Pumped storage power station unit shafting safety early warning method and system based on digital twinning

Abstract

The application discloses a pumped storage power station unit shaft system safety early warning method based on digital twinning, comprising the following steps: S1, obtaining operation data of a pumped storage power station unit shaft system and preprocessing; S2, based on the preprocessed operation data, establishing a pumped storage power station unit shaft system digital twinning model and a shaft system posture prediction model; S3, comparing the difference between the output of the shaft system digital twinning model and the output of the posture prediction model, and correcting the digital twinning model; S4, realizing pumped storage power station unit safety early warning according to the corrected digital twinning model and the shaft system posture prediction model. The application can realize intelligent and real-time vibration monitoring and safety early warning of the pumped storage power station unit shaft system.

Description

Technical Field

[0001] This invention relates to the technical field of shaft system safety early warning for pumped storage power station units. More specifically, this invention relates to a method and system for shaft system safety early warning based on digital twins for pumped storage power station units. Background Technology

[0002] With the increasing proportion of new energy power generation such as wind and solar power, the safety and stability of the power grid face new challenges. As an important regulator, pumped storage hydropower has undergone significant changes in its functional positioning, operating mode, and operating intensity. The frequency of unit start-ups and shutdowns has increased exponentially, and the units often operate at the boundary of the stable region. During frequent start-ups, shutdowns, and switching of operating conditions, the shaft system of pumped storage units is susceptible to factors such as hydraulic impact, mechanical failure, and electromagnetic imbalance, which may induce abnormal vibrations, structural fatigue, electrical faults, and other problems, seriously threatening the safe operation of the units.

[0003] Against the backdrop of rapid development in the digital economy, the deep integration of new technologies such as digital twins with various industries has provided a powerful impetus for the digitalization, networking, and intelligentization of industries. Digital twins, as a new generation of digital technology, have the potential to drive new economic development. Currently, digital twin models and systems for pumped storage units are still in the planning and initial stages. While establishing digital twin models of these units is of great value, standardization is challenging due to geographical limitations and system differences between different power stations. Furthermore, the operation of pumped storage units is greatly affected by water flow patterns, resulting in complex operating conditions that further increase the difficulty of modeling.

[0004] The main research challenges include: the urgent need for breakthroughs in intelligent sensing and monitoring technology for the vibration state of rotating shaft systems in pumped-storage units. Current condition monitoring technologies cannot meet the demands of intelligent sensing and information processing in a big data environment, leading to wasted computing and network resources and hindering the rationality and real-time nature of data processing; constructing a parameterized rotor dynamics model of the shaft system and accurately describing its time-varying external excitations is crucial. The vibration of pumped-storage units during operation is affected by multiple factors, including hydraulic, mechanical, and electromagnetic factors, and simulation analysis based on a single factor is insufficient to accurately describe shaft motion; the practical application of a digital twin integrated system for pumped-storage units requires comprehensive and long-term experimental verification of cutting-edge scientific theories to ensure the robustness and stability of the methods, thereby achieving comprehensive health management of hydropower units. Summary of the Invention

[0005] One objective of this invention is to provide a method and system for safety early warning of pumped storage power station unit shaft system based on digital twin, so as to realize intelligent and real-time vibration monitoring and safety early warning of pumped storage power station unit shaft system.

[0006] To achieve these objectives and other advantages of the present invention, according to one aspect of the present invention, a method for safety early warning of pumped storage power station unit shaft system based on digital twin is provided, comprising: S1: acquiring operating data of the pumped storage power station unit shaft system and preprocessing it; S2: establishing a digital twin model of the pumped storage power station unit shaft system and a shaft system attitude prediction model based on the preprocessed operating data; S3: comparing the difference between the output of the shaft system digital twin model and the output of the attitude prediction model, and correcting the digital twin model; S4: realizing safety early warning of the pumped storage power station unit based on the corrected digital twin model and the shaft system attitude prediction model.

[0007] Furthermore, the formula for the digital twin model of the shaft system is as follows:

[0008]

[0009] Among them, M DT (t) represents the output of the digital twin model of the axis system at time t, w i Let x be the weight of the i-th feature. i (τ) represents the preprocessed value of the i-th feature at time t, τ is the full width at half maximum (FWHM) of the Lorentz distribution, λ is the time decay coefficient, m is the number of factors affecting the axis attitude, and k j Let y be the growth rate of the j-th influencing factor. i (t) represents the value of the j-th influencing factor at time t, y j0 Let be the baseline value of the j-th influencing factor, and n be the number of features in the shaft system operating data.

[0010] Furthermore, the formula for the shaft system attitude prediction model is as follows:

[0011]

[0012] Among them, M AP (t) represents the output of the axis attitude prediction model at time t, N is the size of the historical data window, τ is the full width at half maximum (FWHM) of the Lorentz distribution, h is the bandwidth of the Gaussian kernel, and M DT (t) represents the output of the digital twin model of the pumped storage unit shaft system at time t.

[0013] Furthermore, in S3, the method for calibrating the digital twin model includes:

[0014] The attitude data of each key point of the axis system is obtained in real time from the digital twin model, and the predicted attitude data at the corresponding time point is obtained from the axis system attitude prediction model.

[0015] The two sets of data are aligned according to time series, a difference analysis is performed, and the parameter space of the digital twin model of the axis system is established. The objective of parameter optimization is determined with the goal of minimizing the difference score.

[0016] Based on the parameter optimization objective, a heuristic algorithm is used for global parameter search, while an iterative optimization algorithm of gradient descent is used for local search in local regions to obtain the optimal parameters.

[0017] The digital twin model is corrected based on the optimal parameters.

[0018] Furthermore, the difference analysis includes the following steps:

[0019] Calculate the difference between the output of the digital twin model and the output of the axis attitude prediction model at each time point, and use the first derivative to determine the rate of change of the attitude data difference, and calculate the first difference. The attitude data includes gradual data and rapid data.

[0020] The differences in the data are characterized by positional and shape variations;

[0021] Empirical modal analysis is used to decompose rapidly changing data into trend and stationary components;

[0022] Based on the first difference, different difference evaluation indicators are used for the trend term and the stationary term respectively;

[0023] The difference evaluation indicators include amplitude difference evaluation indicators and spectrum difference evaluation indicators;

[0024] The various indicators are combined into a comprehensive difference score, which reflects the overall degree of deviation.

[0025] Further, in S4, feature parameters are extracted from the corrected digital twin model, and corresponding prediction feature sequences are obtained from the shaft system attitude prediction model. The feature parameters include the amplitude, displacement and stress of each node of the shaft system.

[0026] A warning threshold is set for the aforementioned feature parameters, and the real-time monitoring data and predicted data are compared to determine whether the warning threshold is exceeded.

[0027] Furthermore, the warning threshold includes a first-level warning threshold and a second-level warning threshold;

[0028] When the bearing vibration amplitude exceeds the first-level warning threshold, an alarm message is generated and the bearing location is marked.

[0029] If the maintenance personnel fail to respond within the specified time, the alarm will be escalated and a notification will be sent to higher-level management personnel.

[0030] When the vibration amplitude of the same bearing continues to rise and exceeds the second-level warning threshold, the fault mode library is invoked to compare the current vibration characteristics with typical fault modes and generate preliminary diagnostic opinions.

[0031] Furthermore, the formula for the first-level warning threshold is as follows:

[0032]

[0033] Where S1 is the first-level warning threshold, T is the evaluation time window, κ1 is the adjustment parameter of the first-level warning threshold, and M... DT (t) represents the output of the digital twin model of the pumped-storage unit shaft system at time t, M AP (t) is the output of the shaft system attitude prediction model at time t;

[0034] The specific formula for the second-level warning threshold is as follows:

[0035]

[0036] Where S2 is the second-level warning threshold, T is the evaluation time window, κ2 is the adjustment parameter of the second-level warning threshold, and M... DT (t) represents the output of the digital twin model of the pumped-storage unit shaft system at time t, M AP (t) is the output of the shaft system attitude prediction model at time t.

[0037] According to another aspect of the present invention, a pumped-storage power station unit shaft system safety early warning system based on digital twin is also provided, comprising: a collection module for acquiring and preprocessing operating data of the pumped-storage power station unit shaft system; a modeling module for establishing a digital twin model of the pumped-storage power station unit shaft system and a shaft system attitude prediction model based on the preprocessed operating data; a correction module for comparing the difference between the output of the shaft system digital twin model and the output of the attitude prediction model, and correcting the digital twin model; and an early warning module for implementing a safety early warning for the pumped-storage power station unit based on the corrected digital twin model and the shaft system attitude prediction model.

[0038] According to another aspect of the invention, a computer device is also provided, including a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement any of the described digital twin pumped storage power station unit shaft system safety early warning methods.

[0039] The present invention has at least the following beneficial effects:

[0040] This invention employs a multi-physics coupled finite element full-order model, considering mechanical vibration and incorporating the effects of hydraulic and electromagnetic forces to improve model accuracy and reveal typical fault mechanisms of hydraulic imbalance. It introduces Fourier transform for time-frequency analysis, extracting time-domain and frequency-domain features to capture the dynamic characteristics of faults and achieve early and accurate fault type identification. By combining a real-time calibrated pumped-storage unit shaft system digital twin model and shaft system attitude prediction model, a dynamic early warning threshold is set to improve warning sensitivity. A graded response mechanism reduces false alarms and downtime losses, enabling safe early warning for pumped-storage power station units.

[0041] Other advantages, objectives and features of the present invention will become apparent in part from the following description, and in part from those skilled in the art through study and practice of the invention. Attached Figure Description

[0042] Figure 1 This is a flowchart of the digital twin-based pumped storage power station unit shaft system safety early warning method in Example 1;

[0043] Figure 2 This is a flowchart illustrating the construction of a digital twin model of the pumped storage power station unit shaft system for the early warning method of digital twin in Example 1.

[0044] Figure 3 This is a flowchart illustrating the correction process for the digital twin-based pumped storage power station unit shaft system safety early warning method in Example 1. Detailed Implementation

[0045] The present invention will now be described in further detail with reference to the accompanying drawings, so that those skilled in the art can implement it based on the description.

[0046] It should be understood that terms such as "having," "comprising," and "including" used in the embodiments of this application do not exclude the presence or addition of one or more other elements or combinations thereof. All directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of this application are only used to explain the relative positional relationship and movement of components in a specific posture. If the specific posture changes, the directional indication will also change accordingly. When an element is referred to as "fixed to" or "set on" another element, it can be directly on the other element or may have an intervening element present. When an element is referred to as "connected to" another element, it can be directly connected to the other element or indirectly connected to the other element through an intervening element. Descriptions involving "first," "second," etc., in the embodiments of this application are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of those features.

[0047] It should be noted that the technical solutions of the various embodiments of this application can be combined with each other, but only if they are based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such combination of technical solutions does not exist and is not within the scope of protection claimed by this application.

[0048] Example 1:

[0049] Reference Figure 1-3 This is the first embodiment of the present application, which provides a method for early warning of shaft system safety in pumped storage power station units based on digital twins, including:

[0050] S1: Collect and preprocess the operating data of the pumped storage unit shaft system.

[0051] Specifically, sensors are installed at key locations in the pumped-storage unit's shaft system, such as the main bearing, guide bearing, and generator bearing. The bearing locations require triaxial accelerometers and displacement sensors, with the accelerometers measuring radial and axial vibrations and the displacement sensors measuring shaft displacement. T-type thermocouple temperature sensors are installed at the shaft system, rotor, and stator, and magnetoelectric tachometer speed sensors are installed at the shaft system locations.

[0052] Furthermore, a data acquisition system is set up, using a high-speed data acquisition card with a sampling rate at least twice the highest frequency of interest. GPS clocks are used to synchronize the clocks of all sensors and acquisition devices to ensure that the timestamps of all data points are accurate to the millisecond level.

[0053] Furthermore, the collected operational data is preprocessed: data exceeding the physical range is marked as anomalies, outliers are detected using statistical methods, short-term missing data is filled with linear interpolation, and long-term missing data is filled with multivariate interpolation; vibration data is filtered with a low-pass filter to remove high-frequency noise, and wavelet transform is used for multi-scale decomposition to remove noise from different frequency bands; temperature data can be reduced by using moving average or exponential smoothing methods.

[0054] S2: Based on the preprocessed operating data, establish a digital twin model of the pumped storage unit shaft system and a shaft system attitude prediction model.

[0055] Specifically, the construction process of the digital twin model of the pumped-storage unit shaft system includes the following steps: Based on the finite element analysis method, a multi-physics coupled finite element full-order model of the pumped-storage unit shaft system is built; Typical fault conditions are simulated based on the multi-physics coupled finite element full-order model to obtain sample data and classification results, solve the dynamic response under typical fault conditions, and analyze and store the shaft vibration factors; The order of the multi-physics coupled finite element full-order model is reduced to obtain the time-domain and frequency-domain characteristics of the shaft vibration response under the coupling of hydraulic and electromagnetic forces, and the calculation errors and time of each reduced-order model are compared; Based on the time-domain and frequency-domain characteristics, the fault mechanism of the pumped-storage unit is analyzed, and the parameters under the fault mechanism condition are calculated; Through each reduced-order model, the accuracy and timeliness of the reduced-order model are analyzed and verified, and a digital twin model of the pumped-storage unit shaft system for real-time simulation of typical operating conditions is constructed. The specific formulas are as follows:

[0056]

[0057] Among them, M DT (t) represents the output of the digital twin model of the axis system at time t, w i Let x be the weight of the i-th feature. i (τ) represents the preprocessed value of the i-th feature at time t, τ is the full width at half maximum (FWHM) of the Lorentz distribution, λ is the time decay coefficient, m is the number of factors affecting the axis attitude, and k j Let y be the growth rate of the j-th influencing factor. i (t) represents the value of the j-th influencing factor at time t, y j0 Let be the baseline value of the j-th influencing factor, and n be the number of features in the shaft system operating data.

[0058] It should be noted that the digital twin model of the pumped-storage unit shaft system includes excessive main shaft bending, excessive turning gear data, main shaft not perpendicular to the mirror plate, thrust head loosening, thrust bearing not level, bearing misalignment, bearing rubbing, and rotation center offset. The formula of the pumped-storage unit shaft system digital twin model integrates physical insights and data-driven approaches. Based on the Lorentz distribution, it describes the decay effect of various shaft system characteristics over time, capturing the time-varying characteristics and frequency distribution of vibration amplitude. It introduces a logarithmic growth model of shaft attitude influencing factors to reflect the nonlinear influence of hydraulic and electromagnetic forces on shaft dynamics. The optimization process of weights and parameters enables the model to adapt to actual operating conditions, while the selection of features and influencing factors reflects an understanding of complex failure mechanisms, providing real-time and accurate shaft state prediction. It reveals potential failures through changes in its structure and parameters, providing a solid foundation for predictive maintenance and safety early warning.

[0059] Furthermore, relevant data on shaft attitude are acquired through the pumped-storage unit monitoring system and preprocessed. The relevant data includes vibration, temperature, and rotational speed. The preprocessed relevant data is segmented according to time series alignment and subjected to Fourier transform time-frequency analysis to extract time-domain and frequency-domain features. Features of non-stationary signals are extracted using wavelet analysis algorithm, and key features are screened using correlation analysis dimensionality reduction technology to construct a feature matrix containing key features, which serves as input data for the shaft attitude prediction model.

[0060] Furthermore, a lightweight axisymmetric attitude prediction model is developed. The feature matrix is ​​proportionally divided into training, validation, and test sets. The training set is used to train the axisymmetric attitude prediction model and adjust its parameters. The validation set is used to evaluate the model, and the test set is used to evaluate its generalization ability. A knowledge distillation strategy is employed to use the prediction results of the trained axisymmetric attitude prediction model on the training data as soft labels, and the evaluation results on the test set as hard labels. The model is then trained by combining the hard and soft labels, as detailed in the following formula:

[0061]

[0062] Among them, M AP (t) represents the output of the axis attitude prediction model at time t, N is the size of the historical data window, τ is the full width at half maximum (FWHM) of the Lorentz distribution, h is the bandwidth of the Gaussian kernel, and M DT (t) represents the output of the digital twin model of the pumped storage unit shaft system at time t.

[0063] Specifically, the digital twin model of the pumped storage unit's shaft system is corrected in real time based on the shaft system attitude prediction model.

[0064] It should be noted that the shaft attitude prediction model formula integrates time series analysis and machine learning. The first part, based on the Lorentz distribution, captures the trend of shaft attitude change over time, which is particularly suitable for describing the decay of vibration amplitude. A Gaussian kernel is introduced to quantify the similarity between the output of the pumped-storage unit shaft digital twin model and the output of the shaft attitude prediction model, enabling the model to remember historical data and pay attention to abnormal states. The size of the historical window, the range of time influence, and the similarity sensitivity are controlled respectively. Through optimization, the model can adapt to different operating conditions and provide accurate short-term attitude prediction. Through interaction with the pumped-storage unit shaft digital twin model, the performance of the entire early warning system is continuously corrected and improved.

[0065] S3: By comparing the differences between the output of the pumped-storage unit shaft system digital twin model and the output of the attitude prediction model, the model parameters are optimized and the pumped-storage unit shaft system digital twin model is corrected in real time.

[0066] Specifically, the calibration includes the following steps: Real-time attitude data of key points in the pumped-storage unit shaft system are acquired from the digital twin model of the shaft system, and predicted attitude data at corresponding time points are acquired from the shaft system attitude prediction model; the attitude data includes displacement, velocity, and acceleration; the two sets of data are aligned according to time series, a difference analysis is performed, and a parameter space for the digital twin model of the pumped-storage unit shaft system is established, with the objective function being the minimization of the difference score; based on the parameter optimization objective, a heuristic algorithm is used for global parameter search, while an iterative optimization algorithm using gradient descent is used for local search in local areas to obtain the optimal parameters; based on the optimal parameters, the digital twin model of the pumped-storage unit shaft system is corrected.

[0067] Furthermore, the parameter space includes adjustable parameters and their value ranges; the difference analysis includes the following steps: calculating the difference between the output of the digital twin model and the output of the prediction model at each time point, and determining the rate of change of the attitude data difference using the first derivative, and calculating the first difference; the attitude data includes gradually changing data and rapidly changing data; the difference between the gradually changing data is characterized by positional difference and shape difference; the rapidly changing data is decomposed into trend term and stationary term using empirical modes; based on the first difference, different difference evaluation indicators are applied to the trend term and stationary term respectively; the difference evaluation indicators include amplitude difference evaluation indicators and spectral difference evaluation indicators; all indicators are integrated into a comprehensive difference score to reflect the overall degree of deviation and realize the safety early warning of pumped storage power station units.

[0068] Furthermore, feature parameters are extracted from the corrected digital twin model, and corresponding prediction feature sequences are obtained from the shaft system attitude prediction model. The time history and prediction trend of the feature parameters are then visualized. The feature parameters include the amplitude, displacement, and stress of each node in the shaft system. For the dynamic changes of the feature parameters, an early warning threshold is set, and the real-time monitoring data and prediction data are compared to determine whether the early warning threshold is exceeded.

[0069] S4: Implement safety early warning for pumped storage power station units based on the corrected digital twin model of the pumped storage unit shaft system and the shaft system attitude prediction model.

[0070] Specifically, the warning thresholds include a first-level warning threshold and a second-level warning threshold. When the bearing vibration amplitude exceeds the first-level warning threshold, an alarm is generated and the bearing location is marked in yellow on the monitoring interface. If the maintenance personnel do not respond within the specified time, the alarm is escalated and a notification is sent to higher-level management personnel.

[0071] Furthermore, the specific formula for the first-level warning threshold is as follows:

[0072]

[0073] Where S1 is the first-level warning threshold, T is the evaluation time window, κ1 is the adjustment parameter of the first-level warning threshold, and M... DT (t) represents the output of the digital twin model of the pumped-storage unit shaft system at time t, M AP (t) is the output of the shaft system attitude prediction model at time t;

[0074] Furthermore, when the vibration amplitude of the same bearing continues to rise and exceeds the second-level warning threshold, the fault mode library is invoked to compare the current vibration characteristics with typical fault modes and generate preliminary diagnostic opinions. If the diagnosis result indicates that the main shaft is misaligned, the main shaft load is reduced and closely monitored. If an abnormal peak appears in the vibration spectrum of the generator bearing, this frequency component is analyzed to determine the fault type. If the phase angle of the shaft vibration changes abruptly, the hydraulic parameters are abnormal, and the guide vanes or impeller are checked. The hydraulic parameters include head and flow rate. When periodic impact components appear in the shaft vibration waveform, there is a bearing fault or gear fault. If the impact frequency matches the characteristic frequency of the bearing or gear, the inner ring or rolling element of the output bearing has a defect.

[0075] Specifically, the formula for the second-level warning threshold is as follows:

[0076]

[0077] Where S2 is the second-level warning threshold, T is the evaluation time window, κ2 is the adjustment parameter of the second-level warning threshold, and M... DT (t) represents the output of the digital twin model of the pumped-storage unit shaft system at time t, M AP (t) is the output of the shaft system attitude prediction model at time t.

[0078] Furthermore, this embodiment also provides a pumped-storage power station unit shaft system safety early warning system based on digital twin, including: a collection module for collecting operating data of the pumped-storage unit shaft system and performing preprocessing;

[0079] The system consists of a model building module, which establishes a digital twin model of the pumped-storage unit's shaft system and a shaft system attitude prediction model based on preprocessed operating data; a calibration module, which optimizes model parameters and calibrates the pumped-storage unit's shaft system digital twin model in real time by comparing the differences between the outputs of the pumped-storage unit's shaft system digital twin model and the attitude prediction model; and an early warning module, which provides safety early warnings for the pumped-storage power station units based on the calibrated pumped-storage unit's shaft system digital twin model and shaft system attitude prediction model.

[0080] This embodiment also provides a computer device applicable to the digital twin pumped storage power station unit shaft system safety early warning method, including a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the digital twin pumped storage power station unit shaft system safety early warning method as proposed in the above embodiment.

[0081] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0082] In summary, this invention employs a multi-physics coupled finite element full-order model, considering mechanical vibration and incorporating the effects of hydraulic and electromagnetic forces to improve model accuracy and reveal complex fault mechanisms such as hydraulic imbalance. It introduces Fourier transform for time-frequency analysis, extracting time-domain and frequency-domain features to capture the dynamic characteristics of faults and achieve early and accurate fault type identification. By combining a real-time calibrated pumped-storage unit shaft system digital twin model and shaft system attitude prediction model, and setting dynamic early warning thresholds, it improves alarm sensitivity and reduces false alarms and downtime losses through a graded response mechanism, thus achieving safety early warning for pumped-storage power station units.

[0083] Example 2:

[0084] Referring to Table 1, which is the second embodiment of this application, this embodiment provides a method for early warning of shaft system safety of pumped storage power station units based on digital twins. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiment.

[0085] Specifically, this invention underwent a one-month field test at a large pumped-storage power station. This power station has six pumped-storage units, each with a capacity of 300MW. Before the test, engineers collected operating data from these six units over the past year, including shaft vibration, bearing temperature, axial displacement, and shaft power curves. Additionally, extra sensors were installed to collect environmental data, such as unit room temperature, humidity, and cooling water flow rate.

[0086] Furthermore, the collected historical data underwent preprocessing. This included data denoising, missing value imputation, and feature extraction. For shaft vibration data, wavelet transform was used for denoising to effectively remove the influence of high-frequency noise. Some missing values ​​in the bearing temperature data were imputed using Gaussian process regression to ensure data continuity. For axial displacement data, statistical features such as mean, variance, and peak value were extracted to capture key characteristics of shaft motion. The preprocessed data provided high-quality input for model building.

[0087] Furthermore, based on the preprocessed data, digital twin models M of the shaft systems were established for each of the six generating units. DT (t) and shaft attitude prediction model M AP (t). During the model construction process, it was found that traditional methods often only consider a single feature, such as the root mean square value of shaft vibration. The model proposed in this invention can comprehensively consider the dynamic influence of multiple features and environmental factors, capturing the time-varying nature of feature influences and the spectral characteristics of power fluctuations through parameters such as the time decay coefficient λ and the full width at half maximum (FWHM) τ of the Lorentz distribution. This makes the model's description of the shaft state more accurate and comprehensive.

[0088] Specifically, during the model training phase, the objective function minimization method is used to optimize the model parameters. After the model is built, a one-month real-time monitoring and early warning phase begins. Every 5 minutes, the system automatically collects real-time data and inputs it into the shaft system digital twin model M. DT (t) and shaft attitude prediction model M AP In (t), by comparing the output differences between the two models, the digital twin model M of the shaft system is corrected in real time. DT (t). For example: when Unit 1 experiences a slight bearing temperature rise on day 15, M DT (t) quickly captures this change and, through M AP By comparing (t), minor anomalies in shaft attitude can be accurately predicted.

[0089] Furthermore, to quantify the model's performance, a safety warning index was defined, and two levels of warning thresholds were set. During the experiment, the S-values ​​of six units were recorded, along with the warning status based on a fixed threshold method using the root mean square value of shaft vibration.

[0090] Furthermore, as shown in Table 1, on day 15, the traditional method failed to detect the shaft system anomaly in Unit 1 because the shaft vibration (1.2 mm / s) did not exceed the fixed threshold (usually set at 1.8-2.0 mm / s). However, the safety warning index S of this invention reached 0.42, exceeding the first-level warning threshold S1 (0.35), successfully providing an early warning. This invention not only considers shaft system vibration but also comprehensively considers factors such as increased bearing temperature (72°C) and increased power fluctuation (±8 MW).

[0091] Specifically, on day 20, Unit 2 triggered an early warning using conventional methods due to shaft vibration (1.8 mm / s) exceeding a threshold. However, the unit was actually still operating well at this time, with power fluctuations (±6 MW) within the normal range. The safety warning index S of this invention was 0.38, lower than the second-level warning threshold S2 (0.65), thus avoiding unnecessary shutdowns for inspection. This demonstrates that this invention effectively reduces the false alarm rate by comprehensively evaluating multiple parameters.

[0092] Furthermore, on day 18, Unit 5's safety warning index S reached 0.51, triggering a Level 2 warning. Although the shaft vibration (1.3 mm / s) was far below the conventional threshold, this invention detected abnormal growth trends in bearing temperature (73°C) and power fluctuations (±9 MW). In fact, subsequent inspections revealed slight damage to the bearing oil film, and timely preventative maintenance prevented more serious failures. This highlights the advantages of this invention in early failure prevention.

[0093] Table 1 Experimental Data

[0094]

[0095] Furthermore, by comparing the data from Unit 3 on days 5 and 30, it can be observed that although the shaft parameters fluctuated slightly (e.g., shaft vibration increased from 0.9 mm / s to 1.0 mm / s), the safety warning index S did not change significantly (from 0.15 to 0.32). Unit 6 operated smoothly throughout the entire test period, with the safety warning index consistently remaining at a low level (e.g., 0.08 on day 8). This indicates that the invention not only performs excellently under abnormal conditions but also maintains stability during normal operation, avoiding frequent warnings triggered by minor fluctuations, which is beneficial for the continuous and stable operation of the unit.

[0096] Although embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. They can be applied to various fields suitable for the present invention. For those skilled in the art, other modifications can be easily made. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to the specific details and illustrations shown and described herein.

Claims

1. A method for early warning of shaft system safety in pumped-storage power station units based on digital twins, characterized in that, include: S1: Obtain the operating data of the pumped storage power station unit shaft system and perform preprocessing; S2: Based on the preprocessed operating data, establish a digital twin model of the pumped storage power station unit shaft system and a shaft system attitude prediction model; S3: Compare the difference between the output of the axis digital twin model and the output of the attitude prediction model, and correct the digital twin model; S4: Implement safety early warning for pumped storage power station units based on the corrected digital twin model and the shaft system attitude prediction model; The formula for the digital twin model of the shaft system is as follows: Among them, M DT (t) represents the output of the digital twin model of the axis system at time t, w i Let x be the weight of the i-th feature. i (τ) represents the preprocessed value of the i-th feature at time t, τ is the full width at half maximum (FWHM) of the Lorentz distribution, λ is the time decay coefficient, m is the number of factors affecting the axis attitude, and k j Let y be the growth rate of the j-th influencing factor. i (t) represents the value of the j-th influencing factor at time t, y j0 Let j be the baseline value of the j-th influencing factor. The number of features in the shaft system operating data; The formula for the shaft system attitude prediction model is as follows: Among them, M AP (t) represents the output of the axis attitude prediction model at time t, N is the size of the historical data window, τ is the full width at half maximum (FWHM) of the Lorentz distribution, h is the bandwidth of the Gaussian kernel, and M DT (t) represents the output of the digital twin model of the pumped-storage unit shaft system at time t; In S3, the method for calibrating the digital twin model includes: The attitude data of each key point of the axis system is obtained in real time from the digital twin model, and the predicted attitude data at the corresponding time point is obtained from the axis system attitude prediction model. The two sets of data are aligned according to time series, a difference analysis is performed, and the parameter space of the digital twin model of the axis system is established. The objective of parameter optimization is determined with the goal of minimizing the difference score. Based on the parameter optimization objective, a heuristic algorithm is used for global parameter search, while an iterative optimization algorithm of gradient descent is used for local search in local regions to obtain the optimal parameters. The digital twin model is corrected based on the optimal parameters; The difference analysis includes the following steps: Calculate the difference between the output of the digital twin model and the output of the axis attitude prediction model at each time point, and use the first derivative to determine the rate of change of the attitude data difference, and calculate the first difference. The attitude data includes gradual data and rapid data. The differences in the data are characterized by positional and shape variations; Empirical modal analysis is used to decompose rapidly changing data into trend and stationary components; Based on the first difference, different difference evaluation indicators are used for the trend term and the stationary term respectively; The difference evaluation indicators include amplitude difference evaluation indicators and spectrum difference evaluation indicators; The various indicators are combined into a comprehensive difference score, which reflects the overall degree of deviation. In step S4, feature parameters are extracted from the corrected digital twin model, and corresponding prediction feature sequences are obtained from the shaft system attitude prediction model. The feature parameters include the amplitude, displacement, and stress of each node in the shaft system. A warning threshold is set for the aforementioned feature parameters, and the real-time monitoring data and predicted data are compared to determine whether the warning threshold is exceeded.

2. The method for early warning of shaft system safety of pumped storage power station units based on digital twin as described in claim 1, characterized in that, The warning thresholds include a first-level warning threshold and a second-level warning threshold; When the bearing vibration amplitude exceeds the first-level warning threshold, an alarm message is generated and the bearing location is marked. If the maintenance personnel fail to respond within the specified time, the alarm will be escalated and a notification will be sent to higher-level management personnel. When the vibration amplitude of the same bearing continues to rise and exceeds the second-level warning threshold, the fault mode library is invoked to compare the current vibration characteristics with typical fault modes and generate preliminary diagnostic opinions.

3. The digital twin-based pumped storage power station unit shaft system safety early warning method as described in claim 2, characterized in that, The formula for the first-level warning threshold is as follows: Where S1 is the first-level warning threshold, T is the evaluation time window, κ1 is the adjustment parameter of the first-level warning threshold, and M... DT (t) represents the output of the digital twin model of the pumped-storage unit shaft system at time t, M AP (t) is the output of the shaft system attitude prediction model at time t; The specific formula for the second-level warning threshold is as follows: Where S2 is the second-level warning threshold, T is the evaluation time window, κ2 is the adjustment parameter of the second-level warning threshold, and M... DT (t) represents the output of the digital twin model of the pumped-storage unit shaft system at time t, M AP (t) is the output of the shaft system attitude prediction model at time t.

4. A system for implementing the method of claim 1, characterized in that, include: The data collection module is used to acquire and preprocess the operating data of the pumped storage power station unit shaft system. A module is established to build a digital twin model of the pumped storage power station unit shaft system and a shaft system attitude prediction model based on the preprocessed operating data. A correction module is used to compare the difference between the output of the axis digital twin model and the output of the attitude prediction model, and to correct the digital twin model. The early warning module is used to provide safety early warning for pumped storage power station units based on the corrected digital twin model and the shaft attitude prediction model.

5. A computer device, comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the digital twin pumped storage power station unit shaft system safety early warning method according to any one of claims 1-3.

Images