Methods, systems, equipment and storage media for predicting vibration of turbine generator shaft systems
By combining K-Means clustering, PCA, and VMD with a sparrow search-optimized XGBoost model, the accuracy problem of turbine generator vibration monitoring was solved, enabling efficient prediction and real-time monitoring of shaft vibration.
Patent Information
- Application Number
- CN202311065804.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-08-23
- Publication Date
- 2025-10-31
- Estimated Expiration
- 2043-08-23
AI Technical Summary
In the current technology, vibration monitoring of steam turbine generators mainly relies on manual experience, which makes it difficult to accurately judge the operating status of the unit under deep peak shaving conditions, and lacks forward-looking anomaly warning and maintenance plans.
An XGBoost model optimized by the Sparrow Search algorithm was built using the K-Means data clustering algorithm and the support vector machine model, combined with principal component analysis (PCA) and variational mode decomposition (VMD), to predict the vibration parameters of the turbine generator shaft system.
It enables accurate prediction of turbine generator shaft vibration, improves the accuracy and efficiency of operating status monitoring, and is suitable for real-time monitoring in complex operating environments.
Smart Images
Figure CN117150330B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of steam turbines, and more particularly to a method, system, device, and storage medium for predicting shaft vibration of a steam turbine generator. Background Technology
[0002] As a core piece of equipment in thermal power generation systems, steam turbine generators undertake crucial power generation tasks. Ensuring the stable and safe operation of steam turbine generators remains one of the most important tasks of the power system. During the process of deep peak shaving by steam turbine generators, the operating conditions of large-capacity thermal power units become increasingly complex, which places higher demands on ensuring the safe and stable operation of steam turbine generators.
[0003] The operation monitoring system for steam turbine generators is becoming increasingly mature, enabling the accurate acquisition of real-time operational data through computer and signal processing. Vibration core parameters such as shaft vibration and bearing vibration, as the main protection parameters of the steam turbine generator, can reflect its operating characteristics and, in most cases, infer the unit's operating status. Accurate prediction of steam turbine generator vibration parameters can achieve the goal of accurate monitoring of its operating status. In practical engineering applications, abnormal vibration warnings can alert operators to the current state of the steam turbine generator, allowing them to proactively assess any abnormalities. Furthermore, by comparing the abnormal values with the actual values of vibration parameters, the health status of the steam turbine generator can be determined, providing data support for scheduling appropriate maintenance plans.
[0004] Currently, in the actual engineering operation of steam turbine generators, the traditional method of judging the status of steam turbine generators, which relies on manual monitoring data, is mostly adopted. This method is not only subjective, meaning that the judgment results mostly depend on the experience and knowledge of experts, but also difficult to cope with the complex operating conditions of steam turbine generators under the current industrial background of deep peak shaving.
[0005] Therefore, the above problems urgently need to be solved. Summary of the Invention
[0006] Purpose of the invention: The first objective of this invention is to provide a method for accurately predicting the vibration of a steam turbine generator shaft system.
[0007] The second objective of this invention is to provide a turbine generator shaft vibration prediction system.
[0008] A third objective of this invention is to provide an electronic device.
[0009] A fourth objective of this invention is to provide a computer storage medium.
[0010] Technical solution: To achieve the above objectives, this invention discloses a method for predicting vibration of a steam turbine generator shaft system, comprising the following steps:
[0011] (1) Obtain historical data of turbine generator operating parameters;
[0012] (2) The collected historical data of turbine generator operating parameters are preprocessed to eliminate the influence of the difference in the dimensions of variables on the results;
[0013] (3) The operating status of the steam turbine generator is identified. The K-Means data clustering algorithm is used to divide the operating data of the steam turbine generator into different operating modes based on two pattern recognition feature parameters: ambient temperature and steam turbine active power. A support vector machine (SVM) steam turbine operating mode recognition model is constructed.
[0014] (4) Use principal component analysis (PCA) to extract features from the historical operating parameter data of the steam turbine generator to obtain state feature values;
[0015] (5) Divide the state characteristic parameters of the steam turbine generator after dimensionality reduction by principal component analysis (PCA) into test set, validation set and test set in a ratio of 8:1:1;
[0016] (6) The shaft vibration parameter sequence described in step 1 is decomposed by variational mode decomposition algorithm (VMD), and the final number of sub-signal modes K is determined by the center frequency method. The shaft vibration parameter sequence is decomposed into K IMF sequences by variational mode decomposition algorithm (VMD).
[0017] (7) Build and optimize the XGBoost model according to the Sparrow Search Algorithm (SSA). Set the optimal parameters for the sequences decomposed by VMD. For the K IMF sequences obtained by decomposition, build K XGBoost models respectively. Use the optimized XGBoost model to predict the vibration parameters of the steam turbine generator.
[0018] (8) The state characteristic parameters of the steam turbine generator after dimensionality reduction by principal component analysis (PCA) are used as input features and input into the optimized K XGBoost prediction models respectively. The predicted values of each IMF sequence are superimposed and reconstructed to finally obtain the vibration parameter prediction values.
[0019] In step (1), the historical data of turbine generator operating parameters obtained include historical data of turbine generator power parameters, historical data of thermodynamic parameters, and historical data of shaft vibration parameters.
[0020] The thermal parameters of the turbine generator shaft system include the main steam valve steam pressure, the steam pressure after the high-pressure bypass valve, the pressure of the high-pressure cylinder blade stage, the steam pressure after the regulating valve, the steam pressure after the make-up air valve, the steam inlet pressure of the intermediate-pressure cylinder, the steam outlet pressure of the intermediate-pressure cylinder, the steam outlet temperature of the intermediate-pressure cylinder, the steam outlet temperature of the high-pressure cylinder, the vacuum of the low-pressure cylinder, the deaerator pressure, and the main steam flow rate.
[0021] Preferably, the specific steps of the preprocessing operation in step (2) are as follows:
[0022] All historical data of operating parameters are normalized and mapped to the range [0, 1] as sample data. The normalization formula is shown below:
[0023]
[0024] Where X and A represent the data before and after normalization, respectively. max With X min These represent the maximum and minimum values of a sequence before normalization, respectively.
[0025] Set up a historical dataset of steam turbine generator operating parameters. It has n sets of data, and the feature dimension of each set of data is m;
[0026] Then the normalized data matrix of dataset X is
[0027] Furthermore, the principal component analysis (PCA) feature extraction method in step (4) specifically includes the following steps:
[0028] (4.1) Process the normalized data matrix A to obtain the covariance matrix L;
[0029] (4.2) Calculate the eigenvalues and eigenvectors of the covariance matrix L;
[0030] (4.3) Calculate the cumulative contribution rate of the principal components and select the feature values with a cumulative contribution rate of 90% as the input for prediction.
[0031] Furthermore, in step (7), the hyperparameters of the XGBoost prediction model are tuned and optimized using the Sparrow Search Algorithm (SSA). The hyperparameter optimization process includes:
[0032] (7.1) The optimal hyperparameter combination is obtained by using the Sparrow Search Algorithm (SSA) as the hyperparameter input of the XGBoost model. Initial values are set for the number of intelligent algorithm populations and the number of intelligent algorithm evolutions in the Sparrow Search Algorithm.
[0033] (7.2) Optimize the hyperparameters of the prediction model, i.e. set them as the search dimension of the SSA sparrow search algorithm and set the corresponding search range;
[0034] (7.3) Construct a decision tree model, gradually increase the number of trees by using the evaluation criteria of feature values, and use the new trees to fit the residuals of the previous prediction;
[0035] (7.4) Calculate the fitness of each sparrow to classify discoverers and followers, update their positions, and compare the current fitness of each sparrow with the current best fitness:
[0036] (7.5) If the fitness of the current individual is higher than the current best fitness, then update the current best fitness and use its corresponding position as the fitness and position of the current individual;
[0037] (7.6) If the fitness of an individual is lower than the current best fitness, then the current best fitness and the current best position remain unchanged;
[0038] (7.7) When the number of iterations reaches the global maximum number of iterations, end the iteration and record the current hyperparameter value;
[0039] (7.8) The recorded hyperparameter values are the global optimal parameters, and an XGBoost prediction model is built.
[0040] This invention discloses a turbine generator shaft vibration prediction system based on signal decomposition and reconstruction, characterized in that it comprises:
[0041] The data acquisition module is used to acquire historical data of turbine generator operating parameters. The historical data of turbine generator operating parameters acquired by the data acquisition module includes historical data of turbine generator power parameters, historical data of thermal parameters, and historical data of shaft vibration parameters. Among them, the turbine generator shaft thermal parameters include main steam valve steam pressure, high-pressure bypass valve downstream steam pressure, high-pressure cylinder blade stage pressure, regulating valve downstream steam pressure, make-up air valve downstream steam pressure, intermediate-pressure cylinder inlet steam pressure, intermediate-pressure cylinder exhaust steam pressure, intermediate-pressure exhaust steam temperature, high-pressure exhaust steam temperature, low-pressure cylinder vacuum, deaerator pressure, and main steam flow rate.
[0042] The preprocessing module is used to preprocess the collected historical data of turbine generator operating parameters to eliminate the influence of differences in variable dimensions on the results.
[0043] The status recognition module is used to identify the operating status of the steam turbine generator. It uses the K-Means data clustering algorithm to divide the operating data of the steam turbine generator into different operating modes based on two pattern recognition feature parameters: ambient temperature and steam turbine active power. It then constructs a support vector machine (SVM) steam turbine operating mode recognition model.
[0044] The eigenvalue extraction module is used to extract features from the historical operating parameter data of the steam turbine generator using Principal Component Analysis (PCA) to obtain state eigenvalues. The PCA feature extraction method within the eigenvalue extraction module is specifically executed as follows:
[0045] The normalized data matrix A is processed to obtain the covariance matrix L;
[0046] Calculate the eigenvalues and eigenvectors of the covariance matrix L;
[0047] Calculate the cumulative contribution rate of the principal components, and select the feature values with a cumulative contribution rate of 90% as the input for prediction;
[0048] The data partitioning module is used to divide the state characteristic parameters of the steam turbine generator after dimensionality reduction by principal component analysis (PCA) into test set, validation set and test set in a ratio of 8:1:1.
[0049] The decomposition processing module is used to decompose the acquired shaft vibration parameter sequence using the variational mode decomposition algorithm (VMD), determine the final number of sub-signal modes K using the center frequency method, and decompose the above shaft vibration parameter sequence into K IMF sequences using the variational mode decomposition algorithm (VMD).
[0050] The XGBoost model optimization module is used to build and optimize the XGBoost model according to the Sparrow Search Algorithm (SSA). It sets the optimal parameters for the sequences decomposed by VMD, builds K XGBoost models for the K IMF sequences obtained by decomposition, and uses the optimized XGBoost model to predict the vibration parameters of the steam turbine generator.
[0051] The vibration parameter prediction module is used to take the state characteristic parameters of the steam turbine generator after dimensionality reduction by principal component analysis (PCA) as input features, input them into K optimized XGBoost prediction models respectively, and superimpose and reconstruct the predicted values of each IMF sequence to finally obtain the vibration parameter prediction values.
[0052] The specific steps of the preprocessing operation in the preprocessing module are as follows:
[0053] All historical data of operating parameters are normalized and mapped to the range [0, 1] as sample data. The normalization formula is shown below:
[0054]
[0055] Where X and A represent the data before and after normalization, respectively. max With X min These represent the maximum and minimum values of a sequence before normalization, respectively.
[0056] Set up a historical dataset of steam turbine generator operating parameters. It has n sets of data, and the feature dimension of each set of data is m;
[0057] Then the normalized data matrix of dataset X is
[0058] Preferably, the hyperparameter tuning and optimization of the XGBoost prediction model is performed in the XGBoost model optimization module using the Sparrow Search Algorithm (SSA). The specific execution process of its hyperparameter optimization is as follows:
[0059] The optimal hyperparameter combination was obtained by using the Sparrow Search Algorithm (SSA) as the hyperparameter input for the XGBoost model. Initial values were set for the number of intelligent algorithm populations and the number of intelligent algorithm evolutions in the Sparrow Search Algorithm.
[0060] The hyperparameters of the prediction model are optimized, that is, set as the search dimension of the SSA sparrow search algorithm and set the corresponding search range.
[0061] Build a decision tree model, gradually increase the number of trees by using the evaluation criteria of feature values, and use the new trees to fit the residuals of the previous prediction;
[0062] Calculate the fitness of each sparrow to classify them as discoverers and followers, update their positions, and compare the current fitness of each sparrow with the current best fitness:
[0063] If the fitness of the current individual is higher than the current best fitness, then update the current best fitness and use its corresponding position as the fitness and position of the current individual.
[0064] If the fitness of an individual is lower than the current best fitness, then the current best fitness and the current best position remain unchanged.
[0065] When the number of iterations reaches the global maximum number of iterations, the iteration ends and the current hyperparameter value is recorded;
[0066] The recorded hyperparameter values are the globally optimal parameters, and an XGBoost prediction model is built.
[0067] This invention discloses an electronic device, including a processor and a memory.
[0068] The memory is used to store a computer program, which is executed by the processor to perform the above-described method for predicting the vibration of a turbine generator shaft system.
[0069] The present invention discloses a computer storage medium storing a computer program, which is executed by a processor to perform the above-described method for predicting the vibration of a turbine generator shaft system.
[0070] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages: The present invention performs modal decomposition of vibration target parameters based on variational mode decomposition, and combines the sparrow search algorithm to optimize the hyperparameters of the XGBoost model to further improve the prediction accuracy. It solves the inaccuracy caused by manual parameter tuning and improves the efficiency of model optimization, making it suitable for solving practical engineering problems. By establishing vibration prediction models for different operating modes of steam turbine generators and combining real-time operating parameters under different operating modes, long-term accurate and efficient monitoring of steam turbine vibration can be achieved. Attached Figure Description
[0071] Figure 1 This is a schematic diagram of the process of the present invention;
[0072] Figure 2 This is a schematic diagram of the turbine shaft system in an embodiment of the present invention;
[0073] Figure 3 This is a diagram showing the VMD decomposition results in this invention;
[0074] Figure 4 This is a flowchart of the XGBoost hyperparameter optimization based on SSA in this invention;
[0075] Figure 5 This is a vibration prediction diagram for the turbine generator shaft system of the present invention. Detailed Implementation
[0076] The technical solution of the present invention will be further described below with reference to the accompanying drawings.
[0077] Example 1
[0078] Extreme Gradient Boosting (XGBoost) is an efficient and effective gradient boosting algorithm that can be used for classification or regression prediction modeling problems. Because it combines a linear model solver and a tree learning algorithm, XGBoost outperforms algorithms such as Random Forest (RF), Gradient Boosting Machine (GBM), and Gradient Boosting Decision Tree (GBDT) in terms of execution speed and model performance. In particular, its cross-cluster parallelism makes it several times faster than other prediction algorithms, making it suitable for practical steam turbine generator engineering applications.
[0079] Variational Mode Decomposition (VMD) is a signal decomposition and estimation method proposed by Dragomiretskiy and Zosso in 2014. It assumes that the original signal f is composed of intrinsic mode functions (IMFs). By iteratively searching for the optimal solution of the variational model, the frequency center and bandwidth of each component are determined. By continuously updating each mode function and its center frequency, several mode functions with a certain bandwidth are obtained, which have excellent noise resistance, decomposition performance and stability.
[0080] like Figure 1 and Figure 2 As shown, the present invention provides a method for predicting vibration of a turbine generator shaft system, comprising the following steps:
[0081] (1) Obtain historical operating parameter data of the steam turbine generator; the historical operating parameter data of the steam turbine generator includes historical data of steam turbine generator power parameters, historical data of thermal parameters, and historical data of shaft vibration parameters.
[0082] The thermal parameters of the turbine generator shaft system include the main steam valve steam pressure, the steam pressure after the high pressure bypass valve, the pressure of the high pressure cylinder blade stage, the steam pressure after the regulating valve, the steam pressure after the make-up air valve, the steam inlet pressure of the intermediate pressure cylinder, the steam outlet pressure of the intermediate pressure cylinder, the steam outlet temperature of the intermediate pressure cylinder, the steam outlet temperature of the high pressure cylinder, the vacuum of the low pressure cylinder, the deaerator pressure, and the main steam flow rate.
[0083] (2) The collected historical data of turbine generator operating parameters are preprocessed to eliminate the influence of the difference in the dimensions of variables on the results;
[0084] Common data standardization methods mainly include min-max standardization and z-score standardization. This invention selects min-max as the data standardization method. The specific steps of the preprocessing operation are as follows:
[0085] All historical data of operating parameters are normalized and mapped to the range [0, 1] as sample data. The normalization formula is shown below:
[0086]
[0087] Where X and A represent the data before and after normalization, respectively. max With X min These represent the maximum and minimum values of a sequence before normalization, respectively.
[0088] Set up a historical dataset of steam turbine generator operating parameters. It has n sets of data, and the feature dimension of each set of data is m;
[0089] Then the normalized data matrix of dataset X is
[0090] (3) The operating status of the steam turbine generator is identified. The K-Means data clustering algorithm is used to divide the operating data of the steam turbine generator into different operating modes based on two pattern recognition feature parameters: ambient temperature and steam turbine active power. A support vector machine (SVM) steam turbine operating mode recognition model is constructed.
[0091] (4) Use principal component analysis (PCA) to extract features from the historical operating parameter data of the steam turbine generator to obtain state feature values;
[0092] Principal Component Analysis (PCA) feature extraction method specifically includes the following steps:
[0093] (4.1) Process the normalized data matrix A to obtain the covariance matrix L;
[0094] (4.2) Calculate the eigenvalues and eigenvectors of the covariance matrix L;
[0095] (4.3) Calculate the cumulative contribution rate of the principal components and select the feature values with a cumulative contribution rate of 90% as the input for prediction. The cumulative contribution rate of PCA is shown in Table 1.
[0096] Number of principal components Eigenvalues Variance contribution rate (%) Cumulative variance contribution rate (%) 1 13.77851 52.99427 52.99427 2 6.376174 24.52375 77.51801 3 1.823306 7.012717 84.53073 4 1.087509 4.182727 88.71346 5 0.907131 3.488967 92.20243
[0097] As shown in Table 1, the cumulative contribution rate at the fifth principal element is 92.20243%, which is greater than 90%. At this point, each principal element represents the information of the original data.
[0098] (5) Divide the state characteristic parameters of the steam turbine generator after dimensionality reduction by principal component analysis (PCA) into test set, validation set and test set in a ratio of 8:1:1;
[0099] (6) The shaft vibration parameter sequence described in step 1 is decomposed by variational mode decomposition algorithm (VMD), and the final number of sub-signal modes K is determined by the center frequency method. The center frequencies corresponding to different K values are shown in Table 2. The above shaft vibration parameter sequence is decomposed into K IMF sequences by variational mode decomposition algorithm (VMD).
[0100] Modal number Mode 1 Mode 2 Mode 3 Mode 4 Mode 5 Modal 6 Modal 7 Modal 8 Modal 9 Modal 10 Modal 11 K＝2 0.091875806 2693.451 K＝3 0.072372458 1501.777 3848.509 K＝4 0.07365692 1560.612 2675.426 3901.661 K＝5 0.0681098 1259.632 2224.582 3175.546 4333.246 K＝6 0.0614107 933.6154 1846.835 2787.704 3670.359 4554.492 K＝7 0.0565453 683.2139 1439.261 2209.647 3025.252 3809.816 4621.507 K＝8 0.0553798 638.6014 1335.295 2013.86 2699.274 3364.463 4023.69 4688.179 K＝9 0.052985365 530.9096 1137.117 1729.055 2324.802 2921.462 3520.222 4121.166 4719.883 K＝10 0.050408206 421.9069 956.6788 1493.746 2048.007 2604.204 3154.724 3694.263 4236.298 4753.79 K＝11 0.049250024 380.6877 877.279 1372.95 1864.489 2348.645 2836.29 3319.283 3801.842 4301.011 4769.796
[0101] Table 2 shows the center frequencies of each modal component when K is set to different values. The data indicates that when K is 11, components with similar center frequencies appear, suggesting that VMD over-decomposes the signal. Therefore, the final value of K is set to 10, decomposing the vibration target parameters into 10 IMF sequences using VMD (Variational Mode Decomposition), as follows: Figure 3 As shown;
[0102] (7) Build and optimize the XGBoost model according to the Sparrow Search Algorithm (SSA). Set the optimal parameters for the sequences decomposed by VMD. For the K IMF sequences obtained by decomposition, build K XGBoost models respectively. Use the optimized XGBoost model to predict the vibration parameters of the steam turbine generator.
[0103] like Figure 4 As shown, the XGBoost model is trained using the divided training set, tested using the test set, and evaluated using multiple prediction evaluation metrics.
[0104] The hyperparameter tuning and optimization of the XGBoost prediction model was performed using the Sparrow Search Algorithm (SSA). The hyperparameter optimization process included:
[0105] (7.1) The optimal hyperparameter combination is obtained by using the Sparrow Search Algorithm (SSA) as the hyperparameter input of the XGBoost model. Initial values are set for the number of intelligent algorithm populations and the number of intelligent algorithm evolutions in the Sparrow Search Algorithm.
[0106] (7.2) Optimize the hyperparameters of the prediction model, i.e. set them as the search dimension of the SSA sparrow search algorithm and set the corresponding search range;
[0107] (7.3) Construct a decision tree model, gradually increase the number of trees by using the evaluation criteria of feature values, and use the new trees to fit the residuals of the previous prediction;
[0108] (7.4) Calculate the fitness of each sparrow to classify discoverers and followers, update their positions, and compare the current fitness of each sparrow with the current best fitness:
[0109] (7.5) If the fitness of the current individual is higher than the current best fitness, then update the current best fitness and use its corresponding position as the fitness and position of the current individual;
[0110] (7.6) If the fitness of an individual is lower than the current best fitness, then the current best fitness and the current best position remain unchanged;
[0111] (7.7) When the number of iterations reaches the global maximum number of iterations, end the iteration and record the current hyperparameter value;
[0112] (7.8) The recorded hyperparameter values are the global optimal parameters, and an XGBoost prediction model is built.
[0113] (8) Using the state characteristic parameters of the steam turbine generator after dimensionality reduction by principal component analysis (PCA) as input features, these parameters are input into the optimized K XGBoost prediction models. The predicted values of each IMF sequence are then superimposed and reconstructed to finally obtain the predicted values of the vibration parameters, such as... Figure 5 As shown.
[0114] The purpose of this invention is to predict the vibration parameter values at the next moment. The vibration parameters are the objects of prediction. The current thermal parameters, power parameters, and vibration parameters are known variables. After PCA feature extraction, they are used as input variables of the prediction model and input into the XGBoost model to obtain the vibration parameter values at the next moment.
[0115] Example 2
[0116] This invention discloses a turbine generator shaft vibration prediction system based on signal decomposition and reconstruction, comprising:
[0117] The data acquisition module is used to acquire historical data of turbine generator operating parameters. The historical data of turbine generator operating parameters acquired by the data acquisition module includes historical data of turbine generator power parameters, historical data of thermal parameters, and historical data of shaft vibration parameters. Among them, the turbine generator shaft thermal parameters include main steam valve steam pressure, high-pressure bypass valve downstream steam pressure, high-pressure cylinder blade stage pressure, regulating valve downstream steam pressure, make-up air valve downstream steam pressure, intermediate-pressure cylinder inlet steam pressure, intermediate-pressure cylinder exhaust steam pressure, intermediate-pressure exhaust steam temperature, high-pressure exhaust steam temperature, low-pressure cylinder vacuum, deaerator pressure, and main steam flow rate.
[0118] The preprocessing module is used to preprocess the collected historical operating parameter data of the steam turbine generator to eliminate the influence of differences in variable dimensions on the results. The specific steps of the preprocessing operation in the preprocessing module are as follows:
[0119] All historical data of operating parameters are normalized and mapped to the range [0, 1] as sample data. The normalization formula is shown below:
[0120]
[0121] Where X and A represent the data before and after normalization, respectively. max With X min These represent the maximum and minimum values of a sequence before normalization, respectively.
[0122] Set up a historical dataset of steam turbine generator operating parameters. It has n sets of data, and the feature dimension of each set of data is m;
[0123] Then the normalized data matrix of dataset X is
[0124] The status recognition module is used to identify the operating status of the steam turbine generator. It uses the K-Means data clustering algorithm to divide the operating data of the steam turbine generator into different operating modes based on two pattern recognition feature parameters: ambient temperature and steam turbine active power. It then constructs a support vector machine (SVM) steam turbine operating mode recognition model.
[0125] The eigenvalue extraction module is used to extract features from the historical operating parameter data of the steam turbine generator using Principal Component Analysis (PCA) to obtain state eigenvalues. The PCA feature extraction method within the eigenvalue extraction module is specifically executed as follows:
[0126] The normalized data matrix A is processed to obtain the covariance matrix L;
[0127] Calculate the eigenvalues and eigenvectors of the covariance matrix L;
[0128] Calculate the cumulative contribution rate of the principal components, and select the feature values with a cumulative contribution rate of 90% as the input for prediction;
[0129] The data partitioning module is used to divide the state characteristic parameters of the steam turbine generator after dimensionality reduction by principal component analysis (PCA) into test set, validation set and test set in a ratio of 8:1:1.
[0130] The decomposition processing module is used to decompose the acquired shaft vibration parameter sequence using the variational mode decomposition algorithm (VMD), determine the final number of sub-signal modes K using the center frequency method, and decompose the above shaft vibration parameter sequence into K IMF sequences using the variational mode decomposition algorithm (VMD).
[0131] The XGBoost model optimization module is used to build and optimize XGBoost models using the Sparrow Search Algorithm (SSA). It sets optimal parameters for each sequence decomposed by VMD, and builds K XGBoost models for each of the K IMF sequences obtained from the decomposition. The optimized XGBoost models are then used to predict the vibration parameters of the steam turbine generator. The module also performs hyperparameter tuning and optimization of the XGBoost prediction model using the SSA algorithm. The specific hyperparameter optimization process is as follows:
[0132] The optimal hyperparameter combination was obtained by using the Sparrow Search Algorithm (SSA) as the hyperparameter input for the XGBoost model. Initial values were set for the number of intelligent algorithm populations and the number of intelligent algorithm evolutions in the Sparrow Search Algorithm.
[0133] The hyperparameters of the prediction model are optimized, that is, set as the search dimension of the SSA sparrow search algorithm and set the corresponding search range.
[0134] Build a decision tree model, gradually increase the number of trees by using the evaluation criteria of feature values, and use the new trees to fit the residuals of the previous prediction;
[0135] Calculate the fitness of each sparrow to classify them as discoverers and followers, update their positions, and compare the current fitness of each sparrow with the current best fitness:
[0136] If the fitness of the current individual is higher than the current best fitness, then update the current best fitness and use its corresponding position as the fitness and position of the current individual.
[0137] If the fitness of an individual is lower than the current best fitness, then the current best fitness and the current best position remain unchanged.
[0138] When the number of iterations reaches the global maximum number of iterations, the iteration ends and the current hyperparameter value is recorded;
[0139] The recorded hyperparameter values are the globally optimal parameters, and an XGBoost prediction model is built accordingly.
[0140] The vibration parameter prediction module is used to take the state characteristic parameters of the steam turbine generator after dimensionality reduction by principal component analysis (PCA) as input features, input them into K optimized XGBoost prediction models respectively, and superimpose and reconstruct the predicted values of each IMF sequence to finally obtain the vibration parameter prediction values.
[0141] Example 3
[0142] Corresponding to the method of Embodiment 1 of the present invention, Embodiment 3 of the present invention also provides an electronic device.
[0143] In this embodiment 3, the electronic device includes: at least one communication bus, at least one processor, at least one memory, at least one network interface, and at least one peripheral interface. The memory contains programs and data.
[0144] A communication bus can be a communication device that transmits data between components within an electronic device, such as an internal bus (CPU and memory bus) or an external bus (Universal Serial Bus port, Peripheral Component Interconnect Fast Port, etc.).
[0145] The memory may include high-speed RAM, and may also include non-volatile memory, such as at least one disk storage device.
[0146] The processor calls the program and data stored in the memory to execute the turbine generator shaft vibration prediction method provided in Embodiment 1 of the present invention.
[0147] Peripheral interfaces are used to connect to peripherals, which are external devices. External devices may include, but are not limited to, keyboards, monitors, cursor control devices (such as mice, touchpads or touch screens), video input devices, etc.
[0148] A network interface provides wired or wireless communication with external networks (e.g., the Internet, intranets, local area networks, mobile communication networks, etc.).
[0149] Example 4
[0150] Corresponding to the method of Embodiment 1 of the present invention, Embodiment 4 of the present invention also provides a computer storage medium for data acquisition and reception. The computer storage medium stores a computer program, which is run by a processor to execute the turbine generator shaft vibration prediction method provided in Embodiment 1 of the present invention.
[0151] The functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware, as a software functional unit, or in a combination of software and hardware.
[0152] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as portable hard drives, USB flash drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0153] This invention provides a method, system, device, and storage medium for predicting shaft vibration of a steam turbine generator. The above description is merely a preferred embodiment of the invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of this invention, and these improvements and modifications should also be considered within the scope of protection of this invention. All components not explicitly stated in this embodiment can be implemented using existing technology.
Claims
1. A method for predicting vibration of a steam turbine generator shaft system, characterized in that, Includes the following steps: (1) Obtain historical data of turbine generator operating parameters; (2) The collected historical data of turbine generator operating parameters are preprocessed to eliminate the influence of the difference in the dimensions of variables on the results; (3) The operating status of the steam turbine generator is identified. The K-Means data clustering algorithm is used to divide the operating data of the steam turbine generator into different operating modes based on two pattern recognition feature parameters: ambient temperature and steam turbine active power. A support vector machine (SVM) steam turbine operating mode recognition model is constructed. (4) Use principal component analysis (PCA) to extract features from the historical operating parameter data of the steam turbine generator to obtain state feature values; (5) Divide the state characteristic parameters of the steam turbine generator after dimensionality reduction by principal component analysis (PCA) into test set, validation set and test set in a ratio of 8:1:1; (6) The shaft vibration parameter sequence described in step 1 is decomposed by variational mode decomposition algorithm (VMD), and the final number of sub-signal modes K is determined by the center frequency method. The shaft vibration parameter sequence is decomposed into K IMF sequences by variational mode decomposition algorithm (VMD). (7) Build and optimize the XGBoost model according to the Sparrow Search Algorithm (SSA). Set the optimal parameters for the sequences decomposed by VMD. For the K IMF sequences obtained by decomposition, build K XGBoost models respectively. Use the optimized XGBoost model to predict the vibration parameters of the steam turbine generator. (8) The state characteristic parameters of the steam turbine generator after dimensionality reduction by principal component analysis (PCA) are used as input features and input into the optimized K XGBoost prediction models respectively. The predicted values of each IMF sequence are superimposed and reconstructed to finally obtain the vibration parameter prediction values.
2. The method for predicting vibration of a steam turbine generator shaft system according to claim 1, characterized in that: The historical data of turbine generator operating parameters obtained in step (1) includes historical data of turbine generator power parameters, historical data of thermal parameters, and historical data of shaft vibration parameters. The thermal parameters of the turbine generator shaft system include the main steam valve steam pressure, the steam pressure after the high-pressure bypass valve, the pressure of the high-pressure cylinder blade stage, the steam pressure after the regulating valve, the steam pressure after the make-up air valve, the steam inlet pressure of the intermediate-pressure cylinder, the steam outlet pressure of the intermediate-pressure cylinder, the steam outlet temperature of the intermediate-pressure cylinder, the steam outlet temperature of the high-pressure cylinder, the vacuum of the low-pressure cylinder, the deaerator pressure, and the main steam flow rate.
3. The method for predicting vibration of a steam turbine generator shaft system according to claim 2, characterized in that: The specific steps of the preprocessing operation in step (2) are as follows: All historical data of operating parameters are normalized and mapped to the range [0, 1] as sample data. The normalization formula is shown below: Where X and A represent the data before and after normalization, respectively. max With X min These represent the maximum and minimum values of a sequence before normalization, respectively. Set up a historical dataset of steam turbine generator operating parameters. It has n sets of data, and the feature dimension of each set of data is m; Then the normalized data matrix of dataset X is 4. The method for predicting shaft vibration of a steam turbine generator according to claim 3, characterized in that: The principal component analysis (PCA) feature extraction method in step (4) specifically includes the following steps: (4.1) Process the normalized data matrix A to obtain the covariance matrix L; (4.2) Calculate the eigenvalues and eigenvectors of the covariance matrix L; (4.3) Calculate the cumulative contribution rate of the principal components and select the feature values with a cumulative contribution rate of 90% as the input for prediction.
5. The method for predicting vibration of a steam turbine generator shaft system according to claim 4, characterized in that: In step (7), the hyperparameters of the XGBoost prediction model are tuned and optimized using the Sparrow Search Algorithm (SSA). The hyperparameter optimization process includes: (7.1) The optimal hyperparameter combination is obtained by using the Sparrow Search Algorithm (SSA) as the hyperparameter input of the XGBoost model. Initial values are set for the number of intelligent algorithm populations and the number of intelligent algorithm evolutions in the Sparrow Search Algorithm. (7.2) Optimize the hyperparameters of the prediction model, i.e. set them as the search dimension of the SSA sparrow search algorithm and set the corresponding search range; (7.3) Construct a decision tree model, gradually increase the number of trees by using the evaluation criteria of feature values, and use the new trees to fit the residuals of the previous prediction; (7.4) Calculate the fitness of each sparrow to classify discoverers and followers, update their positions, and compare the current fitness of each sparrow with the current best fitness: (7.5) If the fitness of the current individual is higher than the current best fitness, then update the current best fitness and use its corresponding position as the fitness and position of the current individual; (7.6) If the fitness of an individual is lower than the current best fitness, then the current best fitness and the current best position remain unchanged; (7.7) When the number of iterations reaches the global maximum number of iterations, end the iteration and record the current hyperparameter value; (7.8) The recorded hyperparameter values are the global optimal parameters, and an XGBoost prediction model is built.
6. A turbine generator shaft vibration prediction system, characterized in that, include: The data acquisition module is used to acquire historical data of turbine generator operating parameters. The historical data of turbine generator operating parameters acquired by the data acquisition module includes historical data of turbine generator power parameters, historical data of thermal parameters, and historical data of shaft vibration parameters. Among them, the turbine generator shaft thermal parameters include main steam valve steam pressure, high-pressure bypass valve downstream steam pressure, high-pressure cylinder blade stage pressure, regulating valve downstream steam pressure, make-up air valve downstream steam pressure, intermediate-pressure cylinder inlet steam pressure, intermediate-pressure cylinder exhaust steam pressure, intermediate-pressure exhaust steam temperature, high-pressure exhaust steam temperature, low-pressure cylinder vacuum, deaerator pressure, and main steam flow rate. The preprocessing module is used to preprocess the collected historical data of turbine generator operating parameters to eliminate the influence of differences in variable dimensions on the results. The status recognition module is used to identify the operating status of the steam turbine generator. It uses the K-Means data clustering algorithm to divide the operating data of the steam turbine generator into different operating modes based on two pattern recognition feature parameters: ambient temperature and steam turbine active power. It then constructs a support vector machine (SVM) steam turbine operating mode recognition model. The eigenvalue extraction module is used to extract features from the historical operating parameter data of the steam turbine generator using Principal Component Analysis (PCA) to obtain state eigenvalues. The PCA feature extraction method within the eigenvalue extraction module is specifically executed as follows: The normalized data matrix A is processed to obtain the covariance matrix L; Calculate the eigenvalues and eigenvectors of the covariance matrix L; Calculate the cumulative contribution rate of the principal components, and select the feature values with a cumulative contribution rate of 90% as the input for prediction; The data partitioning module is used to divide the state characteristic parameters of the steam turbine generator after dimensionality reduction by principal component analysis (PCA) into test set, validation set and test set in a ratio of 8:1:
1. The decomposition processing module is used to decompose the acquired shaft vibration parameter sequence using the variational mode decomposition algorithm (VMD), determine the final number of sub-signal modes K using the center frequency method, and decompose the above shaft vibration parameter sequence into K IMF sequences using the variational mode decomposition algorithm (VMD). The XGBoost model optimization module is used to build and optimize the XGBoost model according to the Sparrow Search Algorithm (SSA). It sets the optimal parameters for the sequences decomposed by VMD, builds K XGBoost models for the K IMF sequences obtained by decomposition, and uses the optimized XGBoost model to predict the vibration parameters of the steam turbine generator. The vibration parameter prediction module is used to take the state characteristic parameters of the steam turbine generator after dimensionality reduction by principal component analysis (PCA) as input features, input them into K optimized XGBoost prediction models respectively, and superimpose and reconstruct the predicted values of each IMF sequence to finally obtain the vibration parameter prediction values.
7. The turbine generator shaft vibration prediction system according to claim 6, characterized in that, The specific steps of the preprocessing operation in the preprocessing module are as follows: All historical data of operating parameters are normalized and mapped to the range [0, 1] as sample data. The normalization formula is shown below: Where X and A represent the data before and after normalization, respectively. max With X min These represent the maximum and minimum values of a sequence before normalization, respectively. Set up a historical dataset of steam turbine generator operating parameters. It has n sets of data, and the feature dimension of each set of data is m; Then the normalized data matrix of dataset X is 8. The turbine generator shaft vibration prediction system according to claim 7, characterized in that, The XGBoost model optimization module performs hyperparameter tuning and optimization of the XGBoost prediction model using the Sparrow Search Algorithm (SSA). The specific execution process for hyperparameter optimization is as follows: The optimal hyperparameter combination was obtained by using the Sparrow Search Algorithm (SSA) as the hyperparameter input for the XGBoost model. Initial values were set for the number of intelligent algorithm populations and the number of intelligent algorithm evolutions in the Sparrow Search Algorithm. The hyperparameters of the prediction model are optimized, that is, set as the search dimension of the SSA sparrow search algorithm and set the corresponding search range. Build a decision tree model, gradually increase the number of trees by using the evaluation criteria of feature values, and use the new trees to fit the residuals of the previous prediction; Calculate the fitness of each sparrow to classify them as discoverers and followers, update their positions, and compare the current fitness of each sparrow with the current best fitness: If the fitness of the current individual is higher than the current best fitness, then update the current best fitness and use its corresponding position as the fitness and position of the current individual. If the fitness of an individual is lower than the current best fitness, then the current best fitness and the current best position remain unchanged. When the number of iterations reaches the global maximum number of iterations, the iteration ends and the current hyperparameter value is recorded; The recorded hyperparameter values are the globally optimal parameters, and an XGBoost prediction model is built.
9. An electronic device, characterized in that, Including processor and memory, A memory for storing a computer program that is executed by a processor to perform the method described in any one of claims 1-5.
10. A computer storage medium, characterized in that, The computer storage medium stores a computer program that is executed by a processor to perform the method described in any one of claims 1-5.