A hydroelectric generating set operating condition monitoring system and apparatus

Abstract

The application provides a hydroelectric generating set operation state monitoring system and equipment, and relates to the technical field of hydroelectric generating set monitoring.The system comprises: an operation record acquisition module, which is used to acquire historical health operation records of similar units of a hydroelectric generating set; an interval prediction model generation module, which is used to obtain an interval prediction model; a real-time operation data acquisition module, which is used to acquire real-time operation data of the hydroelectric generating set; a health prediction interval acquisition module, which is used to obtain a dynamic health prediction interval; a real-time vibration coverage rate acquisition module, which is used to obtain a real-time vibration coverage rate; and an abnormal early warning module, which is used to perform vibration abnormal early warning on the hydroelectric generating set.The application solves the technical problems in the prior art that hydroelectric generating set monitoring relies on fixed thresholds, is easily disturbed by working conditions, and results in frequent false alarms and insufficient abnormal recognition accuracy, and achieves working condition self-adaptive and accurate monitoring of the operation state of the hydroelectric generating set and intelligent early warning of abnormalities, thereby improving the technical effect of vibration abnormal recognition accuracy.

Description

Technical Field

[0001] This invention relates to the field of hydropower unit monitoring technology, specifically to a hydropower unit operation status monitoring system and equipment. Background Technology

[0002] Currently, most hydropower unit operation status monitoring methods use fixed threshold judgments, judging abnormalities solely by whether the vibration amplitude exceeds a preset value. This fails to consider the normal fluctuation differences of the unit under different operating conditions, different years of operation, and special scenarios such as transient overvoltages, leading to false alarms and missed alarms in vibration monitoring. At the same time, traditional models are unable to dynamically predict healthy vibration ranges based on operating conditions and parameters, failing to achieve accurate adaptive monitoring of the unit's operating status and affecting the safe and stable operation of hydropower units and the effectiveness of early fault warnings.

[0003] Existing technologies suffer from technical problems such as reliance on fixed thresholds for monitoring hydropower units, susceptibility to operating condition interference leading to frequent false alarms, and insufficient accuracy in anomaly identification. Summary of the Invention

[0004] This application provides a hydropower unit operation status monitoring system and equipment, which aims to solve the technical problems in the prior art where hydropower unit monitoring relies on fixed thresholds, is easily affected by operating conditions leading to frequent false alarms, and has insufficient anomaly identification accuracy.

[0005] The first aspect disclosed in this application provides a hydropower unit operation status monitoring system, the system comprising: The system includes the following modules: an operation record acquisition module for acquiring historical healthy operation records of similar hydropower units, wherein the historical healthy operation records include multiple records with operating condition type identifiers; an interval prediction model generation module for training a long short-term memory network using the multiple records with operating condition type identifiers as reference data to obtain an interval prediction model; a real-time operation data acquisition module for acquiring real-time operation data of the hydropower unit, wherein the real-time operation data includes real-time operating condition type, real-time key parameters, and real-time vibration data; a health prediction interval acquisition module for predicting and analyzing the real-time operating condition type and the real-time key parameters using the interval prediction model to obtain a dynamic health prediction interval; a real-time vibration coverage acquisition module for introducing a sliding time window to compare and analyze the real-time vibration data with the dynamic health prediction interval to obtain a real-time vibration coverage rate; and an anomaly warning module for issuing a vibration anomaly warning for the hydropower unit when the real-time vibration coverage rate is not at a preset threshold.

[0006] A second aspect of this application discloses an electronic device comprising: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to execute the digital twin-based survey data analysis method provided in this application.

[0007] One or more technical solutions provided in this application have at least the following beneficial effects: The system comprises the following modules: an operation record acquisition module for acquiring historical healthy operation records of similar hydropower units; an interval prediction model generation module for training a long short-term memory network using multiple records with operating condition type identifiers as benchmark data to obtain an interval prediction model; a real-time operation data acquisition module for acquiring real-time operation data of the hydropower unit; a health prediction interval acquisition module for predicting and analyzing the real-time operating condition type and the real-time key parameters using the interval prediction model to obtain a dynamic health prediction interval; a real-time vibration coverage acquisition module for introducing a sliding time window to compare and analyze the real-time vibration data with the dynamic health prediction interval to obtain a real-time vibration coverage rate; and an anomaly warning module for issuing a vibration anomaly warning for the hydropower unit when the real-time vibration coverage rate is not at a preset threshold. This achieves the technical effect of realizing adaptive and accurate monitoring of the hydropower unit's operating status and intelligent anomaly warning, improving the accuracy of vibration anomaly identification. Attached Figure Description

[0008] Figure 1 This is a schematic diagram of a hydropower unit operation status monitoring system provided in an embodiment of this application.

[0009] Figure 2 This is a schematic diagram of the structure of an electronic device provided in this application.

[0010] Explanation of reference numerals in the attached figures: 10 for operation record acquisition module, 20 for interval prediction model generation module, 30 for real-time operation data acquisition module, 40 for health prediction interval acquisition module, 50 for real-time vibration coverage acquisition module, 60 for anomaly warning module, 21 for processor, 22 for memory, 23 for input device, and 24 for output device. Detailed Implementation

[0011] This application provides a hydropower unit operation status monitoring system and equipment to address the technical problems in the prior art where hydropower unit monitoring relies on fixed thresholds, is easily affected by operating conditions leading to frequent false alarms, and has insufficient accuracy in anomaly identification.

[0012] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0013] Example 1, as Figure 1 As shown in the figure, this application provides a hydropower unit operation status monitoring system, the system comprising: The operation record acquisition module 10 is used to acquire historical healthy operation records of similar hydropower units, wherein the historical healthy operation records include multiple records with operating condition type identifiers.

[0014] Specifically, the operation record acquisition module 10 is used to acquire historical health operation records of similar units of the same model, structure, and operating condition range as the hydropower unit to be monitored, under healthy conditions. Among them, similar units refer to hydropower units of the same series as the hydropower unit to be monitored, which have the same capacity level, runner type, unit structure, and installation conditions. Historical health operation records refer to historical operation data collected by similar units under normal healthy operation conditions without faults, abnormalities, or alarms. The operating condition type identifier is a predefined operating condition classification label, including power generation operating condition, phase adjustment operating condition, transient overvoltage operating condition, etc. The operation record acquisition module filters and extracts the above-mentioned health data from the unit's historical operation database and automatically labels each historical record with the corresponding operating condition type identifier, forming multiple benchmark data records with clear operating condition classifications that can be used for model training.

[0015] The interval prediction model generation module 20 is used to train a long short-term memory network using the multiple records with working condition type identifiers as reference data to obtain an interval prediction model.

[0016] Specifically, the interval prediction model generation module 20 uses multiple historical healthy operation records with operating condition type identifiers output by the operation record acquisition module as training benchmark data, and constructs and trains a dedicated interval prediction model for hydropower units based on a Long Short-Term Memory (LSTM) network. The benchmark data consists of cleaned, normalized, and labeled healthy operation data of similar units, including historical key parameters, historical healthy vibration intervals, and operating condition type identifiers. The LSTM network adopts a three-layer structure: input layer, hidden layer, and output layer. The input layer incorporates operating condition type features and historical key parameter sequences, and the hidden layer uses LST... The M-memory unit captures the long-term dependencies and fluctuation patterns of time-series operational data, and the output layer outputs the corresponding upper and lower limits of healthy vibration. During training, supervised learning is carried out using the operating condition type and historical key parameters as inputs and the historical healthy vibration interval as supervision labels. The prediction interval error is calculated through forward propagation, and the network weights are iteratively updated through backpropagation until convergence. The interval prediction model is a multi-branch prediction model trained and integrated according to different operating conditions. Each branch corresponds to an interval prediction layer for a certain operating condition type. It can accurately output a dynamic health prediction interval that matches the current operating state based on the input real-time operating condition type and real-time key parameters.

[0017] The real-time operation data acquisition module 30 is used to acquire the real-time operation data of the hydropower unit, wherein the real-time operation data includes real-time operating condition type, real-time key parameters and real-time vibration data.

[0018] Specifically, the real-time operation data acquisition module 30 acquires the current operating data of the monitored hydropower unit in real time through sensor acquisition, communication interface parsing, and background data interaction, forming real-time operation data. Among them, the real-time operating condition type refers to the current operating condition of the unit, including power generation, phase adjustment, and special conditions such as transient overvoltage, which are output in real time by the unit control system and acquired and identified by this module; the real-time key parameters refer to the electrical and mechanical parameters reflecting the current operating status of the unit, including time-series data such as unit load, speed, excitation voltage, guide vane opening, and head flow; the real-time vibration data refers to the time-domain data such as swing, vibration velocity, and vibration amplitude collected in real time by the unit vibration sensors; this module synchronously acquires the above three types of data, aligns the timestamps, and performs normalization preprocessing to form standardized real-time operation data that can be directly input into the interval prediction model, providing data support for subsequent dynamic health prediction interval calculation.

[0019] The health prediction interval acquisition module 40 is used to perform predictive analysis on the real-time operating condition type and the real-time key parameters through the interval prediction model to obtain the dynamic health prediction interval.

[0020] Specifically, the health prediction interval acquisition module 40 inputs the real-time operating condition type and real-time key parameters output by the real-time operation data acquisition module into the trained interval prediction model. The model matches the corresponding prediction layer according to the current operating condition, performs time-series prediction calculations in combination with the real-time key parameters, and outputs a dynamic health prediction interval that adapts to the current unit status. At the same time, it matches the corresponding health confidence factor according to the unit's service life and adaptively corrects the dynamic health prediction interval, ultimately obtaining an upper and lower limit interval of vibration health that accurately fits the current operating condition and equipment status.

[0021] The real-time vibration coverage acquisition module 50 is used to introduce a sliding time window to compare and analyze the real-time vibration data with the dynamic health prediction interval to obtain the real-time vibration coverage.

[0022] Specifically, the real-time vibration coverage acquisition module 50 introduces a sliding time window mechanism to compare and statistically analyze the dynamic health prediction interval output by the health prediction interval acquisition module with the real-time vibration data output by the real-time operation data acquisition module point by point, and calculate the real-time vibration coverage. The sliding time window is a time-series data extraction unit with adjustable width and adjustable step size. It segments continuous real-time vibration data according to a set window length and step interval, forming a time-series continuous real-time vibration window data. The dynamic health prediction interval is the normal fluctuation range of vibration corresponding to the current operating condition and key parameters, including the upper and lower limits of predicted vibration. The real-time vibration coverage refers to the ratio of the number of effective vibration sampling points within the dynamic health prediction interval to the total number of vibration sampling points within the window within the current sliding time window. The module first randomly samples the vibration data within the window to form a set of real-time vibration sampling points, then judges point by point whether the sampling point falls within the health interval and marks the covered point. Finally, the ratio of the number of covered points to the total number of sampling points is used as the real-time vibration coverage, which characterizes the degree of consistency between the current real-time vibration and the health status.

[0023] The abnormal warning module 60 is used to provide an abnormal vibration warning for the hydropower unit when the real-time vibration coverage rate is not at a preset threshold.

[0024] Specifically, the abnormal early warning module 60 first compares the real-time vibration coverage rate output by the real-time vibration coverage rate acquisition module 50 with the preset threshold set by the system in real time. When it is determined that the real-time vibration coverage rate is not within the normal range defined by the preset threshold, the module immediately performs graded vibration abnormality early warning and corresponding emergency response for the hydropower unit. The preset threshold is a normal coverage rate judgment threshold determined based on the statistical analysis of historical healthy operation data of similar units, and can be adaptively expanded and adjusted according to special operating conditions such as transient overvoltage. The real-time vibration coverage rate not being within the preset threshold means that the real-time vibration coverage rate is lower than the lower limit of the preset threshold, indicating that the current vibration data deviates from the healthy fluctuation range to the level of abnormality judgment. Vibration abnormality early warning includes alarm methods such as local audible and visual alarms, pop-up prompts in the background monitoring system, and remote push of abnormal information. At the same time, the module will extract and record information such as vibration abnormality duration and vibration abnormality degree, and complete the abnormality level judgment by combining it with the preset abnormality level table, and perform emergency handling operations such as load reduction and protective shutdown according to the level, thereby realizing timely identification and safe handling of vibration abnormalities of hydropower units.

[0025] In one possible implementation, the interval prediction model generation module 20 further includes: The first record extraction unit is used to extract the first record from the plurality of records with working condition type identifiers, and the first record corresponds to the first working condition type.

[0026] The historical health interval determination unit is used to determine the first historical health interval based on the first historical vibration data in the first record.

[0027] The historical key parameter extraction unit is used to extract the first historical key parameter from the first record and form a first data group with the first historical health interval.

[0028] The interval prediction layer acquisition unit is used to perform supervised learning on the first data group to obtain the first interval prediction layer corresponding to the first working condition type.

[0029] An interval prediction model building unit is used to build the interval prediction model based on the first interval prediction layer.

[0030] Specifically, the first record extraction unit is used to perform targeted filtering and extraction from multiple historical healthy operation records with operating condition type identifiers output by the operation record acquisition module 10, according to the first operating condition type, to obtain the first record uniquely corresponding to the operating condition. Here, the record with operating condition type identifier refers to standardized healthy operation data that has been classified and labeled and contains historical key parameters and historical vibration data under the corresponding operating condition. The first operating condition type is a specific operating condition that has been preset, including any one of the power generation condition, phase adjustment condition, or transient overvoltage condition. The first record is a single-condition historical healthy operation dataset that belongs only to the first operating condition type and is not mixed with other operating conditions. This dataset removes abnormal data and redundant data and retains pure time-series operating characteristics, which is used to provide standard benchmark data under a single operating condition for subsequent historical health interval determination and model training.

[0031] The historical health interval determination unit is used to perform statistical calculations on the first historical vibration data in the first record output by the first record extraction unit to determine the first historical health interval. The first record is pure historical healthy operation data belonging to the first working condition type. The first historical vibration data is the time-series vibration values ​​such as vibration velocity, vibration amplitude, and swing collected in the first record under a healthy state without faults, abnormalities, or alarms. The first historical health interval is the normal fluctuation range formed after statistical processing of the first historical vibration data. This unit determines the corresponding upper and lower vibration limits by calculating the mean, fitting the standard deviation, and selecting the confidence interval of the first historical vibration data, thereby forming the first historical health interval that is only applicable to the first working condition type, which serves as the standard output label for subsequent model training.

[0032] The historical key parameter extraction unit extracts the first historical key parameter strongly correlated with vibration from the first record output by the first record extraction unit, and aligns and combines this parameter with the first historical health interval output by the historical health interval determination unit to form the first data set for model training. The first record is pure single-condition historical healthy operation data belonging to the first operating condition type. The first historical key parameter refers to the operating characteristic quantity in the first record that can characterize the unit's operating status and is highly correlated with vibration changes, including time-series data such as unit load, speed, excitation voltage, guide vane opening, working head, and flow rate. The first historical health interval is the standard vibration interval formed by the statistically determined upper and lower limits of healthy vibration under the first operating condition. The first data set is a standardized training sample pair consisting of the first historical key parameter as input features and the first historical health interval as output labels. This data set is time-complete and feature-matched, and is used to provide training data for the supervised learning of the subsequent interval prediction layer.

[0033] The interval prediction layer acquisition unit uses the first data set output by the historical key parameter extraction unit as training samples. Supervised learning is performed based on a Long Short-Term Memory (LSTM) network to train a first interval prediction layer uniquely corresponding to the first operating condition type. The first data set consists of standardized training samples with the first historical key parameters as input features and the first historical healthy interval as the output label. The first operating condition type is a specific condition among power generation, phase regulation, and transient overvoltage. Supervised learning is a deep learning training method based on known inputs and corresponding standard outputs, and the training process includes forward propagation and backward propagation. The Long Short-Term Memory (LSTM) network... The system employs a four-layer structure consisting of an input layer, an LSTM hidden layer, a fully connected layer, and an output layer. The input layer receives the first set of key historical parameters, the LSTM hidden layer captures the temporal dependency between the operating parameters and the vibration interval through memory units, the fully connected layer is used for feature fusion, and the output layer outputs the upper and lower limits of vibration. During training, the first set of data is used as input, and the error between the predicted interval and the true healthy interval is used as the loss function to iteratively update the network weights until convergence. The first interval prediction layer is an independent prediction branch dedicated to the first working condition type, which can stably output the corresponding upper and lower limits of healthy vibration based on the key parameters under this working condition, serving as the core component of the interval prediction model.

[0034] The interval prediction model building unit is used to integrate, splice, and structurally encapsulate the first interval prediction layer and multiple independent interval prediction layers corresponding to different operating conditions, ultimately constructing a complete interval prediction model. The first interval prediction layer refers to a neural network branch trained with supervised learning using a Long Short-Term Memory (LSTM) network for a specific operating condition type, capable of independently predicting vibration health intervals. The interval prediction layers corresponding to other operating conditions refer to dedicated prediction branches trained sequentially for various operating conditions such as power generation, phase modulation, and transient overvoltage, using the same training method. The interval prediction model is a multi-branch parallel deep learning ensemble model, employing an overall operating condition routing structure. During use, it automatically matches and calls the corresponding interval prediction layer based on the input real-time operating condition type, and then that layer calculates and outputs a dynamic health prediction interval based on real-time key parameters. Each operating condition branch within the model is independent and does not interfere with each other, enabling high-precision interval prediction with full operating condition coverage and adaptive switching.

[0035] In one possible implementation, the interval prediction model generation module 20 further includes: The first historical health interval refers to the vibration interval formed by the first historical vibration upper limit and the first historical vibration lower limit based on the first historical vibration data.

[0036] Specifically, the first historical health interval refers to the range of vibration values ​​determined by the historical health interval determination unit based on the first historical vibration data and calculated using statistical methods, consisting of the first historical vibration upper limit and the first historical vibration lower limit. The first historical vibration data consists of raw time-series data such as vibration velocity, vibration amplitude, and sway collected from similar units under the first operating condition in a healthy and fault-free state. The first historical vibration upper limit is the maximum permissible vibration value determined by fitting a confidence interval to the set of healthy vibration data. The first historical vibration lower limit is the minimum reasonable vibration value determined by fitting a confidence interval to the set of healthy vibration data. The continuous vibration interval defined by the aforementioned upper and lower limits is the first historical health interval applicable only to the first operating condition, used as the standard output label for model training and the vibration health judgment benchmark.

[0037] In one possible implementation, the health prediction interval acquisition module 40 further includes: The service life acquisition unit is used to acquire the service life of the hydropower unit.

[0038] The health and self-confidence factor matching unit is used to match the health and self-confidence factors corresponding to the years of work experience.

[0039] The prediction interval adjustment unit is used to adjust the dynamic health prediction interval based on the health confidence factor.

[0040] Specifically, the working years acquisition unit is used to acquire and determine the working years of the hydropower unit to be monitored. The hydropower unit refers to the target hydropower main equipment monitored by this system. The working years refer to the cumulative actual operating years of the hydropower unit from the date of its commissioning and first grid connection to the current monitoring time. This unit automatically extracts and calculates the actual working years of the target unit by reading the unit's factory file, commissioning date, operation log, or internal timing data of the unit's control system, providing a basis for subsequent health confidence factor matching.

[0041] The health confidence factor matching unit employs a combination of table lookup mapping and segmented threshold determination. It matches the acquired service life of the hydropower unit with a pre-set database to determine the corresponding health confidence factor. The service life refers to the cumulative actual operating time of the unit from its commissioning date to the current moment, and the health confidence factor is a dimensionless coefficient characterizing the reliability of the unit's operating status and the confidence level of vibration data. The unit pre-stores a one-to-one mapping table between service life segments and health confidence factors. By substituting the service life value into the segmented intervals for comparison and positioning, the corresponding coefficient in the table is directly retrieved to complete the matching output. This mapping relationship can be adaptively configured according to the unit model, design life, and operation and maintenance experience, thereby achieving rapid and accurate matching of the health confidence factor of hydropower units with different service durations.

[0042] The prediction interval adjustment unit employs a numerical scaling correction method to adaptively adjust the dynamic health prediction interval based on the health confidence factor. The health confidence factor is a dimensionless confidence coefficient calculated comprehensively based on the hydropower unit's operating years and the stability of historical vibration data; a higher value indicates stronger reliability of the unit's health status. The dynamic health prediction interval is a judgment interval formed by the upper and lower vibration limits output by the interval prediction model based on real-time key parameters. This unit multiplies the upper and lower limits of the dynamic health prediction interval by the scaling factor corresponding to the health confidence factor. When the health confidence factor is high, the scaling factor is decreased to narrow the interval and improve monitoring sensitivity; when the health confidence factor is low, the scaling factor is increased to widen the interval and reduce the false alarm rate. This numerical correction method achieves precise adjustment of the dynamic health prediction interval, making the vibration judgment standard more closely match the actual aging state and operating characteristics of the unit.

[0043] In one possible implementation, the real-time vibration coverage acquisition module 50 further includes: The vibration window data acquisition unit is used to perform window sampling on the real-time vibration data based on the sliding time window to obtain real-time vibration window data.

[0044] The sampling point set acquisition unit is used to randomly sample the real-time vibration window data to obtain a real-time vibration sampling point set.

[0045] The health prediction interval judgment unit is used to determine whether the first real-time sampling point in the set of real-time vibration sampling points is within the dynamic health prediction interval.

[0046] The coverage point determination unit is used to record the first real-time sampling point as a coverage point when the first real-time sampling point is within the dynamic health prediction range.

[0047] The real-time vibration coverage determination unit is used to take the ratio of the coverage point to the sampling point of the real-time vibration sampling point set as the real-time vibration coverage.

[0048] Specifically, the vibration window data acquisition unit is used to perform time-series segmented sampling on the real-time vibration data output by the real-time operation data acquisition module 30 based on a sliding time window, thereby obtaining real-time vibration window data. The sliding time window refers to a time-series data extraction rule with adjustable window width and adjustable sliding step size. The window width is the duration of a single segment of vibration data, and the sliding step size is the time interval for each forward movement of the window, which can be dynamically adjusted according to monitoring needs. Real-time vibration data refers to continuous time-series data such as vibration velocity, vibration amplitude, and sway, collected in real-time by vibration sensors and aligned with timestamps. Real-time vibration window data refers to a fixed-length local vibration data segment obtained by segmenting from continuous real-time vibration data according to the sliding time window rule. This segment retains its original time-series characteristics and is used for subsequent sampling and coverage calculation.

[0049] The sampling point set acquisition unit performs equal-probability random sampling on the real-time vibration window data output by the vibration window data acquisition unit, extracting several representative vibration data points to form a real-time vibration sampling point set. The real-time vibration window data refers to a continuous time-series vibration data segment of fixed duration obtained by truncating a sliding time window, containing continuous vibration values ​​and timestamps. Random sampling refers to unbiased sampling within the time series of the real-time vibration window data, according to a preset sampling quantity or sampling ratio, avoiding data redundancy and preserving vibration distribution characteristics. The real-time vibration sampling point set is a dataset composed of multiple randomly extracted independent vibration sampling points. Each sampling point contains the vibration amplitude, vibration velocity, or sway value at the corresponding time. This set serves as the basis for subsequent interval judgment and coverage calculation, improving processing efficiency without reducing computational accuracy.

[0050] The health prediction interval judgment unit uses numerical interval comparison to determine whether the vibration value corresponding to the first real-time sampling point in the real-time vibration sampling point set is within the dynamic health prediction interval adjusted by the health confidence factor. The real-time vibration sampling point set is a collection of vibration sampling points randomly sampled from real-time vibration window data. The first real-time sampling point is any independent sampling point to be judged within this set, containing the vibration amplitude or vibration velocity value at the corresponding time. The dynamic health prediction interval is the normal value range formed by the upper and lower limits of vibration output by the interval prediction model and adaptively adjusted. This unit compares the vibration value of the first real-time sampling point with the upper and lower limits of the dynamic health prediction interval sequentially. If the vibration value is greater than or equal to the lower limit and less than or equal to the upper limit, it is determined to be within the interval; otherwise, it is determined to be outside the interval. Finally, the judgment result is output point by point, providing a basis for subsequent calculation of real-time vibration coverage.

[0051] The coverage point determination unit classifies and marks the first real-time sampling point according to the interval judgment result output by the health prediction interval judgment unit. The first real-time sampling point is a single sampling point in the real-time vibration sampling point set that has undergone interval judgment. The dynamic health prediction interval is the normal allowable range of the unit's current vibration. If the vibration value of the first real-time sampling point is determined to be between the upper and lower limits of the dynamic health prediction interval, the first real-time sampling point is marked as a covered point, indicating that the vibration state of the sampling point is normal. If the vibration value of the first real-time sampling point is determined to exceed the upper limit or fall below the lower limit of the dynamic health prediction interval, the first real-time sampling point is marked as an uncovered point, indicating that the vibration state of the sampling point is abnormal. The unit forms a classification result of covered and uncovered points by marking each point, providing a statistical basis for subsequent calculation of the real-time vibration coverage rate.

[0052] The real-time vibration coverage determination unit uses a ratio calculation method to calculate the ratio between the number of covered points and the total number of sampling points in the real-time vibration sampling point set, and determines the real-time vibration coverage rate as the result. Here, the covered points are normal vibration sampling points determined to be within the dynamic health prediction range, the real-time vibration sampling point set is the total number of sampling points obtained by randomly sampling real-time vibration window data, and the sampling point ratio is the ratio obtained by dividing the number of covered points by the total number of sampling points in the real-time vibration sampling point set. The real-time vibration coverage rate is a quantitative indicator characterizing the degree to which the unit's real-time vibration state conforms to the dynamic health prediction range; a higher ratio indicates a higher proportion of normal vibration points and a more stable unit vibration state.

[0053] In one possible implementation, the real-time vibration coverage acquisition module 50 further includes: The sliding time window has an adjustable width and an adjustable step size.

[0054] Specifically, the sliding time window is the core processing unit for time-series segmentation of real-time vibration data, possessing two key attributes: adaptive adjustable width and adjustable step size. The sliding time window refers to a time-series data processing mechanism that segments continuous vibration data according to a set time length and moves sequentially along the time axis. The adjustable width is the duration of a single segment of vibration data captured by the window, supporting adaptive adjustment based on the unit's operating status. When the unit's operating status is stable, the window width automatically narrows to improve monitoring sensitivity and analysis accuracy; when the unit's operating status changes drastically, such as during transient overvoltages or sudden load changes, the window width automatically widens to cover the complete fluctuation process and avoid misjudgments. The adjustable step size is the time interval for each forward movement of the window along the time axis, which can be flexibly set according to real-time monitoring requirements. Through the coordinated adjustment of width and step size, the sliding time window can adapt to the vibration analysis needs of hydropower units under different operating conditions, balancing monitoring accuracy and robustness.

[0055] In one possible implementation, the anomaly warning module 60 further includes: The special working condition judgment unit is used to determine whether the first working condition type is a predetermined special working condition.

[0056] An expansion adjustment unit is used to expand and adjust the preset threshold if the first working condition type is the predetermined special working condition.

[0057] Specifically, the special operating condition judgment unit adopts a combination of operating condition identifier matching and logical judgment to identify the operating condition attributes of the first operating condition type in order to determine whether the first operating condition type is a predetermined special operating condition predefined by the system. The first operating condition type is a target operating condition classification identifier extracted from historical healthy operation records with operating condition type identifiers and used to construct the corresponding interval prediction layer, including power generation operating condition, phase adjustment operating condition, and transient overvoltage operating condition, etc. The predetermined special operating condition specifically refers to the non-steady-state operating condition triggered by a transient overvoltage event and a drastic change in the unit's operating state, which corresponds to a unique preset operating condition identifier. The unit internally stores a predetermined special operating condition identifier library. By comparing the identifier character or code corresponding to the first operating condition type with the preset identifier in the library field by field, if the two are completely consistent, it is determined to belong to the predetermined special operating condition; otherwise, it is determined to be a normal operating condition. The unit finally outputs a Boolean judgment result, which provides a judgment basis for the adaptive adjustment of subsequent interval prediction model training parameters, sliding time window width, and warning threshold.

[0058] The expansion adjustment unit is used to automatically perform adaptive expansion adjustment on the preset threshold if the first operating condition type is determined to belong to the predetermined special operating condition after receiving the judgment result output by the special operating condition judgment unit. The first operating condition type is the target operating condition type identifier currently participating in interval prediction and vibration judgment. The predetermined special operating condition specifically refers to a non-steady-state operating condition caused by a transient overvoltage event, where the unit vibration will exhibit normal large fluctuations. The preset threshold is a benchmark threshold used to determine whether the real-time vibration coverage is normal. The expansion adjustment refers to relaxing the coverage judgment standard by a preset ratio or fixed value based on the original threshold, appropriately reducing the normal coverage requirement to adapt to the characteristics of reasonable vibration fluctuations under special operating conditions, avoiding false warnings caused by normal drastic changes such as transient overvoltage, thereby improving the accuracy and anti-interference capability of unit anomaly judgment.

[0059] In one possible implementation, the anomaly warning module 60 further includes: The predetermined special operating condition refers to the special operating condition under the influence of transient overvoltage events.

[0060] Specifically, the predetermined special operating condition refers to the short-term unsteady-state operating condition that a hydropower unit enters when a transient overvoltage event is triggered by factors such as grid disturbances, excitation switching, closing operations, or fault tripping during operation. Among them, a transient overvoltage event refers to an electrical transient process in which the voltage at the stator terminal, excitation system, or bus side of the unit deviates from the rated value and fluctuates rapidly within a short period of time. The predetermined special operating condition specifically refers to the fact that the vibration, sway, and electrical parameters of the unit will fluctuate significantly within the normal range during the occurrence of this event, which is not an abnormality of mechanical fault. Therefore, it is necessary to adaptively expand and adjust the preset threshold to avoid false alarms during this normal unsteady-state process of transient overvoltage and to ensure the accuracy of monitoring and judgment.

[0061] In one possible implementation, the anomaly warning module 60 further includes: Vibration anomaly information acquisition unit, used to acquire vibration anomaly information.

[0062] The real-time vibration anomaly level acquisition unit is used to match the vibration anomaly duration and vibration anomaly degree in the vibration anomaly information with a preset anomaly level table to obtain the real-time vibration anomaly level.

[0063] An emergency response unit is used to perform emergency response on the hydropower unit based on the real-time vibration anomaly level.

[0064] Specifically, the vibration anomaly information acquisition unit is used to collect and summarize vibration anomaly information in real time when the anomaly early warning module determines that the hydropower unit has a vibration anomaly. The vibration anomaly information is a multi-dimensional anomaly description information including the time of anomaly occurrence, real-time vibration coverage value, vibration sampling point data that deviates from the dynamic health prediction range, current unit operating condition type, service life, and sliding time window parameters. This unit integrates and encapsulates anomaly-related feature information by calling the intermediate calculation results of each module and real-time operating data to form standardized anomaly data records, providing complete data support for subsequent anomaly early warning push, fault location analysis, and historical anomaly tracing.

[0065] The real-time vibration anomaly level acquisition unit adopts a dual-index interval matching judgment method: first, it extracts two quantitative features, vibration anomaly duration and vibration anomaly degree, from the vibration anomaly information; then, it compares these two indicators with the predefined duration threshold interval and degree threshold interval in the preset anomaly level table segment by segment to determine their respective level numbers; then, it selects the higher level of the two levels as the comprehensive judgment result according to the higher priority rule, and finally outputs the corresponding real-time vibration anomaly level.

[0066] The emergency response unit is used to determine the real-time vibration anomaly level output by the anomaly level determination unit, and execute a graded response logic that matches the severity of the anomaly to implement corresponding emergency response for the hydropower unit. The real-time vibration anomaly level is a classification based on the degree to which the real-time vibration coverage deviates from a preset threshold, the amplitude of the vibration exceeding limits, and the duration of the anomaly, typically including minor anomalies, general anomalies, and severe anomalies. The emergency response includes a series of safety control measures such as graded alarm prompts, optimized adjustment of operating parameters, limiting unit load, disconnecting auxiliary equipment, and even triggering protective shutdowns. Through preset level-response mapping rules, the unit automatically selects the corresponding response strategy, ensuring unit safety while avoiding unnecessary shutdowns, thus achieving intelligent and graded emergency response to vibration anomalies.

[0067] Example 2, Figure 2 This is a schematic diagram of the electronic device provided by the digital twin-based survey data analysis method of the present invention, showing an exemplary electronic device suitable for implementing the embodiments of the present invention. Figure 2 The electronic device shown is merely an example and should not be construed as limiting the functionality or scope of the embodiments of the present invention. Figure 2 As shown, the electronic device includes a processor 21, a memory 22, an input device 23, and an output device 24; the number of processors 21 in the electronic device can be one or more. Figure 2 Taking a processor 21 as an example, the processor 21, memory 22, input device 23, and output device 24 in an electronic device can be connected via a bus or other means. Figure 2 Taking the example of a connection between China and Israel via a bus.

[0068] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0069] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0070] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A hydropower unit operation status monitoring system, characterized in that, include: The operation record acquisition module is used to acquire historical healthy operation records of similar hydropower units, wherein the historical healthy operation records include multiple records with operating condition type identifiers; The interval prediction model generation module is used to train a long short-term memory network using the multiple records with working condition type identifiers as reference data to obtain an interval prediction model. The real-time operation data acquisition module is used to acquire the real-time operation data of the hydropower unit, wherein the real-time operation data includes real-time operating condition type, real-time key parameters and real-time vibration data; The health prediction interval acquisition module is used to predict and analyze the real-time operating condition type and the real-time key parameters through the interval prediction model to obtain the dynamic health prediction interval. The real-time vibration coverage acquisition module is used to introduce a sliding time window to compare and analyze the real-time vibration data with the dynamic health prediction interval to obtain the real-time vibration coverage. The abnormality warning module is used to provide an abnormality warning for the hydropower unit when the real-time vibration coverage rate is not at a preset threshold.

2. The hydropower unit operation status monitoring system as described in claim 1, characterized in that, The interval prediction model generation module also includes: The first record extraction unit is used to extract the first record from the plurality of records with working condition type identifiers, and the first record corresponds to the first working condition type. The historical health interval determination unit is used to determine the first historical health interval based on the first historical vibration data in the first record. A historical key parameter extraction unit is used to extract the first historical key parameter from the first record and form a first data group with the first historical health interval. An interval prediction layer acquisition unit is used to perform supervised learning on the first data group to obtain the first interval prediction layer corresponding to the first working condition type. An interval prediction model building unit is used to build the interval prediction model based on the first interval prediction layer.

3. The hydropower unit operation status monitoring system as described in claim 2, characterized in that, The first historical health interval refers to the vibration interval formed by the first historical vibration upper limit and the first historical vibration lower limit based on the first historical vibration data.

4. The hydropower unit operation status monitoring system as described in claim 1, characterized in that, The health prediction interval acquisition module also includes: A service life acquisition unit is used to acquire the service life of the hydropower unit. A health confidence factor matching unit is used to match the health confidence factor corresponding to the years of service. The prediction interval adjustment unit is used to adjust the dynamic health prediction interval based on the health confidence factor.

5. The hydropower unit operation status monitoring system as described in claim 1, characterized in that, The real-time vibration coverage acquisition module also includes: The vibration window data acquisition unit is used to perform window sampling on the real-time vibration data based on the sliding time window to obtain real-time vibration window data. The sampling point set acquisition unit is used to randomly sample the real-time vibration window data to obtain a real-time vibration sampling point set. A health prediction interval determination unit is used to determine whether the first real-time sampling point in the set of real-time vibration sampling points is within the dynamic health prediction interval. The coverage point determination unit is used to record the first real-time sampling point as a coverage point when the first real-time sampling point is within the dynamic health prediction range. The real-time vibration coverage determination unit is used to take the ratio of the coverage point to the sampling point of the real-time vibration sampling point set as the real-time vibration coverage.

6. A hydropower unit operation status monitoring system as described in claim 1 or 5, characterized in that, The sliding time window has an adjustable width and an adjustable step size.

7. The hydropower unit operation status monitoring system as described in claim 2, characterized in that, The anomaly warning module also includes: A special working condition determination unit is used to determine whether the first working condition type is a predetermined special working condition; An expansion adjustment unit is used to expand and adjust the preset threshold if the first working condition type is the predetermined special working condition.

8. The hydropower unit operation status monitoring system as described in claim 7, characterized in that, The predetermined special operating condition refers to the special operating condition under the influence of transient overvoltage events.

9. The hydropower unit operation status monitoring system as described in claim 1, characterized in that, The anomaly warning module also includes: Vibration anomaly information acquisition unit, used to acquire vibration anomaly information; The real-time vibration anomaly level acquisition unit is used to match the vibration anomaly duration and vibration anomaly degree in the vibration anomaly information with a preset anomaly level table to obtain the real-time vibration anomaly level. An emergency response unit is used to perform emergency response on the hydropower unit based on the real-time vibration anomaly level.

10. An electronic device, characterized in that, The electronic device includes: processor; Memory used to store the processor's executable instructions; The processor is used to execute the hydropower unit operation status monitoring system according to any one of claims 1 to 9.

Images