Hydropower station electromechanical equipment operation state monitoring method and system based on digital twinning

Abstract

The application provides a hydropower station electromechanical equipment operation state monitoring method and system based on digital twinning, belonging to the technical field of hydropower station electromechanical equipment operation and maintenance monitoring, which comprises the following steps: inputting monitoring data into a digital twinning model to obtain state mapping results and perform state evaluation; if any state mapping parameter is greater than a safe operation threshold, then fault mode deduction is performed; if no fault is detected, then an action data submodel is called based on the state mapping results to calculate a health index; a physical mechanism submodel is called to virtually sense and invert non-measuring point parts of the equipment that are not equipped with physical sensors to obtain holographic state evaluation results and fuse them with the health index to obtain a comprehensive health degree; the comprehensive health degree is time-series predicted to obtain a degradation trend prediction result; if the health degree is less than a health degree threshold, then fault mode deduction is performed based on the digital twinning model; and if the health degree is greater than or equal to the health degree threshold, then the electromechanical equipment is allowed to continue operating, and the monitoring frequency of the next cycle is adjusted based on the degradation trend prediction result.

Description

Technical Field

[0001] This application relates to the field of hydropower station electromechanical equipment operation and maintenance monitoring technology, and in particular to a method and system for monitoring the operating status of hydropower station electromechanical equipment based on digital twins. Background Technology

[0002] Hydropower station electromechanical equipment is a core asset in power production, and its operational status directly affects grid security and power generation efficiency. Currently, hydropower stations mainly monitor equipment using a combination of threshold alarms and regular manual inspections. Alarm mechanisms based on fixed thresholds only trigger when parameters exceed limits, failing to capture early, subtle deterioration characteristics. By the time potential problems are discovered, the equipment is already on the verge of failure. Sensor deployment points are limited, lacking direct monitoring methods for critical non-measuring points such as the root of the runner blades and the internal shaft system, resulting in sensing blind spots. Vibration, temperature, and operating condition data are scattered across different systems, lacking multi-dimensional fusion analysis capabilities and failing to accurately pinpoint the root cause of faults. Existing methods also suffer from fixed monitoring frequencies, unable to dynamically adjust according to the actual health status of the equipment, leading to wasted computing resources or insufficient collection of critical data.

[0003] Therefore, there is an urgent need for a method and system for monitoring the operational status of hydropower station electromechanical equipment based on digital twins. Summary of the Invention

[0004] To address the aforementioned technical issues, this application provides a method and system for monitoring the operational status of electromechanical equipment in hydropower stations based on digital twins.

[0005] A first aspect of this application provides a method for monitoring the operational status of electromechanical equipment in a hydropower station based on digital twins, comprising: The monitoring data of the hydropower station's electromechanical equipment is input into a preset digital twin model to obtain the state mapping results of the hydropower station's electromechanical equipment in multiple dimensions; wherein, the digital twin model includes a physical mechanism sub-model based on finite element analysis and a behavioral data sub-model based on temporal neural network. A first state assessment is performed on the state mapping result. If any state mapping parameter in the state mapping result is found to be greater than the preset safe operation threshold, then a fault mode inference is performed based on the digital twin model to obtain the fault root cause analysis result. If no state mapping parameter is detected to be greater than the preset safe operation threshold, then based on the state mapping result, the behavior data sub-model is called to calculate the health index of the hydropower station's electromechanical equipment; the physical mechanism sub-model is called to perform virtual sensing inversion on the non-measuring point parts of the hydropower station's electromechanical equipment where no physical sensors are installed, and obtain the holographic state assessment result. The health index is fused with the holographic status assessment result to obtain a comprehensive health score; Based on the behavioral data sub-model, a time-series prediction of the overall health is performed to obtain a deterioration trend prediction result. If the overall health score is less than the preset health score threshold, then the fault mode is deduced based on the digital twin model to obtain the root cause analysis results. If the overall health status is greater than or equal to the preset health status threshold, the electromechanical equipment is allowed to continue operating, and the monitoring frequency for the next cycle is adjusted based on the predicted deterioration trend of the overall health status.

[0006] A second aspect of this application provides a digital twin-based hydropower station electromechanical equipment operation status monitoring system, comprising: The twin mapping module is used to input the monitoring data of the hydropower station's electromechanical equipment into a preset digital twin model to obtain the state mapping results of the hydropower station's electromechanical equipment in multiple dimensions; wherein, the digital twin model includes a physical mechanism sub-model based on finite element analysis and a behavioral data sub-model based on a temporal neural network; The safety assessment module is used to perform a first state assessment on the state mapping result. If any state mapping parameter in the state mapping result is detected to be greater than the preset safe operation threshold, then the fault mode inference is performed based on the digital twin model to obtain the fault root cause analysis result. The deep evaluation module is used to calculate the health index of the hydropower station's electromechanical equipment based on the state mapping result if no state mapping parameter is detected to be greater than the preset safe operation threshold; and to call the physical mechanism sub-model to perform virtual sensing inversion on the non-measuring point parts of the hydropower station's electromechanical equipment where no physical sensors are installed, so as to obtain the holographic state evaluation result. The comprehensive health module is used to fuse the health index with the holographic status assessment result to obtain a comprehensive health score; The degradation prediction module is used to perform time-series prediction of the overall health based on the behavioral data sub-model to obtain degradation trend prediction results. The fault inference module is used to perform fault mode inference based on the digital twin model and obtain the fault root cause analysis results if the overall health is less than the preset health threshold. The decision adjustment module is used to allow the electromechanical equipment to continue operating if the overall health level is greater than or equal to the preset health level threshold, and to adjust the monitoring frequency for the next cycle based on the predicted deterioration trend of the overall health level.

[0007] A third aspect of this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of the above-described method for monitoring the operating status of hydropower station electromechanical equipment based on digital twins.

[0008] In a fourth aspect of this application, a computer-readable storage medium is provided, the computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described method for monitoring the operating status of hydropower station electromechanical equipment based on digital twins.

[0009] The beneficial effects of the digital twin-based hydropower station electromechanical equipment operation status monitoring method and system provided in this application are as follows: This application achieves accurate multi-dimensional state mapping of hydropower station electromechanical equipment by constructing a physical mechanism and behavioral data model. Through two state assessments (state mapping parameter detection and comprehensive health judgment), it achieves the goals of timely early warning of abnormal faults and accurate assessment of normal states. Based on virtual sensing inversion, health index calculation, degradation trend prediction, and dynamic monitoring frequency adjustment, it comprehensively covers the entire process of hydropower station electromechanical equipment monitoring, assessment, early warning, and operation and maintenance. This solves the shortcomings of existing monitoring methods, such as insufficient holographic perception and lack of in-depth extrapolation, improves the comprehensiveness and accuracy of hydropower station electromechanical equipment operation status monitoring, reduces the failure rate and operation and maintenance costs, and enables hydropower station electromechanical equipment to operate safely, stably, and efficiently. Attached Figure Description

[0010] Figure 1 A flowchart illustrating a method for monitoring the operational status of hydropower station electromechanical equipment based on digital twins, provided in an embodiment of this application; Figure 2 A structural block diagram of a hydropower station electromechanical equipment operation status monitoring system based on digital twin provided in an embodiment of this application; Figure 3 This is a schematic block diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0011] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0012] To make the purpose, technical solution, and advantages of this application clearer, the following will be described in conjunction with the appendix. Figure 1-3 The following is an explanation using specific examples.

[0013] Please refer to Figure 1 , Figure 1 This is a flowchart illustrating a method for monitoring the operational status of hydropower station electromechanical equipment based on digital twins, as provided in an embodiment of this application. The method includes: S101: Input the monitoring data of the hydropower station's electromechanical equipment into a preset digital twin model to obtain the state mapping results of the hydropower station's electromechanical equipment in multiple dimensions; wherein, the digital twin model includes a physical mechanism sub-model based on finite element analysis and a behavioral data sub-model based on temporal neural networks.

[0014] In this embodiment, the electromechanical equipment of a hydropower station refers to all equipment involved in the generation, transmission, and regulation of electrical energy in the hydropower station, including turbines (runners, main shafts), generators (stator, rotors), governors, excitation devices, top cover bolts, flange connection surfaces, etc., which are the objects of the monitoring method in this embodiment. The monitoring data is real-time data collected by physical sensors at key measuring points of the equipment, including mechanical parameters, electrical parameters, environmental parameters, and operating condition parameters, which form the basis for the input of the digital twin model. Among these, mechanical parameters include vibration, sway, and stress; electrical parameters include voltage and current; environmental parameters include temperature and oil pressure; and operating condition parameters include load and speed.

[0015] In this embodiment, the preset digital twin model is a virtual model built on a 1:1 scale based on the physical entity of the equipment. It adopts a dual-driven architecture of physical mechanism and behavioral data. Its function is to input monitoring data, simulate equipment operation, and output state mapping results, connecting the physical entity and virtual monitoring. The state mapping results are virtual data output by the digital twin model that correspond to the actual operating state of the equipment, including multi-dimensional information such as stress, temperature, and vibration of each component, realizing a virtual replication of the physical state and providing support for subsequent evaluation.

[0016] In this embodiment, the physical mechanism sub-model is constructed using the finite element method based on the equipment's physical structure and material properties. It simulates the mechanical and thermal characteristics of the equipment under the coupling effects of hydraulics, electromagnetics, and mechanics, enabling virtual inversion at non-measuring points. The behavioral data sub-model is constructed using a temporal neural network with historical monitoring data and fault data of the unit as the training set. It mines the temporal patterns of the data for health index calculation and trend prediction. Finite element analysis discretizes the complex structure of the equipment into finite elements and solves for parameters such as stress and strain of each element according to the laws of mechanics and fluid mechanics; this is the technique for constructing the physical mechanism sub-model. The temporal neural network is a deep learning model adept at processing time-series data, capturing the temporal variation patterns of the data, and is used to fit and predict the operating state of the equipment; this is the technique for constructing the behavioral data sub-model.

[0017] Specifically, the pre-designed digital twin model is a fusion model of physical mechanism and behavioral data, comprising two major sub-models. These sub-models are clearly hierarchical and tightly integrated, adaptable to the monitoring scenarios of hydropower station electromechanical equipment. The physical mechanism sub-model consists of three layers: a geometric modeling layer, which reproduces the three-dimensional structure of the electromechanical equipment and maps the dimensions and connections of core components such as the impeller and main shaft; a physical field calculation layer, including modules for mesh generation, boundary condition loading, and solving dynamic equations, simulating changes in physical fields such as equipment stress and vibration; and a result output layer, which outputs mapping results such as physical field distribution and component state parameters. The behavioral data sub-model consists of four layers: an input preprocessing layer, which performs denoising and normalization on the monitoring data; a temporal feature extraction layer, composed of three convolutional sub-layers and two pooling sub-layers, which extracts temporal and frequency domain features of the monitoring data; a feature fusion layer, which integrates multi-dimensional temporal features and eliminates redundancy; and an output layer, which outputs results such as health indices and state trend predictions. The two sub-models are connected through a data interaction interface to achieve collaborative mapping.

[0018] The training process is divided into two main sub-models, trained separately and then collaboratively debugged. First, the physical mechanism sub-model is trained: the model is initialized based on unit design parameters and historical operation and maintenance data. Boundary conditions and physical properties are calibrated using measured data from physical sensors (vibration, temperature, stress, etc.) to ensure the simulation results have an error of less than 5% compared to the measured data, thus completing the calibration. Second, the behavioral data sub-model is trained: a sample set is constructed using monitoring data and fault data from the hydropower station's electromechanical equipment over the past five years, divided into training, validation, and test sets in a 7:2:1 ratio. Using the cross-entropy loss function as the objective, the Adam optimizer (learning rate 0.001, batch size 32) is used for 100 training rounds, with validation every 10 rounds. Training is terminated if the accuracy does not improve after five consecutive validation rounds, ensuring the model adapts to the target equipment. Finally, the two sub-models are collaboratively debugged to ensure smooth data interaction and consistent mapping results, completing the digital twin model training. The model input consists of real-time monitoring data of hydropower station electromechanical equipment, including multi-dimensional data such as vibration amplitude, temperature, pressure, current, and voltage; the model output consists of multi-dimensional state mapping results of the equipment, including physical field distribution, component operating parameters, and health index.

[0019] S102: Perform the first state assessment on the state mapping results. If any state mapping parameter in the state mapping results is found to be greater than the preset safe operation threshold, then perform fault mode deduction based on the digital twin model to obtain the root cause analysis results of the fault.

[0020] In this embodiment, the first state assessment is a preliminary health judgment of the state mapping results: whether any parameter exceeds a preset safe operating threshold, such as excessive vibration, excessive stress, or excessive temperature. This is a first-level warning and serves as the entry point for determining whether to trigger fault deduction. State mapping parameters are specific indicators in the state mapping results, such as: spindle runout value, impeller blade stress, stator winding temperature, peak vibration acceleration, and pressure pulsation amplitude. These are key parameters used to determine whether the equipment is exceeding its limits.

[0021] In this embodiment, the preset safe operating threshold is a critical value set based on design specifications, manufacturing standards, historical operating data, and maintenance experience. For example, if the spindle vibration amplitude is less than or equal to 0.25 mm, the rotor blade stress is less than or equal to 80% of the design allowable stress, or the stator temperature is less than or equal to 130°C, and any parameter of the equipment exceeds this safe operating threshold, fault simulation needs to be initiated. Fault mode simulation is a process of reverse simulating possible faults, simulating fault propagation paths, and matching fault characteristics within a digital twin model. The aim is to quickly identify which fault mode best matches the current data. The root cause analysis result is the final fault cause and confidence level output after the simulation, such as rotor imbalance causing excessive vibration or rotor blade cavitation causing abnormal pressure pulsation, which can directly guide maintenance and repair.

[0022] S103: If no state mapping parameter is detected to be greater than the preset safe operation threshold, then based on the state mapping result, the behavior data sub-model is called to calculate the health index of the hydropower station's electromechanical equipment; the physical mechanism sub-model is called to perform virtual sensing inversion on the non-measuring point parts of the hydropower station's electromechanical equipment where no physical sensors are installed, and obtain the holographic state assessment result.

[0023] In this embodiment, the absence of any state mapping parameter exceeding the preset safe operating threshold after the first state assessment indicates that all state mapping parameters are within the safe threshold range, the equipment shows no obvious abnormalities, and fault simulation is not triggered. State mapping parameters include vibration, stress, and temperature. "Invocation" refers to actively starting the corresponding sub-model within the digital twin model to perform a specified function, such as calculating the health index or virtual inversion; essentially, it's like starting the sub-model to perform its tasks. The health index is a numerical value representing the overall health status of the hydropower station's electromechanical equipment, ranging from 0 to 100 points. A higher score indicates better equipment health; it is calculated based on monitoring data through a behavioral data sub-model and intuitively represents the current health level of the hydropower station's electromechanical equipment.

[0024] In this embodiment, the physical mechanism sub-model is constructed based on finite element analysis to simulate the physical characteristics of the equipment. Its function is to complete virtual sensing inversion and compensate for the shortcomings of physical sensors. Non-measuring points where no physical sensors are installed are critical areas on hydropower station electromechanical equipment where physical sensors cannot be deployed due to limitations in installation space and operating environment (e.g., inside the runner blades, deep within the stator windings). Examples include the unmeasured area at the root of the runner blades, the unmeasured location at the top cover bolt connection, and the blind zone of the generator stator end winding. Virtual sensing inversion, without physical sensors, relies on the physical mechanism sub-model and existing measured data from measuring points to calculate the state data (stress, temperature, etc.) of the non-measuring points. Essentially, it uses virtual sensors to replace physical sensors to monitor the blind zone areas.

[0025] In this embodiment, the holographic status assessment result includes complete status data of all measuring points and non-measuring points of the equipment, including real-time status values ​​of each part, comparison with historical data of the same period, and remaining life consumption rate, so as to realize the status assessment of the equipment in all dimensions and without blind spots, rather than looking at only a certain measuring point or part.

[0026] S104: Integrate the health index with the holographic status assessment results to obtain the overall health score.

[0027] In this embodiment, fusion combines the health index with the holographic status assessment results. Through a preset weighting strategy, a more comprehensive and accurate equipment health evaluation is calculated, avoiding the limitations of single-indicator assessment. The comprehensive health score, obtained after fusion, is the final health evaluation index that takes into account both the overall macroscopic state of the equipment and the microscopic state of each part. It ranges from 0 to 100 points and is more comprehensive than the health index. It reflects both the overall health level and the state of non-measuring points, serving as the basis for subsequent degradation trend prediction and fault diagnosis.

[0028] S105: Based on the behavioral data sub-model, time-series prediction of overall health is performed to obtain the prediction results of the deterioration trend.

[0029] In this embodiment, time-series prediction is based on historical time-series data of overall health (overall health records over a past period), combined with the time-series analysis capabilities of the behavioral data sub-model, to predict the trend of overall health changes over a future period, capturing the pattern of overall health changes over time. The degradation trend prediction result is the output obtained after time-series prediction, including the predicted value of overall health within a preset future time window (e.g., 1 month, 3 months), changes in health status level, and degradation rate (e.g., how many points the overall health decreases each month), intuitively representing the future health status trend of the device (whether it deteriorates slowly, remains stable, or deteriorates rapidly).

[0030] S106: If the overall health level is less than the preset health level threshold, then the fault mode is deduced based on the digital twin model to obtain the root cause analysis results.

[0031] In this embodiment, the overall health score is used to determine whether the electromechanical equipment of a hydropower station requires in-depth fault diagnosis. The preset health score threshold is a set critical value for the overall health score. If the overall health score is lower than this preset health score threshold, it indicates that the equipment has potential serious hidden dangers, and fault simulation must be initiated. The preset health score threshold is set using a cross-validation combined with an operation and maintenance requirement calibration method. For example, it is based on the weighted fusion result of three evaluation indicators: fault missed detection rate, false detection rate, and equipment operational stability. The minimum health score threshold that meets the operation and maintenance requirements of the hydropower station's electromechanical equipment is determined by comparing and statistically analyzing the simulation test results of a large number of health status monitoring cases.

[0032] In this embodiment, fault mode deduction (FMD) involves simulating the occurrence process, propagation path, and characteristic responses of different fault types within a digital twin model to identify fault symptoms. This process is then used to reverse-verify and match the fault mode that best matches the measured data. The root cause analysis result is the final conclusion output after the FMD is completed. It clearly indicates the root cause of the fault (e.g., component wear, assembly error, material fatigue, etc.) and its corresponding confidence level, providing accurate direction for maintenance.

[0033] S107: If the overall health status is greater than or equal to the preset health status threshold, the electromechanical equipment is allowed to continue operating, and the monitoring frequency for the next cycle is adjusted based on the predicted deterioration trend of the overall health status.

[0034] In this embodiment, a comprehensive health score greater than or equal to the preset health score threshold indicates that the equipment's health status meets the standards and there are no serious potential faults. Allowing the electromechanical equipment to continue operating is based on an assessment that the hydropower station's electromechanical equipment is currently in good health, poses no threat to safe operation, requires no shutdown for maintenance, and can maintain normal operation under current conditions (e.g., load, speed). The degradation trend prediction result is a prediction of the comprehensive health score's change trend over a future period using a behavioral data sub-model, including the predicted value, degradation rate (e.g., how many points it decreases by per month), and health status trend.

[0035] In this embodiment, the monitoring frequency for the next cycle is the time interval between the next monitoring of the hydropower station's electromechanical equipment's operational status. For example, if the original monitoring cycle was 1 hour / time, it is adjusted to 2 hours / time or 30 minutes / time. This is dynamically adapted based on the deterioration trend to avoid data redundancy or missed data collection. Adjusting the monitoring frequency is based on the deterioration trend prediction results, optimizing the original monitoring frequency. If the hydropower station's electromechanical equipment deteriorates slowly and its condition is stable, the monitoring frequency can be reduced (interval extended); if there are slight signs of deterioration, the monitoring frequency can be increased (interval shortened), balancing monitoring accuracy and efficiency.

[0036] As can be seen from the above, this application achieves accurate multi-dimensional state mapping of hydropower station electromechanical equipment by constructing a physical mechanism and behavioral data model. Through two state assessments (state mapping parameter detection and comprehensive health assessment), it achieves the goals of timely early warning of abnormal faults and accurate assessment of normal states. Based on virtual sensing inversion, health index calculation, degradation trend prediction, and dynamic monitoring frequency adjustment, it comprehensively covers the entire process of monitoring, assessment, early warning, and operation and maintenance of hydropower station electromechanical equipment. This solves the shortcomings of existing monitoring systems, such as insufficient holographic perception and lack of in-depth extrapolation, improves the comprehensiveness and accuracy of hydropower station electromechanical equipment operation status monitoring, reduces the failure rate and operation and maintenance costs, and enables the safe, stable, and efficient operation of hydropower station electromechanical equipment.

[0037] In one embodiment of this application, a physical mechanism sub-model is invoked to perform virtual sensing inversion on non-measuring point parts of the hydropower station's electromechanical equipment where no physical sensors are installed, to obtain a holographic state assessment result, including: Identify the non-measuring points in the electromechanical equipment of the hydropower station that are not equipped with physical sensors, and use them as the target set for virtual sensing inversion. Acquire measured data from deployed physical sensors as boundary conditions for the physical mechanism sub-model; Based on the physical mechanism sub-model, with boundary conditions as constraints, the finite element inverse problem solution method is used to perform inversion calculation on the target part set to obtain the state data of non-measuring point parts; Align and fuse state data with measured data to construct a holographic perception data layer; Based on the holographic perception data layer, holographic state assessment results are generated.

[0038] In this embodiment, the finite element inverse problem solution method is a numerical calculation method that uses known measured data from measurement points to infer the state parameters (e.g., stress, temperature) of non-measurement points. The holographic sensing data layer aggregates the state data from all measurement points and non-measurement points, forming a complete equipment operation data system to provide data support for the holographic state assessment results.

[0039] As can be seen from the above, this embodiment achieves accurate virtual sensing inversion of non-measuring points by overcoming the limitations of physical sensor deployment, filling the monitoring gaps in key areas such as the root of the runner blades and the stator end windings. Using measured data as boundary conditions, the accuracy of the inversion is improved through solving the finite element inverse problem, constructing a holographic sensing data layer that integrates measuring and non-measuring point data to achieve comprehensive perception of multiple physical fields. The holographic state assessment results include real-time values, historical comparisons, and remaining lifespan consumption rates, providing rich evidence for health assessment, improving the comprehensiveness and refinement of the assessment, and providing reliable support for fault early warning and maintenance decisions.

[0040] In one embodiment of this application, after calling the physical mechanism sub-model to perform virtual sensing inversion on non-measuring point parts of the hydropower station's electromechanical equipment where no physical sensors are installed, and obtaining the holographic state assessment result, the method further includes: Temporary calibration sensors are deployed at pre-defined calibration locations on the electromechanical equipment of hydropower stations; The measured value of the temporary calibration sensor is compared with the virtual sensing inversion value at the corresponding position in the holographic state evaluation result, and the inversion error is calculated. When the inversion error exceeds a preset accuracy threshold, a parameter correction process is performed on the physical mechanism sub-model. The parameter correction process includes: The Bayesian parameter estimation method is adopted to reduce the inversion error. The boundary conditions, material property parameters and contact stiffness coefficients in the physical mechanism sub-model are back-calibrated to obtain the corrected physical mechanism sub-model parameters. The revised physical mechanism sub-model parameters are updated in the digital twin model for use in the virtual sensing inversion calculation of the next cycle.

[0041] In this embodiment, the preset verification location is a key location used to verify the accuracy of virtual sensing inversion. It needs to be easy to deploy temporary sensors and be able to represent the core accuracy of non-measuring point inversion, such as the middle of the runner blade or the shallow area at the end of the stator. The temporary verification sensor is a sensor temporarily deployed at the preset verification location. It is only used to verify the accuracy of the virtual inversion value and does not participate in regular monitoring. After verification, it can be removed or kept for future use.

[0042] In this embodiment, the inversion error is the difference between the measured value (true value) of the temporary verification sensor and the virtual inversion value at the corresponding position in the holographic state assessment result. It is used to represent the accuracy of virtual sensing inversion; the smaller the error, the more accurate the inversion. The preset accuracy threshold is a set critical value for inversion error, used to determine whether the virtual inversion accuracy meets the standard. The preset accuracy threshold is determined through multi-dimensional statistical analysis and scenario-based verification. For example, firstly, a large number of historical virtual sensing inversion data and physical sensor measured data error statistical samples are collected to establish an inversion error distribution model; secondly, combined with the safety operation and maintenance level of the hydropower station's electromechanical equipment and the allowable on-site measurement error, an upper limit for error tolerance is set; finally, through cross-validation, the error value that minimizes the overall false alarm rate and ensures the consistency between the virtual sensing inversion result and the measured result meets the engineering early warning requirements is selected as the preset accuracy threshold.

[0043] In this embodiment, the parameter correction process involves a series of steps to adjust and calibrate the parameters of the physical mechanism sub-model when the inversion error exceeds a preset accuracy threshold, thereby reducing the inversion error and improving the inversion accuracy of the physical mechanism sub-model. The Bayesian parameter estimation method is a probability-statistical parameter calibration method that can accurately correct the parameters of the physical mechanism sub-model based on prior information (original parameters of the physical mechanism sub-model and historical inversion data) and current measured data, thus minimizing the inversion error.

[0044] In this embodiment, boundary conditions are the input constraints of the physical mechanism sub-model, such as measured vibration and temperature data, which form the basis for the inversion calculation of the physical mechanism sub-model and require calibration to match the actual operating state. Material property parameters are parameters that simulate the material properties of equipment components in the physical mechanism sub-model, such as the elastic modulus of the impeller material and the thermal conductivity of the stator material, which directly affect the inversion accuracy. Contact stiffness coefficient is a parameter that simulates the contact characteristics at the connections between various components of the equipment (e.g., the main shaft and bearing, the blade and impeller) in the physical mechanism sub-model, used to accurately simulate the mechanical transmission between components, and requires calibration to reduce errors. The corrected physical mechanism sub-model parameters are model parameters that have been calibrated by Bayesian parameter estimation to reduce inversion errors. After updating, they are used for the next cycle of virtual sensing inversion to improve inversion accuracy.

[0045] As can be seen from the above, this embodiment verifies the virtual sensing inversion accuracy through a temporary calibration sensor, establishes a comprehensive physical mechanism sub-model parameter correction mechanism, and solves the shortcomings of digital twin models such as easy drift and decreased inversion accuracy. When the inversion error exceeds the standard, the Bayesian parameter estimation method is used to accurately calibrate the physical mechanism sub-model parameters, improving the reliability of the inversion results, providing reliable support for health assessment and fault diagnosis, extending the effective service life of the model, and reducing maintenance costs.

[0046] In one embodiment of this application, a comprehensive health score is obtained by fusing a health index with a holographic status assessment result, including: The fusion weight allocation strategy is determined based on the weight adjustment factor of the current operating condition of the hydropower station's electromechanical equipment and the weight coefficient determined based on the reliability risk level of the hydropower station's electromechanical equipment. Using the health index as the first fusion dimension, the holographic status assessment results are decomposed into the second and third fusion dimensions; Among them, the second fusion dimension is the health sub-item corresponding to the virtual sensing inversion value of each non-measuring point; The third fusion dimension is the field deviation between the distribution of stress field, temperature field, vibration field and pressure field in the holographic perception data layer and the design reference value; Based on the fusion weight allocation strategy, the weighted comprehensive evaluation method is used to calculate the first fusion dimension, the second fusion dimension, and the third fusion dimension to obtain the comprehensive health score.

[0047] In this embodiment, the weight adjustment factor for operating conditions is a coefficient set based on the current operating status of the hydropower station's electromechanical equipment, such as steady state, start-up / shutdown, and load adjustment. This coefficient is used to adjust the weights of each fusion dimension to adapt to the health assessment requirements under different operating conditions. The reliability risk level is a classification based on the importance of equipment components and the scope of failure impact, used to determine the weight coefficients for each fusion dimension. The fusion weight allocation strategy is based on the weight adjustment factor and weight coefficients to formulate a weight allocation scheme for each fusion dimension, clarifying the proportion of each dimension in the overall health calculation.

[0048] In this embodiment, the first fusion dimension represents the overall macroscopic health status of the device, serving as the foundational dimension for comprehensive health calculation. The second fusion dimension is the dimension derived from the decomposition of the holographic status assessment result, corresponding to the health sub-items of the virtual sensing inversion values ​​for each non-measuring point. For example, the health score corresponding to the stress and temperature of each non-measuring point represents the microscopic state of the non-measuring point. The third fusion dimension, also derived from the decomposition of the holographic status assessment result, refers to the degree of deviation between the actual distribution of the device's stress field, temperature field, vibration field, and pressure field in the holographic sensing data layer and the design baseline values. The smaller the deviation, the better the health status.

[0049] In this embodiment, the health sub-item is a virtual inversion value for each non-measuring point, and a single health score given according to a standard threshold, which is a component of the second fusion dimension. Field deviation is the percentage difference between the actual field distribution of the equipment and the design baseline value, used to represent the degree of conformity between the equipment's operating status and the design standard, and is an indicator of the third fusion dimension. The weighted comprehensive evaluation method assigns corresponding weights to the three fusion dimensions according to a fusion weight allocation strategy, and calculates the comprehensive health score through weighted calculation, taking into account the influence of each dimension.

[0050] As can be seen from the above, this embodiment achieves accurate quantification of the comprehensive health status of hydropower station electromechanical equipment by integrating the health index with the holographic status assessment results from multiple dimensions. The decomposition and fusion dimensions take into account the overall health of the hydropower station's electromechanical equipment, the status of non-measuring points, and deviations in the physical field distribution, comprehensively representing the operating status. The fusion weights are determined based on the operating condition weight adjustment factor and the reliability risk level, making the calculation more closely aligned with actual operating scenarios and avoiding assessment biases caused by fixed weights. The weighted comprehensive evaluation makes the fusion results scientific and reasonable, accurately representing the health status, providing a reliable basis for fault early warning and maintenance strategy adjustments, and improving the scientific rigor and relevance of health assessments.

[0051] In one embodiment of this application, the health index is fused with the holographic status assessment result to obtain a comprehensive health score, and the method further includes: Obtain the twin confidence index; When the twin confidence index is less than the preset first confidence threshold, in the fusion weight allocation strategy, the weight of the first fusion dimension is reduced, and the weight of the non-measurement point parts with high confidence in the second fusion dimension is increased. When the twin confidence index is greater than or equal to the first confidence threshold, the fusion weight allocation strategy is maintained; Based on the adjusted fusion weight allocation strategy, the weighted comprehensive evaluation method is used to calculate the first fusion dimension, the second fusion dimension, and the third fusion dimension to obtain the comprehensive health score.

[0052] In this embodiment, the twin confidence index is a comprehensive index representing the degree to which the digital twin model matches the actual operating state of the device. It includes three sub-indices: data consistency, model structure, and operating condition coverage, and serves as the basis for adjusting the fusion weights. The preset first confidence threshold is a set twin confidence critical value used to determine the credibility of the digital twin model and decide whether to adjust the fusion weight allocation strategy. The fusion weight allocation strategy is a weight allocation scheme for the first, second, and third fusion dimensions determined by combining the operating condition weight adjustment factor and the device reliability risk level.

[0053] The method for determining the first confidence threshold is as follows: First, collect historical operating data, twin confidence assessment records, and model calibration data of the target hydropower station's digital twin model, and statistically analyze core indicators such as operational stability and failure rate corresponding to different twin confidence levels; second, combine the accuracy requirements of the digital twin model, the calculation logic of the twin confidence index, and industry standards for similar hydropower stations to determine a reasonable range for confidence assessment; finally, based on the principle of effectively distinguishing the reliability levels of the twin model while avoiding excessive stringency or leniency, take a reasonable critical value for confidence assessment as the first confidence threshold.

[0054] In this embodiment, the first fusion dimension, namely the health index, is mainly calculated based on the behavioral data sub-model of the digital twin model. The model's credibility (twin confidence level) affects its data reliability. The second fusion dimension is the dimension decomposed from the holographic state assessment result, corresponding to the health sub-items of the virtual sensing inversion values ​​of each non-measurement point. Each health sub-item corresponds to a confidence level (representing inversion accuracy). Non-measurement points with high confidence levels are those with high virtual sensing inversion accuracy and high credibility in the second fusion dimension, such as those with small inversion errors and qualified verification. The corresponding health sub-item weights need to be increased. The adjusted fusion weight allocation strategy is a new strategy obtained by adjusting the original fusion weights based on the comparison results of the twin confidence index and the first confidence threshold, thereby improving the accuracy of the comprehensive health score calculation. The weighted comprehensive evaluation method is a method that assigns corresponding weights to the three fusion dimensions based on the adjusted fusion weight allocation strategy, and calculates the comprehensive health score through weighted calculation. This method adapts to the model's credibility and improves the assessment accuracy.

[0055] As can be seen from the above, this embodiment integrates the twin confidence index into the comprehensive health calculation, achieving dynamic adjustment of the fusion weights and solving the problems of fixed weights and neglect of model reliability. When the twin confidence is low, the weight of the health index is reduced and the weight of high-confidence non-measurement points is increased to avoid evaluation bias caused by unreliable models; when the confidence is high, the original weights are maintained to ensure evaluation efficiency. Dynamic weight adjustment makes the comprehensive health calculation more closely match the actual reliability level of the model, improving the accuracy and credibility of the evaluation results, providing more reliable support for fault early warning and operation and maintenance decisions, and further improving the equipment health assessment system.

[0056] In one embodiment of this application, a fusion weight allocation strategy is determined based on a weight adjustment factor for the current operating condition of the hydropower station's electromechanical equipment and a weight coefficient determined based on the reliability risk level of the hydropower station's electromechanical equipment, including: Obtain the current operating status of the electromechanical equipment in the hydropower station; Based on the preset working condition-weight adjustment factor mapping table, determine the weight adjustment factor corresponding to the current operating condition; Obtain the reliability risk level of electromechanical equipment in hydropower stations; The weight coefficients of the monitoring indicators corresponding to the preset components in the electromechanical equipment of the hydropower station are determined according to the reliability risk level. Among them, the weight coefficients of the monitoring indicators corresponding to the preset components with higher risk levels are larger. Multiplying the weight adjustment factor by the weight coefficient yields the fusion weight allocation strategy.

[0057] In this embodiment, the operating condition refers to the current actual operating state of the hydropower station's electromechanical equipment, such as steady-state operation, start-up and shutdown, load adjustment, and low-load standby, which directly affects the weight allocation of each monitoring indicator. The weight adjustment factor is a coefficient set according to the equipment's operating condition, used to adjust the weight coefficients to adapt to the health integration requirements under different operating conditions, and is determined by the operating condition-weight adjustment factor mapping table. The operating condition-weight adjustment factor mapping table is a pre-defined correspondence table that clearly defines the matching relationship between different operating conditions and corresponding weight adjustment factors; for example, steady-state operation corresponds to 0.95, and start-up and shutdown correspond to 1.1, eliminating the need for repeated calculations and allowing direct querying and retrieval. The reliability risk level is a classification (high, medium, low) based on the importance of each pre-defined component of the equipment and the scope of fault impact (e.g., whether the fault leads to shutdown or affects power generation), used to determine the weight coefficients.

[0058] In this embodiment, the weighting coefficient is the weight percentage of the corresponding preset component monitoring indicators set according to the reliability risk level. The higher the risk level, the larger the weighting coefficient, highlighting the impact of high-risk components. Preset components are selected core components crucial to the safe operation of the equipment, such as impellers, stators, guide bearings, and spindles. Their monitoring indicators are key bases for health fusion. Monitoring indicators are specific parameters used to assess the health status of preset components, such as stress, temperature, vibration, and sway, and are the corresponding objects of the weighting coefficient. The fusion weight allocation strategy is the final determined weight scheme for each monitoring indicator (corresponding to each fusion dimension), obtained by multiplying the weight adjustment factor and the weighting coefficient, and used for the subsequent weighted calculation of the overall health score.

[0059] As can be seen from the above, this embodiment solves the problems of blind and unfounded weight setting by clearly defining the specific method for determining the fusion weight allocation strategy. Through the operating condition-weight adjustment factor mapping table, the weights are dynamically adapted to the current operating condition, improving the accuracy of assessments under different operating conditions. By combining weight allocation with equipment reliability risk levels, key component monitoring points are highlighted, enhancing the relevance of the assessment. Multiplying the weight adjustment factor by the weight coefficient achieves a dual consideration of operating condition adaptation and risk control, making the comprehensive health calculation more closely aligned with the actual operating status and maintenance needs of the equipment, providing precise weight support for health assessment and maintenance decisions.

[0060] In one embodiment of this application, if the overall health score is less than a preset health score threshold, a fault mode deduction is performed based on a digital twin model, and the root cause analysis results are output, including: If the overall health score is less than the preset health score threshold, extract the fault symptom feature vector from the current monitoring data; The fault symptom feature vectors are matched and reasoned with the fault patterns in the preset fault knowledge graph to obtain a list of candidate fault causes and their corresponding initial confidence levels. Candidate fault causes are input into the physical mechanism sub-model for simulation verification. The initial confidence level is corrected based on the degree of agreement between the simulation results and the measured data, and the final root cause analysis results are output.

[0061] In this embodiment, fault mode deduction is based on a digital twin model, simulating the fault occurrence process and analyzing the fault propagation path to locate the fault cause. The fault symptom feature vector is a set of parameters extracted from current monitoring data that can represent fault signs, such as abnormal vibration and excessive temperature, and serves as the basis for fault matching. The fault knowledge graph is a database storing the correspondence between various fault types, symptoms, and root causes, used for quickly matching candidate fault causes. The candidate fault cause list is a set of possible fault causes matched from the knowledge graph based on the fault symptom feature vector, including various potential fault types.

[0062] In this embodiment, the initial confidence level is the credibility of the candidate fault causes, set initially and subsequently adjusted based on simulation results. The root cause analysis result is the final determined root cause of the fault, such as component wear, assembly deviation, etc., including specific rectification directions to guide maintenance.

[0063] As can be seen from the above, this embodiment achieves accurate root cause localization of faults through fault knowledge graphs and physical mechanism simulations, solving the problems of fault diagnosis relying solely on data matching, lacking physical logic verification, and inaccurate root cause localization. By matching fault symptoms with the knowledge graph, candidate faults are quickly screened, improving diagnostic efficiency; candidate faults are input into the physical mechanism sub-model for simulation verification, and the confidence level is adjusted based on the consistency between simulation and measured data, making the root cause analysis reliable. It balances the efficiency of data-driven approaches with the rigor of physical mechanisms, accurately locating the root cause of faults, providing a clear direction for rapid fault handling, shortening processing time, and reducing economic losses caused by faults.

[0064] In one embodiment of this application, candidate fault causes are input into a physical mechanism sub-model for simulation verification. The initial confidence level is corrected based on the degree of agreement between the simulation results and measured data, and the final fault root cause analysis results are output, including: For each candidate fault cause in the candidate fault cause list, determine the corresponding physical mechanism sub-model parameter adjustment scheme to obtain the adjusted physical mechanism sub-model parameters; The adjusted physical mechanism sub-model parameters are input into the physical mechanism sub-model for simulation to obtain simulation feature data under candidate fault causes; Calculate the feature similarity between the simulated feature data and the measured feature data of the corresponding measurement points in the monitoring data; The initial confidence score is weighted and corrected based on feature similarity to obtain the corrected final confidence score; The candidate fault cause with the highest confidence after correction is taken as the final root cause analysis result.

[0065] In this embodiment, the candidate fault cause list is a set of potential faults that may lead to equipment health exceeding the standard, obtained after fault spectrum matching, such as guide bearing wear, blade cavitation, and excessive spindle runout. These are the objects of simulation verification. The physical mechanism sub-model parameter adjustment scheme is a scheme for adjusting the model parameters for each candidate fault cause to simulate the fault state. For example, to simulate guide bearing wear, parameters such as bearing clearance and contact stiffness need to be adjusted to allow the physical mechanism sub-model to reproduce the fault condition. The adjusted physical mechanism sub-model parameters are obtained by adjusting the original basic parameters (such as material properties and boundary conditions) of the physical mechanism sub-model according to the parameter adjustment scheme, and are used to simulate equipment operation under fault conditions.

[0066] In this embodiment, simulation operation involves inputting adjusted model parameters into the physical mechanism sub-model to simulate the equipment's operation under the influence of the candidate fault, thus recreating the dynamic characteristics of the fault state. Simulation feature data refers to the parameter data output after simulation operation that represents the characteristics of the candidate fault, such as vibration frequency, stress value, and temperature changes under the fault state, corresponding to measured data. Monitoring data refers to the real-time data (i.e., measured data) collected by deployed physical sensors during equipment operation, serving as the benchmark for comparing simulation results and judging the rationality of the fault. Measured feature data consists of parameters extracted from the monitoring data that correspond to the simulation feature data, used to calculate feature similarity. Feature similarity is the degree of agreement between the simulation feature data and the measured feature data. Weighted correction uses feature similarity as a weight to adjust the initial confidence of the candidate fault (higher similarity results in higher corrected confidence), improving the accuracy of root cause identification. Initial confidence is the initial credibility assigned to each candidate fault cause after spectrum matching, without simulation verification and calibration. The final confidence score, obtained after feature similarity weighting correction, is the final credibility of the candidate fault cause and serves as the basis for determining the root cause of the fault. The final root cause analysis result selects the fault with the highest final confidence score from all candidate fault causes as the root cause of the equipment health exceeding the standard, and includes the confidence score, simulation verification basis, and maintenance recommendations.

[0067] As can be seen from the above, this embodiment further improves the accuracy and reliability of root cause localization by refining the simulation verification and confidence correction process of root cause analysis. Specific parameter adjustment schemes are developed for different candidate faults to ensure that the simulation scenario matches the actual fault, improving the realism of the simulation feature data. The similarity between simulation and measured data features is calculated to scientifically correct the confidence level, avoiding misjudgments caused by initial confidence level deviations. The root cause of the fault with the highest confidence level is output, clarifying the essence of the fault, providing accurate guidance for fault handling, avoiding blind repairs, reducing maintenance costs, improving the reproducibility and scientific rigor of fault diagnosis, and meeting the needs of refined fault handling.

[0068] In one embodiment of this application, if the overall health level is greater than or equal to a preset health level threshold, the electromechanical equipment is allowed to continue operating, and the monitoring frequency for the next cycle is adjusted based on the predicted degradation trend of the overall health level, including: Obtain the comprehensive health status time series of the hydropower station's electromechanical equipment under the current operating conditions. The comprehensive health status time series includes the health status records for each cycle within the past preset time window. Based on the comprehensive health time series, a deterioration trend prediction model is used to predict the health change trend within a preset time window in the future, and the health decay rate is calculated. Based on the current value of the overall health status and the rate of health status decay, the first monitoring frequency of the hydropower station's electromechanical equipment is determined. Based on the current operating conditions of the hydropower station's electromechanical equipment, the monitoring bias of the hydropower station's electromechanical equipment is determined. The monitoring bias may include structural fatigue bias, hydraulic stability bias, or electrical aging bias. Based on the monitoring bias, the first monitoring frequency is adjusted to obtain the second monitoring frequency; Based on the second monitoring frequency, the monitoring strategy for the next cycle of the hydropower station's electromechanical equipment is determined.

[0069] In this embodiment, the comprehensive health time series is a collection of comprehensive health data recorded periodically within a preset time window, such as health records once a day for the past 7 days. This serves as the foundational data for trend prediction. The preset time window is a pre-defined time range for collecting time series data or predicting trends, such as the past 7 days or the next 3 days, and can be adjusted according to the operation and maintenance needs of the hydropower station's electromechanical equipment. The degradation trend prediction model is a model used to predict the trend of equipment health changes, capable of calculating the health decay rate based on the time series.

[0070] Specifically, the degradation trend prediction model adopts a time-series prediction network structure, which is divided into four layers: the input layer is responsible for receiving comprehensive health time-series data and standardizing the data format; the time-series feature extraction layer uses gated recurrent units to capture the time-series change patterns of health; the feature fusion layer integrates health data and operating parameters to eliminate data redundancy; and the output layer outputs the health degradation trend and corresponding monitoring frequency adjustment suggestions. Each layer works collaboratively through parameter transmission to improve prediction accuracy.

[0071] The training process is as follows: First, collect the comprehensive health status time-series data of the hydropower station's electromechanical equipment over the past 3-5 years, and divide it into training, validation, and test sets in an 8:1:1 ratio. With the goal of minimizing the health status prediction error, use the Adam optimizer (learning rate 0.001) for iterative training until the validation set error is less than 5%. During training, adjust network parameters in real time to ensure the prediction results are consistent with the actual degradation trend. The inputs are the comprehensive health status time-series sequence and operating parameters; the output is the health status degradation trend curve and corresponding monitoring frequency adjustment suggestions. After training, the system can directly receive comprehensive health status data and output degradation trends and monitoring frequency adjustment schemes.

[0072] In this embodiment, the health decay rate is the rate of decrease in the overall health score per unit time, such as a decrease of 0.2 points per day or 0.5 points per month, indicating the speed at which the health of the hydropower station's electromechanical equipment deteriorates. The first monitoring frequency is the preliminary monitoring frequency for the next cycle determined based on the current overall health score and the health decay rate, such as collecting data once every hour.

[0073] In this embodiment, the operating condition characteristics refer to the current specific operating state of the hydropower station's electromechanical equipment. For example, high-load steady-state operation, low-load standby, and load fluctuation adjustment determine the selection of monitoring priorities. The monitoring priorities are determined based on the operating conditions and are categorized into three types: structural fatigue, hydraulic stability, and electrical aging. The second monitoring frequency is the final monitoring frequency obtained by adjusting the first monitoring frequency based on the monitoring priorities, thus better aligning with the current operating needs of the hydropower station's electromechanical equipment. The monitoring strategy is a complete monitoring plan based on the second monitoring frequency, including monitoring time intervals, monitoring points, and data acquisition methods, used to guide equipment monitoring in the next cycle.

[0074] As can be seen from the above, this embodiment breaks through the limitations of fixed monitoring frequencies and optimizes the allocation of operation and maintenance resources by dynamically adjusting the monitoring frequency. By comprehensively predicting the degradation trend and decay rate based on the health time series, the initial monitoring frequency is determined according to the current health status, achieving monitoring adaptation to the equipment's health status. Monitoring priorities are determined based on operating conditions, and the monitoring frequency and channels are adjusted accordingly, improving the targeting and efficiency of monitoring. This avoids the waste of resources from over-monitoring when the equipment is in good health, and prevents the omission of faults due to insufficient monitoring when the equipment is in poor health, achieving a reasonable allocation of operation and maintenance resources, reducing operation and maintenance costs, and improving the timeliness and targeting of monitoring.

[0075] In one embodiment of this application, adjusting the first monitoring frequency based on monitoring bias to obtain a second monitoring frequency includes: If the monitoring focuses on structural fatigue, then based on the first step, the acquisition frequency of structural vibration parameters and stress parameters is increased, the acquisition frequency of vibration characteristic values ​​in the first monitoring frequency is increased by a first preset multiple, and strain sensor data acquisition channels are added to obtain the second monitoring frequency; If the monitoring bias is towards hydraulic stability, the sampling frequency of pressure pulsation parameters and swing parameters is increased based on the second step size. The sampling rate of the original waveform of pressure pulsation in the first monitoring frequency is increased by a second preset multiple, and the cavitation acoustic emission monitoring channel is enabled to obtain the second monitoring frequency. If the monitoring bias is towards electrical aging, then based on the third step, the acquisition frequency of partial discharge parameters and insulation parameters is increased, the partial discharge monitoring cycle in the first monitoring frequency is shortened to the third preset ratio, and a stator end vibration monitoring channel is added to obtain the second monitoring frequency.

[0076] In this embodiment, structural fatigue bias is a type of monitoring bias, focusing on the fatigue loss of structural components of equipment, such as impellers and main shafts, with core monitoring of structural vibration and stress parameters. The first step is to adjust the frequency range set for structural fatigue bias, such as the amount of time interval reduction or frequency increase, to uniformly adjust the acquisition frequency of corresponding parameters. Structural vibration parameters are parameters representing the vibration state of structural components of hydropower station electromechanical equipment, such as vibration frequency, amplitude, and peak value, and are the core parameters for structural fatigue monitoring. Stress parameters are parameters representing the stress state of structural components of hydropower station electromechanical equipment, such as stress value and strain rate, used to determine the degree of structural fatigue. The first preset multiple is the multiple by which the vibration characteristic value acquisition frequency is increased for structural fatigue bias, i.e., new frequency = first monitoring frequency × first preset multiple. The strain sensor data acquisition channel is used to acquire data from strain sensors (monitoring stress parameters); adding channels can improve the comprehensiveness and accuracy of stress data acquisition.

[0077] Specifically, the first step length is determined comprehensively based on the structural fatigue characteristics of hydropower station electromechanical equipment, historical fault data, and operation and maintenance safety requirements. The specific method is as follows: First, collect the material fatigue strength, design life, and historical structural fatigue fault records of the target electromechanical equipment structural components (spindle, impeller, flange, etc.), and statistically analyze the correlation between structural fatigue faults and the acquisition frequency of vibration and stress parameters. Second, combined with the structural fatigue early warning sensitivity requirements, determine the minimum acquisition frequency increase that can timely capture structural fatigue deterioration signals, as the initial value of the first step length. Finally, through field test verification and combined with the experience of structural fatigue monitoring of similar units, calibrate the initial value to ensure that after the acquisition frequency of vibration and stress parameters corresponding to the first step length is increased, it can accurately capture the trend of structural fatigue deterioration without wasting operation and maintenance resources due to excessive frequency increase. Finally, determine the first step length suitable for the target hydropower station.

[0078] The first preset multiple is determined comprehensively based on the structural fatigue monitoring requirements of hydropower station electromechanical equipment, the accuracy of vibration characteristic value acquisition, and the constraints of operation and maintenance costs. The specific method is as follows: First, historical structural fatigue failure data of the target equipment structural components (spindle, impeller, flange, etc.) are collected, and the correlation between the vibration characteristic value acquisition frequency and the structural fatigue signal capture accuracy is analyzed to determine the minimum acquisition frequency increase multiple that can accurately identify early structural fatigue deterioration, which is used as the initial value of the first preset multiple. Second, the initial value is initially calibrated by combining the monitoring accuracy requirements of structural vibration parameters and stress parameters, as well as the load-bearing capacity of the strain sensor data acquisition channel. Finally, through field test verification and combined with the experience of structural fatigue monitoring of similar units, the monitoring accuracy and operation and maintenance costs are balanced to avoid data redundancy and increased costs due to excessively high multiples. Ultimately, the first preset multiple suitable for the target hydropower station is determined so that it can effectively increase the vibration characteristic value acquisition frequency and accurately capture structural fatigue deterioration signals.

[0079] In this embodiment, hydraulic stability bias is a type of monitoring bias that focuses on the stability of the hydraulic operation of the equipment, with core monitoring parameters including pressure pulsation and swing. The second step is the frequency adjustment range set for the hydraulic stability bias, adapted to the hydraulic parameter monitoring requirements. Pressure pulsation parameters represent the fluctuations in water flow pressure inside the hydropower station's electromechanical equipment, such as pressure pulsation amplitude and frequency, and are key indicators for judging hydraulic stability. Swing parameters represent the swing amplitude of rotating components (e.g., main shaft, impeller) in the hydropower station's electromechanical equipment, used to assess equipment vibration caused by hydraulic imbalance. The second preset multiplier is the factor that increases the sampling rate of the original pressure pulsation waveform set for the hydraulic stability bias, improving the accuracy of hydraulic parameter monitoring. The cavitation acoustic emission monitoring channel is used to monitor the acoustic emission signals of cavitation phenomena (bubble generation and collapse) in the water flow inside the equipment, and its activation can supplement the hydraulic stability monitoring data.

[0080] Specifically, the second step length is determined comprehensively based on the hydraulic stability characteristics of the hydropower station's electromechanical equipment, historical hydraulic fault data, and the accuracy requirements for hydraulic parameter monitoring. The specific method is as follows: First, the design parameters and historical hydraulic fault records (such as cavitation and excessive pressure pulsation) of the target electromechanical equipment's hydraulic system (runner, guide vanes, water intake pipes, etc.) are collected. The correlation between hydraulic stability anomalies and the acquisition frequency of pressure pulsation and swing parameters is statistically analyzed to determine the minimum acquisition frequency increase required to capture hydraulic anomaly signals. Second, combining the hydraulic stability early warning requirements and the accuracy requirements for analyzing the original waveform of pressure pulsation, the initial value of the second step length is determined so that the initial value can meet the early identification of hydraulic anomalies such as cavitation and pressure pulsation. Finally, through on-site hydraulic condition tests and combined with the hydraulic stability monitoring experience of similar units, the initial value is calibrated to avoid data redundancy and increased operation and maintenance costs caused by excessively high acquisition frequency increases. Ultimately, a second step length suitable for the target hydropower station is determined so that the corresponding pressure pulsation and swing parameter acquisition frequency can accurately capture the trend of hydraulic stability deterioration after increase.

[0081] The second preset multiplier is determined comprehensively based on the hydraulic stability monitoring requirements of hydropower station electromechanical equipment, the accuracy of pressure pulsation signal identification, and the constraints of operation and maintenance costs. The specific method is as follows: First, historical pressure pulsation and swing data of the target equipment are collected. The identification requirements of pressure pulsation signals under different operating conditions are statistically analyzed to determine the minimum sampling frequency enhancement multiplier that can accurately capture pressure pulsation anomalies, which serves as an initial reference value. Combining hydraulic stability monitoring standards and referring to industry specifications for pressure pulsation monitoring in similar hydropower stations, the initial multiplier is adjusted to effectively identify abnormal signals such as pressure pulsation and cavitation. Through on-site testing and calibration, the identification accuracy of pressure pulsation data under different multipliers is compared. Multipliers that are too high or too low, resulting in data redundancy and identification omissions, are eliminated. Finally, the second preset multiplier suitable for the current hydropower station scenario is determined, which ensures the monitoring accuracy of core parameters such as pressure pulsation and swing, avoids oversampling and increased operation and maintenance costs, and matches the first preset multiplier and operating condition requirements, thus ensuring coordination with the overall monitoring process.

[0082] In this embodiment, electrical aging bias is a type of monitoring bias that focuses on the aging degree of electrical components of the equipment, such as the stator and windings, with core monitoring of partial discharge and insulation parameters. The third step is the frequency adjustment range set for electrical aging bias to adapt to the monitoring needs of electrical parameters. Partial discharge parameters represent the discharge parameters of insulation defects in electrical components (e.g., discharge quantity, discharge frequency), and are indicators for judging electrical aging. Insulation parameters represent the insulation performance parameters of electrical components, such as insulation resistance and dielectric loss, and are used to assess the aging degree of electrical components. The third preset ratio is the percentage reduction in the partial discharge monitoring cycle set for electrical aging bias (e.g., 20%, 30%), i.e., new cycle = first monitoring cycle × (1 - third preset ratio). The stator end vibration monitoring channel is used to collect stator end vibration data. Stator end vibration is related to electrical aging, and adding channels can improve the comprehensiveness of electrical aging monitoring.

[0083] Specifically, the third step length is determined comprehensively based on the electrical aging characteristics of hydropower station electromechanical equipment, historical electrical fault data, and the accuracy requirements for monitoring electrical parameters. The specific method is as follows: First, the design parameters and historical electrical aging fault records (such as excessive partial discharge and decreased insulation performance) of the target electromechanical equipment's electrical components (stator, windings, insulation structure, etc.) are collected. The correlation between electrical aging anomalies and the acquisition frequency of partial discharge and insulation parameters is statistically analyzed to determine the minimum acquisition frequency increase required to capture electrical aging signals, which serves as the initial value for the third step length. Second, the initial value is preliminarily calibrated by combining the electrical aging early warning requirements and the monitoring accuracy standards for partial discharge and insulation parameters, and referring to the electrical aging monitoring experience of similar units. Finally, through on-site electrical condition tests, the monitoring accuracy and operation and maintenance costs are balanced to avoid data redundancy and increased operation and maintenance burden due to excessively high step lengths. Ultimately, a third step length suitable for the target hydropower station is determined, ensuring that the corresponding partial discharge and insulation parameter acquisition frequencies can accurately capture the trend of electrical aging degradation, providing reliable support for monitoring frequency adjustments.

[0084] The third preset ratio (corresponding to the frequency adjustment ratio for electrical aging related monitoring) is determined as follows: First, collect relevant data on the electrical system of the target hydropower station's electromechanical equipment, including design parameters of electrical components (e.g., stator, windings) and historical electrical aging fault records. Analyze the relationship between partial discharge, insulation performance, and monitoring frequency to clarify the range of monitoring frequency increase required for effective capture of electrical aging signals. Second, based on the accuracy requirements of electrical aging monitoring and referring to industry standards for monitoring frequency adjustment of similar electrical equipment, initially determine the initial value of the third preset ratio. Subsequently, through on-site testing, compare the monitoring accuracy of electrical parameters (e.g., partial discharge detection accuracy, insulation performance monitoring effect) under different ratios to select a ratio that ensures effective capture of electrical aging signals without causing data redundancy or increased maintenance costs. Finally, combining the previously trained model data and actual maintenance experience, calibrate the initial ratio to ultimately determine the third preset ratio, ensuring it aligns with the overall monitoring process and health assessment, meeting the accuracy requirements of electrical aging monitoring while avoiding resource waste from over-monitoring.

[0085] As can be seen from the above, this embodiment further enhances the pertinence and practicality of dynamic monitoring strategies by formulating differentiated monitoring frequency adjustment schemes based on different monitoring emphases. For the three core emphases of structural fatigue, hydraulic stability, and electrical aging, the acquisition frequency of corresponding key parameters is increased and monitoring channels are added to achieve focused and precise monitoring. For example, the structural fatigue emphase focuses on increasing the acquisition frequency of vibration and stress parameters, while the hydraulic stability emphase focuses on monitoring pressure pulsation and swing parameters. This concentrates monitoring resources on key aspects, improves the pertinence and effectiveness of monitoring data, provides precise support for fault early warning and health assessment, and optimizes the allocation of operation and maintenance resources.

[0086] In one embodiment of this application, a method for monitoring the operating status of electromechanical equipment in a hydropower station based on digital twins further includes: The twin confidence index of the digital twin model is calculated based on logical rules; the twin confidence index includes a data consistency confidence index, a model structure confidence index, and a working condition coverage confidence index. The data consistency confidence sub-index is calculated based on the multi-source error propagation analysis strategy of the measured values ​​of physical sensors and the simulation output values ​​of the corresponding positions of the digital twin model. The model structure confidence sub-index is determined by the error decay curve constructed based on the historical inversion error of the temporary calibration sensor. When the inversion error of the most recent N calibrations shows a continuous increasing trend, the model structure confidence sub-index is determined to decrease. The operating condition coverage confidence sub-index is determined based on the sample coverage density in the historical model correction data under the current operating conditions. For marginal operating conditions with a sample coverage density less than a preset density threshold, an operating condition coverage confidence sub-index less than a preset benchmark value is assigned. The data consistency confidence sub-index, model structure confidence sub-index, and operating condition coverage confidence sub-index are weighted and fused to obtain the twin confidence index.

[0087] In this embodiment, the logical rules are pre-defined judgment rules used to calculate the twin confidence index, such as error calculation rules and sample coverage judgment rules, making the calculation process standardized and repeatable. The twin confidence index is a comprehensive index used to measure the degree of fit between the digital twin model and the actual operating state of the hydropower station's electromechanical equipment. It includes three sub-indices and is the core evaluation indicator of model reliability. The data consistency confidence sub-index is one of the sub-items of the twin confidence index, representing the consistency between the measured values ​​of the physical sensors and the simulation output values ​​of the corresponding positions in the digital twin model. It is calculated through a multi-source error propagation analysis strategy. The model structure confidence sub-index is one of the sub-items of the twin confidence index, representing the rationality of the structural parameters of the digital twin model (the core of which is the physical mechanism sub-model). It is determined based on the error decay curve constructed from the historical inversion error history. The operating condition coverage confidence sub-index is one of the sub-items of the twin confidence index, representing the degree of adaptation of the digital twin model to the current operating conditions of the equipment. It is determined based on the sample coverage density of the current operating conditions in the historical model correction data.

[0088] In this embodiment, the multi-source error propagation analysis strategy is a method for calculating the data consistency confidence sub-index. It quantifies the consistency between the measured values ​​from physical sensors and the simulated values ​​from the model by analyzing the sources of error (e.g., sensor error, model parameter error) and the propagation patterns of the error. The historical inversion error record of the temporary calibration sensor is a record of the error between the measured values ​​and the virtual inversion values ​​collected by the temporary calibration sensor (e.g., error data from the past N calibrations), forming the basis for constructing the error decay curve. The error decay curve, plotted based on the historical inversion error record, represents the trend of inversion error over time or the number of calibrations (e.g., decay, stabilization, increase), and is used to determine the magnitude of the model structure confidence sub-index. The most recent N calibrations are a preset number of calibrations used to analyze the trend of inversion error; the stability of the model structure is determined by the number of consecutive N error changes.

[0089] In this embodiment, historical model correction data is a collection of relevant data recorded during past parameter corrections of the digital twin model, including operating conditions, correction parameters, and inversion errors. Sample coverage density represents the frequency and completeness of the current operating condition in the historical model correction data, indicating the model's adaptation experience to the current condition (higher density indicates better adaptation). The preset density threshold is a set critical value for sample coverage density, used to determine whether the current operating condition is an edge condition; a density less than the preset density threshold indicates an edge condition. Edge conditions are operating conditions with low frequency and incomplete data coverage in the historical model correction data, such as abnormally low unit load or sudden load fluctuations. The model's simulation accuracy for these conditions is low. The preset benchmark value is a pre-set benchmark for the operating condition coverage confidence sub-index; the operating condition coverage confidence sub-index for edge conditions must be less than this benchmark value.

[0090] The preset density threshold is determined by statistically analyzing the operating condition sample data of the target hydropower station over the past 3-5 years. First, the distribution patterns of all historical operating conditions are sorted out, the sample coverage density under different operating conditions is statistically analyzed, the mean and fluctuation range of the sample coverage density under normal operating conditions are calculated, and the lower limit of the sample coverage density under normal operating conditions is taken as the initial threshold. Then, based on the hydropower station's operation and maintenance needs, data acquisition costs, and model calibration accuracy requirements, the final preset density threshold is calibrated to ensure that the preset density threshold can effectively distinguish between marginal operating conditions and normal operating conditions.

[0091] In this embodiment, the weighted fusion is to calculate the weighted index of the three sub-indices using preset weights to obtain the final twin confidence index. The weights are set according to the importance of each sub-indice, for example, the weights of the three are all 1 / 3, or the weight of data consistency is higher.

[0092] As can be seen from the above, this embodiment achieves a quantitative assessment of the reliability of digital twin models by constructing a twin confidence index, solving the problems of lack of reliability assessment and inability to detect deviations in a timely manner. Through three sub-indices, it comprehensively covers the core dimensions of data consistency, model structure, and operational condition coverage, providing a holistic assessment of model reliability. Combined with methods such as error propagation analysis, error decay curves, and sample coverage density, the scientific calculation of the sub-indices is made achievable. The twin confidence index provides an important basis for health fusion and model optimization, promptly identifying problems such as model drift and data inconsistency, ensuring the long-term stable operation of the model, and improving the reliability and accuracy of equipment monitoring and assessment.

[0093] Corresponding to the above embodiment of the method for monitoring the operating status of hydropower station electromechanical equipment based on digital twins, Figure 2 This is a structural block diagram of a hydropower station electromechanical equipment operation status monitoring system based on digital twins, provided as an embodiment of this application. For ease of explanation, only the parts relevant to the embodiment of this application are shown. References Figure 2The hydropower station electromechanical equipment operation status monitoring system 20 based on digital twins includes: a twin mapping module 21, a safety assessment module 22, a deep assessment module 23, a comprehensive health module 24, a deterioration prediction module 25, a fault inference module 26, and a decision adjustment module 27.

[0094] The twin mapping module 21 is used to input the monitoring data of the hydropower station's electromechanical equipment into a preset digital twin model to obtain the state mapping results of the hydropower station's electromechanical equipment in multiple dimensions. The digital twin model includes a physical mechanism sub-model based on finite element analysis and a behavioral data sub-model based on a temporal neural network. The safety assessment module 22 is used to perform the first state assessment on the state mapping results. If any state mapping parameter in the state mapping results is detected to be greater than the preset safe operation threshold, the fault mode inference is performed based on the digital twin model to obtain the root cause analysis results of the fault. The deep evaluation module 23 is used to calculate the health index of hydropower station electromechanical equipment based on the state mapping result if no state mapping parameter is detected to be greater than the preset safe operation threshold; and to call the physical mechanism sub-model to perform virtual sensing inversion on non-measuring point parts of hydropower station electromechanical equipment where no physical sensors are installed, so as to obtain holographic state evaluation results. The comprehensive health module 24 is used to integrate the health index with the holographic status assessment results to obtain the comprehensive health score; The degradation prediction module 25 is used to perform time-series prediction of the overall health based on the behavioral data sub-model to obtain the degradation trend prediction result. The fault simulation module 26 is used to perform fault mode simulation based on the digital twin model and obtain the root cause analysis results if the overall health is less than the preset health threshold. The decision adjustment module 27 is used to allow the electromechanical equipment to continue operating if the overall health level is greater than or equal to the preset health level threshold, and to adjust the monitoring frequency for the next cycle based on the prediction results of the overall health level deterioration trend.

[0095] See Figure 3 , Figure 3 This is a schematic block diagram of an electronic device provided according to an embodiment of this application. Figure 3The electronic device 300 in this embodiment may include one or more processors 301, one or more input devices 302, one or more output devices 303, and one or more memories 304. The processors 301, input devices 302, output devices 303, and memories 304 communicate with each other via a communication bus 305. The memories 304 store computer programs, including program instructions. The processors 301 execute the program instructions stored in the memories 304. Specifically, the processors 301 are configured to invoke the program instructions to perform the functions of the modules in the aforementioned device embodiments, for example... Figure 2 The functions of the twin mapping module 21, security assessment module 22, in-depth assessment module 23, comprehensive health module 24, degradation prediction module 25, fault inference module 26, and decision adjustment module 27 are shown.

[0096] It should be understood that, in the embodiments of this application, the processor 301 may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0097] Input device 302 may include a touchpad, a fingerprint sensor (for collecting the user's fingerprint information and fingerprint orientation information), a microphone, etc., and output device 303 may include a display (LCD, etc.), a speaker, etc.

[0098] The memory 304 may include read-only memory and random access memory, and provides instructions and data to the processor 301. A portion of the memory 304 may also include non-volatile random access memory. For example, the memory 304 may also store device type information.

[0099] In specific implementations, the processor 301, input device 302, and output device 303 described in the embodiments of this application can execute the implementation methods described in any embodiment of the hydropower station electromechanical equipment operation status monitoring method based on digital twin provided in the embodiments of this application, or they can execute the implementation methods of the electronic devices described in the embodiments of this application, which will not be repeated here.

[0100] In another embodiment of this application, a computer-readable storage medium is provided. This computer-readable storage medium stores a computer program, which includes program instructions. When executed by a processor, the program instructions implement all or part of the processes in the methods described above. Alternatively, the computer program can instruct related hardware to complete the process. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include any entity or device capable of carrying computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0101] The computer-readable storage medium can be an internal storage unit of the electronic device in any of the foregoing embodiments, such as a hard disk or memory of the electronic device. The computer-readable storage medium can also be an external storage device of the electronic device, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc., equipped on the electronic device. Furthermore, the computer-readable storage medium can include both internal and external storage units of the electronic device. The computer-readable storage medium is used to store computer programs and other programs and data required by the electronic device. The computer-readable storage medium can also be used to temporarily store data that has been output or will be output.

[0102] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this application.

[0103] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the electronic devices and units described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0104] In the several embodiments provided in this application, it should be understood that the disclosed electronic devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces or units, or it may be an electrical, mechanical, or other form of connection.

[0105] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments of this application, depending on actual needs.

[0106] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0107] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for monitoring the operating status of electromechanical equipment in a hydropower station based on digital twins, characterized in that, include: The monitoring data of the hydropower station's electromechanical equipment is input into a preset digital twin model to obtain the state mapping results of the hydropower station's electromechanical equipment in multiple dimensions; wherein, the digital twin model includes a physical mechanism sub-model based on finite element analysis and a behavioral data sub-model based on temporal neural network. A first state assessment is performed on the state mapping result. If any state mapping parameter in the state mapping result is found to be greater than the preset safe operation threshold, then a fault mode inference is performed based on the digital twin model to obtain the fault root cause analysis result. If no state mapping parameter is detected to be greater than the preset safe operation threshold, then based on the state mapping result, the behavior data sub-model is called to calculate the health index of the hydropower station's electromechanical equipment; the physical mechanism sub-model is called to perform virtual sensing inversion on the non-measuring point parts of the hydropower station's electromechanical equipment where no physical sensors are installed, and obtain the holographic state assessment result. The health index is fused with the holographic status assessment result to obtain a comprehensive health score; Based on the behavioral data sub-model, a time-series prediction of the overall health is performed to obtain a deterioration trend prediction result. If the overall health score is less than the preset health score threshold, then the fault mode is deduced based on the digital twin model to obtain the root cause analysis results. If the overall health status is greater than or equal to the preset health status threshold, the electromechanical equipment is allowed to continue operating, and the monitoring frequency for the next cycle is adjusted based on the predicted deterioration trend of the overall health status.

2. The method for monitoring the operating status of electromechanical equipment in a hydropower station based on digital twins according to claim 1, characterized in that, The step of calling the physical mechanism sub-model to perform virtual sensing inversion on non-measuring point parts of the hydropower station's electromechanical equipment where no physical sensors are installed, and obtaining holographic state assessment results, includes: The non-measuring points in the electromechanical equipment of the hydropower station that do not have physical sensors installed are identified as the target set for virtual sensing inversion. Acquire measured data from deployed physical sensors as boundary conditions for the physical mechanism sub-model; Based on the physical mechanism sub-model, with the boundary conditions as constraints, the finite element inverse problem solution method is used to perform inversion calculation on the target part set to obtain the state data of the non-measuring point parts; Align and fuse the state data with the measured data to construct a holographic perception data layer; Based on the holographic perception data layer, the holographic state evaluation result is generated.

3. The method for monitoring the operating status of electromechanical equipment in a hydropower station based on digital twins according to claim 2, characterized in that, After calling the physical mechanism sub-model to perform virtual sensing inversion on the non-measuring points of the hydropower station's electromechanical equipment where no physical sensors are installed, and obtaining the holographic state assessment result, the process further includes: Temporary calibration sensors are deployed at predetermined calibration locations on the electromechanical equipment of the hydropower station. The measured value of the temporary verification sensor is compared with the virtual sensing inversion value at the corresponding position in the holographic state evaluation result, and the inversion error is calculated. When the inversion error exceeds a preset accuracy threshold, a parameter correction process is performed on the physical mechanism sub-model. The parameter correction process includes: Using the Bayesian parameter estimation method to reduce the inversion error, the boundary conditions, material property parameters, and contact stiffness coefficients in the physical mechanism sub-model are reverse-calibrated to obtain the corrected physical mechanism sub-model parameters. The corrected physical mechanism sub-model parameters are updated in the digital twin model for use in the virtual sensing inversion calculation of the next cycle.

4. The method for monitoring the operating status of electromechanical equipment in a hydropower station based on digital twins according to claim 2, characterized in that, The process of fusing the health index with the holographic status assessment result to obtain a comprehensive health score includes: Based on the weight adjustment factor of the current operating condition of the hydropower station's electromechanical equipment and the weight coefficient determined based on the reliability risk level of the hydropower station's electromechanical equipment, a fusion weight allocation strategy is determined. Using the health index as the first fusion dimension, the holographic state assessment result is decomposed into a second fusion dimension and a third fusion dimension; Among them, the second fusion dimension is the health sub-item corresponding to the virtual sensing inversion value of each non-measuring point; The third fusion dimension is the field deviation between the distribution of stress field, temperature field, vibration field and pressure field in the holographic perception data layer and the design reference value; Based on the fusion weight allocation strategy, the first fusion dimension, the second fusion dimension, and the third fusion dimension are calculated using a weighted comprehensive evaluation method to obtain the comprehensive health score.

5. The method for monitoring the operating status of electromechanical equipment in a hydropower station based on digital twins according to claim 4, characterized in that, The step of determining the fusion weight allocation strategy based on the weight adjustment factor of the current operating condition of the hydropower station's electromechanical equipment and the weight coefficient determined based on the reliability risk level of the hydropower station's electromechanical equipment includes: Obtain the current operating status of the electromechanical equipment of the hydropower station; Based on the preset working condition-weight adjustment factor mapping table, determine the weight adjustment factor corresponding to the current operating condition; Obtain the reliability risk level of the electromechanical equipment of the hydropower station; The weight coefficients of the monitoring indicators corresponding to the preset components in the electromechanical equipment of the hydropower station are determined according to the reliability risk level, wherein the weight coefficient of the monitoring indicator corresponding to the preset component is larger for the higher the risk level. The fusion weight allocation strategy is obtained by multiplying the weight adjustment factor by the weight coefficient.

6. The method for monitoring the operating status of electromechanical equipment in a hydropower station based on digital twins according to claim 1, characterized in that, If the overall health score is less than a preset health score threshold, then based on the digital twin model, fault mode deduction is performed, and the root cause analysis results are output, including: If the overall health score is less than a preset health score threshold, extract the fault symptom feature vector from the current monitoring data; The fault symptom feature vectors are matched and reasoned with the fault patterns in the preset fault knowledge graph to obtain a list of candidate fault causes and their corresponding initial confidence levels. The candidate fault causes are input into the physical mechanism sub-model for simulation verification. The initial confidence level is corrected based on the degree of agreement between the simulation results and the measured data, and the final root cause analysis results are output.

7. The method for monitoring the operating status of electromechanical equipment in a hydropower station based on digital twins according to claim 6, characterized in that, The process of inputting the candidate fault causes into the physical mechanism sub-model for simulation verification, correcting the initial confidence level based on the degree of agreement between the simulation results and the measured data, and outputting the final root cause analysis results includes: For each candidate fault cause in the candidate fault cause list, determine the corresponding physical mechanism sub-model parameter adjustment scheme to obtain the adjusted physical mechanism sub-model parameters; The adjusted physical mechanism sub-model parameters are input into the physical mechanism sub-model for simulation to obtain simulation feature data under the candidate fault causes. Calculate the feature similarity between the simulated feature data and the measured feature data of the corresponding measurement points in the monitoring data; The initial confidence score is weighted and corrected based on the feature similarity to obtain the corrected final confidence score; The candidate fault cause with the highest confidence after correction is taken as the final root cause analysis result.

8. The method for monitoring the operating status of electromechanical equipment in a hydropower station based on digital twins according to claim 1, characterized in that, If the overall health level is greater than or equal to the preset health level threshold, the electromechanical equipment is allowed to continue operating, and the monitoring frequency for the next cycle is adjusted based on the predicted degradation trend of the overall health level, including: Obtain the comprehensive health status time series of the electromechanical equipment of the hydropower station under the current operating conditions. The comprehensive health status time series includes the health status records of each cycle within a preset time window in the past. Based on the comprehensive health time series, a deterioration trend prediction model is used to predict the health change trend within a preset time window in the future, and the health decay rate is calculated. Based on the current value of the overall health status and the rate of health status decay, the first monitoring frequency of the hydropower station's electromechanical equipment is determined; Based on the current operating conditions of the hydropower station's electromechanical equipment, the monitoring bias of the hydropower station's electromechanical equipment is determined. The monitoring bias includes structural fatigue bias, hydraulic stability bias, or electrical aging bias. Based on the aforementioned monitoring bias, the first monitoring frequency is adjusted to obtain the second monitoring frequency; Based on the second monitoring frequency, the monitoring strategy for the next cycle of the hydropower station's electromechanical equipment is determined.

9. A method for monitoring the operating status of electromechanical equipment in a hydropower station based on digital twins according to claim 8, characterized in that, The step of adjusting the first monitoring frequency based on the monitoring bias to obtain the second monitoring frequency includes: If the monitoring bias is the structural fatigue bias, then based on the first step of increasing the acquisition frequency of structural vibration parameters and stress parameters, the acquisition frequency of vibration characteristic values ​​in the first monitoring frequency is increased by a first preset multiple, and strain sensor data acquisition channels are added to obtain the second monitoring frequency; If the monitoring bias is towards hydraulic stability, then based on the second step size, the sampling frequency of pressure pulsation parameters and swing parameters is increased, the sampling rate of the original waveform of pressure pulsation in the first monitoring frequency is increased by a second preset multiple, and the cavitation acoustic emission monitoring channel is enabled to obtain the second monitoring frequency; If the monitoring bias is the electrical aging bias, then based on the third step, the acquisition frequency of partial discharge parameters and insulation parameters is increased, the partial discharge monitoring cycle in the first monitoring frequency is shortened by a third preset ratio, and a stator end vibration monitoring channel is added to obtain the second monitoring frequency.

10. A monitoring system for the operational status of electromechanical equipment in a hydropower station based on digital twins, characterized in that, include: The twin mapping module is used to input the monitoring data of the hydropower station's electromechanical equipment into a preset digital twin model to obtain the state mapping results of the hydropower station's electromechanical equipment in multiple dimensions; wherein, the digital twin model includes a physical mechanism sub-model based on finite element analysis and a behavioral data sub-model based on a temporal neural network; The safety assessment module is used to perform a first state assessment on the state mapping result. If any state mapping parameter in the state mapping result is detected to be greater than the preset safe operation threshold, then the fault mode inference is performed based on the digital twin model to obtain the fault root cause analysis result. The deep evaluation module is used to calculate the health index of the hydropower station's electromechanical equipment based on the state mapping result if no state mapping parameter is detected to be greater than the preset safe operation threshold; and to call the physical mechanism sub-model to perform virtual sensing inversion on the non-measuring point parts of the hydropower station's electromechanical equipment where no physical sensors are installed, so as to obtain the holographic state evaluation result. The comprehensive health module is used to fuse the health index with the holographic status assessment result to obtain a comprehensive health score; The degradation prediction module is used to perform time-series prediction of the overall health based on the behavioral data sub-model to obtain degradation trend prediction results. The fault inference module is used to perform fault mode inference based on the digital twin model and obtain the fault root cause analysis results if the overall health is less than the preset health threshold. The decision adjustment module is used to allow the electromechanical equipment to continue operating if the overall health level is greater than or equal to the preset health level threshold, and to adjust the monitoring frequency for the next cycle based on the predicted deterioration trend of the overall health level.

Images