Pump turbine plant vibration prediction method and monitoring and early warning system

Abstract

The invention belongs to the field of water turbine vibration monitoring, and particularly provides a pump turbine plant vibration prediction method and a monitoring and early warning system. According to the collected data, a three-dimensional model is established, and the three-dimensional model comprises a unit flow field full-three-dimensional model and a plant structure field full-three-dimensional model; carrying out fluid calculation on the pump turbine by utilizing CFD software, and simulating fluid dynamic characteristics under different working conditions; fluid pressure data obtained through CFD simulation is used as a boundary condition, and fluid-solid coupling analysis is carried out in combination with a plant structure model; preprocessing the collected vibration data; constructing a vibration prediction model; monitoring the vibration condition of the factory building in real time, comparing and analyzing the pressure pulsation data obtained through real-time monitoring with the CFD simulation data, and inputting the data into the vibration prediction model to obtain the prediction condition of the vibration of the factory building. According to the prediction method and the monitoring and early warning system, the vibration condition of the plant in actual operation can be accurately simulated and predicted, and the prediction accuracy is improved.

Description

Technical Field

[0001] The invention belongs to the field of turbine vibration monitoring, and in particular relates to a pump turbine plant vibration prediction method and a monitoring and early warning system. Background Art

[0002] With the rapid development of my country's economy, energy demand is increasing day by day. Hydropower, as a clean and renewable energy form, has received widespread attention and application. As the core part of hydropower equipment, the safe and stable operation of pump-turbine is crucial to the stability of the entire pumped-storage power station. However, in the actual operation process, there is a vibration problem in the pump-turbine plant. Excessive vibration will affect the normal operation of the equipment and even cause equipment damage. Therefore, it is of great practical significance to predict and alarm the vibration of the pump-turbine plant. In the prior art, the research on the vibration of the pump-turbine plant is mainly based on empirical formulas and field tests, lacking systematic theoretical analysis and prediction methods. Therefore, it is necessary to study a pump-turbine plant vibration prediction method to improve the safety and stability of the pumped-storage system. Summary of the invention

[0003] The technical problem to be solved by the present invention is to provide a pump turbine plant vibration prediction method and a monitoring and early warning system, which can accurately simulate and predict the vibration conditions of the plant in actual operation, thereby improving the accuracy of the prediction.

[0004] In order to solve the above technical problems, the technical solution adopted by the present invention is: a method for predicting vibration of a pump turbine plant, comprising the following steps: S1. Data collection; S2. Establishing a three-dimensional model: establishing a three-dimensional model based on the collected data, wherein the three-dimensional model includes a full three-dimensional model of the unit flow field and the plant structure field; S3, CFD simulation and fluid-solid coupling analysis: Use CFD simulation software to perform fluid calculations on the pump turbine and simulate the fluid dynamic characteristics under different working conditions; use the fluid pressure data obtained by CFD simulation as boundary conditions, combine with the plant structure model, perform fluid-solid coupling analysis, and obtain CFD simulation data; S4, data preprocessing: preprocessing the vibration data collected in step S1; S5, constructing a vibration prediction model; S6. Real-time monitoring of the vibration of the plant is performed to obtain real-time monitoring data, which includes plant vibration data and pressure pulsation data. The real-time monitoring data is compared and analyzed with the CFD simulation data, and input into the vibration prediction model to obtain the prediction of the plant vibration.

[0005] In a preferred solution, in step S1, data collection includes collecting vibration data and fluid data of the pump turbine plant.

[0006] In a preferred solution, in step S4, data preprocessing includes denoising and normalizing the vibration data set.

[0007] In a preferred solution, in step S5, constructing a vibration prediction model includes the following steps: S501, applying a variational mode decomposition algorithm to the preprocessed vibration data to decompose and obtain an intrinsic mode function, and extracting effective vibration feature information therefrom; S502: Using the data obtained from the data preprocessing and variational mode decomposition steps, train an artificial neural network and / or a support vector machine model to establish a vibration prediction model.

[0008] In a preferred solution, step S501 includes the following steps: 1) Apply the variational mode decomposition algorithm to the preprocessed vibration data to decompose it into several intrinsic mode functions (IMFs), each of which represents a vibration mode in the signal; It is expressed in the following mathematical formula: ; in: x ( t ) is the preprocessed vibration data, IMF 1, IMF 2, ..., IMF n is the intrinsic mode function obtained by decomposition; 2) Extract effective vibration feature information from each IMF, the vibration feature information includes: Energy characteristics: Calculate the energy of each IMF, that is, the sum of the squares of the IMFs, expressed as: ; in: and Respectively represent the initial time and the end time of data collection; Center frequency: Calculate the center frequency of each IMF by performing a fast Fourier transform on the IMF and finding the frequency with the maximum energy. The expression is: ; Spectral entropy: Calculate the spectral entropy of IMF, the expression is: ; in: P k is the normalized probability distribution of the spectrum; Waveform Characteristics: Calculates the waveform characteristics of the IMF.

[0009] In a preferred solution, in step S6, a vibration threshold is set, and when the vibration prediction result exceeds the set threshold, an alarm is triggered.

[0010] In a preferred solution, in step S6, comparing and analyzing the real-time monitoring data with the CFD simulation data includes the following steps: S601, data synchronization: aligning real-time monitoring data with CFD simulation data in time; S602, data comparison: compare and analyze the real-time monitoring data with the data obtained by CFD simulation, use correlation analysis, error analysis or frequency analysis, and input the results of the comparison analysis as features into the vibration prediction model.

[0011] In a preferred embodiment, in step S602, the correlation analysis is to calculate the correlation coefficient between two sets of data.

[0012] In a preferred solution, in step S602, the error analysis is to calculate the mean square error or the root mean square error.

[0013] The present invention also provides a pump-turbine plant vibration monitoring and early warning system, comprising a real-time monitoring system including a signal receiving module, a data analysis module, a data storage module, a real-time early warning module and a threshold setting module; the signal receiving module is used to receive vibration data collected by a sensor and CFD simulation data; the data analysis module is used to store a vibration prediction model to obtain plant vibration prediction conditions; the data storage module is used to store real-time collected vibration data and historical data; the threshold setting module is used to set an alarm threshold; the real-time early warning module is used to trigger an alarm when the vibration prediction result exceeds the set threshold.

[0014] The present invention provides a pump turbine plant vibration prediction method and monitoring and early warning system, which has the following beneficial effects: 1. Through CFD simulation and fluid-solid coupling analysis, the vibration conditions of the plant during actual operation can be accurately simulated and predicted, improving the accuracy of the prediction.

[0015] 2. Using VMD decomposition technology, useful information in vibration signals can be effectively extracted, providing high-quality input data for subsequent feature analysis and modeling.

[0016] 3. Combining the artificial neural network and support vector machine models, the advantages of each are brought into play, improving the generalization ability and robustness of the vibration prediction model.

[0017] 4. The introduction of a real-time monitoring and early warning system enables abnormal vibration in the factory to be discovered and handled in a timely manner, greatly reducing the risk of equipment damage and safety accidents caused by vibration. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The present invention will be further described below in conjunction with the accompanying drawings and embodiments: Figure 1 is a flow chart of an embodiment of the present invention; Figure 2 The measurement point arrangement of the plant vibration source test in the embodiment of the present invention; Figure 3 It is a real-time monitoring and early warning system. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0020] Embodiment 1: A method for predicting vibration of a pump-turbine plant, such as Figure 1 As shown, the following steps are included: S1. Data collection.

[0021] Data collection includes collecting vibration data and fluid data of the pump turbine plant. These data can be obtained through field testing, sensor collection and historical data recording. The collected data includes but is not limited to vibration acceleration, displacement, fluid pressure, flow rate, etc. The collected data is stored in the database.

[0022] Figure 2 The arrangement of measuring points for the plant vibration source test of the embodiment of the present invention is shown. In actual operation, the measuring points can be reasonably arranged according to the structural characteristics of the plant and the working characteristics of the pump turbine to ensure that the vibration data can be collected comprehensively and accurately.

[0023] S2. Establishing a three-dimensional model: establishing a three-dimensional model based on the collected data, wherein the three-dimensional model includes a full three-dimensional model of the unit flow field and the plant structure field.

[0024] The model can be constructed using professional 3D modeling software such as ANSYS, CATIA, etc. to ensure that the model can accurately reflect the plant structure and flow field characteristics.

[0025] S3, CFD (Computational Fluid Dynamics) simulation and fluid-solid coupling analysis: Use CFD software to perform fluid calculations on pump turbines and simulate the fluid dynamic characteristics under different working conditions. Use CFD software (such as FLUENT, OpenFOAM, etc.) to perform fluid calculations on pump turbines and simulate fluid dynamic characteristics under different working conditions. Parameters such as pressure distribution and velocity distribution in the flow field are obtained through simulation to provide input for subsequent fluid-solid coupling analysis.

[0026] Fluid-solid coupling analysis: The fluid pressure data obtained from CFD simulation is used as the boundary condition. Combined with the plant structure model, fluid-solid coupling analysis software (such as the FSI module in ANSYS Fluent) is used to perform fluid-solid coupling analysis to obtain CFD simulation data, which are then stored in the database.

[0027] The CFD simulation data includes plant vibration simulation data and unit pressure pulsation simulation data.

[0028] S4. Data preprocessing: preprocess the vibration data collected in step S1.

[0029] The purpose of this step is to improve data quality and ensure the accuracy of subsequent analysis.

[0030] Data preprocessing usually includes two main processes: denoising and normalization to extract characteristic parameters reflecting vibration characteristics.

[0031] 1) Denoising: Remove random noise from vibration data through low-pass filtering, wavelet transform and other methods to extract useful signals.

[0032] 2) Normalization: Convert the data to the same scale for easier analysis and modeling. For example, the following formula can be used for normalization: ; in, x is the original data, and are the minimum and maximum values ​​of the data respectively.

[0033] S5. Construct a vibration prediction model.

[0034] Building a vibration prediction model involves the following steps: S501, applying a variational mode decomposition (VMD) algorithm to the preprocessed vibration data to decompose and obtain intrinsic mode functions (IMFs), and extracting effective vibration feature information therefrom.

[0035] 1) VMD algorithm is a technique for signal decomposition, which can decompose the signal into several intrinsic mode functions (IMFs), each of which represents a vibration mode in the signal.

[0036] It is expressed in the following mathematical formula: ; in: x ( t ) is the preprocessed vibration data, IMF 1, IMF 2, ..., IMF n is the decomposed intrinsic mode function.

[0037] 2) Extract effective vibration feature information from each IMF, for example: Energy feature: calculate the energy of each IMF to describe the intensity of vibration; Frequency feature: analyze the frequency component of each IMF to identify the frequency characteristics of vibration.

[0038] Vibration signature information includes: ① Energy characteristics: Calculate the energy of each IMF to describe the intensity of vibration, that is, the sum of the squares of the IMFs. The expression is: ; in: and Respectively represent the initial time and the end time of data collection; ② Center frequency: Calculate the center frequency of each IMF by performing a fast Fourier transform on the IMF and finding the frequency with the maximum energy. The expression is: ; ③ Spectral entropy: Calculating the spectral entropy of IMF can reflect the complexity and irregularity of the signal; The expression is: ; in: P k is the normalized probability distribution of the spectrum; ④ Waveform characteristics: Calculate the waveform characteristics of IMF, such as kurtosis, skewness and other statistics.

[0039] In the specific implementation, according to the requirements of the vibration prediction model, the features that best reflect the vibration characteristics are selected. These features will be used as input for training the vibration prediction model.

[0040] Through the above steps, effective vibration feature information is extracted from the intrinsic mode function decomposed by the VMD algorithm, providing an important data basis for the subsequent construction of the vibration prediction model. These feature information can help the model better understand and predict the vibration of the plant.

[0041] S502: Using the data obtained from the data preprocessing and VMD decomposition steps, train an artificial neural network or / and a support vector machine model to establish a vibration prediction model.

[0042] Artificial Neural Network (ANN): Use the preprocessed vibration data and feature information obtained from VMD decomposition as input to train one or more ANN models.

[0043] The ANN model can be a feedforward neural network (FFN) or a recurrent neural network (RNN), etc. Its basic structure includes an input layer, a hidden layer, and an output layer. The model can be expressed by the following formula: ; in, y is the output of the model, W is the weight matrix, x is the input data, b is the bias vector, f is the activation function.

[0044] Support Vector Machine Model (SVM): Train one or more SVM models using the same data. SVM is a powerful tool for classification and regression that can find an optimal hyperplane to separate data. The SVM model can be expressed by the following formula: ; in, is the weight vector, w and input data, x The dot product of b is the bias term and sign is the sign function.

[0045] The vibration prediction model can be a separate ANN or SVM model, or a combination of them. In practical applications, the advantages of these two models are usually combined, for example, using ANN for feature extraction and preliminary prediction, and then using SVM for final classification or regression prediction.

[0046] To supplement the key steps of the present invention in detail, the following is an example of a possible model formula for building a vibration prediction model: (1) Feature extraction: The VMD algorithm is used to decompose the vibration data and extract the features of IMF.

[0047] (2) ANN model: Establish a FFN whose input layer receives the features extracted by VMD, the hidden layer uses the ReLU activation function, and the output layer uses the linear activation function.

[0048] ; in, is the feature extracted by VMD, and are the weights and biases of the input layer, and are the weights and biases of the ANN.

[0049] (3) SVM model: Use the extracted features to train an SVM regression model.

[0050] ; in, and are the weights and biases of the SVM.

[0051] The final vibration prediction model can be a combination of ANN and SVM models, and the best model or model combination is selected to predict the vibration situation according to the actual situation.

[0052] In the process of building a vibration prediction model, methods such as cross-validation and grid search can be used to optimize model parameters and improve the prediction accuracy of the model.

[0053] S6. Monitor the vibration of the plant in real time, compare and analyze the real-time monitoring data with the CFD simulation data, and input them into the vibration prediction model to obtain the prediction of the plant vibration. The comparative analysis can be completed through specialized software tools or custom scripts.

[0054] For example, the comparative analysis of pressure pulsation data obtained from real-time monitoring and CFD simulation data includes the following steps: S601, data synchronization: align the pressure pulsation data obtained by real-time monitoring with the CFD simulation data in time to facilitate accurate comparative analysis.

[0055] S602, data comparison: perform comparative analysis on the real-time monitored pressure pulsation data and the pressure distribution obtained by CFD simulation, use correlation analysis, error analysis or frequency analysis, and input the comparative analysis results into the vibration prediction model as features.

[0056] Correlation analysis: Calculate the correlation coefficient between two sets of data to evaluate the linear relationship between them.

[0057] Error analysis: Calculate metrics such as mean square error (MSE) or root mean square error (RMSE) to quantify the difference between two sets of data.

[0058] Frequency analysis: Compare the frequency components and amplitudes of two sets of data through spectrum analysis.

[0059] For example, in this embodiment, the correlation coefficient is Pearson's correlation coefficient. r , used to measure the linear relationship between two sets of data, Pearson correlation coefficient r The calculation formula is as follows: ; in, and There are two sets of data, and are the means of the two sets of data respectively.

[0060] Error indicators such as mean square error (MSE) or root mean square error (RMSE) are used to quantify the difference between the predicted value and the actual value. The formula is as follows: ; ; in, is the actual value, is the predicted value.

[0061] The results of the comparative analysis are used as features to input into the vibration prediction model. These results can include error indicators, correlation coefficients, etc. obtained from the comparative analysis, which will be used as inputs to the vibration prediction model to predict the vibration of the plant.

[0062] Through this process, the results of the analysis are indeed input into the vibration prediction model, which will predict the vibration conditions of the factory building based on these input data.

[0063] Preferably, a vibration threshold is set, and when the vibration prediction result exceeds the set threshold, the alarm system is triggered to alarm. The alarm system can remind the operation and maintenance personnel to take corresponding measures through sound and light alarms, SMS notifications, etc.

[0064] Establish a database and store the collected data and calculation results in the database. The database can use a database management system such as MySQL, Oracle, etc. to facilitate data management and query.

[0065] Embodiment 2: In this embodiment, if Figure 3 As shown, a vibration monitoring and early warning system for a pump turbine plant includes a real-time monitoring system including a signal receiving module, a data analysis module, a data storage module, a real-time early warning module and a threshold setting module; the signal receiving module is used to receive vibration data collected by the sensor and CFD simulation data; the data analysis module is used to store the vibration prediction model to obtain the vibration prediction of the plant; the data storage module is used to store the real-time collected vibration data and historical data; the threshold setting module is used to set the alarm threshold; the real-time early warning module is used to trigger an alarm when the vibration prediction result exceeds the set threshold.

[0066] The system collects data through the real-time monitoring module. After processing, the data is input into the vibration prediction model of the data analysis module. The prediction results are displayed in real time through the early warning module. When the prediction value exceeds the set threshold, an alarm is triggered.

[0067] The present invention provides a pump-turbine plant vibration prediction method and monitoring and early warning system, which can effectively improve the safety and stability of the power generation system and reduce the risk of equipment damage and safety accidents.

[0068] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for predicting vibration of a pump-turbine plant, characterized in that: The following steps are involved: S1. Data collection; S2. Establishing a three-dimensional model: establishing a three-dimensional model based on the collected data, wherein the three-dimensional model includes a full three-dimensional model of the unit flow field and the plant structure field; S3, CFD simulation and fluid-solid coupling analysis: Use CFD simulation software to perform fluid calculations on the pump turbine and simulate the fluid dynamic characteristics under different working conditions; use the fluid pressure data obtained by CFD simulation as boundary conditions, combine with the plant structure model, perform fluid-solid coupling analysis, and obtain CFD simulation data; S4, data preprocessing: preprocessing the vibration data collected in step S1; S5, constructing a vibration prediction model; S6. Real-time monitoring of the vibration of the plant is performed to obtain real-time monitoring data, which includes plant vibration data and pressure pulsation data. The real-time monitoring data is compared and analyzed with the CFD simulation data, and input into the vibration prediction model to obtain the prediction of the plant vibration.

2. A pump-turbine plant vibration prediction method according to claim 1, characterized in that: In the step S1, data collection includes collecting vibration data and fluid data of the pump turbine plant.

3. A pump-turbine plant vibration prediction method according to claim 1, characterized in that: In step S4, data preprocessing includes denoising and normalizing the vibration data set.

4. A pump-turbine plant vibration prediction method according to claim 1, characterized in that: In step S5, constructing a vibration prediction model includes the following steps: S501, applying a variational mode decomposition algorithm to the preprocessed vibration data to decompose and obtain an intrinsic mode function, and extracting effective vibration feature information therefrom; S502: Using the data obtained from the data preprocessing and variational mode decomposition steps, train an artificial neural network and / or a support vector machine model to establish a vibration prediction model.

5. A pump-turbine plant vibration prediction method according to claim 4, characterized in that: The step S501 includes the following steps: 1) Apply the variational mode decomposition algorithm to the preprocessed vibration data to decompose it into several intrinsic mode functions (IMFs), each of which represents a vibration mode in the signal; It is expressed in the following mathematical formula: ； in: x ( t ) is the preprocessed vibration data, IMF 1, IMF 2, ..., IMF n is the intrinsic mode function obtained by decomposition; 2) Extract effective vibration feature information from each IMF, the vibration feature information includes: Energy characteristics: Calculate the energy of each IMF, that is, the sum of the squares of the IMFs, expressed as: ； in: and Respectively represent the initial time and the end time of data collection; Center frequency: Calculate the center frequency of each IMF by performing a fast Fourier transform on the IMF and finding the frequency with the maximum energy. The expression is: ； Spectral entropy: Calculate the spectral entropy of IMF, the expression is: ； in: P k is the normalized probability distribution of the spectrum; Waveform Characteristics: Calculates the waveform characteristics of the IMF.

6. A pump-turbine plant vibration prediction method according to claim 1, characterized in that: In step S6, a vibration threshold is set, and when the vibration prediction result exceeds the set threshold, an alarm is triggered.

7. A pump-turbine plant vibration prediction method according to claim 1, characterized in that: In step S6, comparing and analyzing the real-time monitoring data with the CFD simulation data includes the following steps: S601, data synchronization: aligning real-time monitoring data with CFD simulation data in time; S602, data comparison: compare and analyze the real-time monitoring data with the data obtained by CFD simulation, use correlation analysis, error analysis or frequency analysis, and input the results of the comparison analysis as features into the vibration prediction model.

8. A pump-turbine plant vibration prediction method according to claim 7, characterized in that: In step S602, the correlation analysis is to calculate the correlation coefficient between two sets of data.

9. A pump-turbine plant vibration prediction method according to claim 7, characterized in that: In step S602, the error analysis is to calculate the mean square error or the root mean square error.

10. A pump turbine plant vibration monitoring and early warning system, characterized in that: The real-time monitoring system includes a signal receiving module, a data analysis module, a data storage module, a real-time warning module and a threshold setting module; the signal receiving module is used to receive the vibration data collected by the sensor and the CFD simulation data; the data analysis module is used to store the vibration prediction model and obtain the vibration prediction of the factory building; the data storage module is used to store the real-time collected vibration data and historical data; the threshold setting module is used to set the alarm threshold; the real-time warning module is used to trigger an alarm when the vibration prediction result exceeds the set threshold.

Images