Water turbine vibration control method and device based on shaft seal air supply amount detection

By obtaining the turbine shaft seal gas volume data in real time and using machine learning algorithms to establish a mapping relationship model, the problem of direct measurement difficulties and low prediction accuracy in turbine vibration monitoring and control is solved, and high-precision vibration control and unit stability are achieved. .

CN118934529APending Publication Date: 2024-11-12THREE GORGES JINSHAJIANG CHUANYUN HYDROPOWER DEV CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411218642.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-02
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

The existing turbine vibration monitoring and control technologies have problems such as direct measurement difficulties, low indirect prediction accuracy, and lack of active control and closed-loop adjustment mechanisms.

Method used

By obtaining the turbine shaft seal gas volume data in real time, pre-processing the data, and using machine learning algorithms to establish a mapping relationship model between the shaft seal gas volume and vibration, designing a feedback controller to achieve closed-loop control.

Benefits of technology

It improves the safety and reliability of vibration measurement, realizes high-precision vibration prediction and control, and significantly improves the stability and safety of the unit.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118934529A_ABST
    Figure CN118934529A_ABST
Patent Text Reader

Abstract

The invention provides a water turbine vibration control method and device based on shaft seal air supply amount detection, and relates to the technical field of water turbine control. The method comprises the steps that shaft seal air supply amount data at a water turbine shaft seal is obtained in real time; obtaining vibration data of the water turbine synchronized with the shaft seal air supply amount data, and performing data preprocessing on the shaft seal air supply amount data and the vibration data; a machine learning algorithm is utilized to establish and train a mapping relation model between the shaft seal air supply amount and the water turbine vibration based on the preprocessed shaft seal air supply amount data and vibration data, and a GBRT integrated model is obtained; verifying the GBRT integrated model, and optimizing and improving model parameters according to a verification result; an adjustable water turbine operation parameter is determined as a control quantity, and a feedback controller based on the optimized and improved GBRT integrated model is designed; and controlling the water turbine to vibrate based on the feedback controller. According to the invention, accurate detection and effective control of the vibration state of the water turbine are realized.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the technical field of water turbine control, and in particular to a water turbine vibration control method and device based on shaft seal air supply volume detection. Background Art

[0002] The turbine is the core equipment of the power plant, and its vibration state is crucial to the safe and stable operation of the unit. The traditional vibration control monitoring method is to place vibration sensors on the turbine rotor to directly measure the vibration of the rotor. However, this approach has certain risks. The vibration sensor is difficult to install and maintain, and is easily affected by the high-speed rotation of the rotor, resulting in poor measurement accuracy and reliability.

[0003] Another vibration monitoring method is to use hydraulic measurement parameters to indirectly reflect the vibration state, such as inlet and outlet pressure, flow, etc. This method avoids the defects of direct measurement, but due to the complex mapping relationship between hydraulic parameters and vibration, it is difficult to establish a high-precision mapping model, and the prediction accuracy is generally low. At the same time, the existing vibration control methods mostly use active control or vibration elimination technology, lacking effective means of actively suppressing vibration such as closed-loop feedback. Summary of the invention

[0004] The embodiments of the present application provide a turbine vibration control method and device based on shaft seal air supply volume detection to solve the problems of existing vibration monitoring and control technologies such as difficulty in direct measurement, low indirect prediction accuracy, lack of active control and closed-loop adjustment mechanism.

[0005] Other features and advantages of the present application will become apparent from the following detailed description, or may be learned in part by the practice of the present application.

[0006] According to a first aspect of an embodiment of the present application, a method for controlling vibration of a hydraulic turbine based on shaft seal air supply amount detection is provided, comprising:

[0007] Obtain the shaft seal air supply volume data at the turbine shaft seal in real time;

[0008] Acquire vibration data of the hydraulic turbine synchronized with the shaft seal air supply volume data, and perform data preprocessing on the shaft seal air supply volume data and the vibration data;

[0009] Using machine learning algorithms, based on the preprocessed shaft seal air supply volume data and vibration data, a mapping relationship model between the shaft seal air supply volume and turbine vibration is established and trained to obtain a GBRT integrated model.

[0010] Verify the GBRT integrated model and optimize and improve the model parameters according to the verification results;

[0011] Determine the adjustable turbine operating parameters as the control variables and design a feedback controller based on the optimized and improved GBRT integrated model;

[0012] The turbine vibration is controlled based on the feedback controller.

[0013] In some embodiments of the present application, based on the above scheme, the data preprocessing of the shaft seal air supply volume data and the vibration data includes:

[0014] Clean the shaft seal air supply volume data and vibration raw data;

[0015] Wavelet transform or EMD decomposition is used to denoise the cleaned shaft seal air supply volume data and vibration data;

[0016] Perform data standardization on the denoised shaft seal air supply volume data and vibration data;

[0017] According to actual needs, feature engineering processing is performed on the standardized shaft seal air supply volume data and vibration data.

[0018] In some embodiments of the present application, based on the aforementioned scheme, the mapping relationship model between the shaft seal air supply volume and the turbine vibration is established and trained based on the preprocessed shaft seal air supply volume data and vibration data using a machine learning algorithm to obtain a GBRT integrated model, including:

[0019] According to the characteristics of the problem, the gradient boosting regression tree GBRT is selected as the machine learning algorithm for training the mapping relationship model;

[0020] A mapping relationship model is constructed, and the pre-processed shaft seal air supply volume data is used as the model input feature vector, and the vibration data is used as the model output target value to form a training data set;

[0021] Randomly extract a part of the data from the training data set as the test set, and the remaining part as the first training validation set;

[0022] Perform grid search or random search on the hyperparameters of the mapping relationship model to find the optimal parameter combination;

[0023] Use mean square error (MSE) as the loss function, select stochastic gradient descent or Adam optimization algorithm, and iteratively train the mapping relationship model;

[0024] Set the model evaluation period parameter eval_freq, evaluate the current model performance on the first training validation set every eval_freq iterations, and calculate the validation root mean square error RMSE and decision coefficient R 2 , when the error no longer decreases, stop training;

[0025] Calculate the root mean square error RMSE and decision coefficient R of the final mapping relationship model on the test set 2 , mean absolute error MAE and median absolute deviation MedAE, and use these parameters as indicators to evaluate its generalization ability;

[0026] Use Bagging or Boosting integration strategies to combine the output results of multiple mapping relationship models to build a GBRT integration model.

[0027] In some embodiments of the present application, based on the above solution, the grid search or random search is performed on the hyperparameters of the mapping relationship model to find the optimal parameter combination, including:

[0028] Determine the GBRT model hyperparameters that need to be optimized and their candidate value ranges;

[0029] Separate a portion of the training data set as the second training validation set;

[0030] Construct a grid search or random search estimator, and set the parameter grid for the grid search or random search, depending on the selected search strategy;

[0031] Perform grid search or random search on the second training validation set;

[0032] Get the best performing parameter instance during the search process, and re-instantiate the GBRT model based on this parameter instance as the final model instance for subsequent training and evaluation.

[0033] In some embodiments of the present application, based on the above scheme, the calculation formula of the root mean square error RMSE of the final mapping relationship model is as follows:

[0034]

[0035] In formula (1), n ​​represents the number of test set samples, y i represents the true vibration value of the i-th sample, f(x i ) represents the model's input x for the i-th shaft seal air supply volume i The predicted vibration value;

[0036] The decision coefficient R of the final mapping relationship model 2 The calculation formula is as follows:

[0037]

[0038] In formula (2), y mean Represents the mean of the true vibration value y, y i represents the true vibration value of the i-th sample, f(x i ) represents the model's input x for the i-th shaft seal air supply volumei The predicted vibration value;

[0039] The calculation formula of the mean absolute error MAE of the final mapping relationship model is as follows:

[0040]

[0041] In formula (3), n represents the number of test set samples, y i represents the true vibration value of the i-th sample, f(x i ) represents the model's input x for the i-th shaft seal air supply volume i The predicted vibration value;

[0042] The calculation formula of the median absolute deviation MedAE of the final mapping relationship model is as follows:

[0043] MedAE=median(y i -f(x i )); (4)

[0044] In formula (4), y i represents the true vibration value of the i-th sample, f(x i ) represents the model's input x for the i-th shaft seal air supply volume i The predicted vibration value of , median represents the median.

[0045] In some embodiments of the present application, based on the above-mentioned solution, the GBRT integrated model is verified, and the model parameters are optimized and improved according to the verification results, including:

[0046] Prepare a new test dataset;

[0047] Using the new test data set, calculate the root mean square error RMSE and determination coefficient R of the GBRT integrated model on the new test data set. 2 , mean absolute error MAE, median absolute deviation MedAE, and use these indicators to judge whether the prediction performance of the model on new data meets the standards;

[0048] If the target is not met, a root cause analysis is conducted to analyze the potential causes of the performance degradation;

[0049] According to the analysis results, the GBRT integrated model is optimized;

[0050] Re-evaluate the performance of the optimized GBRT integrated model on a new test dataset, and repeat the verification and improvement process until the performance of the GBRT integrated model meets the requirements.

[0051] In some embodiments of the present application, based on the above scheme, the step of determining the adjustable turbine operating parameters as the control quantity and designing a feedback controller based on the optimized and improved GBRT integrated model includes:

[0052] Analyze the operating mechanism of the turbine and determine the set of adjustable operating parameters;

[0053] Conduct sensitivity analysis to evaluate the sensitivity of each operating parameter to the vibration value and select a subset of parameters that have a greater impact on vibration as potential control operations;

[0054] The control framework of the feedback controller is constructed based on the operating parameters, the optimized and improved GBRT integrated model and the control operation quantity;

[0055] Designing a control algorithm of a feedback controller and tuning parameters of the control algorithm;

[0056] Modeling and simulation analysis are performed on the feedback controller to evaluate whether the control performance meets the requirements under different working conditions. If the performance does not meet the requirements, the control parameters are repeatedly optimized, the control algorithm is improved, and the control framework design is adjusted until the control performance meets the requirements.

[0057] In some embodiments of the present application, based on the above solution, it also includes:

[0058] An adaptive mechanism is set in the feedback controller to correct the output of the GBRT integrated model in real time.

[0059] In some embodiments of the present application, based on the above solution, the adaptive mechanism is set in the feedback controller to correct the output of the GBRT integrated model in real time, including:

[0060] Pre-train multiple GBRT integrated models, each GBRT integrated model corresponds to a typical working condition of a turbine;

[0061] The shaft seal air supply volume data is acquired in real time and input into the currently selected GBRT integrated model after data preprocessing to obtain the vibration value predicted by the model;

[0062] The actual measured vibration value is obtained synchronously, and the prediction error is calculated based on the predicted vibration value and the actual measured vibration value;

[0063] Use adaptive filtering algorithm to model and track prediction errors and obtain compensation terms;

[0064] Add the vibration value predicted by the model to the compensation term to obtain the final vibration prediction value for control decision-making;

[0065] Monitor the operating condition changes of the turbine in real time. When the operating condition changes significantly, switch to the corresponding GBRT integrated model for prediction and adaptation.

[0066] According to a second aspect of an embodiment of the present application, a hydraulic turbine vibration control device based on shaft seal air supply amount detection is provided, comprising:

[0067] The first acquisition unit is used to acquire the shaft seal air supply volume data at the turbine shaft seal in real time;

[0068] A second acquisition unit is used to acquire vibration data of the hydraulic turbine synchronized with the shaft seal air supply volume data;

[0069] A preprocessing unit, used for preprocessing the shaft seal air supply volume data and vibration data;

[0070] A model building unit is used to establish and train a mapping relationship model between the shaft seal air supply volume and the turbine vibration based on the preprocessed shaft seal air supply volume data and vibration data by using a machine learning algorithm to obtain a GBRT integrated model;

[0071] A verification and optimization unit, used to verify the GBRT integrated model and optimize and improve the model parameters according to the verification results;

[0072] A design unit is used to determine adjustable turbine operating parameters as control quantities and design a feedback controller based on the optimized and improved GBRT integrated model;

[0073] A control unit is used for controlling the vibration of the turbine based on the feedback controller.

[0074] According to a third aspect of an embodiment of the present application, a computer-readable storage medium is provided, wherein the storage medium stores computer instructions, and when the computer instructions are executed on a computer, the computer executes the method according to the first aspect.

[0075] According to a fourth aspect of the embodiments of the present application, an electronic device is provided, the electronic device comprising: a memory and a processor;

[0076] The memory is used to store computer instructions;

[0077] The processor is used to call the computer instructions stored in the memory so that the electronic device executes the method as described in the first aspect.

[0078] The technical solution of the present application utilizes the shaft seal air supply as a new monitoring parameter reflecting the vibration state of the turbine, avoiding the direct placement of vibration sensors on the high-speed rotor, thereby improving the safety and reliability of the measurement. By establishing a high-precision mapping model between the shaft seal air supply and the vibration value, the real-time vibration level of the rotor can be accurately predicted, providing strong support for vibration control. An innovative model framework based on the gradient boosting regression tree (GBRT) integrated algorithm is proposed, which can efficiently handle the complex nonlinear mapping relationship between the air supply and vibration data. The model framework significantly improves the generalization ability and prediction accuracy of the mapping model through strategies such as automatic feature extraction and parameter optimization. A feedback controller is designed to compare the predicted vibration state with the target value, and automatically adjust the operating parameters of the turbine to form a closed-loop control loop. This innovative design enables the system to actively suppress vibration problems, thereby significantly improving the stability and safety of the unit operation.

[0079] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] The drawings herein are incorporated into the specification and constitute a part of the specification, showing embodiments consistent with the present application, and together with the specification, are used to explain the principles of the present application. Obviously, the drawings described below are only some embodiments of the present application, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0081] Figure 1 A schematic flow chart of a hydraulic turbine vibration control method based on shaft seal air supply amount detection according to an embodiment of the present application is shown;

[0082] Figure 2 A block diagram of a hydraulic turbine vibration control device based on shaft seal air supply amount detection according to an embodiment of the present application is shown;

[0083] Figure 3 A block diagram of an electronic device according to an embodiment of the present application is shown;

[0084] Figure 4 A schematic diagram of the structure of a computer system suitable for implementing an electronic device of an embodiment of the present application is shown. DETAILED DESCRIPTION

[0085] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this application will be more comprehensive and complete and fully convey the concept of the example embodiments to those skilled in the art.

[0086] In addition, described feature, structure or characteristic can be combined in one or more embodiments in any suitable manner. In the following description, many specific details are provided to provide a full understanding of the embodiments of the present application. However, those skilled in the art will appreciate that the technical scheme of the present application can be put into practice without one or more of the specific details, or other methods, components, devices, steps, etc. can be adopted. In other cases, known methods, devices, realizations or operations are not shown or described in detail to avoid blurring the various aspects of the application.

[0087] The block diagrams shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically independent entities. That is, these functional entities may be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices.

[0088] The flowcharts shown in the accompanying drawings are only exemplary and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps can be decomposed, and some operations / steps can be combined or partially combined, so the actual execution order may change according to actual conditions.

[0089] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the objects used in this way can be interchanged where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those shown or described.

[0090] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is only a part of the embodiment of the present invention, not all the embodiments. Based on the embodiment of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0091] Some embodiments of the present application will be described in detail below in conjunction with the accompanying drawings. In the absence of conflict, the following embodiments and features in the embodiments may be combined with each other.

[0092] See also Figure 1 , shows a flow chart of a turbine vibration control method based on shaft seal air supply volume detection according to an embodiment of the present application.

[0093] like Figure 1 As shown, a turbine vibration control method based on shaft seal air supply amount detection is presented, which specifically includes steps S100 to S600.

[0094] refer to Figure 1 , step S100, obtaining the shaft seal air supply volume data at the turbine shaft seal in real time.

[0095] It should be noted that, in this embodiment, a shaft seal pressure / flow sensor is used to measure the shaft seal air supply volume data at the turbine shaft seal in real time as a parameter indication for monitoring the vibration state of the turbine.

[0096] Continue to refer Figure 1 , step S200, obtaining vibration data of the turbine synchronized with the shaft seal air supply volume data, and performing data preprocessing on the shaft seal air supply volume data and the vibration data.

[0097] It should be noted that, in this embodiment, a shaft seal pressure / flow sensor is used to synchronously collect the original vibration data measured by the vibration sensor.

[0098] In some feasible embodiments, based on the above solution, the data preprocessing of the shaft seal air supply volume data and the vibration data includes:

[0099] Data cleaning is performed on the shaft seal air supply volume data and vibration raw data; data cleaning includes removing abnormal values, interpolating missing values, and segmented splicing processing;

[0100] Wavelet transform or EMD decomposition is used to denoise the cleaned shaft seal air supply volume data and vibration data;

[0101] Perform data standardization on the denoised shaft seal air supply volume data and vibration data;

[0102] According to actual needs, feature engineering processing is performed on the standardized shaft seal air supply volume data and vibration data.

[0103] It should be noted that the data after feature engineering processing will be used as input data for mapping relationship model training.

[0104] Specifically, in this embodiment, according to actual needs, the process of performing feature engineering processing on the standardized shaft seal air supply volume data and vibration data is as follows:

[0105] If secondary optimization of the feature vector is required, a kernel space is constructed to map the original data to a high-dimensional feature space, and the kernel technique is used to extract nonlinear features. If the shaft seal air supply volume data and vibration data have nonlinear or non-stationary characteristics, deep learning automatic feature extraction technology is used to learn effective feature representation. If the shaft seal air supply volume data and vibration data have multimodal or heterogeneous characteristics, a multi-view feature fusion framework is constructed to jointly optimize heterogeneous features. Feature engineering processing includes at least time domain statistical feature extraction, frequency domain spectral feature extraction, wavelet or EMD time-frequency feature extraction, principal component analysis or other dimensionality reduction techniques.

[0106] Continue to refer Figure 1 , step S300, using a machine learning algorithm, based on the preprocessed shaft seal air supply volume data and vibration data, establish and train a mapping relationship model between the shaft seal air supply volume and the turbine vibration to obtain a GBRT integrated model.

[0107] In some feasible embodiments, based on the above solution, step S300 specifically includes:

[0108] Step S301, according to the problem characteristics, select the gradient boosted regression tree GBRT as the machine learning algorithm for training the mapping relationship model.

[0109] It should be noted that GBRT is a powerful integrated algorithm that can efficiently handle nonlinear problems and has good prediction performance and robustness.

[0110] Step S302, constructing a mapping relationship model, taking the pre-processed shaft seal air supply volume data as the model input feature vector, and the vibration data as the model output target value, to form a training data set.

[0111] Step S303: randomly extract test_ratio data from the training data set as a test set, and the remaining 1-test_ratio part is used as a first training verification set.

[0112] Among them, test_ratio is a decimal between 0 and 1, indicating the proportion of the test set. Usually the value of test_ratio is in the range of 0.2 to 0.3, for example, it can be set to 0.25. Correspondingly, the ratio of the first training validation set is (1-test_ratio) = 0.75. In practical applications, the value of test_ratio can be appropriately adjusted according to factors such as the amount of data and the complexity of the model to obtain better model performance and generalization ability.

[0113] Step S304, performing grid search or random search on the hyper parameters of the mapping relationship model including learning rate, maximum tree depth, and number of trees to find the optimal parameter combination.

[0114] In some feasible embodiments, step S304 is specifically as follows:

[0115] Determine the hyperparameters of the GBRT model that need to be optimized and their candidate value ranges, where the hyperparameters include learning rate, maximum tree depth, and the number of trees. The candidate value range of the learning rate is 0.01 to 0.3, the candidate value range of the maximum tree depth is 3 to 10, and the candidate value range of the number of trees is 100 to 1000;

[0116] A portion of the training data set is again separated as the second training validation set to evaluate the performance of different parameter combinations during grid search;

[0117] According to the selected search strategy, construct a grid search or random search estimator and set the parameter grid of the grid search or random search, that is, specify a series of preset candidate values ​​for all hyperparameters to be optimized;

[0118] Implement grid search or random search on the second training validation set. For each parameter combination, evaluate the average performance on the second validation set by K-fold cross validation to find the best parameter combination.

[0119] Get the best performing parameter instance during the search process, and re-instantiate the GBRT model based on this parameter instance as the final model instance for subsequent training and evaluation.

[0120] Step S305, using mean square error MSE as the loss function, selecting stochastic gradient descent or Adam optimization algorithm, and iteratively training the mapping relationship model.

[0121] Specifically, the specific formula of the loss function is as follows:

[0122]

[0123] Among them, n represents the number of test set samples, y i represents the true vibration value of the i-th sample, f(x i ) represents the model's input x for the i-th shaft seal air supply volume i The predicted vibration value.

[0124] It is understandable that the mean square error (MSE) is used as the loss function of the GBRT model for the following reasons: this case is a typical regression problem, which requires learning the mapping relationship between the shaft seal air supply volume data and the vibration value; the vibration value is a continuous numerical variable, and the MSE can well measure the difference between the predicted value and the true value, which is convenient for model optimization, and for continuous value prediction problems, the MSE is usually better than the absolute value loss function (such as MAE); the optimization goal of the GBRT algorithm itself is to minimize the sum of squares of the residuals between the predicted value and the true value, so using the MSE as the loss function can naturally coincide with the optimization process of the GBRT.

[0125] Step S306, set the periodic parameter eval_freq of the model evaluation, evaluate the current model performance on the first training validation set every eval_freq iterations, and calculate the validation root mean square error RMSE and decision coefficient R 2 , when the error no longer decreases, stop training.

[0126] It should be noted that the periodic parameter eval_freq of the model evaluation, that is, the number of training iterations after which the mapping relationship model is evaluated. Usually, eval_freq can be set to a smaller integer, such as 10, 20, or 50, so that the model can be evaluated more frequently and the changing trend of the model performance can be captured in time. At the same time, an early stopping mechanism can also be set. If the model error no longer decreases in several consecutive evaluations (for example, 10 times), the training is terminated to prevent overfitting. In this way, through the eval_freq parameter and the early stopping mechanism, the model training process can be controlled more flexibly, and the training can be terminated when the model performance tends to be stable, thereby improving the training efficiency.

[0127] Step S307: Calculate the root mean square error RMSE and decision coefficient R of the final mapping relationship model on the reserved test set. 2 , mean absolute error MAE and median absolute deviation MedAE, which are used as indicators to evaluate its generalization ability.

[0128] Specifically, the calculation formula of the root mean square error RMSE of the final mapping relationship model is as follows:

[0129]

[0130] In formula (2), n represents the number of test set samples, y i represents the true vibration value of the i-th sample, f(x i ) represents the model's input x for the i-th shaft seal air supply volume i The predicted vibration value;

[0131] The decision coefficient R of the final mapping relationship model 2The calculation formula is as follows:

[0132]

[0133] In formula (3), y mean Represents the mean of the true vibration value y, y i represents the true vibration value of the i-th sample, f(x i ) represents the model's input x for the i-th shaft seal air supply volume i The predicted vibration value;

[0134] The calculation formula of the mean absolute error MAE of the final mapping relationship model is as follows:

[0135]

[0136] In formula (4), n represents the number of test set samples, y i represents the true vibration value of the i-th sample, f(x i ) represents the model's input x for the i-th shaft seal air supply volume i The predicted vibration value;

[0137] The calculation formula of the median absolute deviation MedAE of the final mapping relationship model is as follows:

[0138] MedAE=median(y i -f(x i )); (5)

[0139] In formula (5), y i represents the true vibration value of the i-th sample, f(x i ) represents the model's input x for the i-th shaft seal air supply volume i The predicted vibration value of , median represents the median.

[0140] Step S308, using Bagging or Boosting integration strategy, combining the output results of multiple mapping relationship models, to build a GBRT integration model.

[0141] Step S309, save the trained GBRT integrated model parameters and weights as a model file, and save the model's evaluation indicators, training process and other metadata for subsequent loading, analysis and optimization.

[0142] Continue to refer Figure 1 , step S400, verifying the GBRT integrated model and optimizing and improving the model parameters according to the verification results.

[0143] In some feasible embodiments, based on the above solution, step S400 specifically includes:

[0144] Step S401, prepare a new test data set, which contains shaft seal air supply volume data and corresponding measured vibration data, and the data distribution is somewhat different from the training data, so as to evaluate the generalization ability of the model.

[0145] Step S402: Using the new test data set, calculate the root mean square error RMSE and determination coefficient R of the GBRT integrated model on this data set. 2 , mean absolute error MAE, median absolute deviation MedAE, and use these indicators to judge whether the prediction performance of the model on new data meets the standards.

[0146] Step S403: If the prediction performance of the model on the new data does not meet the standard, a root cause analysis is performed to analyze the potential reasons for the performance degradation, such as changes in data distribution, data quality issues, or deviations in the model itself.

[0147] Step S404, based on the analysis results, optimize and improve the GBRT integrated model, including adjusting the data preprocessing process (improving feature engineering), adjusting GBRT model hyperparameters (such as learning rate, number of trees, etc.), expanding training data (to cover more data distribution), and trying to replace the core algorithm.

[0148] Step S405, re-evaluate the performance of the optimized model on a new test data set, repeat the model verification and improvement process until the model performance meets the requirements; save the optimized GBRT integrated model parameters and replace the original model parameters.

[0149] Continue to refer Figure 1 , step S500, determining the adjustable turbine operating parameters as the control quantity, and designing a feedback controller based on the optimized and improved GBRT integrated model.

[0150] In some feasible embodiments, based on the above solution, step S500 specifically includes:

[0151] Step S501: Analyze the operating mechanism of the turbine and determine the adjustable operating parameter set P = {p1, p2, ..., pn}; perform sensitivity analysis to evaluate the sensitivity of each operating parameter pi to the vibration value and select the parameter subset P that has a greater impact on the vibration. * As a potential control operation quantity.

[0152] Step S502, constructing a control framework of a feedback controller based on operating parameters, the optimized and improved GBRT integrated model and control operation quantities, wherein the control framework of the feedback controller includes three parts: a state observer, a controller and an actuator; wherein the state observer uses the GBRT integrated model and real-time shaft seal air supply volume data to predict the vibration state, the controller calculates the increment of the control quantity according to the vibration state and the target value, and the actuator adjusts the operating parameters to achieve control.

[0153] Step S503, designing a control algorithm of a feedback controller, which may adopt algorithms such as classic PID, feedforward-feedback, extreme value auto-disturbance rejection control, etc., and the control goal is to make the actual vibration value approach the ideal vibration value.

[0154] Step S504, controller parameter tuning is performed on the parameters of the selected control algorithm, such as PID parameters, to ensure the response speed, steady-state accuracy and anti-disturbance capability of the controller.

[0155] Step S505, modeling and simulation analysis are performed on the feedback controller to evaluate whether the control performance meets the requirements under different working conditions. If the performance does not meet the requirements, the control parameters are optimized, the control algorithm is improved, and the control framework design is adjusted.

[0156] Step S506, repeat the above steps until the control performance meets the requirements, and save the final feedback controller design scheme; encode the final feedback controller design scheme and implement it on the actual control hardware.

[0157] It should be noted that the adjustable operating parameter set, the adjustment of these parameters can affect the vibration state of the turbine, common adjustable parameters include pilot valve opening, rotation speed, water inlet guide ring opening, etc.

[0158] Continue to refer Figure 1 , step S600, controlling the vibration of the turbine based on the feedback controller.

[0159] It should be noted that, in this embodiment, the vibration control of the turbine is achieved by applying the feedback controller to the turbine site.

[0160] In some feasible embodiments, based on the above solution, the method further includes:

[0161] An adaptive mechanism is set in the feedback controller to correct the output of the GBRT integrated model in real time.

[0162] It can be understood that this step can improve control accuracy and adapt to changes in operating conditions.

[0163] In some feasible embodiments, based on the above solution, the adaptive mechanism is set in the feedback controller to correct the output of the GBRT integrated model in real time, including:

[0164] Pre-train multiple GBRT integrated models {M1, M2, ..., Mk}, each model Mi corresponds to a typical working condition of the turbine;

[0165] The shaft seal air supply volume data is obtained in real time, and after data preprocessing, it is input into the currently selected GBRT integrated model Mi to obtain the vibration value y_pred predicted by the model; the actual measured vibration value y_real is obtained simultaneously, and the prediction error e=y_real-y_pred is calculated;

[0166] Using an adaptive filtering algorithm, such as recursive least squares RLS or least mean squares algorithm LMS, the prediction error e is modeled and tracked to obtain the compensation term c;

[0167] Add the vibration value y_pred predicted by the model to the compensation term c to obtain the final vibration prediction value y_final = y_pred + c, which is used for control decision-making;

[0168] Monitor the operating condition changes of the turbine in real time. When the operating condition changes significantly, switch to the corresponding GBRT integrated model Mi for prediction and adaptation.

[0169] In summary, the present invention uses the turbine shaft seal air supply volume data as a new parameter indication for monitoring the vibration state, which indirectly reflects the vibration condition and solves the problem that it is difficult to directly and accurately measure the vibration in the machine operation environment.

[0170] The present invention proposes a comprehensive data preprocessing process, including data cleaning, denoising, standardization and feature engineering steps, to provide high-quality input data for modeling.

[0171] The present invention adopts a gradient boosted regression tree (GBRT) algorithm and trains a mapping relationship model based on shaft seal air supply volume data and vibration data, which can efficiently handle nonlinear problems.

[0172] The present invention proposes a search strategy for hyperparameter optimization of the GBRT model, including grid search and random search, to find the best parameter combination.

[0173] The present invention designs a model evaluation mechanism including indicators such as root mean square error and decision coefficient to achieve model performance monitoring and early stopping.

[0174] The present invention adopts an integrated strategy to construct a GBRT integrated model to improve the model robustness and prediction accuracy.

[0175] The present invention proposes a feedback process for model verification and optimization improvement, which optimizes the model in a targeted manner and improves the generalization ability on new data.

[0176] The present invention designs a closed-loop vibration control framework based on the GBRT integrated model to achieve automatic adjustment and control of adjustable parameters.

[0177] The present invention introduces an adaptive mechanism into the control framework to correct the model prediction value in real time, improve the control accuracy and adapt to changes in working conditions.

[0178] In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments. Below, a specific application example of the present invention is provided.

[0179] Specifically, the experiment uses a 105MW hydro-turbine generator set in a hydropower station as the controlled object. First, the shaft seal pressure sensor and vibration sensor are installed to measure the shaft seal air supply volume and shaft vibration data respectively. Preliminary data analysis found that there is a significant correlation between the shaft seal air supply volume and shaft vibration.

[0180] The first step is to pre-process the raw data. The shaft seal air supply volume data is processed by data cleaning, wavelet transform denoising, and standardization. The vibration data is also processed similarly, and then feature vectors such as statistical features, frequency domain features, and wavelet coefficients are extracted.

[0181] The second step is to train the GBRT integrated model, using the processed gas volume data as input features and the vibration data as the target value. 25% of the data is selected as the test set, and the remaining 75% of the data is used as the training set. Grid search is used to tune the hyperparameters of GBRT, with the best learning rate of 0.08, the maximum tree depth of 8, and the number of trees of 250. MSE is selected as the training loss function, and Adam is the optimization algorithm. The root mean square error RMSE and decision coefficient R are used. 2 The evaluation model has an RMSE of 0.212 on the training set and R 2 The RMSE on the test set is 0.287, and the R 2 It is 0.912, and the generalization performance is good.

[0182] The third step is to build a feedback controller based on the GBRT model. Determine the pilot valve opening and rotation speed as the control operation quantity, the PID algorithm controls the pilot valve opening, and the feedforward controls the rotation speed. The PID parameter is set to K p =1.5, K i =0.03, K d =2.0, test response time 2 seconds, maximum overshoot 3%.

[0183] The fourth step is to introduce an adaptive mechanism. Four GBRT models were trained, corresponding to different working conditions. The RLS adaptive algorithm was introduced with a step factor of 0.95. The adaptive controller was updated every 10 milliseconds to correct the predicted value of the GBRT model in real time.

[0184] Finally, the complete control system was tested on the simulation platform. When the simulated vibration amplitude exceeded 1.2 times the rated value, the adaptive controller could control the vibration within a safe range within 3 seconds, while the traditional PID controller took 8 seconds to suppress the vibration. At the same time, the adaptive controller can maintain good control performance under different working conditions. Comprehensively considering various indicators, this control scheme is significantly better than conventional practices and has strong engineering application value.

[0185] The comparison indexes of the present invention and the traditional method are shown in Table 1.

[0186] Table 1 Comparison indexes of the present invention and the traditional method

[0187]

[0188] Specifically, it can be seen from Table 1 that the method of the present invention is significantly superior to the traditional vibration control method in multiple key indicators such as modeling efficiency, system robustness, scalability, control accuracy, generalization ability, response speed and adaptability. Therefore, the method has strong practical value and application prospects.

[0189] The following describes an embodiment of the device of the present application, which can be used to implement a turbine vibration control method based on shaft seal air supply volume detection in the above embodiment of the present application. For details not disclosed in the embodiment of the device of the present application, please refer to the embodiment of the above method of the present application.

[0190] Reference Figure 2 As shown, according to an embodiment of the present application, a turbine vibration control device 200 based on shaft seal air supply amount detection includes:

[0191] The first acquisition unit 201 is used to acquire the shaft seal air supply volume data at the turbine shaft seal in real time;

[0192] A second acquisition unit 202 is used to acquire vibration data of the hydraulic turbine synchronized with the shaft seal air supply volume data;

[0193] A preprocessing unit 203 is used to preprocess the shaft seal air supply volume data and the vibration data;

[0194] The model building unit 204 is used to establish and train a mapping relationship model between the shaft seal air supply volume and the turbine vibration based on the preprocessed shaft seal air supply volume data and the vibration data by using a machine learning algorithm to obtain a GBRT integrated model;

[0195] A verification and optimization unit 205 is used to verify the GBRT integrated model and optimize and improve the model parameters according to the verification results;

[0196] A design unit 206 is used to determine adjustable turbine operating parameters as control quantities and design a feedback controller based on the optimized and improved GBRT integrated model;

[0197] The control unit 207 is used to control the vibration of the turbine based on the feedback controller.

[0198] like Figure 3 As shown, an embodiment of the present application also provides an electronic device 300, including a memory 310, a processor 320, and a computer program 311 stored in the memory 310 and executable on the processor. When the processor 320 executes the computer program 311, the steps of the above-mentioned turbine vibration control method based on shaft seal air supply amount detection are implemented.

[0199] Since the electronic device introduced in this embodiment is a device used to implement a turbine vibration control device based on shaft seal air supply volume detection in the embodiment of the present application, based on the method introduced in the embodiment of the present application, the technical personnel in this field can understand the specific implementation mode of the electronic device of this embodiment and its various variations. Therefore, how the electronic device implements the method in the embodiment of the present application is not introduced in detail here. As long as the equipment used by the technical personnel in this field to implement the method in the embodiment of the present application is within the scope of protection of this application.

[0200] During the specific implementation process, when the computer program 311 is executed by a processor, any implementation method in the embodiments corresponding to the first aspect can be implemented.

[0201] Figure 4 A schematic diagram of the structure of a computer system suitable for implementing an electronic device of an embodiment of the present application is shown.

[0202] It should be noted that Figure 4 The computer system 400 of the electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0203] like Figure 4As shown, the computer system 400 includes a central processing unit (CPU) 401, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 402 or the program loaded from the storage part 408 to the random access memory (RAM) 403, such as executing the method described in the above embodiment. In the RAM 403, various programs and data required for system operation are also stored. The CPU 401, ROM 402 and RAM 403 are connected to each other through a bus 404. An input / output (I / O) interface 405 is also connected to the bus 404.

[0204] The following components are connected to the I / O interface 405: an input section 406 including a keyboard, a mouse, etc.; an output section 407 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 408 including a hard disk, etc.; and a communication section 409 including a network interface card such as a LAN (Local Area Network) card, a modem, etc. The communication section 409 performs communication processing via a network such as the Internet. A drive 410 is also connected to the I / O interface 405 as needed. A removable medium 411, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 410 as needed so that a computer program read therefrom is installed into the storage section 408 as needed.

[0205] In particular, according to an embodiment of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present application includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication section 409, and / or installed from a removable medium 411. When the computer program is executed by a central processing unit (CPU) 401, various functions defined in the system of the present application are executed.

[0206] It should be noted that the computer-readable medium shown in the embodiment of the present application may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, - but not limited to - an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, a computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by an instruction execution system, device or device or used in combination with it. In the present application, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, wherein a computer-readable program code is carried. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, which may send, propagate, or transmit programs for use by or in conjunction with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0207] The flowchart and block diagram in the accompanying drawings illustrate the possible architecture, functions and operations of the system, method and computer program product according to various embodiments of the present application. Wherein, each box in the flowchart or block diagram can represent a module, a program segment, or a part of the code, and the above-mentioned module, program segment, or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flowchart, and the combination of boxes in the block diagram or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0208] The units involved in the embodiments described in this application may be implemented by software or hardware, and the units described may also be set in a processor. The names of these units do not, in some cases, constitute limitations on the units themselves.

[0209] As another aspect, the present application further provides a computer program product or a computer program, the computer program product or the computer program including computer instructions, the computer instructions being stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes a hydraulic turbine vibration control method based on shaft seal air supply amount detection described in the above embodiment.

[0210] As another aspect, the present application also provides a computer-readable medium, which may be included in the electronic device described in the above embodiment; or may exist independently without being assembled into the electronic device. The above computer-readable medium carries one or more programs, and when the above one or more programs are executed by an electronic device, the electronic device implements a turbine vibration control method based on shaft seal air supply volume detection described in the above embodiment.

[0211] It should be noted that, although several modules or units of the equipment for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiments of the present application, the features and functions of two or more modules or units described above can be embodied in one module or unit. On the contrary, the features and functions of one module or unit described above can be further divided into being embodied by multiple modules or units.

[0212] Through the description of the above implementation methods, it is easy for those skilled in the art to understand that the example implementation methods described here can be implemented by software or by combining software with necessary hardware. Therefore, the technical solution according to the implementation methods of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which can be a personal computer, a server, a touch terminal, or a network device, etc.) to execute the method according to the implementation methods of the present application.

[0213] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the embodiments disclosed herein. The present application is intended to cover any variations, uses or adaptations of the present application, which follow the general principles of the present application and include common knowledge or customary techniques in the art that are not disclosed in the present application. It should be understood that the present application is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from the scope thereof. The scope of the present application is limited only by the appended claims.

Claims

1. A method for controlling vibration of a hydraulic turbine based on detection of shaft seal air supply volume, characterized in that: include: Obtain the shaft seal air supply volume data at the turbine shaft seal in real time; Acquire vibration data of the hydraulic turbine synchronized with the shaft seal air supply volume data, and perform data preprocessing on the shaft seal air supply volume data and the vibration data; Using machine learning algorithms, based on the preprocessed shaft seal air supply volume data and vibration data, a mapping relationship model between the shaft seal air supply volume and turbine vibration is established and trained to obtain a GBRT integrated model. Verify the GBRT integrated model and optimize and improve the model parameters according to the verification results; Determine the adjustable turbine operating parameters as the control variables and design a feedback controller based on the optimized and improved GBRT integrated model; The turbine vibration is controlled based on the feedback controller.

2. The method according to claim 1, characterized in that The data preprocessing of the shaft seal air supply volume data and the vibration data includes: Clean the shaft seal air supply volume data and vibration raw data; Wavelet transform or EMD decomposition is used to denoise the cleaned shaft seal air supply volume data and vibration data; Perform data standardization on the denoised shaft seal air supply volume data and vibration data; According to actual needs, feature engineering processing is performed on the standardized shaft seal air supply volume data and vibration data.

3. The method according to claim 1, characterized in that The machine learning algorithm is used to establish and train a mapping relationship model between the shaft seal air supply volume and the turbine vibration based on the pre-processed shaft seal air supply volume data and vibration data, and obtain a GBRT integrated model, including: According to the characteristics of the problem, the gradient boosting regression tree GBRT is selected as the machine learning algorithm for training the mapping relationship model; A mapping relationship model is constructed, and the pre-processed shaft seal air supply volume data is used as the model input feature vector, and the vibration data is used as the model output target value to form a training data set; Randomly extract a part of the data from the training data set as the test set, and the remaining part as the first training validation set; Perform grid search or random search on the hyperparameters of the mapping relationship model to find the optimal parameter combination; Use mean square error (MSE) as the loss function, select stochastic gradient descent or Adam optimization algorithm, and iteratively train the mapping relationship model; Set the model evaluation period parameter eval_freq, evaluate the current model performance on the first training validation set every eval_freq iterations, and calculate the validation root mean square error RMSE and decision coefficient R 2 , when the error no longer decreases, stop training; Calculate the root mean square error RMSE and decision coefficient R of the final mapping relationship model on the test set 2 , mean absolute error MAE and median absolute deviation MedAE, and use these parameters as indicators to evaluate its generalization ability; Use Bagging or Boosting integration strategies to combine the output results of multiple mapping relationship models to build a GBRT integration model.

4. The method according to claim 3, characterized in that The grid search or random search of the hyperparameters of the mapping relationship model to find the optimal parameter combination includes: Determine the GBRT model hyperparameters that need to be optimized and their candidate value ranges; Separate a portion of the training data set as the second training validation set; Construct a grid search or random search estimator, and set the parameter grid for the grid search or random search, depending on the selected search strategy; Perform grid search or random search on the second training validation set; Get the best performing parameter instance during the search process, and re-instantiate the GBRT model based on this parameter instance as the final model instance for subsequent training and evaluation.

5. The method according to claim 3, characterized in that: The calculation formula of the root mean square error RMSE of the final mapping relationship model is as follows: In formula (1), n ​​represents the number of test set samples, y i represents the true vibration value of the i-th sample, f(x i ) represents the model's input x for the i-th shaft seal air supply volume i The predicted vibration value; The decision coefficient R of the final mapping relationship model 2 The calculation formula is as follows: In formula (2), y mean Represents the mean of the true vibration value y, y i represents the true vibration value of the i-th sample, f(x i ) represents the model's input x for the i-th shaft seal air supply volume i The predicted vibration value; The calculation formula of the mean absolute error MAE of the final mapping relationship model is as follows: In formula (3), n represents the number of test set samples, y i represents the true vibration value of the i-th sample, f(x i ) represents the model's input x for the i-th shaft seal air supply volume i The predicted vibration value; The calculation formula of the median absolute deviation MedAE of the final mapping relationship model is as follows: MEdAE=median(y i -f(x i )): (4) In formula (4), y i represents the true vibration value of the i-th sample, f(x i ) represents the model's input x for the i-th shaft seal air supply volume i The predicted vibration value of , median represents the median.

6. The method according to claim 1, characterized in that The GBRT integrated model is verified, and model parameters are optimized and improved according to the verification results, including: Prepare a new test dataset; Using the new test data set, calculate the root mean square error RMSE and determination coefficient R of the GBRT integrated model on the new test data set. 2 , mean absolute error MAE, median absolute deviation MedAE, and use these indicators to judge whether the prediction performance of the model on new data meets the standards; If the target is not met, a root cause analysis is conducted to analyze the potential causes of performance degradation; According to the analysis results, the GBRT integrated model is optimized; Re-evaluate the performance of the optimized GBRT integrated model on a new test dataset, and repeat the verification and improvement process until the performance of the GBRT integrated model meets the requirements.

7. The method according to claim 1, characterized in that The step of determining the adjustable turbine operating parameters as control quantities and designing a feedback controller based on the optimized and improved GBRT integrated model includes: Analyze the operating mechanism of the turbine and determine the set of adjustable operating parameters; Conduct sensitivity analysis to evaluate the sensitivity of each operating parameter to the vibration value and select a subset of parameters that have a greater impact on vibration as potential control operations; The control framework of the feedback controller is constructed based on the operating parameters, the optimized and improved GBRT integrated model and the control operation quantity; Designing a control algorithm of a feedback controller and tuning parameters of the control algorithm; Modeling and simulation analysis are performed on the feedback controller to evaluate whether the control performance meets the requirements under different working conditions. If the performance does not meet the requirements, the control parameters are repeatedly optimized, the control algorithm is improved, and the control framework design is adjusted until the control performance meets the requirements.

8. The method according to any one of claims 1 to 7, characterized in that Also includes: An adaptive mechanism is set in the feedback controller to correct the output of the GBRT integrated model in real time.

9. The method according to claim 8, characterized in that The step of setting an adaptive mechanism in the feedback controller to correct the output of the GBRT integrated model in real time includes: Pre-train multiple GBRT integrated models, each GBRT integrated model corresponds to a typical working condition of a turbine; The shaft seal air supply volume data is acquired in real time and input into the currently selected GBRT integrated model after data preprocessing to obtain the vibration value predicted by the model; The actual measured vibration value is obtained synchronously, and the prediction error is calculated based on the predicted vibration value and the actual measured vibration value; Use adaptive filtering algorithm to model and track prediction errors and obtain compensation terms; Add the vibration value predicted by the model to the compensation term to obtain the final vibration prediction value for control decision-making; Monitor the operating condition changes of the turbine in real time. When the operating condition changes significantly, switch to the corresponding GBRT integrated model for prediction and adaptation.

10. A turbine vibration control device based on shaft seal air supply volume detection, characterized in that: include: The first acquisition unit is used to acquire the shaft seal air supply volume data at the turbine shaft seal in real time; A second acquisition unit is used to acquire vibration data of the hydraulic turbine synchronized with the shaft seal air supply volume data; A preprocessing unit, used for preprocessing the shaft seal air supply volume data and vibration data; A model building unit is used to establish and train a mapping relationship model between the shaft seal air supply volume and the turbine vibration based on the preprocessed shaft seal air supply volume data and vibration data by using a machine learning algorithm to obtain a GBRT integrated model; A verification and optimization unit, used to verify the GBRT integrated model and optimize and improve the model parameters according to the verification results; A design unit is used to determine adjustable turbine operating parameters as control quantities and design a feedback controller based on the optimized and improved GBRT integrated model; A control unit is used for controlling the vibration of the turbine based on the feedback controller.