Method for evaluating dynamic reliability of high-speed rotor of variable-speed pumped storage unit

Abstract

The present application relates to pumped storage technology field, especially a kind of variable speed pumped storage unit high-speed rotor dynamic reliability evaluation method, it is unique in the arrangement of sensor, carefully install multiple groups of wireless strain sensors in the end of rotor retainer, accurately capture retainer strain;High-speed camera group is reasonably arranged on test chamber side wall, and radial deformation gradient is effectively obtained;Various types of sensors are scientifically arranged in bearing seat, and shaft vibration displacement and bearing seat vibration acceleration are comprehensively monitored;This series of measures build multi-source data acquisition system, completely overcome the malpractice of unreasonable sensor arrangement and incomplete data acquisition in prior art;Based on these comprehensive and accurate data, with the aid of electromagnetic-mechanical-thermal coupling finite element model, digital twin model and the like, dynamic reliability index can be generated, residual life probability density function is calculated, and evaluation effect is significantly improved, and technical problems of existing evaluation method are effectively solved.

Description

Technical Field

[0001] This invention relates to the field of pumped storage technology, and in particular to a dynamic reliability assessment method for a high-speed rotor of a variable-speed pumped storage unit. Background Technology

[0002] With the large-scale development and utilization of renewable energy, variable-speed pumped-storage units (VPS) are playing an increasingly crucial role due to their flexible peak-shaving and valley-filling capabilities and their vital support for stable grid operation. Among these components, the high-speed rotor, as the core component of a VPS, directly affects the performance and safety of the entire unit. During operation, the high-speed rotor is subjected to complex electromagnetic, mechanical, and thermal stress coupling effects. Furthermore, due to frequent start-ups and shutdowns and changes in operating conditions, the rotor is constantly under alternating loads, making it highly susceptible to fatigue, creep, and wear damage, which in turn affects the reliable operation of the unit. Therefore, a reliability assessment of the high-speed rotor of a VPS is necessary.

[0003] However, existing evaluation methods are flawed in their sensor placement, leading to incomplete data collection and consequently affecting the evaluation results. Summary of the Invention

[0004] The purpose of this invention is to provide a dynamic reliability assessment method for high-speed rotors of variable-speed pumped storage units, aiming to solve the technical problem that the existing assessment methods are unreasonable in terms of sensor arrangement, resulting in incomplete data collection and thus affecting the assessment effect.

[0005] To achieve the above objectives, the present invention employs a dynamic reliability assessment method for a high-speed rotor of a variable-speed pumped storage unit, comprising the following steps:

[0006] Multiple sets of wireless strain sensors are installed circumferentially at the end of the rotor retaining ring, a high-speed camera group is deployed on the side wall of the test chamber, and an eddy current displacement sensor and a low-frequency velocity sensor are arranged in the bearing housing.

[0007] A three-stage speed-loaded runaway test was conducted, and the strain of the retaining ring, radial deformation gradient, shaft vibration displacement and bearing housing vibration acceleration were collected in real time.

[0008] Based on rotor geometric parameters and material properties, an electromagnetic-mechanical-thermal coupled finite element model is established. The electromagnetic field module calculates the Lorentz force distribution, the structural mechanics module calculates the ring deformation, and the thermal field module generates the temperature field and feeds it back to the material yield strength iterative model.

[0009] By inputting multi-source data into the digital twin model, three types of dynamic reliability indicators are generated: ring variable margin, shaft vibration mutation coefficient, and vibration energy entropy.

[0010] A local microstructure evolution equation for fatigue-creep-wear is established, and the remaining lifetime probability density function is calculated by coupling the failure probability through a time-varying Copula function.

[0011] The remaining lifetime and reliability are pushed to the power plant monitoring system using the OPC-UA protocol, and an early warning is triggered when the risk factor exceeds the threshold.

[0012] The method for assessing the dynamic reliability of the high-speed rotor of the variable-speed pumped storage unit also includes a data backup mechanism.

[0013] During data acquisition, a real-time incremental backup algorithm is used to back up multi-source data such as strain gauges to prevent accidental loss. A hash-based redundant storage algorithm is used to store the backup data in a distributed storage system, and the data integrity is checked periodically using a cyclic redundancy check (CRC) algorithm.

[0014] The dynamic reliability assessment method for high-speed rotors of variable-speed pumped storage units also includes log recording and data mining.

[0015] Key evaluation information is recorded using a log recording algorithm, and then the Apriori algorithm for association rule mining and the K-Means algorithm for cluster analysis are used to mine the log data, providing a basis for subsequent model optimization.

[0016] When multiple sets of wireless strain sensors are circumferentially installed at the end of the rotor retaining ring, the specific method is as follows:

[0017] Three sets of wireless strain sensors are symmetrically installed at 120° intervals around the end of the rotor retainer.

[0018] The loading specifications for the three-stage speed-loaded runaway test are as follows:

[0019] Phase 1: 0 → 430 RPM;

[0020] Phase 2: 430 → 622 RPM;

[0021] Phase 3: Maintain at 622 RPM until the deformation rate exceeds 1 micrometer / second, at which point the process terminates.

[0022] In the step of "inputting multi-source data into the digital twin model to generate three types of dynamic reliability indicators: ring variable margin, shaft vibration mutation coefficient, and vibration energy entropy," the digital twin model is updated using a federated learning framework.

[0023] Edge nodes process real-time data to generate initial values ​​for metrics;

[0024] Cloud-based fusion of historical data from multiple units optimizes the prediction weights for ring-shaped transformers.

[0025] Model parameters are synchronized every 24 hours, with a model update delay of less than 5 minutes.

[0026] In the step of "inputting multi-source data into the digital twin model to generate three types of dynamic reliability indicators: ring variable margin, shaft vibration mutation coefficient, and vibration energy entropy", the calculation method for the dynamic reliability indicators is as follows:

[0027] Protective ring deformation margin: 1 minus the ratio of the actual maximum deformation to the material's critical yield deformation;

[0028] Shaft vibration mutation coefficient: the ratio of the standard deviation of shaft vibration displacement to its mean;

[0029] Vibration energy entropy: Calculated based on the information entropy of the energy proportion of the vibration frequency band;

[0030] The reliability is deemed insufficient when the deformation margin is <0.8, the mutation coefficient is >2.5, or the energy entropy is >4.0.

[0031] In the step "Establishing the local microstructure evolution equation of fatigue-creep-wear, and calculating the remaining lifetime probability density function by coupling the failure probability through a time-varying Copula function", the method for constructing the competitive failure model is as follows:

[0032] The fatigue crack propagation rate is described by the Paris formula;

[0033] The creep strain rate was calculated using the Norton model;

[0034] Wear particle concentration was obtained through online oil monitoring;

[0035] The time-varying Copula correlation coefficient matrix is ​​updated using a sliding window exponentially weighted moving average.

[0036] The method for determining the threshold of the risk factor is as follows:

[0037] Based on offline training using support vector machines, the feature vectors are extracted by fusing the rotor surface infrared thermal image, magnetic flux distribution map, and vibration spectrum through a three-dimensional convolutional neural network.

[0038] The training samples are used to enhance the imbalanced data by generating an adversarial network.

[0039] The evaluation results are visualized using digital twins:

[0040] The SCADA system displays a 3D transparent rotor model, with reliability mapped by color (red: low reliability, blue: high reliability) and remaining life percentage mapped by transparency.

[0041] Dynamically label the coordinates of the critical speed range and stress concentration area.

[0042] This invention discloses a dynamic reliability assessment method for high-speed rotors of variable-speed pumped-storage units. In practical application, multiple sets of wireless strain sensors are circumferentially installed at the rotor retaining ring end, a high-speed camera group is deployed on the side wall of the test chamber, and various types of sensors are arranged in the bearing housing. This scientifically and rationally arranges the sensors to comprehensively collect multi-source data such as retaining ring strain, radial deformation gradient, shaft vibration displacement, and bearing housing vibration acceleration. This overcomes the deficiency of incomplete data collection caused by unreasonable sensor arrangement in existing technologies. It provides comprehensive and accurate data support for subsequent generation of dynamic reliability indicators and calculation of remaining lifetime probability density functions based on electromagnetic-mechanical-thermal coupled finite element models and digital twin models. This effectively improves the dynamic reliability assessment effect of high-speed rotors of variable-speed pumped-storage units. This method solves the technical problem in existing assessment methods where unreasonable sensor arrangement leads to incomplete data collection, thus affecting the assessment effect. Attached Figure Description

[0043] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0044] Figure 1 This is a flowchart illustrating the dynamic reliability assessment method for the high-speed rotor of the variable-speed pumped storage unit according to the present invention. Detailed Implementation

[0045] The embodiments of the present invention are described in detail below. Examples of the embodiments are shown in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, but should not be construed as limiting the present invention.

[0046] Please see Figure 1 , Figure 1 This is a flowchart illustrating the dynamic reliability assessment method for the high-speed rotor of the variable-speed pumped storage unit according to the present invention.

[0047] This invention provides a method for dynamic reliability assessment of high-speed rotors in variable-speed pumped storage units, comprising the following steps:

[0048] S1. Multiple sets of wireless strain sensors are installed circumferentially at the end of the rotor retaining ring, a high-speed camera group is deployed on the side wall of the test chamber, and an eddy current displacement sensor and a low-frequency velocity sensor are arranged in the bearing housing.

[0049] In this specific embodiment, the dynamic reliability assessment method for the high-speed rotor of the variable-speed pumped storage unit also includes a data backup mechanism;

[0050] During data acquisition, a real-time incremental backup algorithm is used to back up multi-source data such as strain gauges to prevent accidental loss. A hash-based redundant storage algorithm is used to store the backup data in a distributed storage system, and the data integrity is checked periodically using a cyclic redundancy check (CRC) algorithm.

[0051] When multiple sets of wireless strain sensors are circumferentially installed at the end of the rotor retaining ring, the specific method is as follows:

[0052] Three sets of wireless strain sensors are symmetrically installed at 120° intervals around the end of the rotor retainer.

[0053] S2. Perform a three-stage speed loading runaway test and collect real-time data on retaining ring strain, radial deformation gradient, shaft vibration displacement and bearing housing vibration acceleration.

[0054] For this specific implementation method, when performing the three-stage speed-loaded runaway test, the loading specifications are as follows:

[0055] Phase 1: 0 → 430 RPM;

[0056] Phase 2: 430 → 622 RPM;

[0057] Phase 3: Maintain at 622 RPM until the deformation rate exceeds 1 micrometer / second, at which point the process terminates.

[0058] S3. Based on the rotor's geometric parameters and material properties, an electromagnetic-mechanical-thermal coupled finite element model is established, in which the electromagnetic field module calculates the Lorentz force distribution, the structural mechanics module calculates the ring deformation, and the thermal field module generates the temperature field and feeds it back to the material yield strength iterative model.

[0059] S4. Input multi-source data into the digital twin model to generate three types of dynamic reliability indicators: ring variable margin, shaft vibration mutation coefficient, and vibration energy entropy.

[0060] In this specific implementation, the digital twin model is updated using a federated learning framework:

[0061] Edge nodes process real-time data to generate initial values ​​for metrics;

[0062] Cloud-based fusion of historical data from multiple units optimizes the prediction weights for ring-shaped transformers.

[0063] Model parameters are synchronized every 24 hours, with a model update delay of less than 5 minutes.

[0064] The calculation method for dynamic reliability index is as follows:

[0065] Protective ring deformation margin: 1 minus the ratio of the actual maximum deformation to the material's critical yield deformation;

[0066] Shaft vibration mutation coefficient: the ratio of the standard deviation of shaft vibration displacement to its mean;

[0067] Vibration energy entropy: Calculated based on the information entropy of the energy proportion of the vibration frequency band;

[0068] The reliability is deemed insufficient when the deformation margin is <0.8, the mutation coefficient is >2.5, or the energy entropy is >4.0.

[0069] S5. Establish the local microstructure evolution equation of fatigue-creep-wear, and calculate the remaining lifetime probability density function by coupling the failure probability through the time-varying Copula function.

[0070] For this specific implementation method, the method for constructing the competition failure model is as follows:

[0071] The fatigue crack propagation rate is described by the Paris formula;

[0072] The creep strain rate was calculated using the Norton model;

[0073] Wear particle concentration was obtained through online oil monitoring;

[0074] The time-varying Copula correlation coefficient matrix is ​​updated using a sliding window exponentially weighted moving average.

[0075] S6. The remaining lifetime and reliability are pushed to the power plant monitoring system using the OPC-UA protocol, and an early warning is triggered when the risk factor exceeds the threshold.

[0076] For this specific implementation method, the threshold for the risk factor is determined as follows:

[0077] Based on offline training using support vector machines, the feature vectors are extracted by fusing the rotor surface infrared thermal image, magnetic flux distribution map, and vibration spectrum through a three-dimensional convolutional neural network.

[0078] The training samples are used to enhance the imbalanced data by generating an adversarial network.

[0079] The evaluation results are visualized using digital twins:

[0080] The SCADA system displays a 3D transparent rotor model, with reliability mapped by color (red: low reliability, blue: high reliability) and remaining life percentage mapped by transparency.

[0081] Dynamically label the coordinates of the critical speed range and stress concentration area.

[0082] The dynamic reliability assessment method for high-speed rotors of variable-speed pumped storage units also includes:

[0083] To ensure that data collected by different sensors are synchronized in time, all data are labeled using a unified time reference so that multi-source data can be accurately fused and analyzed in the digital twin model.

[0084] The multiphysics coupling model was validated using experimental data. The differences between the model calculation results and the experimental data were compared. The model was then corrected based on the validation results until the model calculation results and the experimental data were within a reasonable error range.

[0085] The dynamic reliability assessment method for high-speed rotor of variable-speed pumped storage unit of the present invention, in practical use, collects data such as strain and deformation of retaining ring from different positions through the reasonable layout of multi-source sensors; real-time incremental backup, hash check redundant storage and CRC check are adopted to prevent data loss and ensure integrity, thus laying a solid foundation for assessment.

[0086] By performing a three-stage speed-load runaway test, following a clear loading specification (first stage: 0→430RPM; second stage: 430→622RPM; third stage: maintaining 622RPM until the deformation rate exceeds 1 micrometer / second), the test can simulate the operating state of the variable-speed pumped storage unit at different speeds, more realistically reflecting the dynamic characteristics of the unit in actual operation, and providing accurate test data for dynamic reliability assessment.

[0087] An electromagnetic-mechanical-thermal coupled finite element model, established based on rotor geometry parameters and material properties, comprehensively considers the interactions between electromagnetic fields, structural mechanics, and thermal fields. The electromagnetic field module calculates the Lorentz force distribution, the structural mechanics module calculates the ring deformation, and the thermal field module generates the temperature field and feeds it back to the material yield strength iterative model. This model can more accurately simulate the physical behavior of the rotor under different operating conditions, providing a more precise theoretical basis for dynamic reliability assessment.

[0088] The digital twin model is updated using a federated learning framework. Edge nodes process real-time data to generate initial index values, while the cloud integrates historical data from multiple generating units to optimize the prediction weights for the ring-shaped variable. Model parameters are synchronized every 24 hours with an update latency of less than 5 minutes. This update method ensures both the real-time performance of the model and full utilization of historical data from multiple generating units, improving the model's accuracy and generalization ability.

[0089] The calculation methods for three types of dynamic reliability indicators—strain margin, shaft vibration mutation coefficient, and vibration energy entropy—are clearly defined, and corresponding judgment criteria are set (reliability is deemed insufficient when the strain margin is <0.8, the mutation coefficient is >2.5, or the energy entropy is >4.0). These indicators can reflect the dynamic reliability of the high-speed rotor of variable-speed pumped storage units from different perspectives, providing an intuitive and effective basis for assessment and early warning.

[0090] In establishing the local microstructure evolution equations for fatigue-creep-wear, the fatigue crack propagation rate is described using the Paris formula, the creep strain rate is calculated using the Norton model, the wear particle concentration is obtained through online oil monitoring, and the time-varying Copula correlation coefficient matrix is ​​updated using a sliding window exponentially weighted moving average. This construction method fully considers the interaction between different failure modes and can more accurately simulate the rotor failure process.

[0091] By coupling failure probabilities using time-varying Copula functions and calculating the remaining lifetime probability density function, more accurate remaining lifetime information can be provided for the maintenance and decision-making of variable-speed pumped storage units. This helps to rationally arrange maintenance plans and improve the reliability and availability of the units.

[0092] The intelligent risk early warning mechanism promptly pushes the remaining lifespan and reliability, and issues warnings when thresholds are exceeded; the digital twin visualizes the rotor status, making it easier for operators to understand the situation and take measures.

[0093] This invention uses the OPC-UA protocol to push remaining lifetime and reliability information to the power plant monitoring system, triggering an early warning when risk factors exceed a threshold. The threshold determination method for risk factors is based on offline training using a support vector machine. Feature vectors are extracted by fusing rotor surface infrared thermal images, magnetic flux distribution maps, and vibration spectra through a three-dimensional convolutional neural network. Training samples are enhanced with a generative adversarial network to improve imbalance data. This intelligent risk early warning mechanism can promptly detect potential risks to the unit, providing timely warning information to operators and ensuring the safe operation of the unit.

[0094] This approach solves the technical problem in existing evaluation methods where unreasonable sensor placement leads to incomplete data collection, thus affecting the evaluation results.

[0095] The above description discloses only one preferred embodiment of the present invention, and should not be construed as limiting the scope of the present invention. Those skilled in the art will understand that all or part of the processes of the above embodiments can be implemented, and equivalent changes made in accordance with the claims of the present invention are still within the scope of the invention.

Claims

1. A method for dynamic reliability assessment of a high-speed rotor in a variable-speed pumped storage unit, characterized in that, Includes the following steps: Multiple sets of wireless strain sensors are installed circumferentially at the end of the rotor retaining ring, a high-speed camera group is deployed on the side wall of the test chamber, and an eddy current displacement sensor and a low-frequency velocity sensor are arranged in the bearing housing. A three-stage speed-loaded runaway test was conducted, and the strain of the retaining ring, radial deformation gradient, shaft vibration displacement and bearing housing vibration acceleration were collected in real time. Based on rotor geometric parameters and material properties, an electromagnetic-mechanical-thermal coupled finite element model is established. The electromagnetic field module calculates the Lorentz force distribution, the structural mechanics module calculates the ring deformation, and the thermal field module generates the temperature field and feeds it back to the material yield strength iterative model. By inputting multi-source data into the digital twin model, three types of dynamic reliability indicators are generated: ring variable margin, shaft vibration mutation coefficient, and vibration energy entropy. A local microstructure evolution equation for fatigue-creep-wear is established, and the remaining lifetime probability density function is calculated by coupling the failure probability through a time-varying Copula function. The remaining lifetime and reliability are pushed to the power plant monitoring system using the OPC-UA protocol, and an early warning is triggered when the risk factor exceeds the threshold. When performing a three-stage speed-loaded runaway test, the loading specifications are as follows: Phase 1: 0 → 430 RPM; Phase 2: 430 → 622 RPM; Phase 3: Maintain 622 RPM until the deformation rate exceeds 1 micrometer / second, then terminate. In the step "Inputting multi-source data into the digital twin model to generate three types of dynamic reliability indices: ring variable margin, shaft vibration mutation coefficient, and vibration energy entropy", the calculation method for the dynamic reliability indices is as follows: Protective ring deformation margin: 1 minus the ratio of the actual maximum deformation to the material's critical yield deformation; Shaft vibration mutation coefficient: the ratio of the standard deviation of shaft vibration displacement to its mean; Vibration energy entropy: Calculated based on the information entropy of the energy proportion of the vibration frequency band; The reliability is deemed insufficient when the deformation margin is <0.8, the mutation coefficient is >2.5, or the energy entropy is >4.

0.

2. The dynamic reliability assessment method for the high-speed rotor of a variable-speed pumped storage unit as described in claim 1, characterized in that, The dynamic reliability assessment method for high-speed rotors of variable-speed pumped storage units also includes a data backup mechanism. During data acquisition, a real-time incremental backup algorithm is used to back up multi-source data such as strain gauges to prevent accidental loss; a hash-based redundant storage algorithm is used to store the backup data in a distributed storage system, and a cyclic redundancy check algorithm is used periodically to check the data integrity.

3. The dynamic reliability assessment method for the high-speed rotor of a variable-speed pumped storage unit as described in claim 2, characterized in that, The dynamic reliability assessment method for high-speed rotors of variable-speed pumped storage units also includes log recording and data mining. Key evaluation information is recorded using a log recording algorithm, and then the Apriori algorithm for association rule mining and the K-Means algorithm for cluster analysis are used to mine the log data, providing a basis for subsequent model optimization.

4. The dynamic reliability assessment method for the high-speed rotor of a variable-speed pumped storage unit as described in claim 3, characterized in that, When multiple sets of wireless strain sensors are circumferentially installed at the end of the rotor retaining ring, the specific method is as follows: Three sets of wireless strain sensors are symmetrically installed at 120° intervals around the end of the rotor retainer.

5. The dynamic reliability assessment method for the high-speed rotor of a variable-speed pumped storage unit as described in claim 4, characterized in that, In the step "Inputting multi-source data into the digital twin model to generate three types of dynamic reliability indicators: ring variable margin, shaft vibration mutation coefficient, and vibration energy entropy," the digital twin model is updated using a federated learning framework. Edge nodes process real-time data to generate initial values ​​for metrics; Cloud-based fusion of historical data from multiple units optimizes the prediction weights for ring-shaped transformers. Model parameters are synchronized every 24 hours, with a model update delay of less than 5 minutes.

6. The dynamic reliability assessment method for the high-speed rotor of a variable-speed pumped storage unit as described in claim 5, characterized in that, In the step "Establishing the local microstructure evolution equation of fatigue-creep-wear, and calculating the remaining lifetime probability density function by coupling the failure probability through a time-varying Copula function", the method for constructing the competing failure model is as follows: The fatigue crack propagation rate is described by the Paris formula; The creep strain rate was calculated using the Norton model; Wear particle concentration was obtained through online oil monitoring; The time-varying Copula correlation coefficient matrix is ​​updated using a sliding window exponentially weighted moving average.

7. The dynamic reliability assessment method for the high-speed rotor of a variable-speed pumped storage unit as described in claim 6, characterized in that, The method for determining the threshold of risk factors is as follows: Based on offline training using support vector machines, the feature vectors are extracted by fusing the rotor surface infrared thermal image, magnetic flux distribution map, and vibration spectrum through a three-dimensional convolutional neural network. The training samples are used to enhance the imbalanced data by generating an adversarial network.

8. The dynamic reliability assessment method for the high-speed rotor of a variable-speed pumped storage unit as described in claim 7, characterized in that, The evaluation results are visualized using digital twins: The SCADA system displays a 3D transparent rotor model, with reliability mapped by color and remaining life percentage mapped by transparency. Dynamically label the coordinates of the critical speed range and stress concentration area.

Images