A method for multi-layer health assessment of hydroelectric generating units with dynamic comprehensive weight and adaptive segmented health degree

Abstract

A kind of dynamic comprehensive weight and adaptive segmented health degree of hydroelectric generating set multilayer health assessment method, it is related to the field of hydraulic power generation equipment condition monitoring and intelligent operation and maintenance technology, to the weight determination single in the existing health assessment of hydroelectric generating set, health degree calculation is difficult to adapt to the change of operating state and the problem of insufficient reliability of evaluation result, the following scheme is proposed: obtaining the hierarchical relationship of hydroelectric generating set equipment and monitoring index data, establishing a multilayer health assessment structure, determining and updating the comprehensive weight based on importance relationship, data change relationship and association relationship, calculating the health degree of monitoring index according to the current monitoring data, threshold interval and operating state, then aggregating the whole machine health degree level by level, and determining the confidence level.The present application is suitable for hydroelectric generating set state assessment, fault early warning and operation and maintenance decision.

Description

Technical Field

[0001] It belongs to the field of intelligent operation and maintenance and intelligent monitoring technology for hydropower equipment, specifically involving multi-level health assessment of hydropower units based on dynamic comprehensive weight and adaptive segmented health. Background Technology

[0002] Hydropower generating units are the core equipment in hydropower generation, and their operating status directly affects the power station's power generation safety and operation and maintenance arrangements. With the application of hydropower station condition monitoring systems and intelligent operation and maintenance systems, operating data of the units are typically collected through measuring points such as temperature, pressure, vibration, displacement, and flow rate. Health assessment models are then established around objects such as turbines, generators, guide vanes, rotating parts, stators, rotors, and bearing systems to conduct hierarchical analysis and health evaluation of the hydropower generating units' operating status, providing a reference for maintenance planning, early warning judgment, and operation and maintenance decisions.

[0003] Existing methods for assessing the health of hydropower units typically begin by establishing an evaluation hierarchy based on the unit's physical structure or the monitored objects. Then, multiple monitoring indicators are selected as the bottom-level evaluation objects. Indicator weights are determined using methods such as the Analytic Hierarchy Process (AHP), entropy weight method, principal component analysis, grey relational analysis, or fuzzy comprehensive evaluation. Finally, preset thresholds, membership functions, or health functions are used to calculate the health status of each measuring point or component. Finally, the health status of the subsystem or the entire unit is obtained according to certain aggregation rules. While these methods can achieve a comprehensive evaluation of the operating status of hydropower units to a certain extent, there is still room for improvement in scenarios involving long-term operation, changing operating conditions, and multi-measuring point coupled evaluation.

[0004] On the one hand, existing health assessment methods often employ a single approach to determining weights. When using expert experience or the analytic hierarchy process (AHP) to determine weights, the evaluation results are easily influenced by expert judgment. When using objective weighting methods such as entropy weighting, weights are typically determined primarily based on the dispersion of monitoring data, making it difficult to simultaneously reflect the structural relationships between indicators and the impact of equipment operating status changes over time on the importance of indicators. Therefore, in the multi-level, multi-indicator health assessment of hydropower units, a single weighting method cannot adequately consider expert experience, data change characteristics, and indicator relationships, easily leading to unreasonable weight matching between different levels and different monitoring points.

[0005] On the other hand, existing health status calculation processes mostly rely on fixed thresholds or piecewise functions with fixed shapes. For hydropower units, the correspondence between the numerical changes of the same monitoring index and the health status is not entirely the same under different operating conditions, different historical states, and different stages of deterioration. If the threshold points, interval shape parameters, or interpolation methods of the health status function remain unchanged for a long time, it is difficult to reflect the impact of changes in equipment operating conditions on the health status calculation rules. Especially when indicators such as temperature and vibration gradually deviate from the normal operating range, fixed functions can easily make health status changes too rigid, which is not conducive to continuous and detailed characterization of the equipment deterioration process.

[0006] Furthermore, existing assessment models often employ fixed weights and evaluation rules after their establishment, lacking a mechanism to update weights based on newly collected operational data. When the operating status of hydropower units, the distribution of monitoring data, or the correlation between indicators change, the original weights may no longer accurately reflect the current equipment status. Simultaneously, some health assessment results only output a single health value, lacking a quantitative assessment of the consistency between the current assessment results and historical operating patterns, making it difficult for maintenance personnel to determine the reliability of the health result under current data conditions.

[0007] In summary, existing technologies suffer from several drawbacks, including difficulty in balancing subjective experience, objective data, and the relationship between indicator structures in the health assessment of hydropower units; difficulty in adaptively adjusting the health function according to the operating status of the equipment; and insufficient quantification of weight updates and assessment results confidence. Summary of the Invention

[0008] To address the shortcomings of existing technologies, such as the difficulty in balancing subjective experience, objective data, and the structural relationship of indicators in the health assessment of hydropower units, the inability of the health function to adaptively adjust with the operating status of the equipment, and insufficient quantification of weight updates and assessment results confidence, the technical solution provided by this invention is as follows: A multi-level health assessment method for hydropower units with dynamic comprehensive weighting and adaptive segmented health status includes: The steps include: obtaining the equipment hierarchy of hydropower units and the monitoring indicator data corresponding to each level; establishing a health assessment structure including at least two evaluation levels; and determining the correspondence between each monitoring indicator and the superior evaluation object. The steps to determine the comprehensive weight of each evaluation object based on the importance relationship of each evaluation object within the same evaluation level, the relationship of changes in monitoring indicator data, and the correlation between evaluation objects; The steps to calculate the health status of each monitoring indicator based on the current monitoring data, corresponding threshold ranges, and the range change relationship with the operating status of each indicator. According to the health assessment structure, the health of the lower-level evaluation objects is aggregated step by step according to the corresponding comprehensive weight to obtain the health of the upper-level evaluation objects, until the health of the entire hydropower unit is obtained. The steps for outputting the overall health status of the hydropower unit and the health status of the corresponding evaluation objects at each level.

[0009] Furthermore, in a preferred embodiment, the health assessment structure is established according to the physical hierarchy of the hydropower unit. The evaluation hierarchy includes a system layer, a subsystem layer, a component layer, and a monitoring indicator layer, with each monitoring indicator in the monitoring indicator layer corresponding to its respective component.

[0010] Furthermore, in a preferred embodiment, the importance relationship is determined by the relative importance of each evaluation object within the same evaluation level, and participates in the determination of the comprehensive weight after satisfying the consistency constraint.

[0011] Furthermore, in a preferred embodiment, the relationship between the changes in the monitoring indicator data is determined based on the degree of change in the historical data of each monitoring indicator, the data time attribute, and the current operating status, and participates in the determination of the comprehensive weight.

[0012] Furthermore, in a preferred embodiment, the correlation between the evaluation objects is determined based on the direct and indirect influence relationships between the evaluation objects within the same evaluation level, and participates in the determination of the comprehensive weight.

[0013] Furthermore, in a preferred embodiment, the health status of each monitoring indicator is determined based on the relative position of the current monitoring data within the corresponding threshold range and the change relationship corresponding to that threshold range.

[0014] A multi-layer health assessment device for hydropower units based on dynamic comprehensive weighting and adaptive segmented health status includes: A module for acquiring the equipment hierarchy of hydropower units and the corresponding monitoring index data for each level, establishing a health assessment structure including at least two evaluation levels, and determining the correspondence between each monitoring index and the higher-level evaluation object; A module that determines the comprehensive weight of each evaluation object based on the importance relationship of each evaluation object within the same evaluation level, the relationship of changes in monitoring indicator data, and the correlation between evaluation objects; A module that calculates the health status of each monitoring indicator based on the current monitoring data, corresponding threshold ranges, and the range change relationship with the operating status. Based on the health assessment structure, the health of the lower-level evaluation objects is aggregated level by level according to the corresponding comprehensive weight to obtain the health of the upper-level evaluation objects, until the module of the overall health of the hydropower unit is obtained; A module that outputs the overall health status of the hydropower unit and the health status of the corresponding evaluation objects at each level.

[0015] A computer storage medium for storing a computer program, which, when read by the computer, is executed by the computer using the method described thereon.

[0016] A computer, including a processor and a storage medium, executes the method when the processor reads a computer program stored in the storage medium.

[0017] A computer program product, which, as a computer program, implements the method when the computer program is executed.

[0018] Compared with the prior art, the advantages of the technical solution provided by the present invention are as follows: The multi-layered tree-structure model of hydropower units organizes the entire system, subsystems, components, and measuring points according to the physical composition of the unit, ensuring that low-level monitoring parameters such as temperature, pressure, vibration, and flow rate can be attributed to specific components and subsystems. Compared to methods that evaluate only a single measuring point or component, this structure establishes a correspondence between low-level monitoring data and the actual structure of the unit. This facilitates the subsequent transmission of health status from the measuring point layer to the component layer, subsystem layer, and system layer, thereby reducing the problems of hierarchical confusion in health assessment and difficulty in locating evaluation results to specific equipment locations.

[0019] The multimodal feature fusion weight calculation model integrates subjective weights, objective weights, and structural weights, ensuring that the weighting results are simultaneously constrained by expert experience, monitoring data change characteristics, and the influence relationships between indicators. Compared to using the analytic hierarchy process (AHP) or entropy weighting method alone, this feature reduces the bias caused by a single weighting method, making the weights of indicators at different measuring points or levels closer to the actual operating state and equipment structure relationships of hydropower units, thereby improving the rationality of weight allocation in multi-indicator health assessments.

[0020] The improved analytic hierarchy process (AHP) introduces a consistency optimization mechanism based on the expert judgment matrix, ensuring that the subjective weights formed by expert experience are subject to consistency constraints before calculation. This feature does not simply adopt expert scores, but rather corrects the logical consistency of the judgment matrix, reducing the impact of inconsistencies in expert comparisons on the weighting results, thereby improving the stability of subjective weights in the health assessment of hydropower units.

[0021] The improved entropy weight method introduces a time decay factor and a state adjustment factor into the calculation of indicator information entropy. This allows the objective weight to not only reflect the dispersion of monitoring data but also the impact of recent data and current operating status on the importance of the indicator. Compared to the traditional entropy weight method, which treats all historical data approximately equally, this feature enables measurement points with abnormal increases, deterioration, or significant recent fluctuations to be reflected more promptly in the weight calculation, thereby improving the responsiveness of the objective weight to changes in equipment status.

[0022] Based on the structural weights obtained from factor analysis and the Dematel method, the mutual influence relationships between measurement points or indicators are transformed into quantitative results such as influence degree, degree of influence, and centrality. This feature makes weight allocation no longer solely dependent on the numerical changes of individual indicators, but further considers the correlation position and transmission influence of indicators in the evaluation system, thereby improving the problem in existing health assessments that neglect the structural relationships of indicators and are difficult to reflect the coupling effects of multiple measurement points.

[0023] The fusion coefficient linearly combines subjective, objective, and structural weights, and can be determined based on historical experience or optimization algorithms. This allows the proportion of different source weights in the final weight to be adjusted according to the actual evaluation results. This feature prevents one type of weight from dominating in the long term, and facilitates adjusting the weight source ratio in different scenarios where expert experience is reliable, data fluctuations are significant, or indicator correlations are strong. This improves the adaptability of the weight model to different power plant units and different monitoring objects.

[0024] The adaptive threshold segmented health function divides the measured indicators into multiple health calculation intervals based on preset threshold points, and calculates the health level within each interval using normalized variables and interpolation factors. This feature allows the measured values ​​to correspond to different health change rules within different threshold intervals. Compared to a single linear function or fixed threshold judgment method, it can more precisely describe the continuous change process of the monitored indicators from normal operation and slight deviation to near alarm state.

[0025] Each sub-interval has an independent shape parameter, allowing the health function to exhibit concave, convex, or linear variations across different intervals. This feature can adjust the rate of health decline or increase based on the characteristics of the indicator's changes at different stages of degradation. For example, it can increase the sensitivity to health changes when approaching the warning interval and reduce the impact of excessive fluctuations on the evaluation results within the stable interval, thereby improving the problem of the rigidity of fixed piecewise functions in characterizing changes in equipment status.

[0026] The optimal operating point is determined based on historical data, and the shape parameters are dynamically adjusted according to the operating conditions, so that the health calculation rules can change with the actual operating status of the equipment. Compared with the long-term use of a fixed health function, this feature enables the same temperature, vibration, or pressure index to have a health mapping relationship that is more consistent with the current operating background under different operating conditions, thereby improving the adaptability of health calculation to the differences in the operating conditions of hydropower units.

[0027] The bottom-up health aggregation algorithm aggregates the health status of lower-level measurement points according to their corresponding weights, progressively up to the component layer, subsystem layer, and system layer. This allows changes in the health status of a single measurement point to influence the evaluation results at higher levels through its structural path. This feature preserves the hierarchical source of the evaluation results while outputting the overall system health status, facilitating further tracing from system anomalies back to relevant subsystems, components, or measurement points, thereby improving the interpretability of health assessment results in operational and maintenance decisions.

[0028] The time-varying weight update mechanism based on sliding window and forgetting factor allows newly collected data to participate in the weight increment calculation, and adjusts the relationship between historical weights and new weights through the forgetting factor. This feature avoids the assessment model from relying on the initial weights in the long term. When the distribution of monitoring data, the fluctuation characteristics of indicators, or the status of equipment changes, the weight coefficients are gradually updated to reflect the current operating status, thereby improving the continuous adaptability of the health assessment model.

[0029] The forgetting factor adaptively adjusts based on data stability, ensuring that the weight update speed matches the degree of change in the monitored data. When data changes are relatively stable, weight updates will not be significantly offset by short-term fluctuations; when data shows continuous changes or abnormal trends, the impact of new data on the weight results can be enhanced. This feature balances the stability and sensitivity of weight updates, thereby reducing the impact of frequent fluctuations or update lags on health assessment results.

[0030] Grey relational analysis calculates the correlation between the current health status result and historical operating patterns, and uses the correlation result to form a confidence level. This feature allows the health assessment result to no longer just output a single health status value, but to further provide the degree of consistency between the current evaluation result and the historical state pattern, thus providing a quantitative basis for operations and maintenance personnel to judge the credibility of the health status result.

[0031] The real-time monitoring data acquisition and preprocessing module, weight calculation module, health status calculation module, aggregation analysis module, dynamic update module, and result visualization output and maintenance decision-making module work together to form a continuous processing chain from data acquisition, weight determination, health status calculation, hierarchical aggregation, weight update, and result output. This feature enables the implementation of multi-layered health assessment methods in intelligent operation and maintenance systems, reducing the problem of fragmentation between different calculation links, thereby facilitating the formation of an executable health assessment process in hydropower unit operation monitoring scenarios. Attached Figure Description

[0032] Figure 1 This is a flowchart illustrating the overall process architecture of a multi-level health assessment method for hydropower units based on dynamic comprehensive weighting and adaptive segmented health.

[0033] Figure 2 A schematic diagram illustrating the control effect of various shape parameter combinations for the adaptive piecewise health function.

[0034] Figure 3 A schematic diagram comparing the influence of sub-interval endpoint division and shape parameters on the adaptive segmented health function.

[0035] Figure 4 This is a schematic diagram showing the simulated temperature rise data of the stator winding and the temperature-health distribution.

[0036] Figure 5 This is a schematic diagram illustrating the implementation process of an intelligent operation and maintenance system for hydropower units based on a multi-layer health assessment method using dynamic comprehensive weights and adaptive segmented health levels. Detailed Implementation

[0037] To make the advantages and benefits of the technical solution provided by the present invention clearer, the technical solution provided by the present invention will now be described in further detail with reference to the accompanying drawings, specifically: Implementation Method 1: This implementation method provides a multi-level health assessment method for hydropower units based on dynamic comprehensive weighting and adaptive segmented health, including: The steps include: obtaining the equipment hierarchy of hydropower units and the monitoring indicator data corresponding to each level; establishing a health assessment structure including at least two evaluation levels; and determining the correspondence between each monitoring indicator and the superior evaluation object. The steps to determine the comprehensive weight of each evaluation object based on the importance relationship of each evaluation object within the same evaluation level, the relationship of changes in monitoring indicator data, and the correlation between evaluation objects; The steps to calculate the health status of each monitoring indicator based on the current monitoring data, corresponding threshold ranges, and the range change relationship with the operating status of each indicator. According to the health assessment structure, the health of the lower-level evaluation objects is aggregated step by step according to the corresponding comprehensive weight to obtain the health of the upper-level evaluation objects, until the health of the entire hydropower unit is obtained. The steps for outputting the overall health status of the hydropower unit and the health status of the corresponding evaluation objects at each level.

[0038] The health assessment structure is established according to the physical hierarchy of the hydropower units. The evaluation hierarchy includes a system layer, a subsystem layer, a component layer, and a monitoring indicator layer. Each monitoring indicator in the monitoring indicator layer corresponds to its respective component.

[0039] The importance relationship is determined by the relative importance of each evaluation object within the same evaluation level, and participates in the determination of the comprehensive weight after the consistency constraint is met.

[0040] The relationship between the changes in the monitoring indicator data is determined based on the degree of change in the historical data of each monitoring indicator, the data time attribute, and the current operating status, and is involved in the determination of the comprehensive weight.

[0041] The relationships between the evaluation objects are determined based on the direct and indirect influence relationships between the evaluation objects within the same evaluation level, and are involved in the determination of the comprehensive weight.

[0042] The health status of each monitoring indicator is determined based on the relative position of the current monitoring data within the corresponding threshold range and the change relationship corresponding to that threshold range.

[0043] A multi-layer health assessment device for hydropower units based on dynamic comprehensive weighting and adaptive segmented health status includes: A module for acquiring the equipment hierarchy of hydropower units and the corresponding monitoring index data for each level, establishing a health assessment structure including at least two evaluation levels, and determining the correspondence between each monitoring index and the higher-level evaluation object; A module that determines the comprehensive weight of each evaluation object based on the importance relationship of each evaluation object within the same evaluation level, the relationship of changes in monitoring indicator data, and the correlation between evaluation objects; A module that calculates the health status of each monitoring indicator based on the current monitoring data, corresponding threshold ranges, and the range change relationship with the operating status. Based on the health assessment structure, the health of the lower-level evaluation objects is aggregated level by level according to the corresponding comprehensive weight to obtain the health of the upper-level evaluation objects, until the module of the overall health of the hydropower unit is obtained; A module that outputs the overall health status of the hydropower unit and the health status of the corresponding evaluation objects at each level.

[0044] A computer storage medium for storing a computer program, which, when read by the computer, is executed by the computer using the method described thereon.

[0045] A computer, including a processor and a storage medium, executes the method when the processor reads a computer program stored in the storage medium.

[0046] A computer program product, which, as a computer program, implements the method when the computer program is executed.

[0047] Implementation Method Two: This implementation method is a further detailed description of the technical solution provided in Implementation Method One, specifically: This embodiment provides a multi-level health assessment method for hydropower units with dynamic comprehensive weighting and adaptive segmented health. The method is executed according to the structural hierarchy of the hydropower unit and the flow of monitoring data, and specifically includes the following steps.

[0048] Step 1: Construct a multi-layered health assessment model for hydropower units.

[0049] The equipment composition and monitoring object information of the hydropower unit to be evaluated are obtained, and the hydropower unit is divided into a system layer, a subsystem layer, a component layer, and a measurement point index layer. The system layer includes the turbine system and the generator system; the subsystem layer includes the water guide mechanism, rotating parts, stator, rotor, and bearing system; the component layer includes components related to the corresponding subsystem; and the measurement point index layer includes monitoring indicators such as temperature, pressure, vibration, displacement, and flow rate.

[0050] Based on the above hierarchical relationship, a tree-like association is established from the system layer to the measurement point index layer, so that each measurement point index corresponds to its corresponding component, each component corresponds to its corresponding subsystem, and each subsystem corresponds to its corresponding system. This multi-layered health assessment model serves as the structural foundation for subsequent weight calculation, bottom-level health degree calculation, and layer-by-layer aggregation calculation.

[0051] Step 2: Read and preprocess the monitoring data of the hydropower unit.

[0052] Monitoring data for various measuring points are read from the real-time monitoring system and historical operation database of the hydropower unit. This monitoring data includes temperature, pressure, vibration, displacement, and flow rate data corresponding to the measuring point indicator layer. The read monitoring data undergoes timestamp alignment, outlier handling, missing value processing, and standardization to create comparable data sequences for different types of measuring point indicators.

[0053] The preprocessed monitoring data is aggregated according to the correspondence between the monitoring point indicators and the multi-layer health assessment model, forming the current operating data sequence and the historical operating data sequence for each monitoring point indicator. The current operating data sequence is used for underlying health degree calculation, while the historical operating data sequence is used for objective weight calculation, operating status identification, and parameter updates.

[0054] Step 3: Determine the comprehensive weight of the evaluation objects at each level.

[0055] For multiple evaluation objects participating in the evaluation within the same level, subjective weight, objective weight, and structural weight are calculated separately.

[0056] Subjective weights are determined based on the comparison of importance among the evaluation objects. First, a judgment relationship between the evaluation objects is constructed based on expert experience or operation and maintenance rules. Then, a consistency check is performed on this judgment relationship. When the judgment relationship does not meet the consistency requirements, it is corrected to satisfy the consistency constraints before the corresponding subjective weight is obtained. The resulting subjective weights are used to characterize the relative importance of each evaluation object in equipment operation and maintenance experience.

[0057] Objective weights are determined based on preprocessed historical operational data. First, information differences are calculated according to the degree of change in the data sequences corresponding to each evaluation object. Then, time decay and state adjustment relationships are introduced during the information difference calculation process to give more influence to recent data and the current operational state on the objective weights. The resulting objective weights are used to characterize the state sensitivity of each evaluation object under the current data conditions.

[0058] Structural weights are determined based on the interrelationships between the evaluation objects. First, the direct influence relationships between the evaluation objects are obtained based on expert judgment or data correlation. Then, these direct influence relationships are standardized, and the comprehensive influence relationships between the evaluation objects are calculated. Based on the comprehensive influence relationships, the degree of influence and the degree of being influenced by each evaluation object are obtained, and the centrality of each evaluation object in the evaluation structure is determined by the degree of influence and the degree of being influenced. The resulting structural weights are used to characterize the structural importance of each evaluation object in the multi-indicator correlation system.

[0059] After obtaining the subjective weight, objective weight, and structural weight, the three types of weights are merged according to a preset or optimized fusion ratio to obtain the comprehensive weight of each evaluation object. This comprehensive weight serves as the weighting basis for subsequent health aggregation calculations and is saved separately for each evaluation level.

[0060] Step 4: Calculate the health of the measurement point index layer.

[0061] For each measurement point indicator, its current operating data, corresponding threshold point, and historical operating status parameters are read. The threshold point is used to divide the indicator into multiple health calculation intervals, and the historical operating status parameters are used to determine the optimal operating point and shape control relationship of each interval for the indicator.

[0062] When the current running data falls within a certain health calculation interval, the current running data is normalized according to the start and end points of the interval to obtain the relative position of the current running data within the interval. Based on the relative position and the shape parameters corresponding to the interval, the interpolation relationship within the interval is calculated, and then combined with the health values ​​corresponding to the two ends of the interval, the current health of the measurement point index is obtained.

[0063] For different health calculation intervals, corresponding shape parameters are set to map health status according to different trends within each interval. When the measured indicator is in the normal operating range, health status changes are mapped in a smooth manner; when the measured indicator is close to the warning range, health status changes are mapped in a more sensitive manner; when the measured indicator is in the linear change range, health status is mapped according to a linear relationship. The current health status calculated for each measured indicator serves as the input for the aggregated health status of its corresponding component layer.

[0064] Step 5: Calculate the health status of the hydropower unit from bottom to top.

[0065] Based on the multi-layered health assessment model, the current health status of each measurement point indicator under the same component is weighted and aggregated with its corresponding comprehensive weight to obtain the health status of that component. After the component health status is calculated, the health status of each component under the same subsystem is weighted and aggregated with its corresponding comprehensive weight to obtain the health status of that subsystem.

[0066] After obtaining the health status of each subsystem, the health status of each subsystem within the same system is weighted and aggregated with its corresponding comprehensive weight to obtain the system-level health status. Then, based on the health status of the turbine system and the generator system and their corresponding comprehensive weights, the overall health status of the hydropower unit is calculated.

[0067] In the above aggregation process, the output health of each level is used as the input for the health calculation of the next level, so that the state changes of the measurement point index layer are transmitted to the overall health result through the multi-layer health assessment model.

[0068] Step 6: Update the overall weight.

[0069] After acquiring new monitoring data, the new data is added to the current data window, and historical data exceeding the window's range is removed, forming an updated sliding data window. Based on the updated sliding data window, the objective weights and structural weights are recalculated and compared with the comprehensive weights of the previous period to obtain the weight change for the current period.

[0070] The degree of forgetting is determined based on data stability. When data changes are relatively stable in the sliding data window, the retention ratio of the previous period's comprehensive weight in the updated result is increased; when data changes are significant in the sliding data window, the influence ratio of the current period's weight change in the updated result is increased. The updated comprehensive weight is obtained based on the previous period's comprehensive weight, the current period's weight change, and the degree of forgetting. The updated comprehensive weight is used for the next period's health aggregation calculation.

[0071] Step 7: Determine the confidence level of the health status results.

[0072] Using the current health status sequence as the reference sequence and the health status sequences from historical operating modes as comparison sequences, the correlation between the current health status sequence and each historical health status sequence is calculated. Based on the correlation between each historical health status sequence, the confidence level of the current health status result is obtained.

[0073] When the confidence level meets the preset requirements, the current overall health, health of each system, health of each subsystem, health of each component, health of the measurement point indicators, and confidence level are output as the health assessment results. When the confidence level does not meet the preset requirements, the current health result is marked as a result to be reviewed, and the relevant measurement point indicators, corresponding data sequences, and changes in health are output together.

[0074] As an optional implementation method, the judgment relationship of subjective weights can be determined by operation and maintenance experts based on the importance of equipment, the degree of impact of failure and maintenance experience, or an initial judgment relationship can be generated based on existing maintenance rules, and then the initial judgment relationship can be corrected by the consistency test results.

[0075] As an optional implementation method, the time decay relationship in the objective weight can be determined based on the distance between the data collection time and the current evaluation time. The closer the monitoring data is to the current evaluation time, the greater its impact on the objective weight calculation. The state adjustment relationship can be determined based on whether the measuring point index is close to the warning range, whether there is a continuous upward trend, or whether it deviates from the historical normal range.

[0076] As an optional implementation method, the direct influence relationship in the structural weights can be determined by expert scoring or by the correlation between historical operating data. When the correlation between historical operating data is used, the degree of data correlation between different measuring point indicators is first calculated, and then the degree of correlation is converted into a direct influence relationship for subsequent comprehensive influence relationship calculation.

[0077] As an optional implementation, the threshold point in the adaptive segmented health function can be determined jointly based on manufacturer specifications, historical statistical results, and operation and maintenance early warning rules; the shape parameters of each health calculation interval can be determined based on historical health status samples, or the existing shape parameters can be adjusted according to the current operating conditions.

[0078] As an optional implementation, health aggregation can be performed separately at the component level, subsystem level, system level, and whole machine level. Alternatively, one intermediate level can be omitted based on the completeness of the actual monitoring object's hierarchy. However, the health of the measurement point index layer is still aggregated according to its superior evaluation object.

[0079] As an optional implementation, the health assessment results can be output through an intelligent operation and maintenance system. This system includes a real-time monitoring data reading, acquisition, and preprocessing module; a weight calculation module; a health status calculation module; an aggregation analysis module; a dynamic update module; and a result visualization output and maintenance decision-making module. The real-time monitoring data reading, acquisition, and preprocessing module outputs preprocessed monitoring data to the weight calculation module and the health status calculation module. The weight calculation module outputs the comprehensive weight to the aggregation analysis module. The health status calculation module outputs the health status of the measurement points to the aggregation analysis module. The aggregation analysis module outputs the health status of each level to the dynamic update module and the result visualization output and maintenance decision-making module. The dynamic update module outputs the updated comprehensive weight to the weight calculation module and the aggregation analysis module.

[0080] Implementation Method 3: This implementation method is described in detail with reference to the accompanying drawings. Specific embodiments are provided to further illustrate the technical solutions offered above. Specifically: A preferred embodiment of this implementation method is the application of health assessment for a single generating unit in a power plant. The specific implementation steps are as follows: Step S1: Construct a multi-layered evaluation structure A four-layer evaluation model is constructed based on the physical structure of the unit: (1) System layer: water turbine, generator; (2) Subsystem layer: water guiding mechanism, rotating parts, stator, bearing system, etc.; (3) Component layer: top cover, guide vanes, impeller, thrust bearing, etc.; (4) Measuring point layer: monitoring parameters such as temperature, pressure, vibration, and flow rate.

[0081] Step S2: Multimodal weight calculation Taking the water guide bearing subsystem as an example, calculate the weight of each measuring point: Step S2.1: Calculation of Subjective Weights Based on Improved AHP The Analytic Hierarchy Process (AHP) constructs a decision matrix and calculates eigenvectors as weights. We employ an improved version of AHP, introducing consistency optimization.

[0082] Step S2.1.1 Determine matrix construction Assume there is a total n Each indicator is used to construct a judgment matrix through expert scoring. ,in Indicators i relative to indicators j The importance of satisfying , ,and .

[0083] Step S2.1.2 Calculate weights using the eigenvector method Calculate the judgment matrix A Maximum eigenvalue and the corresponding feature vector Then, the feature vectors are normalized to obtain the weight vectors. .

[0084] Step S2.1.3 Consistency Check Calculate the consistency index Then query the random consistency index. Calculate the consistency ratio .if If the result is satisfactory, the consistency check is passed; otherwise, the judgment matrix needs to be adjusted.

[0085] Step S2.1.4 Improved Consistency Optimization We introduce an adjustment factor to optimize the consistency ratio, as shown in the following formula:

[0086] This adjustment factor can fine-tune the consistency ratio, making the weight calculation more stable.

[0087] Step S2.2: Calculation of objective weights based on the improved entropy weight method The entropy weight method is an objective method for determining weights based on the degree of variation of indicators. We improve upon it by introducing a time decay factor and a state adjustment factor.

[0088] Step S2.2.1: Data Standardization It has One sample, 1 indicator, construct a data matrix Each indicator is standardized.

[0089]

[0090] Step S2.2.2: Calculate the weight of the indicators The improved formula for calculating information entropy is:

[0091] Step S2.2.3: Calculate the entropy weight

[0092] Step S2.3: Calculation of structural weights based on factor analysis Structural weights take into account the mutual influence relationships between indicators.

[0093] The DEMATE method in decision experimentation and evaluation laboratories calculates centrality and causality for each indicator by measuring its influence and affectedness. The specific steps include: Step S2.3.1: Construct a direct influence matrix through expert scoring or correlation analysis. ; Step S2.3.2: Standardization directly affects moments ; Step S2.3.3: Calculate the total influence moment ; Step S2.3.4: Calculate the impact degree And the degree of influence : , Calculate centrality , reason Structural weights: .

[0094] Step S2.4: Weight Fusion The final weights are obtained through linear fusion:

[0095] Step S3: Adaptive Health Calculation The adaptive threshold segmented health function is .

[0096] For any subinterval Introducing normalized variables , Among them, the definition This is the interpolation factor for the health function.

[0097] The health function Defined in interval Above, among which For a series of threshold points. The health function is included in each sub-interval. Above, assign a health value to the endpoint. and This results in a smooth piecewise interpolation function.

[0098] Segmented interval start endpoint ; Ending point of segmented interval .

[0099] Among them, key parameters are dynamically adjusted: the optimal operating point is determined based on historical data analysis of shape parameters, and then dynamically adjusted according to operating conditions.

[0100] The health function In each sub-interval Above, including the shape parameters of each sub-interval. ,in ,and The interpolation factors in the piecewise function health calculation mathematical model include... , and Three shape control parameters, interpolation factor about The monotonicity and concavity are caused by Decide.

[0101] when At that time, the piecewise function health calculation mathematical model is monotonically increasing, and the function is a concave function, that is, the growth rate is fast at first and then slows down; when At that time, the piecewise function health calculation mathematical model is monotonically increasing, and the function is a convex function, that is, the growth rate is slow at first and then fast. when At this point, the piecewise function health calculation mathematical model degenerates into a linear function with no concavity or convexity. To ensure the continuity of the piecewise function, in When, directly define .

[0102] Step S4: Weight Update Calculate the weight increment based on the newly collected data. Establish a weight update mechanism based on sliding window and forgetting factor:

[0103] Among them, forgetting factor The weights are updated adaptively based on data stability.

[0104] Step S5: Confidence assessment based on grey relational analysis The grey relational analysis method is used to calculate the correlation between the current health status and historical patterns, and the confidence level is obtained. C =0.89, indicating that the evaluation results are highly reliable.

[0105] The confidence level of the health status results is assessed using the grey relational analysis method. Grey relational analysis is used to calculate the similarity between the current data and historical patterns, thereby assessing the confidence level of the health status results.

[0106] Step S5.1 Determine the reference sequence and comparison sequence Let the current data be the reference sequence, and the reference sequence be... The comparison sequence is historical data. Historical data, .

[0107] Step S5.2 Calculate the grey relational coefficient

[0108] in This is the resolution coefficient. It is usually set to 0.5.

[0109] Step S5.3 Calculate the grey relational degree and confidence level. Grey relational degree is The confidence level is the average of the grey relational values ​​of all historical patterns.

[0110]

[0111] The following is an example of the implementation process of this method, using the full process of health assessment for a single generating unit in a power plant as an example. The specific steps are as follows: Step 1: Construct a multi-layered evaluation structure model for the unit Based on the unit's physical structure and monitoring system, a top-down four-layer tree-structured evaluation model is established: (1) Overall system level: turbine system, generator system; (2) Subsystem layer: The turbine subsystem includes systems such as the guide vane mechanism, the rotation system, and the embedded system; the generator subsystem includes systems such as the stator system, the rotor system, and the bearing system. (3) Component layer with water guiding mechanism as a specific embodiment: top cover, guide vane, connecting rod, control ring, etc.; (4) Measurement point index layer corresponding to the monitoring parameters of each component: real-time data such as temperature, pressure, vibration, displacement, and flow rate.

[0112] The embodiment provides specific simulation data and calculates in detail the objective weights of the improved entropy weight method and the DEMATEL structural weights. At the same time, it ensures that the calculation process of the embodiment is standardized. This embodiment takes three temperature measuring points of the generator stator bar as an example, provides simulation data over a time period, and calculates the weights of each item.

[0113] The embodiment simulates temperature data from three measuring points at 10 time points, constructing a data matrix X. The data fluctuates around 85°C, with measuring point 3 showing an abnormal increase at these 10 time points: [86.2, 85.8, 87.1], # t1 [87.5, 86.9, 88.3], # t2 [89.1, 88.2, 89.7], # t3 [91.3, 90.5, 92.1], # t4 [93.8, 92.7, 94.5], # t5 ;[96.2, 95.1, 97.3], # t6 [98.7, 97.4, 99.8], # t7 [101.5, 100.2, 102.9],# t8 [104.3, 103.1, 106.2], # t9 [107.8, 106.5, 116.5] # t10 Step 2: Calculation of Multimodal Feature Fusion Weights Taking generator stator temperature monitoring as an example, the entire process of weight calculation is demonstrated: Step 2.1: Subjective weight calculation based on the improved AHP method By simulating pairwise comparisons of stator winding temperature measurement points by experts in related professional fields, this embodiment focuses on stator winding temperature measurement points. The embodiment includes constructing an expert judgment matrix for stator winding temperature early warning measurement points and adjusting it using a consistency optimization mechanism.

[0114]

[0115]

[0116] The embodiment described above uses the operating data of a single power plant unit to simulate early warning data and obtain an expert judgment matrix of early warning data from three measuring points of the stator bars.

[0117] Using the eigenvector method, the normalized eigenvector corresponding to the largest eigenvalue is calculated to obtain the subjective weight vector, with subjective weights of [0.29696133, 0.16342412, 0.53961455].

[0118]

[0119] Step 2.2: Calculation of objective weights based on the improved entropy weight method Collect historical data, calculate the information entropy of each measuring point, and introduce a time decay factor. With state adjustment factor Then the improved information entropy is constructed as follows: .

[0120] The embodiment calculates entropy weight. The objective weights were calculated to be [0.33382166, 0.33374472, 0.33243362].

[0121]

[0122] Step 2.3: Calculation of DEMATEL structural weights based on factor analysis The DEMATEL method determines weights by calculating the interrelationships between indicators. We assume there is a certain degree of mutual influence among the three temperature measurement points, and construct a direct influence matrix using expert scoring or data correlation. Here, we construct it based on the correlation coefficient matrix of simulated data.

[0123] By analyzing the correlation coefficient matrix among the measurement points, common factors are extracted, and the influence degree of each measurement point is calculated. R And the degree of influence C Data characteristics, etc.

[0124] Calculate the correlation coefficient matrix of the three measurement points, which serves as the base of the direct influence matrix. Then, obtain the direct influence matrix by setting a threshold or scaling. Z The embodiment multiplies the correlation coefficient matrix by a scaling factor so that the matrix elements are between 0 and 1.

[0125] Constructing a standardized direct impact matrix Calculate the total influence matrix , in It is an identity matrix.

[0126] Among them, influence degree of influence .

[0127] Among them, centrality , reason .

[0128] , .

[0129] The specific calculation indicators for the structural weights in the embodiment include: The impact R is [283.10919296, 283.30333314, 281.48358005]; The degree of influence C is [283.10919296, 283.30333314, 281.48358005]; The centrality M is [566.21838592, 566.60666627, 562.96716009]; The structural weights are [0.33389609, 0.33412506, 0.33197886].

[0130] Step 2.4: Weight Fusion The weighted fusion step can optimize the fusion coefficients based on expert historical experience or by using optimization algorithms such as genetic algorithms. This embodiment sets... , , The weights obtained through linear fusion are [0.32278589, 0.28276264, 0.39445147].

[0131]

[0132] Step 3: Adaptive segmented health calculation Taking the "stator core temperature" measuring point as an example: Step 3.1: Threshold setting Three early warning temperature measurement points were selected: generator-stator system-stator conductors. The early warning thresholds are as follows. Threshold parameters were obtained based on historical statistics and manufacturer specifications.

[0133]

[0134] , , .

[0135] Step 3.2: Multi-level early warning health calculation model based on baseline shape parameters The adaptive threshold segmented health function is .

[0136] For any subinterval Introducing normalized variables , Among them, the definition This is the interpolation factor for the health function.

[0137] The health function Defined in interval Above, among which For a series of threshold points. The health function is included in each sub-interval. Above, assign a health value to the endpoint. and This, in turn, forms a smooth piecewise interpolation function, making the starting and ending points of the piecewise intervals... Ending point of segmented interval .

[0138] Among them, key parameters are dynamically adjusted: the optimal operating point is determined based on historical data analysis of shape parameters, and then dynamically adjusted according to operating conditions.

[0139] The final health interval threshold of the piecewise health function calculation model is [70, 80, 85, 90, 100, 115, 130], and the final shape parameter of the piecewise health function calculation model is [0.85, 0.9, 0.95, 1.0, 0.7, 0.3, 0.05].

[0140] Step 3.3: Obtain the current health value of the measuring point. The health scores of each measuring point at time t10 are calculated using an adaptive piecewise function. The health scores of measuring points are as follows: measuring point 1 has a health score of 0.4677, measuring point 2 has a health score of 0.4884, and measuring point 3 has a health score of 0.5488.

[0141] Step 4: Bottom-up health aggregation A weighted average method is used to aggregate layer by layer to obtain the health status of component layers, such as stator components. .

[0142] By aggregating upwards layer by layer to the system layer, the overall system health is obtained. H total .

[0143] Step 5: Confidence Assessment and Weight Update Step 5.1: Calculation of Grey Relational Analysis Confidence Five recent health status sequences were selected as references to calculate the correlation of the current sequence. .

[0144] [108.5, 107.2, 117.8]; [109.3, 108.1, 118.2]; [110.1, 108.9, 118.9]; [111.2, 109.8, 119.3]; [112.0, 110.5, 119.8].

[0145] The grey relational degree is [0.84479267, 0.83303166, 0.82104056, 0.85294239, 0.93709741], and the confidence level of the assessment result is 0.8578. The health confidence level calculation reflects the overall deterioration state, that is, the low relational degree when no warning occurs and the high relational degree when a warning occurs, indicating that the assessment result has a high degree of credibility.

[0146] Step 5.2: Time-varying weight update After obtaining new data at the 11th time point, the data window is slid to retain the most recent 10 points, and the objective weight and comprehensive weight are recalculated. The historical weights are updated using the forgetting factor. The weight increment in the new data window is based on the original weights [0.32278589, 0.28276264, 0.39445147], and the new weights are [0.32381145, 0.28785659, 0.38833196].

[0147] An intelligent operation and maintenance system for implementing a multi-layer health assessment method for hydropower units based on dynamic comprehensive weighting and adaptive segmented health status is disclosed. The system modules specifically include a real-time monitoring data acquisition and preprocessing module 100, a weight calculation module 200, a health status calculation module 300, an aggregation analysis module 400, a dynamic update module 500, and a result visualization output and maintenance decision module 600, used to implement the method described in this embodiment. The specific intelligent operation and maintenance system, including each implementation module, is layered as follows: (1) Data acquisition and timestamp processing layer: For different sensor types, align the timestamps of multi-dimensional data such as temperature, pressure, vibration, displacement, and flow rate.

[0148] (2) Data Acquisition and Feature Processing Layer: In one specific embodiment, data cleaning includes removing outliers, filling in missing values, and standardization.

[0149] (3) Health assessment layer: Weight calculation module: Implements a multimodal fusion weight model; Health Calculation Module: Calculates the adaptive piecewise function; Aggregated assessment module: Bottom-up health aggregation; Dynamic update module: Adjust the sliding window and implement the forgetting factor mechanism.

[0150] (4) Results output and visualization layer: Health report generation; The real-time dashboard displays the health status curve and the confidence level of read faults; Early warning information push; According to the "health function curve," the piecewise function smoothly transitions at the threshold, and the shape parameters are based on expert experience. k 1, k 2, k 3. The function's form is adaptively adjusted to fit the actual changes on site.

[0151] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the multi-layer health assessment method for hydropower units based on dynamic comprehensive weights and adaptive segmented health as described in the embodiments of this invention.

[0152] In summary, through the systematic demonstration of the above embodiments, this implementation method exhibits significant advantages in actual hydropower unit health assessment. The multimodal weighted fusion model considers both subjective and objective information, improving the rationality of weights. Health calculation is accurate, with an adaptive piecewise function dynamically responding to changes in operating conditions, and threshold and shape parameters are flexibly adjustable. The system is highly adaptive, with a time-varying weight update mechanism enabling continuous model optimization to adapt to unit condition evolution. The results have high reliability, with grey relational confidence assessment providing a reliable basis for decision-making. It has strong engineering applicability, with modular design supporting system integration, and visual output facilitating use by operation and maintenance personnel.

[0153] The above description of several specific embodiments further details the technical solution provided by the present invention in order to highlight the advantages and benefits of the technical solution provided by the present invention. However, the above-described specific embodiments are not intended to limit the present invention. Any reasonable modifications and improvements to the present invention, combinations of embodiments, and equivalent substitutions based on the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A multi-level health assessment method for hydropower units with dynamic comprehensive weighting and adaptive segmented health, characterized in that, include: The steps include: obtaining the equipment hierarchy of hydropower units and the monitoring indicator data corresponding to each level; establishing a health assessment structure including at least two evaluation levels; and determining the correspondence between each monitoring indicator and the superior evaluation object. The steps to determine the comprehensive weight of each evaluation object based on the importance relationship of each evaluation object within the same evaluation level, the relationship of changes in monitoring indicator data, and the correlation between evaluation objects; The steps to calculate the health status of each monitoring indicator based on the current monitoring data, corresponding threshold ranges, and the range change relationship with the operating status of each indicator. According to the health assessment structure, the health of the lower-level evaluation objects is aggregated step by step according to the corresponding comprehensive weight to obtain the health of the upper-level evaluation objects, until the health of the entire hydropower unit is obtained. The steps for outputting the overall health status of the hydropower unit and the health status of the corresponding evaluation objects at each level.

2. The multi-level health assessment method for hydropower units with dynamic comprehensive weighting and adaptive segmented health as described in claim 1, characterized in that, The health assessment structure is established according to the physical hierarchy of the hydropower units. The evaluation hierarchy includes a system layer, a subsystem layer, a component layer, and a monitoring indicator layer. Each monitoring indicator in the monitoring indicator layer corresponds to its respective component.

3. The multi-level health assessment method for hydropower units with dynamic comprehensive weighting and adaptive segmented health as described in claim 1, characterized in that, The importance relationship is determined by the relative importance of each evaluation object within the same evaluation level, and participates in the determination of the comprehensive weight after the consistency constraint is met.

4. The multi-level health assessment method for hydropower units with dynamic comprehensive weighting and adaptive segmented health as described in claim 1, characterized in that, The relationship between the changes in the monitoring indicator data is determined based on the degree of change in the historical data of each monitoring indicator, the data time attribute, and the current operating status, and is involved in the determination of the comprehensive weight.

5. The multi-level health assessment method for hydropower units with dynamic comprehensive weighting and adaptive segmented health as described in claim 1, characterized in that, The relationships between the evaluation objects are determined based on the direct and indirect influence relationships between the evaluation objects within the same evaluation level, and are involved in the determination of the comprehensive weight.

6. The multi-level health assessment method for hydropower units with dynamic comprehensive weighting and adaptive segmented health as described in claim 1, characterized in that, The health status of each monitoring indicator is determined based on the relative position of the current monitoring data within the corresponding threshold range and the change relationship corresponding to that threshold range.

7. A multi-layer health assessment device for hydropower units with dynamic comprehensive weighting and adaptive segmented health, characterized in that, include: A module for acquiring the equipment hierarchy of hydropower units and the corresponding monitoring index data for each level, establishing a health assessment structure including at least two evaluation levels, and determining the correspondence between each monitoring index and the higher-level evaluation object; A module that determines the comprehensive weight of each evaluation object based on the importance relationship of each evaluation object within the same evaluation level, the relationship of changes in monitoring indicator data, and the correlation between evaluation objects; A module that calculates the health status of each monitoring indicator based on the current monitoring data, corresponding threshold ranges, and the range change relationship with the operating status. Based on the health assessment structure, the health of the lower-level evaluation objects is aggregated level by level according to the corresponding comprehensive weight to obtain the health of the upper-level evaluation objects, until the module of the overall health of the hydropower unit is obtained; A module that outputs the overall health status of the hydropower unit and the health status of the corresponding evaluation objects at each level.

8. A computer storage medium for storing computer programs, characterized in that, When the computer program is read by the computer, the computer executes the method of claim 1.

9. A computer, comprising a processor and a storage medium, characterized in that, When the processor reads the computer program stored in the storage medium, the computer executes the method of claim 1.

10. A computer program product, as a computer program, is characterized by: When the computer program is executed, it implements the method of claim 1.

Images