagnostics: Gas turbines generate massive amounts of real-time operating data. How to efficiently retrieve the most relevant cases from the historical case database and quickly provide diagnostic suggestions places high demands on database technology and retrieval algorithms.

[0007] Therefore, there is an urgent need for a fault-assisted diagnosis method that can fully consider the operating characteristics of gas turbines, deeply integrate operating background information, and adopt intelligent similarity matching strategies, so as to improve the accuracy, efficiency and intelligence level of gas turbine fault diagnosis. Summary of the Invention

[0008] The purpose of this invention is to provide a method and equipment for auxiliary diagnosis of gas turbine faults, so as to solve the technical problem of poor accuracy in existing gas turbine fault diagnosis.

[0009] To achieve the above objectives, the present invention adopts the following technical solution: A method for auxiliary diagnosis of gas turbine faults includes the following steps: Obtain time series data or feature vectors of the current multi-dimensional operating condition parameters corresponding to the gas turbine when it issues an early warning message, and obtain the current operating background state parameters of the gas turbine; Based on the current operating background parameters of the gas turbine, preliminary screening or determination of context-related feature dimensions and their initial weights are performed on cases in the historical fault case database of the gas turbine. The time series data or feature vector of the current multidimensional operating condition parameters are matched with the time series data or feature vector of the operating condition parameters of the historical fault cases after screening or weighting, and similarity weights are obtained by matching and calculating on multiple preset or adaptively selected feature dimensions. Based on the current operating background status parameters of the gas turbine and the operating background status parameters corresponding to the historical fault cases after screening or confirmation, the similarity weight is adaptively adjusted or weighted twice to generate a comprehensive similarity score. The matched historical fault cases are sorted according to the comprehensive similarity score, and one or more historical fault cases with the highest comprehensive similarity score and their associated handling measures are used as auxiliary diagnostic suggestions for the current gas turbine early warning.

[0010] Furthermore, the gas turbine historical fault case database includes at least time-series data of multi-dimensional operating condition parameters at the time of the fault or extracted steady-state and dynamic feature vectors, alarm information text, verified handling measures, and quantifiable operating background state parameters related to the gas turbine.

[0011] Furthermore, the operating background status parameters include at least: the cumulative operating hours of the gas turbine since the last major overhaul, key indicators of the fuel composition currently used, the equivalent operating time of specific components in the combustion chamber, the intake air temperature and humidity, and the current start-up / shutdown status or load adjustment phase of the unit.

[0012] Furthermore, the feature vector of the multidimensional operating condition parameters includes at least one of the following: parameters related to the stability of the gas turbine combustion system, parameters related to the thermodynamic performance of the turbine system, and parameters related to the response characteristics of the control system.

[0013] Furthermore, the adaptive adjustment or secondary weighting of the similarity weights is achieved through a pre-trained machine learning model or a fuzzy inference rule based on gas turbine expert knowledge. The machine learning model or fuzzy inference rule takes the current operating background state parameters and the operating background state parameters of historical cases as inputs and outputs the adjustment weight system for each feature dimension.

[0014] Furthermore, the system records the user's adoption of the output auxiliary diagnostic suggestions, the final confirmed cause of the fault, and the handling measures. This information is then used as a new case or case feedback for online learning and expansion of the historical fault case database.

[0015] Furthermore, the calculation of the comprehensive similarity score also incorporates the effectiveness assessment level of the handling measures in the historical failure case database or the user's rating information for this case.

[0016] Secondly, a gas turbine fault auxiliary diagnosis system is provided, comprising an acquisition module, a filtering module, a matching module, an adjustment module, and an output module, wherein: Acquisition module: used to acquire time series data or feature vectors of the current multi-dimensional operating condition parameters corresponding to the gas turbine when it issues a warning message, and to acquire the current operating background state parameters of the gas turbine; The filtering module is used to perform preliminary filtering of cases in the gas turbine historical failure case database or to determine context-related feature dimensions and their initial weights based on the current operating background status parameters of the gas turbine. The matching module is used to perform matching calculations on multiple preset or adaptively selected feature dimensions to obtain similarity weights by matching the time series data or feature vectors of the current multidimensional operating condition parameters with the time series data or feature vectors of the operating condition parameters corresponding to the historical fault cases after screening or weighting. The adjustment module is used to adaptively adjust or perform secondary weighting of similarity weights based on the current operating background status parameters of the gas turbine and the operating background status parameters corresponding to the historical fault cases after screening or confirmation, and generate a comprehensive similarity score. The output module is used to sort the matched historical fault cases according to the comprehensive similarity score, and to use one or more historical fault cases with the highest comprehensive similarity score and their associated handling measures as auxiliary diagnostic suggestions for the current gas turbine early warning.

[0017] Thirdly, a terminal device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described method.

[0018] Fourthly, a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described method.

[0019] Compared with the prior art, the present invention has the following beneficial technical effects: This invention provides a gas turbine fault auxiliary diagnosis method. Utilizing intelligent retrieval algorithms and efficient database support, it can quickly match relevant cases from a massive database of historical cases and provide diagnostic suggestions when a gas turbine experiences a warning or anomaly, thus shortening the fault diagnosis time. By characterizing historical fault cases of the gas turbine and combining them with current operating background parameters for multi-dimensional, dynamically weighted similarity matching, it can more accurately identify historical cases most similar to the current abnormal operating conditions, thereby improving the accuracy of fault diagnosis. In the specific application scenario of gas turbine power plants, this method demonstrates unique technical contributions and significant practical value, and is of great significance for ensuring the safe and stable operation of gas turbines, improving power generation efficiency, and reducing operation and maintenance costs.

[0020] This invention fully considers the impact of gas turbine operating background information on fault diagnosis, enabling the diagnostic method to better adapt to the diverse and complex operating conditions of gas turbines; it also solidifies the diagnostic experience and historical fault handling knowledge of experts into the case library and matching algorithm, reducing the over-reliance on the personal experience and skill level of operators, and helping to improve the overall operation and maintenance level and knowledge transfer.

[0021] Through continuous accumulation of user feedback and new cases, the system can learn and iteratively optimize its matching strategy and weight model online, continuously improving diagnostic performance. The diagnostic suggestions are visualized and output through the user monitoring interface, providing operators with intuitive and timely decision support, which helps to quickly respond to and handle abnormal situations. Attached Figure Description

[0022] Figure 1 This is a flowchart of an auxiliary diagnosis method for gas turbine faults according to an embodiment of the present invention; Figure 2 This is a flowchart of an auxiliary diagnosis method for gas turbine faults according to an embodiment of the present invention. Detailed Implementation

[0023] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0024] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0025] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0026] The present invention will now be described in further detail with reference to the accompanying drawings: like Figure 1 As shown, a gas turbine fault auxiliary diagnosis method includes the following steps: Step 1: Obtain the time series data or feature vector of the current multi-dimensional operating condition parameters corresponding to the gas turbine when it issues a warning message, and obtain the current operating background state parameters of the gas turbine: A historical fault case database for gas turbines is constructed and maintained. The historical fault case database contains structured data for the diverse and complex fault modes of gas turbines. Each historical fault case includes: time series data of a set of multi-dimensional operating condition parameters at the time of the fault or its extracted steady-state and dynamic feature vectors, alarm information text, verified handling measures, and a set of quantifiable operating background state parameters related to the gas turbine.

[0027] The operating background status parameters include at least: the cumulative operating hours of the gas turbine since the last major overhaul, key indicators of the current fuel composition, the equivalent operating time of specific components in the combustion chamber, the inlet air temperature and humidity, and the current start-up / shutdown status or load adjustment phase of the unit. These parameters reflect the current macroscopic operating conditions and equipment status of the gas turbine, providing an important basis for subsequent context-aware matching.

[0028] When receiving early warning information or real-time operating data of the gas turbine, the system extracts a set of time series data or feature vectors of current multi-dimensional operating condition parameters corresponding to the early warning information or real-time operating data from the industrial data platform, and obtains the current operating background state parameters of the gas turbine.

[0029] Step two: Based on the current operating background parameters of the gas turbine, perform preliminary screening or determine context-related feature dimensions and their initial weights in the historical fault case database of the gas turbine. Historical failure case databases and real-time operational data are stored on an industrial data platform that supports the fusion and management of diverse heterogeneous data. This platform can collect and store second-level or sub-second-level data from the gas turbine distributed control system (DCS), event sequence data from the safety instrumented system (SIS), and maintenance history records from the equipment maintenance management system (EMS). This robust data platform provides the data foundation for the implementation of this method.

[0030] Step 3: Match the time series data or feature vector of the current multidimensional operating condition parameters with the time series data or feature vector of the operating condition parameters corresponding to the historical fault cases after screening or weighting on multiple preset or adaptively selected feature dimensions to obtain similarity weights. A multi-dimensional, dynamically weighted similarity matching strategy is adopted to match the time series data or feature vector of the current multi-dimensional operating condition parameters with the time series data or feature vector of the corresponding operating condition parameters of historical cases in the historical fault case database after being filtered or weighted in step two, on multiple preset or adaptively selected feature dimensions.

[0031] The feature vector of multidimensional operating condition parameters includes features extracted from at least one of the following categories of parameters: parameters related to the stability of the gas turbine combustion system, such as the fluctuation characteristics of flame monitoring signals, and the frequency and amplitude of combustion pressure pulsations; parameters related to the thermodynamic performance of the turbine system, such as the time-domain statistical characteristics and frequency-domain energy distribution of pressure, temperature, and vibration signals at various levels; and parameters related to the response characteristics of the control system, such as the deviation sequence characteristics between the setpoint and feedback values ​​of key control loops. By extracting these deep features of key systems, the operating status and fault symptoms of the gas turbine can be characterized more accurately.

[0032] Step 4: Based on the current operating background status parameters of the gas turbine and the operating background status parameters corresponding to the historical fault cases after screening or confirmation, the similarity weight is adaptively adjusted or weighted twice to generate a comprehensive similarity score. In the similarity matching strategy, the differences or correlations between the current operating background state parameters and the operating background state parameters in historical failure cases are further utilized to adaptively adjust or perform secondary weighting on the weights corresponding to the matching results of different feature dimensions in step three, so as to generate a comprehensive similarity score. This score can reflect the comprehensive similarity between the current operating condition and historical cases under the specific operating background of the gas turbine.

[0033] Step four involves adaptively adjusting or reweighting the weights corresponding to the matching results of different feature dimensions. This is achieved through a pre-trained machine learning model (such as a gradient boosting tree or neural network) or a set of fuzzy inference rules based on gas turbine expert knowledge. This model or rule takes the current operating background state parameters and the operating background state parameters of historical cases as input and outputs the adjusted weight coefficients for each feature dimension. This approach makes the weight adjustment more intelligent and adaptive.

[0034] Iterative optimization of the dynamic weighting mechanism and similarity matching strategy. Through this closed-loop feedback mechanism, the system can continuously learn and improve, enhancing the accuracy and intelligence of diagnosis.

[0035] Step 5: Sort the matched historical fault cases according to the comprehensive similarity score, and take one or more historical fault cases with the highest comprehensive similarity score and their associated handling measures as auxiliary diagnostic suggestions for the current gas turbine early warning. The matched historical fault cases are sorted according to the comprehensive similarity score, and one or more historical fault cases with the highest comprehensive similarity score and their associated handling measures are used as auxiliary diagnostic suggestions for the current gas turbine early warning or anomaly, and are visualized through the user monitoring interface.

[0036] The calculation of the comprehensive similarity score also incorporates the effectiveness evaluation level of the handling measures in historical failure cases or the user's rating information for the case, so as to prioritize the recommendation of solutions that have been proven to be more effective or have higher ratings in the past, thereby improving the practical value of diagnostic recommendations.

[0037] Record the user's adoption of the output auxiliary diagnostic suggestions, the final confirmed cause of the fault and the handling measures, and use this information as a new case or case feedback for online learning and expansion of the historical fault case database.

[0038] In one optional embodiment, a gas turbine fault auxiliary diagnosis method is provided, such as... Figure 2As shown, its core idea is to build a high-quality database of historical gas turbine fault cases. When a new warning or anomaly occurs, it comprehensively considers the current real-time operating conditions and the gas turbine's operating background information, and uses an intelligent similarity matching algorithm to quickly and accurately retrieve the most relevant historical cases from the case database, providing auxiliary diagnostic suggestions for operators. Specifically, it includes the following steps: Step 1: Construction and characterization of a historical gas turbine failure case database; Extensive historical operating data of gas turbines are collected. This data mainly comes from the power plant's industrial data platform, including but not limited to: various sensor data (such as temperature, pressure, flow, speed, vibration, valve opening, etc.), alarm and event logs recorded by the distributed control system (DCS); action records of the safety instrumented system (SIS); equipment ledgers, maintenance records, fault reports and spare parts replacement records in the equipment management system (EMS) or computerized maintenance management system (CMMS); as well as manually filled operating logs, shift handover records and accident analysis reports.

[0039] The collected raw data undergoes preprocessing operations such as cleaning, noise reduction, missing value handling, and time alignment to ensure data quality and consistency.

[0040] From the preprocessed data, historical gas turbine failures or typical abnormal events are identified and extracted. Each event constitutes a historical failure case.

[0041] Provide a structured description for each case, including at least the following information: Unique Case ID: Used to identify each case; Fault / Abnormality Description: A textual description of the fault phenomenon, type, time of occurrence, scope of impact, etc. Operating parameters: Time-series data of key operating parameters for a period of time prior to the failure (e.g., from one hour before the failure to the moment of failure). These parameters should comprehensively reflect the status of all major systems of the gas turbine (such as the combustion system, turbine system, control system, auxiliary systems, etc.) at the time of the failure. For example, for the combustion system, this may include fuel flow rate, air flow rate, guide vane angle, combustion chamber pressure, exhaust gas temperature, flame monitoring signals, etc.; for the turbine system, this may include inlet and outlet pressures and temperatures of each stage cylinder, bearing vibration, axial displacement, speed, etc. Alarm information: A list of relevant DCS alarms triggered before and after the fault occurred, along with their occurrence times; Corrective measures: Operational procedures, adjustment plans, or repair / replacement records that have been taken and proven effective in response to this fault; Operating background status parameters: This includes the gas turbine's operating background information at the time of the fault, such as: unit model, cumulative operating hours since the last major overhaul, current fuel type (e.g., natural gas, backup fuel and its key component indicators such as calorific value, Wobbe index, etc.), equivalent operating time (EOH) of specific combustion chamber components (e.g., flame tube, transition section), current inlet air temperature, inlet air humidity, atmospheric pressure, and the unit's current start-up / shutdown status (e.g., starting, shutting down, stable combustion, under load, etc.) or load adjustment phase (e.g., rapid load increase, load shedding, etc.). These parameters need to be quantified or standardized. Case source and verification information: Record the data source, verification personnel, verification time, etc. of the case to ensure the reliability of the case.

[0042] To perform similarity calculations more effectively, feature extraction is performed on the time-series data of the operating condition parameters for each case. Extractable features include: Steady-state characteristics: such as the mean, variance, maximum, minimum, median, peak-to-peak value, and other statistical measures of a time series. Dynamic characteristics: such as the trend slope of the time series, fluctuation frequency, autocorrelation coefficient, cross-correlation coefficient (and other related parameters), and spectral characteristics (such as energy spectrum and frequency components obtained through Fourier transform or wavelet transform).

[0043] For specific types of parameters (such as vibration signals), more specialized features can be extracted, such as kurtosis, margin, and side frequencies. The extracted feature vectors will be stored in a case library along with the original time series data (or its compressed representation) for subsequent matching.

[0044] The structured historical fault cases and their feature vectors are stored in a high-performance database. Considering the time-series nature and query requirements of gas turbine data, a time-series database (such as InfluxDB, TimescaleDB), a relational database that supports efficient time-series data processing (such as PostgreSQL with relevant plugins), or a NoSQL database (such as MongoDB) can be selected.

[0045] The database supports fast conditional retrieval, multidimensional indexing, and similarity queries. It provides functions for adding, deleting, modifying, and querying cases, as well as version management, ensuring data consistency and security. The case library is regularly maintained and updated, removing outdated or incorrect cases and adding new, representative examples.

[0046] Step 2: Real-time data acquisition and current state vector generation; The system receives warning signals in real time from the gas turbine's intelligent early warning module, or periodically obtains key real-time operating data of the gas turbine directly from the industrial data platform. Warning information typically includes the warning level, warning parameters, and the time of occurrence.

[0047] The intelligent early warning module itself can monitor the gas turbine's operating status in real time based on big data analysis, artificial intelligence algorithms (such as neural networks, support vector machines, decision trees, etc.) or hybrid modeling technology, and identify weak abnormal signals or parameter deviations in the early stages of a fault (even before the DCS system generates an obvious alarm).

[0048] When an early warning is received or triggered during a routine diagnostic cycle, the system extracts time-series data of relevant operating condition parameters from the industrial data platform based on the early warning parameters or a preset list of key parameters. This data is taken from a period of time prior to the current moment (consistent with the window defined in the case library, such as 1 hour before the early warning occurs).

[0049] For the extracted time series data of current operating condition parameters, the same feature extraction method as the case featureization in step one is used to generate the feature vector of the current operating condition.

[0050] Simultaneously, the system retrieves the current operating background status parameters of the gas turbine from the industrial data platform or related management systems. The types and definitions of these parameters should be consistent with the operating background status parameters stored in the case library. For example, it can retrieve the current load, start-stop status, inlet temperature and humidity, etc., from the DCS; retrieve the last overhaul time from the EMS or maintenance records to calculate the cumulative operating hours; and retrieve the key component indicators of the current fuel from the fuel management system.

[0051] Step 3: Context-aware and dynamically weighted intelligent retrieval; Using the current operating background status parameters obtained in step two, preliminary context-aware processing is performed on the historical fault case database. One approach is preliminary filtering: for example, if the current gas turbine uses natural gas, priority can be given to matching fault cases that occurred in the past when natural gas was also used; if the current unit is in the startup phase, priority can be given to matching fault cases that occurred in the startup phase in the past. This filtering can be achieved by adding conditional filtering based on operating background parameters when querying the database. Another approach is to determine the context-related feature dimensions and their initial weights: under different operating backgrounds, the importance of different operating condition parameters for fault diagnosis may vary. For example, in low-temperature environments, the weight of parameters related to the antifreeze system may need to be increased; in the early stages of overhaul, the weight of parameters related to newly installed components may need to be considered. The system can determine an initial weight allocation scheme for subsequent multi-dimensional feature matching based on the current operating background status parameters, through a preset rule base or a small machine learning model.

[0052] The feature vector of the current operating condition generated in step two is compared with the feature vectors of the operating conditions of historical cases in the historical fault case database after preliminary screening or initial weighting, and similarity is calculated on multiple feature dimensions.

[0053] For each feature dimension, a similarity metric suitable for that feature type can be used. For example, for numerical steady-state features (such as mean and maximum), normalized Euclidean distance, Manhattan distance, or cosine similarity can be used; for time series data itself (if it does not fully depend on feature vectors), Dynamic Time Warping (DTW) or Longest Common Subsequence (LCSS) methods can be used; for text-based alarm information, bag-of-words models, TF-IDF combined with cosine similarity, or more advanced semantic similarity calculation methods based on word embeddings (such as Word2Vec and BERT) can be used. Matching for each feature dimension will yield a similarity score for that dimension.

[0054] The current operating background state parameters are compared with the operating background state parameters of each candidate historical case. A pre-trained machine learning model is used, with the input being the operating background parameter pairs of the current and historical cases, and the output being the adjustment coefficients of the weights of each operating condition feature dimension; or a set of fuzzy inference rules based on gas turbine expert knowledge, to dynamically adjust the weights of the similarity scores of each feature dimension obtained in step two.

[0055] After dynamic weighting, the similarity scores of all weighted feature dimensions are aggregated (e.g., weighted summation, weighted average, or other aggregation functions) to obtain a comprehensive similarity score between the current operating condition and each historical case. This score comprehensively reflects the similarity of the operating condition parameters and the degree of matching under a specific operating context.

[0056] In this step, the effectiveness evaluation level of the handling measures in historical failure cases or user rating information for the case can also be incorporated. For example, if the handling measures of a historical case are rated as "efficient" or have a high user rating, its overall similarity score can be given a certain bonus when calculating its final recommendation ranking.

[0057] Step 4: Diagnostic Assistance and Ranking Recommendation; Based on the comprehensive similarity score calculated in step three, all matching historical fault cases are sorted in descending order, and one or more historical fault cases with the highest comprehensive similarity score (e.g., Top-N, where N can be configured by the user) are selected as the most relevant reference cases.

[0058] Key information of recommended historical fault cases (such as fault description, snapshot or trend of operating parameters at the time, alarm information, verified handling measures, and operating background at the time) and their comprehensive similarity score with the current operating condition are visualized through a specific window or pop-up in the user monitoring interface. Highlighting, comparison and other methods can be used to clearly show the similarities and differences between the current operating condition and the recommended cases to the operators. The handling measures are given in a clear and actionable step-by-step format.

[0059] Users are allowed to confirm, ignore, or mark recommended cases as irrelevant. If a user adopts the handling measures of a case and successfully solves the problem, or determines the cause of the failure and effective measures through other means, the system should provide an interface to record this feedback information. This feedback information will be used for online learning and expansion of the case library: adding confirmed failures and their solutions as new high-quality cases to the database, or supplementing and correcting existing cases.

[0060] Iterative optimization of matching strategies and weight models: By utilizing user feedback data, similarity calculation methods and feature weight allocation models (especially machine learning models) can be retrained or their parameters adjusted, thereby continuously improving the diagnostic performance of the system.

[0061] This invention also provides a gas turbine fault auxiliary diagnosis system, comprising an acquisition module, a filtering module, a matching module, an adjustment module, and an output module, wherein: Acquisition module: used to acquire time series data or feature vectors of the current multi-dimensional operating condition parameters corresponding to the gas turbine when it issues a warning message, and to acquire the current operating background state parameters of the gas turbine; The filtering module is used to perform preliminary filtering of cases in the gas turbine historical failure case database or to determine context-related feature dimensions and their initial weights based on the current operating background status parameters of the gas turbine. The matching module is used to perform matching calculations on multiple preset or adaptively selected feature dimensions to obtain similarity weights by matching the time series data or feature vectors of the current multidimensional operating condition parameters with the time series data or feature vectors of the operating condition parameters corresponding to the historical fault cases after screening or weighting. The adjustment module is used to adaptively adjust or perform secondary weighting of similarity weights based on the current operating background status parameters of the gas turbine and the operating background status parameters corresponding to the historical fault cases after screening or confirmation, and generate a comprehensive similarity score. The output module is used to sort the matched historical fault cases according to the comprehensive similarity score, and to use one or more historical fault cases with the highest comprehensive similarity score and their associated handling measures as auxiliary diagnostic suggestions for the current gas turbine early warning.

[0062] In a preferred embodiment of the present invention, a gas turbine fault auxiliary diagnosis system is provided, which is typically deployed on the industrial internet platform of a power plant or the edge cloud platform of a smart power plant. Its typical architecture includes the following modules: Data access and preprocessing module: responsible for real-time acquisition of gas turbine data from various data sources such as DCS, SIS, and EMS, and for data cleaning, transformation, and preliminary processing; Historical Fault Case Database: The core storage module, used to store structured historical fault cases and their characteristics; Feature engineering module: responsible for feature extraction and vectorization of historical and real-time operating condition data; Background Management Module: Responsible for acquiring, storing, and managing the gas turbine's background status parameters; Intelligent retrieval and matching engine: The core computing module, which realizes functions such as context awareness, multi-dimensional feature matching, dynamic weighting and comprehensive similarity calculation; Diagnosis Results and Recommendation Module: Responsible for sorting the matching results and generating auxiliary diagnostic suggestions; User interface and visualization module: Responsible for displaying diagnostic results, recommended cases, and treatment measures on the monitoring system, and receiving user feedback; Learning and Optimization Module: Responsible for online learning and iterative optimization of the case library and matching model based on user feedback and new cases.

[0063] Deployment considerations: The system should adopt a B / S architecture or a C / S architecture to facilitate user access and management; The database is highly available and scalable to cope with the ever-increasing volume of data and concurrent access demands; The computational efficiency of the intelligent retrieval and matching engine should be able to meet the real-time requirements of fault diagnosis (e.g., second-level response). The system has good compatibility and can be integrated with the power plant's existing industrial data platform, early warning system, monitoring system, etc. Consider data security and access control to ensure the security of sensitive data.

[0064] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can be implemented in one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs) containing computer-usable program code. The form of a computer program product implemented on ROM, optical memory, etc.

[0065] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0066] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0067] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0068] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit its scope of protection. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that after reading the present invention, they can still make various changes, modifications or equivalent substitutions to the specific implementation of the invention, but these changes, modifications or equivalent substitutions are all within the scope of protection of the pending claims of the invention.

Claims

1. A method for auxiliary diagnosis of gas turbine faults, characterized in that, Includes the following steps: Obtain time series data or feature vectors of the current multi-dimensional operating condition parameters corresponding to the gas turbine when it issues an early warning message, and obtain the current operating background state parameters of the gas turbine; Based on the current operating background parameters of the gas turbine, preliminary screening or determination of context-related feature dimensions and their initial weights are performed on cases in the historical fault case database of the gas turbine. The time series data or feature vector of the current multidimensional operating condition parameters are matched with the time series data or feature vector of the operating condition parameters of the historical fault cases after screening or weighting, and similarity weights are obtained by matching and calculating on multiple preset or adaptively selected feature dimensions. Based on the current operating background status parameters of the gas turbine and the operating background status parameters corresponding to the historical fault cases after screening or confirmation, the similarity weight is adaptively adjusted or weighted twice to generate a comprehensive similarity score. The matched historical fault cases are sorted according to the comprehensive similarity score, and one or more historical fault cases with the highest comprehensive similarity score and their associated handling measures are used as auxiliary diagnostic suggestions for the current gas turbine early warning.

2. The auxiliary diagnostic method for gas turbine faults according to claim 1, characterized in that, The gas turbine historical fault case database includes at least time-series data of multi-dimensional operating condition parameters at the time of the fault or extracted steady-state and dynamic feature vectors, alarm information text, verified handling measures, and quantifiable operating background state parameters related to the gas turbine.

3. The auxiliary diagnostic method for gas turbine faults according to claim 1, characterized in that, The operating background status parameters include at least: the cumulative operating hours of the gas turbine since the last major overhaul, key indicators of the fuel composition currently used, the equivalent operating time of specific components in the combustion chamber, the intake air temperature and humidity, and the current start-up / shutdown status or load adjustment phase of the unit.

4. The auxiliary diagnostic method for gas turbine faults according to claim 1, characterized in that, The feature vector of the multidimensional operating condition parameters includes at least one of the following: parameters related to the stability of the gas turbine combustion system, parameters related to the thermodynamic performance of the turbine system, and parameters related to the response characteristics of the control system.

5. The auxiliary diagnostic method for gas turbine faults according to claim 1, characterized in that, The adaptive adjustment or secondary weighting of the similarity weights is achieved through a pre-trained machine learning model or a fuzzy inference rule based on gas turbine expert knowledge. The machine learning model or fuzzy inference rule takes the current operating background state parameters and the operating background state parameters of historical cases as inputs and outputs the adjustment weight coefficients of each feature dimension.

6. The auxiliary diagnostic method for gas turbine faults according to claim 1, characterized in that, Record the user's adoption of the output auxiliary diagnostic suggestions, the final confirmed cause of the fault and the handling measures, and use this information as a new case or case feedback for online learning and expansion of the historical fault case database.

7. The auxiliary diagnostic method for gas turbine faults according to claim 1, characterized in that, The calculation of the comprehensive similarity score also incorporates the effectiveness assessment level of the handling measures in the historical failure case database or the user rating information for this case.

8. A gas turbine fault auxiliary diagnostic system, characterized in that, It includes an acquisition module, a filtering module, a matching module, an adjustment module, and an output module, among which: Acquisition module: used to acquire time series data or feature vectors of the current multi-dimensional operating condition parameters corresponding to the gas turbine when it issues a warning message, and to acquire the current operating background state parameters of the gas turbine; The filtering module is used to perform preliminary filtering of cases in the gas turbine historical failure case database or to determine context-related feature dimensions and their initial weights based on the current operating background status parameters of the gas turbine. The matching module is used to perform matching calculations on multiple preset or adaptively selected feature dimensions to obtain similarity weights by matching the time series data or feature vectors of the current multidimensional operating condition parameters with the time series data or feature vectors of the operating condition parameters corresponding to the historical fault cases after screening or weighting. The adjustment module is used to adaptively adjust or perform secondary weighting of similarity weights based on the current operating background status parameters of the gas turbine and the operating background status parameters corresponding to the historical fault cases after screening or confirmation, and generate a comprehensive similarity score. The output module is used to sort the matched historical fault cases according to the comprehensive similarity score, and to use one or more historical fault cases with the highest comprehensive similarity score and their associated handling measures as auxiliary diagnostic suggestions for the current gas turbine early warning.

9. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements as claimed in claim 1.

7. The steps of any of the methods described.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements as described in claim 1.

7. The steps of any of the methods described.

Images