Operation evaluation method and system for super large bulb tubular turbine generator set

Abstract

The present invention discloses an operation evaluation method and system for a super-large bulb-type tubular turbine generator set, which specifically relates to the field of data processing technology; through high-precision three-dimensional modeling and grid division, the simulation accuracy of computational fluid dynamics is improved; flow velocity monitoring points are arranged in the tailwater area, the turbulence autocorrelation function is calculated, and the vortex evolution time anomaly index is extracted; pressure monitoring points are arranged in key areas, and the pressure pulsation spectrum is analyzed by fast Fourier transform, the pressure pulsation nonlinear growth rate fluctuation index is extracted, and the impact of pressure pulsation on the safety of the unit is quantified. The CFD simulation deviation degree value is calculated in combination with a machine learning model, and the turbulence model parameters are dynamically adjusted based on the Bayesian optimization algorithm to achieve adaptive optimization of CFD calculations. This method effectively improves the simulation accuracy of the turbine, optimizes the unit efficiency evaluation, improves the reliability of resonance risk prediction, and ensures the long-term safe and stable operation of the turbine.

Description

Technical Field

[0001] The invention relates to the technical field of data processing, and in particular to an operation evaluation method and system for a super-large bulb tubular turbine generator set. Background Art

[0002] The operation evaluation of super-large bulb-type tubular turbine generator sets refers to the analysis and evaluation of the performance, efficiency, reliability and economy of this specific type of hydro-turbine generator set during operation. The bulb-type tubular turbine is a type of hydroelectric power generation equipment suitable for low head and large flow. Its key feature is that the water flows axially through the main equipment such as the turbine, generator and governor, which are encapsulated in a streamlined "bulb"-shaped cabin. The operation evaluation usually covers hydraulic performance, electrical performance, unit vibration, efficiency loss and long-term operation stability to ensure the safe and efficient operation of the unit and optimize the operation strategy.

[0003] The prior art has the following deficiencies:

[0004] In the existing technology, the energy conversion efficiency, tailwater return and pressure pulsation of the turbine are analyzed by comparing CFD (computational fluid dynamics) simulation and measured data. Among them, the flow field of the bulb-type tubular turbine is highly unsteady, involving strong turbulence, separated flow and large-scale vortex, while the traditional RANS (Reynolds-averaged Navier-Stokes) model (such as k-ε, SST k-ω) has limitations in dealing with these complex flows, and it is difficult to accurately capture the vortex evolution and pressure pulsation characteristics of the tailwater area. Due to model errors, the simulation may underestimate or overestimate the hydraulic efficiency of the turbine, resulting in the failure of the optimization strategy and the inability to effectively improve the performance of the unit. In addition, if the calculation deviation of the characteristic frequency of pressure pulsation is large, the potential resonance risk may not be identified in advance, causing the blades or tailwater pipes to bear abnormal loads for a long time, accelerating fatigue damage and even causing catastrophic structural failure. Summary of the invention

[0005] The object of the present invention is to provide an operation evaluation method and system for a super-large bulb tubular turbine generator set to overcome the deficiencies in the background technology.

[0006] In order to achieve the above object, the present invention provides the following technical solution: an operation evaluation method for a super-large bulb tubular turbine generator set, comprising the following steps:

[0007] A complete three-dimensional model of the turbine is established through the turbine modeling software, and the three-dimensional model of the turbine is meshed so that the computational fluid dynamics simulation accuracy meets the evaluation requirements;

[0008] Flow field monitoring points are arranged in the tailwater area to obtain flow velocity time series data. Based on the flow velocity time series data at the monitoring points, the turbulence autocorrelation function is calculated, and the time corresponding to when the autocorrelation function drops to 1 / e is used as the vortex evolution time anomaly index;

[0009] Lay out pressure monitoring points, run CFD transient simulation and collect pressure time series data, calculate the frequency spectrum of pressure pulsation through fast Fourier transform, identify the main pulsation frequency, fit the change of pressure amplitude over time, and calculate the nonlinear growth rate fluctuation index of pressure pulsation;

[0010] Based on the calculated vortex evolution time anomaly index and pressure pulsation nonlinear growth rate fluctuation index, the machine learning model is used to calculate the deviation degree of the turbine CFD simulation results. If the deviation exceeds the preset threshold, the turbulence model parameters are dynamically adjusted based on the deviation feedback to optimize the simulation accuracy and improve the accuracy of turbine efficiency evaluation and resonance risk prediction.

[0011] Preferably, the three-dimensional model of the turbine includes an impeller, guide vanes, tailwater pipe and bulb cabin, which is constructed using a parametric modeling method, and is divided into structured grids or hybrid grids to perform grid encryption in the blade, guide vane and tailwater areas to improve the calculation accuracy of the CFD simulation.

[0012] Preferably, multiple monitoring points are selected in the tailwater area to record the flow velocity time series. : ;in: Indicates the monitoring point at time The velocity at the point, N is the total number of sampling data points, and the turbulence autocorrelation function It describes the similarity of flow velocity at the same monitoring point at different time intervals τ. The calculation formula is: ;in: represents the velocity autocorrelation function value corresponding to the time lag τ, U(t) is the velocity time series data, is the mean of the square of the velocity, τ is the time lag, and the calculated In the sequence, find the time when the autocorrelation function drops to 1 / e , the expression is: ;in, and To make From greater than 0.368 to less than 0.368 in adjacent time steps; the vortex evolution time anomaly index is used to measure the deviation between the simulation calculation value and the experimental measurement value, and the calculation expression is: ; Where: VETAI is the vortex evolution time anomaly index, is the vortex evolution time calculated by CFD simulation, is the vortex evolution time obtained from experimental measurements.

[0013] Preferably, the pressure time series collected at a certain monitoring point is set ;in: It's in time The pressure value at , M is the total number of data points; Fourier transform the pressure time series to identify the main pulsation frequency : Main pulsation frequency Depends on the maximum amplitude of the Fourier transform spectrum: ; P(f) is the frequency spectrum function of pressure pulsation in the frequency domain, which describes the distribution of pressure signals at different frequencies; the main pulsation frequency identified by Fourier transform At , calculate the instantaneous amplitude A(t) of the pressure fluctuation: ;in: is the mean of the pressure time series, calculates the nonlinear growth rate λ, performs exponential fitting on A(t), and sets the pressure amplitude to grow with time in accordance with the exponential relationship: ; is the pressure amplitude at the initial moment, the slope λ is solved by the least squares method, and the pressure pulsation nonlinear growth rate fluctuation index NGR is calculated, and the expression is: ;in: is the pressure pulsation growth rate calculated by CFD simulation, is the pressure pulsation growth rate obtained from experimental measurement.

[0014] Preferably, the vortex evolution time anomaly index and the pressure pulsation nonlinear growth rate fluctuation index are converted into a comprehensive feature vector, and the comprehensive feature vector is used as the input of the machine learning model. The machine learning model is trained, and the deviation degree value of the turbine CFD simulation results is determined according to the model output results, wherein the machine learning model is a polynomial regression model.

[0015] Preferably, the obtained deviation degree value of the turbine CFD simulation result is compared with a preset threshold value. If the deviation degree value of the turbine CFD simulation result is greater than or equal to the preset threshold value, it means that the deviation degree of the turbine CFD simulation result is high, and a warning signal is generated at this time, indicating that the turbulence model needs to be adjusted or the grid division needs to be optimized; if the deviation degree value of the turbine CFD simulation result is less than the preset threshold value, it means that the deviation degree of the turbine CFD simulation result is low, and no warning signal is generated at this time.

[0016] Preferably, the method of dynamically adjusting the turbulence model parameters by using deviation feedback adopts Bayesian optimization, and predicts the optimal turbulence parameters through a proxy model to improve the CFD simulation accuracy, and the optimization steps include:

[0017] Define the optimization objective function ;in: is the CFD deviation degree value calculated under the current parameters, is the target CFD error threshold to minimize the CFD simulation error; select the turbulence model parameters to be optimized; use Gaussian process regression as the surrogate model, and use the expected improvement acquisition function to select the optimal parameters; run the CFD simulation and update the surrogate model until the error converges to the set threshold.

[0018] Preferably, for the calculated new deviation degree value DS(Θnew), the reliability of the simulation results is judged by setting an anomaly threshold DSth, where: if DS(Θnew)≥DSth, an anomaly signal is generated to prompt to continue optimizing the turbulence model parameters or adjusting the calculation grid accuracy; if DS(Θnew)<DSth, no anomaly signal is triggered and the current CFD calculation configuration is maintained.

[0019] The present invention also provides an operation evaluation system for a super-large bulb tubular hydro-generating unit, including a grid division module, a vortex analysis module, a pressure pulsation analysis module, and a deviation prediction module;

[0020] Grid division module: establish a complete three-dimensional model of the water turbine through a water turbine modeling software, and perform grid division on the three-dimensional model of the water turbine to make the computational fluid dynamics simulation accuracy meet the evaluation requirements;

[0021] Vortex analysis module: arrange flow field monitoring points in the tail water area to obtain velocity time series data, calculate the turbulence autocorrelation function based on the velocity time series data at the monitoring points, and take the time corresponding to the autocorrelation function dropping to 1 / e as the vortex evolution time anomaly index;

[0022] Pressure pulsation analysis module: arrange pressure monitoring points, run the CFD transient simulation and collect pressure time series data, calculate the spectrum of the pressure pulsation through fast Fourier transform, identify the main pulsation frequency, fit the variation of the pressure amplitude with time, and calculate the pressure pulsation non-linear growth rate fluctuation index;

[0023] Deviation prediction module: based on the calculated vortex evolution time anomaly index and pressure pulsation non-linear growth rate fluctuation index, use a machine learning model to calculate the deviation degree value of the water turbine CFD simulation results. If the deviation exceeds the preset threshold, the turbulence model parameters are dynamically adjusted based on the deviation feedback to optimize the simulation accuracy and improve the accuracy of the water turbine efficiency evaluation and resonance risk prediction.

[0024] In the above technical solution, the technical effects and advantages provided by the present invention:

[0025] 1. The present invention improves the accuracy of CFD calculations by constructing a high-precision three-dimensional model of the turbine and optimizing the grid division; flow field monitoring points and pressure monitoring points are arranged in the tailwater area and key positions to obtain flow velocity and pressure time series data, calculate the vortex evolution time anomaly index and the pressure pulsation nonlinear growth rate fluctuation index, and accurately characterize the flow characteristics and resonance risk of the turbine tailwater area. In addition, the polynomial regression machine learning model is used to calculate the CFD simulation deviation degree value, and the turbulence model parameters are dynamically adjusted based on Bayesian optimization to achieve intelligent optimization of CFD calculations. The technical solution of the present invention not only improves the accuracy of turbine simulation calculations, but also can quickly identify whether the optimization strategy is effective, thereby improving the hydraulic efficiency evaluation and resonance risk prediction capabilities of the unit.

[0026] 2. The present invention realizes adaptive improvement of turbine simulation accuracy through data-driven + intelligent optimization, significantly reduces the problem of underestimation or overestimation of efficiency caused by RANS turbulence model errors, and at the same time improves the accuracy of pressure pulsation prediction, so that resonance risks can be warned in advance. By introducing machine learning models and Bayesian optimization, the system can adjust turbulence parameters based on real-time feedback to ensure that CFD simulation errors converge to an acceptable range, improve computational stability and prediction reliability. Ultimately, the technical solution of the present invention can effectively optimize the operating state of turbines, improve power generation efficiency, reduce operating and maintenance costs, and extend equipment life, providing an efficient and intelligent analysis tool for turbine design and operation optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0028] Figure 1 The figure is a flow chart of the method of the present invention.

[0029] Figure 2 It is a system module diagram of the present invention. DETAILED DESCRIPTION

[0030] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments 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.

[0031] Example 1, please refer to Figure 1 As shown, the operation evaluation method of the super-large bulb tubular turbine generator set described in this embodiment includes the following steps:

[0032] A complete three-dimensional model of the turbine is established through the turbine modeling software, and the three-dimensional model of the turbine is meshed so that the computational fluid dynamics simulation accuracy meets the evaluation requirements;

[0033] Flow field monitoring points are arranged in the tailwater area to obtain flow velocity time series data. Based on the flow velocity time series data at the monitoring points, the turbulence autocorrelation function is calculated, and the time corresponding to when the autocorrelation function drops to 1 / e is used as the vortex evolution time anomaly index;

[0034] Lay out pressure monitoring points, run CFD transient simulation and collect pressure time series data, calculate the frequency spectrum of pressure pulsation through fast Fourier transform, identify the main pulsation frequency, fit the change of pressure amplitude over time, and calculate the nonlinear growth rate fluctuation index of pressure pulsation;

[0035] Based on the calculated vortex evolution time anomaly index and pressure pulsation nonlinear growth rate fluctuation index, the machine learning model is used to calculate the deviation degree of the turbine CFD simulation results. If the deviation exceeds the preset threshold, the turbulence model parameters are dynamically adjusted based on the deviation feedback to optimize the simulation accuracy and improve the accuracy of turbine efficiency evaluation and resonance risk prediction.

[0036] The main purpose of turbine modeling is to establish a complete and accurate geometric model that can be used for CFD calculations, covering the flow path structure and key components to truly reflect the hydrodynamic characteristics of the turbine.

[0037] Commonly used turbine modeling software include: SolidWorks, CATIA, Creo (Pro / E): suitable for detailed structural design and can be used to create accurate turbine geometry models. BladeGen (ANSYS): specially used for parametric modeling of turbine blades, which can optimize blade streamlines. AutoCAD, UG NX: used for overall turbine structure modeling and generating high-precision models of complex components.

[0038] A complete turbine geometry model should include the following key components:

[0039] Impeller (Runner): Generates the blade shape based on the turbine design parameters (number of blades, blade angle, inlet diameter, outlet diameter, etc.).

[0040] The NURBS surface (Non-Uniform Rational B-Spline) technology is used to make the blade surface smoother and reduce discrete errors.

[0041] Bulb Casing: Create a closed streamlined structure to ensure stable flow.

[0042] Internal motors, bearings and other structures need to be considered to ensure the integrity of the hydrodynamic calculations.

[0043] Guide Vane & Stay Vane: includes fixed guide vanes (Stay Vanes) and adjustable guide vanes (Guide Vanes), which are used to guide water flow into the impeller.

[0044] It is necessary to ensure that the geometric parameters of the guide vanes (installation angle, curvature radius, blade spacing) meet the design requirements.

[0045] Draft Tube: It adopts a gradually expanding tube structure to ensure smooth diffusion of water flow and improve water energy conversion efficiency.

[0046] Meshing is a key step in determining the calculation accuracy, convergence and efficiency during CFD simulation. The quality of the mesh directly affects the accuracy of the simulation results and must meet the following requirements:

[0047] It can accurately capture the flow characteristics of key areas (such as boundary layer, vortex, turbulence), and minimize the number of grids to improve computational efficiency while ensuring accuracy.

[0048] Commonly used meshing software include: ICEM CFD (ANSYS): suitable for complex geometric structures, supporting high-quality hexahedral meshing. Gambit (FLUENT): suitable for turbine blade flow channels, supporting multiple mesh types. Pointwise: suitable for high-precision fluid calculations, and can generate high-quality hexahedral and tetrahedral mixed meshes.

[0049] Meshing strategies for key components:

[0050] Impeller (Runner) mesh type: structured hexahedral mesh (Hexahedral Mesh) or unstructured tetrahedral mesh (Tetrahedral Mesh). Boundary layer processing: 10~20 layers of progressively refined boundary layer mesh are added to the blade surface to ensure the accuracy of turbulence model calculation. Mesh density between blades: Locally encrypted mesh is used in the flow channel area between blades to capture high-speed flow and pressure gradient changes.

[0051] Guide Vane & Stay Vane mesh type: unstructured tetrahedral mesh or hybrid mesh (Tetra / Prism). Refinement area: mesh encryption at the guide vane outlet and impeller inlet area to ensure calculation accuracy in the flow interaction area.

[0052] Bulb Casing: Mesh type: Hybrid mesh, hexahedral mesh inside and tetrahedral mesh outside. Optimization strategy: Ensure smooth flow around the cassette and avoid unnecessary vortices.

[0053] Draft Tube mesh type: unstructured mesh, with local mesh refinement in the recirculation area. Boundary layer processing: avoid excessive mesh sparseness to ensure that the turbulence model can accurately calculate the recirculation characteristics in the tailwater area.

[0054] Mesh quality optimization:

[0055] Orthogonal Quality: >0.2 (the closer to 1, the better). AspectRatio: Control between 1 and 10 to avoid extreme distortion. Y+ value control: For turbulence calculations near the wall, Y+ is controlled between 30-100 (standard wall function) or 1-5 (low Reynolds number turbulence model). Use wall refinement, such as Prism Layer in the laminar transition zone, to improve the accuracy of turbulence model calculations.

[0056] When running CFD simulations, the dynamic mesh refinement (AMR) method is used to automatically generate higher resolution meshes in high shear areas (such as blade edges, guide vane outlets, and tailwater pipe recirculation areas), thereby improving calculation accuracy.

[0057] The monitoring points should be located to cover key flow areas to ensure that velocity changes and large-scale eddy evolution in the tailwater area can be captured.

[0058] Tailwater pipe inlet: Monitor the flow velocity near the impeller outlet and analyze the flow pattern of the water discharged from the turbine. Pay attention to the uneven distribution of flow velocity, which may lead to increased tailwater losses.

[0059] Draft tube recirculation area: monitor the location where vortex separation may occur, analyze the low-speed recirculation area inside the draft tube (which may cause hydraulic loss). Calculate the formation and development of vortices through flow velocity data.

[0060] Draft tube outlet: monitor whether the flow rate is uniform and ensure that the water flows steadily into the downstream. Calculate the velocity gradient of the water flow at the outlet and analyze the kinetic energy loss.

[0061] A multi-layer distribution approach is used to capture data at different flow levels:

[0062] Axial direction: Select 3 to 5 monitoring points along the water flow direction to observe the velocity attenuation trend in the tailwater area.

[0063] Radial direction: On the same cross section, monitoring points in the center, near the wall and boundary layer areas are selected to analyze the non-uniformity of flow distribution.

[0064] Vertical direction: arrange upper and lower symmetrical monitoring points in the key sections of the tailwater pipe to observe the symmetry of the flow field and changes in turbulence intensity.

[0065] Depending on the measurement accuracy and application scenario, flow velocity monitoring can be done by physical measurement or numerical simulation. Particle Image Velocimetry (PIV) uses a high-resolution camera and a laser sheet light source to measure the motion trajectory of tracer particles in the water flow and calculate the instantaneous velocity field. It is suitable for laboratory testing and can provide high-precision two-dimensional or three-dimensional flow velocity data. Laser Doppler Velocimetry (LDV) measures the particle velocity in the fluid through laser scattering to obtain high-precision local flow velocity information. It is suitable for high-speed flow areas, such as impeller outlets. The Acoustic Doppler Velocimeter (ADV) uses sound waves to measure the velocity distribution of fluid particles and is suitable for large-scale water flow measurements. It is suitable for field testing, but the measurement accuracy is greatly affected by noise interference.

[0066] CFD (Computational Fluid Dynamics) simulation calculates flow velocity data using transient solutions (such as URANS, LES or DES methods) and extracts flow velocity time series data at monitoring points. Use high-precision time steps (such as 1ms level) to record flow velocity changes to ensure that transient fluid characteristics are obtained. At the deployed monitoring points, record the time series data of water flow velocity changes over time, and use high-frequency sampling (such as 1000Hz) to ensure that high-speed turbulence changes in the flow field are captured.

[0067] Select multiple monitoring points in the tailwater area and record the flow velocity time series : ;in: Indicates the monitoring point at time The velocity at the point (can be axial velocity, radial velocity or overall velocity). N is the total number of sampled data points. Turbulence autocorrelation function It describes the similarity of flow velocity at the same monitoring point at different time intervals τ. The calculation formula is: ;in: represents the velocity autocorrelation function value corresponding to the time lag τ, U(t) is the velocity time series data, is the mean of the square of the velocity, τ is the time lag, and its value range is (The maximum time range is usually from a few seconds to tens of seconds), Δt is the sampling time interval.

[0068] In the calculated In the sequence, find the time when the autocorrelation function drops to 1 / e (i.e. about 0.368) , the expression is: ;in, and To make The adjacent time steps that dropped from greater than 0.368 to less than 0.368.

[0069] The vortex evolution time anomaly index is used to measure the deviation between the simulation calculation value and the experimental measurement value. The calculation expression is: ; Where: VETAI is the vortex evolution time anomaly index, is the vortex evolution time calculated by CFD simulation, is the vortex evolution time obtained from experimental measurement. If VETAI>10%, it means that there is a large error in the vortex simulation in the tailwater area of ​​the CFD simulation, and the turbulence model needs to be adjusted (such as improving from RANS to LES or DES).

[0070] The layout of pressure monitoring points should cover the key flow areas to ensure that the pressure fluctuation characteristics inside the turbine and the tailwater area can be captured. They are mainly distributed in the following areas:

[0071] Monitor the blade surface for uniform pressure distribution and identify possible areas of abnormal pressure.

[0072] Pay attention to the pressure gradient between the suction and pressure surfaces of the blades and analyze the risk of cavitation. Monitor the pressure distribution of water entering the impeller at the guide vane outlet to ensure uniform flow. Pay attention to the high turbulence area and analyze the impact of pressure fluctuations on the blade force. Monitor the flow state of the tailwater at the tailwater diffuser section and analyze the pressure pulsation caused by vortex and backflow. Identify possible low-frequency pressure pulsations to avoid resonance risks. Monitor the pressure changes before the water is discharged at the tailwater outlet to evaluate energy loss. Observe the pressure fluctuation amplitude to determine whether the flow is stable.

[0073] Multi-point monitoring is used to cover different pressure fluctuation characteristic areas: Along the flow direction (axial direction): 3 to 5 monitoring points are arranged to analyze the spatial evolution of pressure fluctuations. In the radial direction: monitoring points are arranged in the center, boundary layer, and vortex core areas on the same cross section to analyze the impact of non-uniform flow on pressure pulsation. Symmetry: monitoring points are arranged in relative positions to analyze whether the pressure changes are periodic to identify the risk of low-frequency resonance.

[0074] Choose CFD simulation software. Commonly used software include: ANSYS Fluent / CFX (commercial software, suitable for complex turbulence modeling). OpenFOAM (open source software, suitable for custom solvers and parallel computing). STAR-CCM+ (suitable for high-precision fluid-solid coupling simulation).

[0075] Turbulence model: Standard k-ω SST (suitable for rotating machinery). DES (Detached Eddy Simulation) (combines RANS and LES to improve transient simulation accuracy). LES (Large Eddy Simulation) (suitable for capturing detailed pressure pulsations, but the computational cost is high). Calculation time step: Use a small time step to ensure that high-frequency pressure fluctuations are captured. The time step is recommended to be smaller than the characteristic time scale of pressure pulsations to prevent data loss. Boundary conditions: Inlet: Set total pressure or velocity inlet (based on flow conditions). Outlet: Set static pressure (based on tailwater depth). Wall: No-slip boundary condition (WallNo-slip).

[0076] Set the total simulation time (e.g. 5-10s) to ensure that sufficient transient flow characteristics are captured. Sample the pressure data once every Δt at the monitoring point to form a pressure time series, and use high-resolution storage (e.g. 1000Hz sampling rate) to ensure data integrity.

[0077] The pressure time series P(t) collected at a certain monitoring point is assumed to be: ;in: It's in time The pressure value at , M is the total number of data points;

[0078] Perform Fourier transform on the pressure time series, and the expression is: ; Where: P(f) is the frequency spectrum, which represents the pressure component at different frequencies f, is the kernel function of Fourier transform, is the time interval; identify the main pulsation frequency : Main pulsation frequency Depends on the maximum amplitude of the Fourier transform spectrum: ; P(f) is the frequency spectrum function of pressure pulsation in the frequency domain, which describes the distribution of pressure signals at different frequencies; that is, find the position with the largest amplitude in the Fourier transform result as the main pressure pulsation frequency.

[0079] Main pulsation frequency identified in Fourier transform At , calculate the instantaneous amplitude A(t) of the pressure fluctuation: ;in: is the mean of the pressure time series, A(t) reflects the amplitude change of pressure pulsation over time, calculate the nonlinear growth rate λ, perform exponential fitting on A(t), and set the pressure amplitude to grow with time in accordance with the exponential relationship: ; is the pressure amplitude at the initial moment. The slope λ is solved by the least square method. If λ>0, it indicates that the pressure pulsation is increasing, which may cause resonance risk. If λ<0, it indicates that the pressure pulsation is attenuating and the system is relatively stable.

[0080] Calculate the pressure pulsation nonlinear growth rate fluctuation index NGR, the expression is: ;in: is the pressure pulsation growth rate calculated by CFD simulation, is the pressure pulsation growth rate obtained from experimental measurements. If NGR>10%, it indicates that there is a significant deviation in the pulsation prediction of the CFD model, and the turbulence model needs to be adjusted (such as increasing the LES calculation accuracy or optimizing the grid division).

[0081] Based on the calculated vortex evolution time anomaly index and pressure pulsation nonlinear growth rate fluctuation index, the machine learning model is used to calculate the deviation degree of the turbine CFD simulation results. If the deviation exceeds the preset threshold, the turbulence model parameters are dynamically adjusted based on the deviation feedback to optimize the simulation accuracy and improve the accuracy of turbine efficiency evaluation and resonance risk prediction.

[0082] The vortex evolution time anomaly index and the pressure pulsation nonlinear growth rate fluctuation index are converted into comprehensive feature vectors, and the comprehensive feature vectors are used as the input of the machine learning model. The machine learning model predicts the deviation degree value label of the turbine CFD simulation results with each group of comprehensive feature vectors as the prediction target, and minimizes the sum of the prediction errors of the deviation degree value labels of all turbine CFD simulation results as the training target. The machine learning model is trained until the sum of the prediction errors converges and the model training is stopped. The deviation degree value of the turbine CFD simulation results is determined according to the model output results, wherein the machine learning model is a polynomial regression model.

[0083] The method for obtaining the deviation value of the turbine CFD simulation results is as follows: from the comprehensive feature vector training data of the trained machine learning model, the corresponding function expression is obtained: ; In the formula, is the output function of the model, VETAI is the vortex evolution time anomaly index, NGR is the pressure pulsation nonlinear growth rate fluctuation index, is the deviation value of the turbine CFD simulation results.

[0084] The obtained deviation degree value of the turbine CFD simulation result is compared with the preset threshold value. If the deviation degree value of the turbine CFD simulation result is greater than or equal to the preset threshold value, it means that the deviation degree of the turbine CFD simulation result is high. At this time, a warning signal is generated, indicating that the turbulence model needs to be adjusted or the grid division needs to be optimized; if the deviation degree value of the turbine CFD simulation result is less than the preset threshold value, it means that the deviation degree of the turbine CFD simulation result is low, and no warning signal is generated.

[0085] When the deviation value of the turbine CFD simulation exceeds the preset threshold, it is necessary to dynamically adjust the turbulence model parameters based on the deviation feedback to optimize the simulation accuracy and improve the accuracy of the turbine efficiency evaluation and resonance risk prediction.

[0086] This method uses Bayesian optimization to dynamically optimize the turbulence model parameters during the CFD calculation process to minimize simulation deviations and improve calculation accuracy.

[0087] Set the turbulence model parameters to be optimized, including: turbulence viscosity parameters (controls the turbulence dissipation rate), the turbulence kinetic energy generation term (affects pressure pulsation calculation), turbulence dissipation term (affects the evolution of large-scale vortices), turbulent wall treatment parameters (Controls boundary layer resolution accuracy).

[0088] The parameter search space is defined as: ;in: Applicable to k-ε or SST k-ω turbulence models, Wall function optimization for LES or DES calculations.

[0089] Define the objective function f(Θ) to evaluate the CFD simulation error of the current parameter configuration: ;in: is the CFD deviation value calculated under the current parameters, is the target CFD error threshold (usually set to within 5% deviation of experimental data). Minimize f(Θ), that is, adjust the parameters to minimize the CFD error.

[0090] Selecting initial turbulence parameters , run CFD simulation . Gaussian process regression is used to build the proxy model: ;in: is the predicted CFD error mean, is the kernel function, which represents the similarity between different parameter configurations. Represents the predicted output value under the input parameter Θ, GP stands for Gaussian process, which is a non-parametric Bayesian model used to represent the distribution of functions. Gaussian process can be regarded as a function space, where each input point Θ corresponds to an output value, which obeys a normal distribution and has a certain covariance structure. Use the acquisition function to select the next set of parameters Θnew to be evaluated. ; Select parameters that maximize error reduction, suitable for CFD optimization. Maximum Probability Improvement (MPI): ; Select the parameters that are most likely to reduce CFD deviations, suitable for robust optimization.

[0091] Run the CFD transient simulation (LES / RANS / DES) with the selected parameter Θnew, and calculate the new deviation degree value DS(Θnew). Update the Gaussian process regression (GPR) model to improve the surrogate function .

[0092] Calculate the current optimal parameter . Continue the search until: the error satisfies the convergence condition: ; where ε is the error threshold (e.g., 0.01). There is no obvious improvement in the acquisition function (i.e., the optimization effect tends to be stable).

[0093] Select the optimal parameter and apply it to the CFD simulation to obtain the optimized calculation result. Record the optimal parameter and store it in the database for future optimization use.

[0094] For the calculated new deviation degree value DS(Θnew), judge the reliability of the simulation result by setting the anomaly threshold DSth, where: if DS(Θnew) ≥ DSth, generate an anomaly signal to prompt to continue optimizing the turbulence model parameters or adjusting the calculation grid accuracy; if DS(Θnew) < DSth, do not trigger an anomaly signal and keep the current CFD calculation configuration.

[0095] Example 2, please refer to Figure 2 As shown, the operation evaluation system for the super-large bulb tubular hydro-generator unit described in this example includes a grid division module, a vortex analysis module, a pressure pulsation analysis module, and a deviation prediction module;

[0096] Grid division module: Establish a complete three-dimensional model of the water turbine through turbine modeling software, and perform grid division on the three-dimensional model of the water turbine to make the computational fluid dynamics simulation accuracy meet the evaluation requirements;

[0097] Vortex analysis module: Arrange flow field monitoring points in the tail water area to obtain velocity time series data. Based on the velocity time series data at the monitoring points, calculate the turbulence autocorrelation function, and take the time corresponding to the autocorrelation function dropping to 1 / e as the vortex evolution time anomaly index;

[0098] Pressure pulsation analysis module: Arrange pressure monitoring points, run the CFD transient simulation and collect pressure time series data, calculate the spectrum of pressure pulsation through fast Fourier transform, identify the main pulsation frequency, and fit the variation of pressure amplitude with time to calculate the pressure pulsation nonlinear growth rate fluctuation index;

[0099] Deviation prediction module: Based on the calculated vortex evolution time anomaly index and pressure pulsation nonlinear growth rate fluctuation index, the machine learning model is used to calculate the deviation degree of the turbine CFD simulation results. If the deviation exceeds the preset threshold, the turbulence model parameters are dynamically adjusted based on the deviation feedback to optimize the simulation accuracy and improve the accuracy of turbine efficiency evaluation and resonance risk prediction.

[0100] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. An operation evaluation method for a super-large bulb tubular turbine generator set, characterized in that: The following steps are involved: A complete three-dimensional model of the turbine is established through the turbine modeling software, and the three-dimensional model of the turbine is meshed so that the computational fluid dynamics simulation accuracy meets the evaluation requirements; Flow field monitoring points are arranged in the tailwater area to obtain flow velocity time series data. Based on the flow velocity time series data at the monitoring points, the turbulence autocorrelation function is calculated. The time corresponding to when the autocorrelation function drops to 1 / e is used as the vortex evolution time anomaly index, which is: Select multiple monitoring points in the tailwater area and record the flow velocity time series ;in: Indicates the monitoring point at time The velocity at the point, N is the total number of sampling data points, and the turbulence autocorrelation function Describes the same monitoring point at different time intervals The velocity similarity at is calculated as: ;in: Indicates time lag The corresponding flow velocity autocorrelation function value is, is the velocity time series data, is the mean of the square of the flow velocity, is the time lag, in the calculated In the sequence, find the time when the autocorrelation function drops to 1 / e , the expression is: ;in, and To make From greater than 0.368 to less than 0.368 in adjacent time steps; the vortex evolution time anomaly index is used to measure the deviation between the simulation calculation value and the experimental measurement value, and the calculation expression is: ; Where: VETAI is the vortex evolution time anomaly index, is the vortex evolution time calculated by CFD simulation, is the vortex evolution time obtained from experimental measurements; Lay out pressure monitoring points, run CFD transient simulation and collect pressure time series data, calculate the frequency spectrum of pressure pulsation through fast Fourier transform, identify the main pulsation frequency, fit the change of pressure amplitude over time, and calculate the nonlinear growth rate fluctuation index of pressure pulsation; Based on the calculated vortex evolution time anomaly index and pressure pulsation nonlinear growth rate fluctuation index, the machine learning model is used to calculate the deviation degree of the turbine CFD simulation results. If the deviation exceeds the preset threshold, the turbulence model parameters are dynamically adjusted based on the deviation feedback to optimize the simulation accuracy and improve the accuracy of turbine efficiency evaluation and resonance risk prediction.

2. The method for evaluating the operation of a super-large bulb tubular turbine generator set according to claim 1 is characterized in that: The three-dimensional model of the turbine includes an impeller, guide vanes, a tailwater pipe and a bulb cabin, is constructed using a parametric modeling method, and is divided into structured grids or hybrid grids to perform grid encryption in the blade, guide vane and tailwater regions to improve the calculation accuracy of the CFD simulation.

3. The method for evaluating the operation of a super-large bulb tubular turbine generator set according to claim 1 is characterized in that: Set the pressure time series collected at a certain monitoring point ;in: It's in time The pressure value at , M is the total number of data points; Fourier transform is performed on the pressure time series to identify the main pulsation frequency : Main pulsation frequency Depends on the maximum amplitude of the Fourier transform spectrum: ; It is the spectrum function of pressure pulsation in the frequency domain, which describes the distribution of pressure signals at different frequencies; the main pulsation frequency identified by Fourier transform Calculate the instantaneous amplitude of the pressure fluctuation at ;in: is the mean of the pressure time series, calculate the nonlinear growth rate λ, Perform exponential fitting and set the pressure amplitude to grow exponentially with time: ; is the pressure amplitude at the initial moment, the slope λ is solved by the least squares method, and the pressure pulsation nonlinear growth rate fluctuation index NGR is calculated, which is expressed as: ;in: is the pressure pulsation growth rate calculated by CFD simulation, is the pressure pulsation growth rate obtained from experimental measurement.

4. The method for evaluating the operation of a super-large bulb tubular turbine generator set according to claim 3 is characterized in that: The vortex evolution time anomaly index and the pressure pulsation nonlinear growth rate fluctuation index are converted into comprehensive feature vectors, which are used as the input of the machine learning model. The machine learning model is trained, and the deviation degree value of the turbine CFD simulation results is determined according to the model output results, wherein the machine learning model is a polynomial regression model.

5. The method for evaluating the operation of a super-large bulb tubular turbine generator set according to claim 4 is characterized in that: The obtained deviation degree value of the turbine CFD simulation result is compared with the preset threshold value. If the deviation degree value of the turbine CFD simulation result is greater than or equal to the preset threshold value, it means that the deviation degree of the turbine CFD simulation result is high. At this time, an early warning signal is generated, indicating that the turbulence model needs to be adjusted or the mesh division needs to be optimized. If the deviation degree value of the turbine CFD simulation result is less than the preset threshold value, it means that the deviation degree of the turbine CFD simulation result is low, and no warning signal is generated at this time.

6. The method for evaluating the operation of a super-large bulb tubular turbine generator set according to claim 5, characterized in that: The method for dynamically adjusting the turbulence model parameters by using deviation feedback adopts Bayesian optimization, and predicts the optimal turbulence parameters through a proxy model to improve the CFD simulation accuracy. The optimization steps include: Define the optimization objective function ;in: is the CFD deviation value calculated under the current parameters, Minimize the CFD simulation error by setting the target CFD error threshold; select the turbulence model parameters to be optimized; use Gaussian process regression as the proxy model and select the optimal parameters using the expected improvement acquisition function; run the CFD simulation and update the proxy model until the error converges to the set threshold.

7. The method for evaluating the operation of a super-large bulb tubular turbine generator set according to claim 6 is characterized in that: The calculated new deviation value , the reliability of the simulation results is judged by setting the abnormal threshold DSth, where: , an abnormal signal is generated, prompting you to continue optimizing the turbulence model parameters or adjusting the computational grid accuracy; if , no abnormal signal is triggered and the current CFD calculation configuration is maintained.

8. An operation evaluation system for a super-large bulb tubular turbine generator set, used to implement the operation evaluation method for a super-large bulb tubular turbine generator set according to any one of claims 1 to 7, characterized in that: It includes meshing module, vortex analysis module, pressure pulsation analysis module and deviation prediction module; Meshing module: Use turbine modeling software to build a complete three-dimensional model of the turbine, and mesh the three-dimensional model of the turbine so that the computational fluid dynamics simulation accuracy meets the evaluation requirements; Vortex analysis module: flow field monitoring points are arranged in the tailwater area to obtain velocity time series data. Based on the velocity time series data at the monitoring points, the turbulence autocorrelation function is calculated, and the time corresponding to when the autocorrelation function drops to 1 / e is used as the vortex evolution time anomaly index; Pressure pulsation analysis module: arrange pressure monitoring points, run CFD transient simulation and collect pressure time series data, calculate the spectrum of pressure pulsation through fast Fourier transform, identify the main pulsation frequency, fit the change of pressure amplitude over time, and calculate the nonlinear growth rate fluctuation index of pressure pulsation; Deviation prediction module: Based on the calculated vortex evolution time anomaly index and pressure pulsation nonlinear growth rate fluctuation index, the machine learning model is used to calculate the deviation degree of the turbine CFD simulation results. If the deviation exceeds the preset threshold, the turbulence model parameters are dynamically adjusted based on the deviation feedback to optimize the simulation accuracy and improve the accuracy of turbine efficiency evaluation and resonance risk prediction.

Images