An intelligent monitoring system for hydraulic parameters of a hydropower station

Abstract

The application belongs to the technical field of hydropower station data monitoring, and provides a kind of intelligent monitoring system for hydropower station hydraulic parameter, comprising: perception subsystem, for collecting the multidimensional monitoring data of hydraulic parameter in each key monitoring area in hydropower station;Processing subsystem, for intelligent fault diagnosis and equipment performance prediction of hydropower station according to multidimensional monitoring data, combined with statistical analysis, signal processing and machine learning strategy, and according to the obtained fault diagnosis result and performance prediction result, generate dynamic maintenance strategy;Visual subsystem, for visualizing multidimensional monitoring data, fault diagnosis result and performance prediction result in parameter monitoring area, and visualizing dynamic maintenance strategy in maintenance decision area.The system realizes the full-process intelligent upgrading of hydraulic parameter monitoring and operation and maintenance decision through the collaborative operation of perception, processing and visualization subsystems, which can significantly improve the safety and stability of hydropower station operation.

Description

Technical Field

[0001] This invention relates to the field of hydropower station data monitoring technology, and in particular to an intelligent monitoring system for hydraulic parameters of a hydropower station. Background Technology

[0002] Against the backdrop of the energy structure transitioning towards clean and low-carbon energy, hydropower stations, as core infrastructure for renewable energy utilization, play a crucial role in ensuring energy supply through their safe, stable operation and efficient power generation. The operational status of hydropower stations heavily relies on the precise monitoring of hydraulic parameters such as dam water level, unit flow rate, spiral casing pressure, tailrace pulsation, and sediment content. These parameters not only directly reflect the hydrological situation of the basin and the operating conditions of the equipment, but also serve as the core basis for judging equipment failures, predicting performance degradation, and formulating maintenance plans.

[0003] However, traditional hydropower station hydraulic parameter monitoring systems have significant technical limitations. On the one hand, in the data acquisition stage, traditional systems mostly use single-type sensors for fixed-point monitoring combined with manual data collection. This results in problems such as incomplete monitoring dimensions, data accuracy being easily affected by environmental interference, and poor real-time performance. It is difficult to capture the dynamic changes of hydraulic parameters under complex operating conditions, leading to a lack of data foundation to support accurate analysis. On the other hand, in the data processing stage, traditional systems rely on simple statistical analysis and manual experience for fault diagnosis. They cannot identify complex faults such as rotor imbalance and shaft misalignment, and the trend prediction function leads to delayed fault warnings and passive maintenance decisions.

[0004] This shows that traditional hydropower station hydraulic parameter monitoring systems suffer from technical problems such as low monitoring accuracy, poor reliability, and incomplete functionality. Summary of the Invention

[0005] This invention provides an intelligent monitoring system for hydraulic parameters of hydropower stations, which addresses the shortcomings of existing traditional hydropower station hydraulic parameter monitoring systems, such as low monitoring accuracy, poor reliability, and insufficient functionality.

[0006] This invention provides an intelligent monitoring system for hydraulic parameters of a hydropower station, comprising: The sensing subsystem is used to collect multi-dimensional monitoring data of hydraulic parameters in key monitoring areas within the hydropower station. The processing subsystem is used to perform intelligent fault diagnosis and equipment performance prediction for hydropower stations based on the multidimensional monitoring data, combined with statistical analysis, signal processing and machine learning strategies, and to generate dynamic maintenance strategies based on the obtained fault diagnosis results and performance prediction results. The visualization subsystem is used to visualize the multidimensional monitoring data, fault diagnosis results, and performance prediction results in the parameter monitoring area, and to visualize the dynamic maintenance strategy in the maintenance decision area.

[0007] According to the intelligent monitoring system for hydraulic parameters of a hydropower station provided by the present invention, the sensing subsystem includes: The first pressure acquisition component is used to collect water level and trash rack pressure difference data within the entire station monitoring area; Ultrasonic flow meters are used to collect water flow data within the monitoring area of ​​the unit section. The second pressure acquisition component is used to collect pressure data of key parts of the generator set within the monitoring area of ​​the generator section. The vibration swing measurement component is used to collect vibration amplitude and swing deviation data of generator sets within the monitoring area of ​​the generator section; Sediment measurement unit is used to collect the sediment content of the water in the forebay.

[0008] According to the intelligent monitoring system for hydraulic parameters of a hydropower station provided by the present invention, the processing subsystem, based on the multidimensional monitoring data and combined with statistical analysis, signal processing, and machine learning strategies, performs intelligent fault diagnosis and equipment performance prediction for the hydropower station, including: Statistical analysis was performed on the key data values ​​in the multidimensional monitoring data to obtain hydraulic parameter statistics; Signal processing is performed on the key raw signals in the multidimensional monitoring data to extract key signal features; Based on the hydraulic parameter statistics and the key signal characteristics, abnormal data in the multidimensional monitoring data are marked as abnormal. The multidimensional monitoring data marked with anomalies are input into the pre-built fault diagnosis model and performance prediction model respectively to obtain fault diagnosis results and performance prediction results.

[0009] According to the intelligent monitoring system for hydraulic parameters of hydropower stations provided by the present invention, the fault diagnosis model includes: The decision tree module is used to perform preliminary fault classification on multidimensional monitoring data after anomaly labeling based on multi-branch decision rules, and output the preliminary classification results; The Support Vector Machine module is used to construct a Support Vector Machine sub-model for each fault type in the preliminary classification results, and to refine the identification of the monitoring data of the corresponding fault type based on the Support Vector Machine sub-model, and output the refined classification results. The fusion and collaboration module is used to fuse and verify the preliminary classification results and the refined classification results, and output the fault diagnosis results.

[0010] According to the intelligent monitoring system for hydraulic parameters of hydropower stations provided by the present invention, the performance prediction model includes: The input layer is used to convert the multidimensional monitoring data after anomaly marking into data feature vectors; The hidden layer is used to extract the hidden features in each dimension of the data feature vector, and to fuse the hidden features in all dimensions to output a comprehensive feature vector. The output layer is used to convert the comprehensive feature vector into performance index values ​​and output the performance prediction results.

[0011] According to the intelligent monitoring system for hydraulic parameters of a hydropower station provided by the present invention, the processing subsystem generates a dynamic maintenance strategy based on the obtained fault diagnosis results and performance prediction results, including: A correlation analysis was performed on the obtained fault diagnosis results and performance prediction results to obtain preliminary correlation results; Based on the preliminary association results, the current fault level is determined, and based on the pre-established mapping relationship between fault levels and maintenance strategies, the preliminary maintenance strategy corresponding to the current fault level is determined. During the execution of the initial maintenance strategy, the initial maintenance strategy is dynamically adjusted according to the measured key parameters of the maintenance location to generate a dynamic maintenance strategy.

[0012] According to the intelligent monitoring system for hydraulic parameters of hydropower stations provided by this invention, a correlation analysis is performed on the obtained fault diagnosis results and performance prediction results to obtain preliminary correlation results, including: The obtained fault diagnosis results and performance prediction results are integrated, and the integrated data is standardized and coded to obtain comprehensive result data. Determine the degree of correlation between fault features and performance indicators in the comprehensive result data, and extract key correlation features based on the degree of correlation; The key association features are used as the preliminary association results.

[0013] According to the intelligent monitoring system for hydraulic parameters of a hydropower station provided by the present invention, the visualization subsystem visualizes the multi-dimensional monitoring data, fault diagnosis results, and performance prediction results in the parameter monitoring area, including: Create a three-dimensional spatial map of the hydropower station; The multidimensional monitoring data, fault diagnosis results, and performance prediction results are annotated on the three-dimensional spatial map to obtain a three-dimensional data annotation map. The three-dimensional data marker map is visualized in the parameter monitoring area.

[0014] According to the intelligent monitoring system for hydraulic parameters of a hydropower station provided by the present invention, the system further includes: an early warning subsystem; The early warning subsystem is used to perform graded early warning tasks when abnormal hydraulic parameters are detected in the hydropower station based on the fault diagnosis results and performance prediction results.

[0015] The intelligent monitoring system for hydraulic parameters of a hydropower station provided by the present invention further includes: a remote control subsystem; The remote control subsystem is used to modify at least some key parameters of the sensing subsystem and / or the processing subsystem in response to the remote control command after receiving and verifying the remote control command initiated by the user.

[0016] The intelligent monitoring system for hydraulic parameters in hydropower stations provided by this invention achieves a fully intelligent upgrade of the entire process of hydraulic parameter monitoring and operation and maintenance decision-making through the coordinated operation of three subsystems: sensing, processing, and visualization. Specifically, the sensing subsystem can comprehensively collect multi-dimensional hydraulic data from key areas, solving the problems of single-dimensional monitoring and poor real-time performance in traditional monitoring, and providing an accurate and complete data foundation for subsequent analysis. The processing subsystem integrates statistical analysis, signal processing, and machine learning strategies, which can accurately diagnose equipment faults, predict performance trends in advance, and generate dynamic maintenance strategies adapted to actual operating conditions, effectively avoiding the lag and blindness of traditional manual experience-based judgments, and reducing fault losses and excessive maintenance costs. The visualization subsystem displays data in partitions, transforming complex data into intuitive visual information, significantly reducing the difficulty of data interpretation for operation and maintenance personnel, and helping them quickly grasp the operating status of equipment and efficiently execute maintenance decisions. Overall, this system can significantly improve the safety and stability of hydropower station operation. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the structure of the intelligent monitoring system for hydraulic parameters of a hydropower station provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the sensing subsystem. Figure 3 This is a schematic diagram of the fault diagnosis model; Figure 4 This is a schematic diagram of the performance prediction model. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0020] The following is combined Figures 1 to 4 This invention describes the detailed scheme of the intelligent monitoring system for hydraulic parameters of hydropower stations provided in an embodiment of the present invention.

[0021] like Figure 1 As shown in the figure, the intelligent monitoring system for hydraulic parameters of a hydropower station provided in this embodiment of the invention mainly includes: The sensing subsystem 110 is used to collect multi-dimensional monitoring data of hydraulic parameters in key monitoring areas within the hydropower station.

[0022] The processing subsystem 120 is used to perform intelligent fault diagnosis and equipment performance prediction for hydropower stations based on multidimensional monitoring data, combined with statistical analysis, signal processing and machine learning strategies, and to generate dynamic maintenance strategies based on the obtained fault diagnosis results and performance prediction results.

[0023] The visualization subsystem 130 is used to visualize multidimensional monitoring data, fault diagnosis results, and performance prediction results in the parameter monitoring area, and to visualize dynamic maintenance strategies in the maintenance decision area.

[0024] In this embodiment, the sensing subsystem 110 is densely distributed in various key parts of the hydropower station, responsible for accurately capturing various hydraulic parameters and equipment operating status information, so as to realize the real-time collection of multi-dimensional monitoring data.

[0025] In one embodiment, such as Figure 2 As shown, the sensing subsystem 110 specifically includes: The first pressure acquisition component 210 is used to acquire water level and trash rack pressure difference data within the entire station monitoring area.

[0026] In full-station measurements, the first pressure acquisition component 210 can be implemented using a submersible pressure sensor, specifically installed at key locations such as the dam water level, the intake trash rack, the upstream and downstream of the sluice gate, and the forebay. The first pressure acquisition component 210 primarily achieves pressure acquisition based on the conversion of pressure into electrical signals. When subjected to water pressure, the sensitive element inside the submersible pressure sensor deforms, leading to changes in electrical quantities such as resistance or capacitance. By accurately measuring these changes in electrical quantities, the water pressure value at the corresponding location can be accurately calculated. Then, based on the relationship between water pressure and water level, the water level height can be calculated. Submersible pressure sensors have significant advantages such as high accuracy, strong stability, and fast response speed, meeting the stringent requirements of hydropower stations for high-precision measurement of water level parameters.

[0027] The ultrasonic flow meter 220 is used to collect water flow in the monitoring area of ​​the unit section.

[0028] In the unit section measurement stage, the ultrasonic flow meter 220 can accurately measure the water flow rate inside the steel pipe. In practical applications, the ultrasonic flow meter 220 can calculate the fluid velocity by utilizing the speed difference of ultrasonic waves propagating in the fluid, and thus obtain the flow rate data. This measurement method is non-contact, has no pressure loss, and has almost no interference with the fluid flow state, providing reliable data support for the flow monitoring of hydropower station units.

[0029] The second pressure acquisition component 230 is used to acquire pressure data of key parts of the generator set within the monitoring area of ​​the generator section.

[0030] The second pressure acquisition component 230 can measure parameters such as the inlet pressure of the spiral casing, the end pressure of the spiral casing, the inlet pressure of the draft tube, the pulsating pressure of the draft tube, and the pressure of the turbine top cover. The second pressure acquisition component 230 can be composed of a pressure gauge and a pressure transmitter. The pressure gauge displays the pressure value intuitively through the deformation of the elastic element, while the pressure transmitter converts the pressure signal into a standard electrical signal, facilitating remote transmission and automated control, and ensuring accurate monitoring of the pressure of key parts of the unit.

[0031] The vibration swing measurement component 240 is used to collect vibration amplitude and swing deviation data of the generator set within the monitoring area of ​​the unit section.

[0032] In this embodiment, to comprehensively monitor the unit's operating status, each power station is also equipped with a vibration and sway measurement component 240, primarily employing eddy current sensors. Eddy current sensors are based on the eddy current effect; when an eddy current sensor approaches a metal conductor, eddy currents are generated on the conductor's surface. The magnitude of these eddy currents is closely related to the distance between the sensor and the conductor. By accurately measuring changes in these eddy currents, real-time information on unit vibration and sway can be obtained, providing crucial data for assessing the unit's operational stability and health status.

[0033] Sediment measurement component 250 is used to collect the sediment content of the water in the forebay.

[0034] In terms of sediment monitoring, the sediment measurement component 250 mainly includes a sludge interface detector and a sediment monitor. The sludge interface detector uses ultrasonic or optical principles to accurately measure the interface position of sludge in the forebay, allowing for timely monitoring of sediment accumulation. The sediment monitor uses various technologies such as optics, acoustics, or electromagnetics to quantitatively analyze the sediment content in the water, providing strong support for the safe operation of hydropower stations and water resource management.

[0035] In this embodiment, the processing subsystem undertakes the important task of in-depth mining and analysis of multi-dimensional monitoring data, which can provide a scientific basis for the operation and management of hydropower stations.

[0036] In some embodiments, before performing fault diagnosis and performance prediction, the multidimensional monitoring data can be preprocessed, starting with data cleaning. Since sensor-collected data may be affected by various interference factors, such as electromagnetic interference and environmental noise, the data may contain noise points, outliers, and missing values. Data cleaning can utilize various algorithms and techniques to remove this noise and outlier data, and to appropriately fill in missing values.

[0037] In practical applications, statistical methods such as mean, median, and mode can be used to impute missing values ​​in continuous data. For outliers, tools like box plots can be used to identify and remove them by setting reasonable threshold ranges. Afterward, data denoising can be performed using filtering algorithms such as low-pass, high-pass, and band-pass filters to remove high-frequency or low-frequency noise, making the data smoother and more accurate.

[0038] Furthermore, data standardization is a crucial step. It transforms data with different dimensions and value ranges into a specific interval, such as [0, 1] or [-1, 1], eliminating dimensional differences between data features and facilitating subsequent data analysis and model training. When analyzing data on different parameters such as water level, flow rate, and pressure, standardization enables these data to be compared and analyzed on the same scale, improving the accuracy and reliability of the analysis results.

[0039] In one embodiment, the processing subsystem, based on multidimensional monitoring data and combined with statistical analysis, signal processing, and machine learning strategies, performs intelligent fault diagnosis and equipment performance prediction for the hydropower station, specifically including: On the one hand, statistical analysis is performed on key data values ​​in the multidimensional monitoring data to obtain hydraulic parameter statistics.

[0040] In this step, by calculating the mean, variance, standard deviation, and other statistical measures of key data values ​​in the multidimensional monitoring data, we can understand the central tendency and dispersion of the data and assess the stability of the hydropower station's operating parameters.

[0041] On the other hand, key raw signals in the multidimensional monitoring data are processed to extract key signal features.

[0042] In this step, for vibration, pressure, and other signal data, signal processing techniques such as Fourier transform and wavelet transform can be used to convert the time-domain signal into a frequency-domain signal, analyze the frequency components of the signal, and extract key signal features related to equipment failure, hydraulic anomalies, etc. By analyzing the frequency of the unit's vibration signal through Fourier transform, if an abnormally large increase in the vibration amplitude at a specific frequency is found, it may indicate problems such as unit imbalance or bearing failure.

[0043] Then, based on hydraulic parameter statistics and key signal characteristics, abnormal data in the multidimensional monitoring data are marked as anomalies.

[0044] In this step, trend change or threshold comparison analysis methods can be used to extract abnormal data from multidimensional monitoring data by combining hydraulic parameter statistics and key signal features. For example, abnormal values ​​of hydraulic parameter statistics exceeding the corresponding thresholds and abnormal feature quantities such as abnormal increases in key signal features can be extracted. Then, corresponding abnormal labels can be established for the abnormal data to achieve anomaly marking.

[0045] Finally, the multidimensional monitoring data marked with anomalies are input into the pre-built fault diagnosis model and performance prediction model to obtain fault diagnosis results and performance prediction results.

[0046] In this embodiment, the fault diagnosis model can perform fault diagnosis on multi-dimensional monitoring data after anomaly marking, thereby obtaining fault diagnosis results; the performance prediction model can use multi-dimensional monitoring data after anomaly marking to predict equipment performance, thereby obtaining performance prediction results.

[0047] In the fault diagnosis stage, the fault diagnosis model adopts a model architecture that combines decision trees and support vector machines. The core is to use the strong interpretability of decision trees to complete the initial classification of fault categories, and then use the high accuracy of support vector machines to achieve detailed fault location, forming an efficient diagnosis process of coarse screening and fine judgment.

[0048] In one embodiment, such as Figure 3 As shown, the fault diagnosis model specifically includes: The decision tree module 310 is used to perform preliminary fault classification on the multidimensional monitoring data after anomaly marking based on multi-branch decision rules, and output the preliminary classification results.

[0049] The decision tree module 310, serving as the front-end classifier of the fault diagnosis model, primarily functions to construct multi-branch decision rules based on historical fault data and multi-dimensional monitoring data after anomaly labeling in hydropower stations. For example, using whether the vibration amplitude exceeds 0.15 mm as the first decision node, faults are initially classified into vibration-related faults and non-vibration-related faults. Then, using whether the inlet pressure deviation of the volute casing is greater than 5% as a secondary node, non-vibration-related faults are further subdivided into major categories such as pressure anomaly faults and flow anomaly faults. The advantage of the decision tree module 310 lies in its intuitive and easy-to-understand decision logic, its ability to quickly eliminate irrelevant fault types, reduce subsequent computational load, and facilitate management personnel's understanding of the fault classification criteria.

[0050] The Support Vector Machine module 320 is used to construct a Support Vector Machine sub-model for each fault type in the preliminary classification results, and to refine the identification of the monitoring data of the corresponding fault type based on the Support Vector Machine sub-model, and output the refined classification results.

[0051] The Support Vector Machine (SVM) module 320 serves as the backend precision analyzer of the model. For each fault category partitioned by the decision tree module 310, it constructs a dedicated SVM sub-model to achieve accurate identification of subdivided faults. For example, for vibration faults, the SVM sub-model uses more granular feature parameters such as radial vibration frequency, axial vibration amplitude, and bearing temperature as input. Through a kernel function, it maps the data to a high-dimensional space to construct the optimal classification hyperplane, thereby distinguishing specific fault types such as rotor imbalance, shaft misalignment, and bearing wear. For pressure anomaly faults, it uses features such as the pressure difference between the inlet and outlet of the volute, the pressure fluctuation period of the top cover, and the vacuum degree of the tailrace pipe to accurately identify subdivided problems such as abnormal guide vane opening and pipe blockage. The core value of the SVM module 320 lies in its ability to effectively improve the identification accuracy of small-sample faults when processing high-dimensional, nonlinear monitoring data, avoiding the classification bias of a single decision tree in complex fault scenarios.

[0052] The fusion and collaboration module 330 is used to fuse and verify the preliminary classification results and the refined classification results, and output the fault diagnosis results.

[0053] The fusion and collaboration module 330 serves as the connection and result verification function between the decision tree module 310 and the support vector machine module 320. On one hand, the decision tree module 310 automatically transmits the preliminary classification results and corresponding feature parameters to the dedicated SVM sub-model in the support vector machine module 320, avoiding redundant data transmission. On the other hand, the refined classification results output by the SVM sub-model are back-matched and verified against the classification rules of the decision tree module 310. If a conflict occurs, such as the decision tree module 310 determining it as an abnormal pressure, but the support vector machine module 320 identifying it as pressure fluctuations caused by vibration, a secondary verification is performed using a historical database of similar fault cases to correct the diagnostic bias. Simultaneously, the fusion and collaboration module 330 records the feature parameters, classification path, and results of each diagnosis, continuously optimizing the branching rules of the decision tree module 310 and the kernel function parameters of the support vector machine module 320, thereby improving the long-term diagnostic stability of the entire fault diagnosis model.

[0054] In one embodiment, such as Figure 4 As shown, the performance prediction model specifically includes: Input layer 410 is used to convert the anomaly-marked multidimensional monitoring data into a data feature vector.

[0055] In this embodiment, the input layer 410 can perform feature filtering on the multidimensional monitoring data after anomaly labeling. Specifically, it uses Pearson correlation coefficient analysis to select parameters with strong correlation to the prediction target as input features, such as guide vane opening, head, and influent flow rate, which have a correlation coefficient higher than 0.7 with turbine efficiency. Irrelevant features are eliminated to reduce the computational load of the model and improve generalization ability. Then, the selected input features can be constructed into a one-dimensional data feature vector. The number of neurons in the input layer 410 is consistent with the dimension of the feature parameters in the data feature vector, with each neuron corresponding to one feature parameter.

[0056] Hidden layer 420 is used to extract the hidden features in each dimension of the data feature vector and fuse the hidden features in all dimensions to output a comprehensive feature vector.

[0057] Understandably, hidden layer 420 is the core of the performance prediction model, enabling feature mapping and performance prediction. Specifically, it uses non-linear computation across multiple neurons to uncover the complex correlation between input features and output performance metrics. This can be divided into feature extraction sub-layers and feature fusion sub-layers. The feature extraction sub-layers can have 2-3 layers, primarily focusing on deep mining of single features. Each feature extraction sub-layer consists of multiple neurons and uses the ReLU activation function to avoid the gradient vanishing problem. Its core function is to deeply process the single feature passed from the input layer, extracting latent features related to the performance metrics.

[0058] Specifically, the first-layer feature extraction sublayer targets the basic features of the input layer and, through linear computation of neurons and ReLU activation, mines local correlations among these basic features. For example, it extracts the product correlation between head and guide vane opening to reflect the impact intensity of water flow on the turbine, and extracts the correlation between flow rate and rotational speed to reflect the matching degree between the turbine's flow capacity and rotational state. The second and third-layer feature extraction sublayers, based on the locally correlated features extracted in the previous layer, further mine the deep interactions among these correlated features, providing crucial support for subsequent fusion calculations.

[0059] The feature fusion sublayer mainly focuses on the collaborative association of multiple features. Specifically, the Sigmoid activation function can be used to control the output in the range of [0,1], which is convenient for matching with the performance index range. The core function is to collaboratively fuse multiple deep features obtained from the feature extraction sublayer to construct a mapping relationship between multiple features and performance index.

[0060] Specifically, in the feature weight allocation stage, backpropagation during model training can be used to assign differentiated weights to different deep features, ensuring that the fusion process focuses on key influencing factors. In the nonlinear fusion calculation stage, nonlinear operations of neurons can be used to fuse multi-dimensional deep features into a comprehensive feature vector, providing a calculation basis for the output layer 430.

[0061] Output layer 430 is used to convert the comprehensive feature vector into performance index values ​​and output the performance prediction results.

[0062] Understandably, the number of neurons in output layer 430 is consistent with the number of prediction targets, and its core function is to transform the comprehensive feature vector into specific performance indicator prediction values.

[0063] Specifically, in the linear mapping calculation stage, a linear activation function can be used to map the comprehensive feature vector into predicted performance index values ​​through linear calculation. In the result standardization and restoration stage, the predicted performance index values ​​can be converted into actual physical units through inverse Min-Max standardization, and the final output is a performance prediction result that can be directly used for hydropower station operation and scheduling.

[0064] In one embodiment, the processing subsystem generates a dynamic maintenance strategy based on the obtained fault diagnosis results and performance prediction results, specifically including: First, a correlation analysis was performed on the obtained fault diagnosis results and performance prediction results to obtain preliminary correlation results.

[0065] In a specific implementation, a correlation analysis is performed on the obtained fault diagnosis results and performance prediction results to obtain preliminary correlation results, which include: The first step is to integrate the obtained fault diagnosis results and performance prediction results, and then standardize and encode the integrated data to obtain comprehensive result data.

[0066] In this step, the fault diagnosis results and performance prediction results can be integrated one-to-one using unique identifiers such as timestamps and unit numbers to form a correlated dataset containing fault characteristics and performance indicators. Then, the integrated data is standardized. For continuous parameters in the performance prediction results, Min-Max standardization is used to map the values ​​to the [0,1] interval to eliminate the influence of differences in the magnitude of different parameters. For categorical data in the fault diagnosis results, label encoding or one-hot encoding is used to convert them into machine-recognizable numerical forms, finally obtaining a unified format of comprehensive result data that can be directly used for analysis.

[0067] The second step is to determine the degree of correlation between fault characteristics and performance indicators in the comprehensive result data, and extract key correlation features based on the degree of correlation.

[0068] In this step, a combination of statistical analysis and feature importance assessment can be used to determine the correlation between fault features and performance indicators. On the one hand, the linear or nonlinear correlation between fault levels and predicted values ​​of various performance indicators is quantified by calculating Pearson correlation coefficients and Spearman rank correlation coefficients. On the other hand, a lightweight machine learning model is constructed, taking fault features as input and abnormal performance indicator states as output. The influence weight of each fault feature on the performance indicator is evaluated by the feature importance score output after model training. Subsequently, by combining the correlation coefficient and feature importance score, a correlation threshold is set to screen out fault features closely related to the performance indicators, i.e., key correlation features.

[0069] The third step is to use key association features as preliminary association results.

[0070] Then, based on the preliminary correlation results, the current fault level is determined, and according to the pre-established mapping relationship between fault level and maintenance strategy, the preliminary maintenance strategy corresponding to the current fault level is determined.

[0071] In this embodiment, the core correlation features corresponding to the current fault can be extracted from the preliminary correlation results. These features are then combined with the fault level determination criteria pre-set by the hydropower station. Specifically, these criteria can be based on historical fault data and equipment safety thresholds. For example, a vibration amplitude of 0.1-0.15 mm and an efficiency decrease of 1%-3% corresponds to fault level 1; a vibration amplitude > 0.15 mm and an efficiency decrease > 3% corresponds to fault level 2. By comparing the correlation feature parameters and determination criteria of the current fault, the current fault level is determined. Then, a pre-established fault level and maintenance strategy mapping relationship library is invoked. This library, built using expert experience and historical maintenance cases, contains information such as maintenance methods, priorities, and required resources corresponding to different fault levels and types. Based on the current fault level and type, the corresponding preliminary maintenance strategy is matched from the mapping relationship library to ensure accurate matching between the strategy and the severity and scope of the fault.

[0072] Finally, during the execution of the initial maintenance strategy, the initial maintenance strategy is dynamically adjusted according to the measured key parameters of the maintenance parts to generate a dynamic maintenance strategy.

[0073] During the initial maintenance strategy execution, key parameters of the maintenance location can be collected in real time, such as the real-time vibration amplitude during rotor dynamic balancing correction, the opening deviation value and corresponding efficiency change value during guide vane adjustment, and the measured key parameters can be compared with the expected parameters in the initial maintenance strategy in real time. If the measured key parameters meet the expectations, the initial maintenance strategy continues to be executed. If the measured parameters deviate from the expectations, the cause of the deviation is analyzed by combining the correlation between faults and performance in the initial correlation results. Subsequently, maintenance measures are adjusted based on the cause of the deviation and the measured parameters, and maintenance resource allocation and maintenance time nodes are updated simultaneously to form a dynamic maintenance strategy adapted to the current actual working conditions. During the dynamic adjustment process, the parameters of the maintenance location need to be continuously monitored until the measured parameters reach the expected standards to ensure the effectiveness and pertinence of the maintenance strategy.

[0074] In one embodiment, the visualization subsystem visualizes multi-dimensional monitoring data, fault diagnosis results, and performance prediction results in the parameter monitoring area, specifically including: First, a three-dimensional spatial map of the hydropower station is created.

[0075] In practical applications, LiDAR can be used to scan the actual scene of a hydropower station to obtain 3D point cloud data of entities such as dams, powerhouses, generating units, and sensor installation points. Simultaneously, the BIM design model of the hydropower station is collected, and the point cloud data is fused and calibrated with the BIM design model to correct deviations between the actual scan and the design model. Then, using a 3D engine such as Cesium or Three.js, the fused model data is imported, high-precision satellite imagery is overlaid as the map base, and terrain elevation data is added to construct a three-level 3D spatial map containing macroscopic, mesoscopic, and microscopic scenes. Finally, basic interactive functions are added to the map, supporting zooming, rotation, and panning operations, as well as quick positioning by region or equipment type, ensuring that staff can easily view the spatial layout of different areas.

[0076] Then, the multidimensional monitoring data, fault diagnosis results, and performance prediction results are annotated on a three-dimensional spatial map to obtain a three-dimensional data annotation map.

[0077] In this step, a mapping relationship between data and spatial location can be established first. Based on the physical installation coordinates of each sensor in the sensing subsystem, the multidimensional monitoring data can be bound to the corresponding device or sensor locations in the 3D map. Then, differentiated annotation formats can be designed. For example, for monitoring data, dynamic dashboards or numerical labels can be used; for fault diagnosis results, icon overlay and flashing effects can be used; and for performance prediction results, trend curve pop-up windows can be used. Finally, the updates of multidimensional monitoring data and fault diagnosis results can be synchronized in real time through the data interface to ensure that the annotation content is dynamically synchronized with the system data, forming a 3D data marker map that includes spatial location and multidimensional data association.

[0078] Finally, the 3D data marker map is visualized in the parameter monitoring area.

[0079] In practical applications, a dedicated display interface can be built in the parameter monitoring area of ​​the visualization subsystem, using a layout of main view and sub-views. The main view embeds the constructed 3D data marker map, while the sub-views display data classification and filtering controls, such as filtering by monitored data type, fault level, and prediction time range. Staff can quickly focus on key information by using the interactive functions of the main view or filtering specific data through the sub-view controls. A data linkage mechanism is also set up; when a marker in the 3D map is clicked, the interface automatically pops up a detailed data panel corresponding to that fault and highlights the associated performance prediction markers. Furthermore, multi-terminal adaptation is supported. The complete 3D scene and detailed data are displayed on the PC, while the scene model is simplified and fault and key monitoring data markers are prioritized on mobile devices, ensuring efficient viewing of the 3D data marker map in different usage scenarios and achieving intuitive monitoring of the hydropower station's operating status.

[0080] In one embodiment, the aforementioned intelligent monitoring system for hydraulic parameters of a hydropower station may further include an early warning subsystem.

[0081] The early warning subsystem is used to perform graded early warning tasks when abnormal hydraulic parameters are detected in a hydropower station based on fault diagnosis results and performance prediction results.

[0082] In this embodiment, the early warning subsystem can first obtain fault diagnosis results and performance prediction results from the processing subsystem in real time, and establish joint fault and performance detection rules. Then, it compares and analyzes the obtained results according to the joint detection rules. If an anomaly that meets the early warning conditions is detected, it calls the pre-built grading standard. The grading standard can be divided into four levels according to the scope of the anomaly's impact and the degree of urgency: Level I may cause equipment shutdown or safety accidents, requiring a response within 1 hour; Level II affects unit efficiency but does not endanger safety for the time being, requiring a response within 6 hours; Level III has parameters that slightly deviate from the standard, requiring a response within 24 hours; and Level IV has performance prediction trends that are close to the threshold, requiring attention within 72 hours. The early warning level is determined by combining the specific parameters of the anomaly.

[0083] Finally, corresponding warning tasks are executed according to the warning level. For example, Level I and Level II warnings can be notified through three methods: system pop-ups, audible and visual alarms, and SMS push notifications to staff mobile phones, along with key information such as the abnormal location, related parameters, and preliminary emergency measures. Level III warnings are notified through system notifications and work group messages. Level IV warnings are only marked and reminded in the warning module of the visualization subsystem. At the same time, all warning information is stored in the warning log database in real time, recording the warning time, level, and processing status, which facilitates subsequent traceability and strategy optimization, and ensures that the hydropower station responds promptly to and manages risks related to abnormal hydraulic parameters.

[0084] In one embodiment, the aforementioned intelligent monitoring system for hydraulic parameters of a hydropower station may further include a remote control subsystem.

[0085] The remote control subsystem is used to modify at least some key parameters in the sensing subsystem and / or processing subsystem in response to a remote control command initiated by a user and after the command has been verified.

[0086] In this embodiment, the remote control command must include key information such as the target object to be operated on, the name of the parameter to be modified, the modified value, and the user's identity identifier.

[0087] Ideally, the verification process can employ a multi-layered verification mechanism. Specifically, the first layer verifies identity and permissions by comparing the user's identity with the system's preset permission list. If permissions do not match, the command is rejected and the reason is provided. The second layer verifies parameter reasonableness by calling a pre-stored parameter safety range library to determine whether the modified value exceeds the safety range. If it does, the user is prompted to adjust the value and given an explanation of the compliant range. The third layer performs a pre-assessment of the operation's impact by simulating the effect of parameter modification on monitoring data collection and fault diagnosis results through the processing subsystem. If the assessment indicates that the modification may cause system instability, a secondary confirmation is triggered, requiring the user to provide additional operational instructions before proceeding.

[0088] After all verifications pass, the remote control subsystem sends an encrypted parameter modification command to the target subsystem, and simultaneously records an operation log containing the operator, time, parameters before and after modification, and verification results. Upon receiving the command, the target subsystem executes the parameter update and feeds back the modification result to the remote control subsystem. The remote control subsystem then pushes the result to the user terminal in real time, and triggers parameter effectiveness detection in the sensing subsystem or processing subsystem to ensure that the modified parameters are accurately applied to the system, achieving safe and controllable remote parameter adjustment.

[0089] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An intelligent monitoring system for hydraulic parameters of a hydropower station, characterized in that, include: The sensing subsystem is used to collect multi-dimensional monitoring data of hydraulic parameters in key monitoring areas within the hydropower station. The processing subsystem is used to perform intelligent fault diagnosis and equipment performance prediction for hydropower stations based on the multidimensional monitoring data, combined with statistical analysis, signal processing and machine learning strategies, and to generate dynamic maintenance strategies based on the obtained fault diagnosis results and performance prediction results. The visualization subsystem is used to visualize the multidimensional monitoring data, fault diagnosis results, and performance prediction results in the parameter monitoring area, and to visualize the dynamic maintenance strategy in the maintenance decision area.

2. The intelligent monitoring system for hydraulic parameters of a hydropower station according to claim 1, characterized in that, The sensing subsystem includes: The first pressure acquisition component is used to collect water level and trash rack pressure difference data within the entire station monitoring area; Ultrasonic flow meters are used to collect water flow data within the monitoring area of ​​the unit section. The second pressure acquisition component is used to collect pressure data of key parts of the generator set within the monitoring area of ​​the generator section. The vibration swing measurement component is used to collect vibration amplitude and swing deviation data of generator sets within the monitoring area of ​​the generator section; Sediment measurement unit is used to collect the sediment content of the water in the forebay.

3. The intelligent monitoring system for hydraulic parameters of a hydropower station according to claim 1, characterized in that, The processing subsystem, based on the multidimensional monitoring data and combined with statistical analysis, signal processing, and machine learning strategies, performs intelligent fault diagnosis and equipment performance prediction for the hydropower station, including: Statistical analysis was performed on the key data values ​​in the multidimensional monitoring data to obtain hydraulic parameter statistics; Signal processing is performed on the key raw signals in the multidimensional monitoring data to extract key signal features; Based on the hydraulic parameter statistics and the key signal characteristics, abnormal data in the multidimensional monitoring data are marked as abnormal. The multidimensional monitoring data marked with anomalies are input into the pre-built fault diagnosis model and performance prediction model respectively to obtain fault diagnosis results and performance prediction results.

4. The intelligent monitoring system for hydraulic parameters of a hydropower station according to claim 3, characterized in that, The fault diagnosis model includes: The decision tree module is used to perform preliminary fault classification on multidimensional monitoring data after anomaly labeling based on multi-branch decision rules, and output the preliminary classification results; The Support Vector Machine module is used to construct a Support Vector Machine sub-model for each fault type in the preliminary classification results, and to refine the identification of the monitoring data of the corresponding fault type based on the Support Vector Machine sub-model, and output the refined classification results. The fusion and collaboration module is used to fuse and verify the preliminary classification results and the refined classification results, and output the fault diagnosis results.

5. The intelligent monitoring system for hydraulic parameters of a hydropower station according to claim 3, characterized in that, The performance prediction model includes: The input layer is used to convert the multidimensional monitoring data after anomaly marking into data feature vectors; The hidden layer is used to extract the hidden features in each dimension of the data feature vector, and to fuse the hidden features in all dimensions to output a comprehensive feature vector. The output layer is used to convert the comprehensive feature vector into performance index values ​​and output the performance prediction results.

6. The intelligent monitoring system for hydraulic parameters of a hydropower station according to claim 1, characterized in that, The processing subsystem generates a dynamic maintenance strategy based on the obtained fault diagnosis results and performance prediction results, including: A correlation analysis was performed on the obtained fault diagnosis results and performance prediction results to obtain preliminary correlation results; Based on the preliminary correlation results, the current fault level is determined, and based on the pre-established mapping relationship between fault levels and maintenance strategies, the preliminary maintenance strategy corresponding to the current fault level is determined. During the execution of the initial maintenance strategy, the initial maintenance strategy is dynamically adjusted according to the measured key parameters of the maintenance location to generate a dynamic maintenance strategy.

7. The intelligent monitoring system for hydraulic parameters of a hydropower station according to claim 6, characterized in that, A correlation analysis was performed on the obtained fault diagnosis results and performance prediction results to obtain preliminary correlation results, including: The obtained fault diagnosis results and performance prediction results are integrated, and the integrated data is standardized and coded to obtain comprehensive result data. Determine the degree of correlation between fault features and performance indicators in the comprehensive result data, and extract key correlation features based on the degree of correlation; The key association features are used as the preliminary association results.

8. The intelligent monitoring system for hydraulic parameters of a hydropower station according to claim 1, characterized in that, The visualization subsystem visualizes the multidimensional monitoring data, fault diagnosis results, and performance prediction results in the parameter monitoring area, including: Create a three-dimensional spatial map of the hydropower station; The multidimensional monitoring data, fault diagnosis results, and performance prediction results are annotated on the three-dimensional spatial map to obtain a three-dimensional data annotation map. The three-dimensional data marker map is visualized in the parameter monitoring area.

9. The intelligent monitoring system for hydraulic parameters of a hydropower station according to claim 1, characterized in that, The system also includes: an early warning subsystem; The early warning subsystem is used to perform graded early warning tasks when abnormal hydraulic parameters are detected in the hydropower station based on the fault diagnosis results and performance prediction results.

10. The intelligent monitoring system for hydraulic parameters of a hydropower station according to claim 1, characterized in that, The system also includes: a remote control subsystem; The remote control subsystem is used to modify at least some key parameters of the sensing subsystem and / or the processing subsystem in response to the remote control command after receiving and verifying the remote control command initiated by the user.

Images