A hydraulic turbine generator bearing oil pump outlet check valve fault diagnosis method and system based on reflux anomaly index

Abstract

The application provides a water turbine generator bearing oil pump outlet check valve fault diagnosis method and system based on a backflow abnormality index, and belongs to the technical field of water turbine generator equipment state monitoring and fault diagnosis, and comprises the following steps: collecting unit operation conditions and bearing oil system key measuring point data; pre-processing the collected data and identifying the unit operation state; verifying the cooling side operation condition by using the related measuring points of the cooling water system; extracting the oil temperature and bushing temperature characteristics in the operation state, and extracting the oil level characteristics in the shutdown state; establishing a normal operation interval according to the evaluation data window and the reference data window, calculating the oil level degradation degree, the oil temperature degradation degree and the bushing temperature degradation degree, and constructing the backflow abnormality index through weighted fusion. According to the backflow abnormality index, the pump outlet check valve blockage or failure to close tightly in the bearing oil system can be identified. The application can realize fault diagnosis by using the existing monitoring data of the unit.

Description

Technical Field

[0001] This invention belongs to the field of hydropower unit equipment condition monitoring and fault diagnosis technology, specifically relating to a fault diagnosis method and system for the outlet check valve of a hydropower unit bearing oil pump based on the backflow anomaly index. Background Technology

[0002] During operation, the bearing oil system of a hydroelectric generator plays a crucial role in lubrication, cooling, and removing frictional heat. The bearing oil system typically uses an oil pump to deliver lubricating oil from the oil pan to the bearing. After absorbing the heat generated by the bearing, the lubricating oil flows back to the oil pan, achieving both cooling and lubrication of the bearing through oil circulation.

[0003] In this oil circulation system, a check valve is typically installed at the oil pump outlet to prevent backflow of lubricating oil when the pump stops. If the check valve becomes stuck or does not close tightly, some high-temperature lubricating oil may flow back into the oil sump through the check valve, causing the oil temperature in the sump to rise. Simultaneously, due to the reduced cooling effect in the oil supply circuit, this may further lead to an increase in the bearing bearing temperature. Furthermore, after the unit is shut down, residual hot oil in the system may also flow back into the oil sump through a loosely closed check valve, causing an abnormal rise in the oil level in the sump.

[0004] In existing technologies, abnormal bearing or oil temperatures in hydropower units are typically diagnosed by operators based on data from a single measuring point or their experience. However, elevated bearing temperatures can also be caused by factors such as cooling system malfunctions, oil deterioration, or load changes, making it difficult to accurately determine the cause of the fault based on a single parameter. Furthermore, existing methods often lack the ability to assess the operating status of the cooling system, making it difficult to effectively distinguish between temperature rises caused by cooling system malfunctions and those caused by oil backflow. Therefore, in actual operation, it is often difficult to promptly identify oil backflow faults caused by check valve blockage or incomplete closure, thus affecting equipment operating safety and maintenance efficiency. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method and system for diagnosing the backflow fault of the outlet check valve of the bearing oil pump of a hydroelectric generator based on the backflow anomaly index. This method can automatically identify the backflow fault of the check valve using the unit's operating data. By verifying the operating conditions of the cooling system to eliminate abnormal factors on the cooling side, and by comprehensively analyzing the characteristics of oil level, oil temperature and bearing temperature under the unit's operating and shutdown states, a backflow anomaly index is constructed to accurately identify the check valve jamming or improper closure fault.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a method for diagnosing the fault of the outlet check valve of a hydroelectric generator bearing oil pump based on the backflow anomaly index, comprising the following steps: S1: Collect unit operating condition data and bearing oil system operating data; S2: Preprocess the collected data and identify the unit's operating status based on the preprocessed data, dividing the data into operating status data and shutdown status data; S3: Use relevant measurement data of the cooling water system to verify the operating status of the cooling system and eliminate abnormal factors of the cooling system; S4: Extract oil temperature and bearing temperature characteristics while the machine is running; extract oil level characteristics while the machine is stopped. S5: Establish a normal state benchmark model based on the feature data within the benchmark data window, and determine the normal benchmark range corresponding to each feature; S6: Calculate the oil level deterioration, oil temperature deterioration, and bearing temperature deterioration based on the degree of deviation of the characteristic data within the evaluation data window from the normal reference range; S7: Based on the oil level deterioration, oil temperature deterioration, and bearing temperature deterioration, and combined with the window type, a normalized weighted fusion is performed to construct the reflux anomaly index BAI; S8: Determine whether the bearing oil pump outlet check valve is stuck or not closing properly based on the backflow anomaly index BAI.

[0007] Preferably, the data collected in step S1 includes the unit's active power, unit speed, guide vane opening, oil level in the oil tank, oil temperature in the oil tank, bearing temperature, cooling water flow rate, cooling water pressure, and cooler outlet water temperature.

[0008] Preferably, the data preprocessing in step S2 includes missing value imputation, outlier removal, and time series data processing; the unit operating status identification is determined jointly by three indicators: unit active power, unit speed, and guide vane opening, specifically: Construct state discriminants separately: ; ; ; in, The active power of the unit. For unit speed, For guide vane opening; , , These are the corresponding thresholds; Define the overall voting result as: ; when When, it is determined to be in running state; when When the machine is in a stopped state, it is determined to be in a stopped state. When the time is right, it is determined to be an unknown state.

[0009] Preferably, the cooling system operating status verification in step S3 includes the following conditions: (1) The maximum value of oil-water mixture satisfies: ; in, This indicates the maximum value of the oil-water mixture within the operating status segment of the evaluation data window. Indicates the permissible threshold for oil-water mixing; (2) The deviation of the cooling water main flow rate from the median of the evaluation window relative to the reference window satisfies: ; in, To assess the median flow rate of the main cooling water pipe inside the window, The median flow rate of the main cooling water pipe within the reference window. This is the flow deviation threshold; (3) The deviation of the cooling water main pressure from the median of the evaluation window relative to the reference window satisfies: ; in, To assess the median pressure in the main cooling water pipe inside the window, The median pressure in the main cooling water pipe within the reference window. This is the pressure deviation threshold; (4) The median difference in cooler outlet water temperature satisfies: ; in, To assess the median outlet water temperature of the in-window cooler, The median outlet water temperature of the reference in-window cooler. The threshold for the outlet water temperature difference; When the above conditions are met simultaneously, the cooling side is determined to be stable, and the subsequent fault characteristic calculation is performed; otherwise, the cooling side is determined to have a suspected abnormality, and the suspected abnormality of the cooling side is output or the matching confidence is reduced.

[0010] Preferably, the normal state baseline model in step S5 is constructed using a statistical distribution model or a machine learning model; the statistical distribution model includes a Gaussian distribution model or a Gaussian mixture distribution model; the machine learning model includes at least one of a single-class support vector machine, an isolated forest, an autoencoder, or a Gaussian process regression model.

[0011] Preferably, when a Gaussian distribution model is used to construct the normal state baseline model, the degradation degree calculation in step S6 is specifically as follows: For any window in the evaluation window eigenvalues The standardized deviation is defined as: ; in, Features within the reference window k The mean, Features within the reference window k standard deviation To prevent the denominator from becoming unstable due to excessively small standard deviations; set up If the cumulative distribution function is a standard normal distribution, then the index The degree of degradation is defined as: ; in, , For normal reference boundary coefficients, This represents the boundary coefficient for significant anomalies.

[0012] In the above-mentioned method for diagnosing check valve faults in hydropower units based on the backflow anomaly index, the steps for constructing and defining the formula of the backflow anomaly index BAI include: When the pump outlet check valve in the bearing oil system becomes stuck or fails to close properly, some high-temperature lubricating oil will flow back into the oil sump through the check valve, causing the oil temperature in the sump to rise and potentially leading to an increase in the bearing bearing temperature. Additionally, after the unit is shut down, hot oil in the system may flow back into the oil sump through a faulty check valve, causing an abnormal rise in the oil level in the sump.

[0013] Therefore, this invention constructs a backflow anomaly index (BAI) by analyzing the variation characteristics of three key operating parameters: oil level, oil temperature, and bearing temperature, to quantify the degree of backflow failure in check valves.

[0014] Preferably, the formula for calculating the reflux anomaly index BAI in step S7 is: ; in, This refers to the degree of oil level deterioration. This refers to the degree of oil temperature degradation. The degree of temperature degradation; , , Let be the weighting coefficient, and satisfy: ; Representing windows respectively Are the internal oil level, oil temperature, and bearing temperature characteristics available? When window When running the window, ; When window When the shutdown window is open, .

[0015] The backflow anomaly index constructed by the above method can comprehensively reflect the degree of anomaly in oil level, oil temperature and bearing temperature. The backflow anomaly index corresponding to each effective time window in the evaluation data window is formed into a time series, and the overall matching degree is calculated based on the series, thereby realizing the identification of pump outlet check valve jamming or incomplete closure faults.

[0016] Preferably, step S8 specifically involves: constructing a time series from the reflux anomaly index corresponding to each effective time window within the evaluation data window, and calculating the high quantile statistic of the series as the overall matching degree; when the matching degree reaches or exceeds the preset diagnostic threshold, determining that the bearing oil pump outlet check valve has a jamming or improper closing fault.

[0017] Another aspect of the present invention provides a fault diagnosis system for the outlet check valve of a hydroelectric generator bearing oil pump based on a backflow anomaly index. The system is implemented based on the aforementioned method and includes: The data acquisition module is used to collect unit operating condition data and bearing oil system operating data; The preprocessing and status identification module is used to preprocess the collected data and identify the unit's operating status based on the preprocessed data, dividing the data into operating status data and shutdown status data. The cooling system verification module is used to verify the operating status of the cooling system using relevant measurement point data of the cooling water system and to eliminate abnormal factors in the cooling system. The feature extraction module is used to extract oil temperature and bearing temperature features when the machine is running, and to extract oil level features when the machine is stopped. The benchmark model construction module is used to build a normal state benchmark model based on the feature data within the benchmark data window and to determine the normal benchmark range corresponding to each feature. The degradation calculation module is used to calculate the oil level degradation, oil temperature degradation, and bearing temperature degradation based on the degree of deviation of the characteristic data in the evaluation data window from the normal reference range. The reflux anomaly index calculation module is used to construct the reflux anomaly index BAI by performing normalized weighted fusion based on the oil level deterioration, oil temperature deterioration, and bearing temperature deterioration, combined with the window type. The fault determination module is used to determine whether there is a jamming or improper closing fault in the bearing oil pump outlet check valve based on the backflow anomaly index (BAI).

[0018] Preferably, the data collected by the data acquisition module includes the unit's active power, unit speed, guide vane opening, oil level in the oil tank, oil temperature in the oil tank, bearing temperature, cooling water flow rate, cooling water pressure, and cooler outlet water temperature.

[0019] Preferably, in the preprocessing and status identification module, data preprocessing includes missing value filling, outlier removal, and time series data processing; the unit operating status identification is determined by a combination of three indicators: unit active power, unit speed, and guide vane opening.

[0020] Preferably, the benchmark model building module constructs a normal state benchmark model through a statistical distribution model or a machine learning model; the statistical distribution model includes a Gaussian distribution model or a Gaussian mixture distribution model; the machine learning model includes at least one of a single-class support vector machine, an isolated forest, an autoencoder, or a Gaussian process regression model.

[0021] The present invention has the following beneficial effects: 1. This invention proposes a new fault evaluation index, the Backflow Anomaly Index (BAI). By comprehensively analyzing operating parameters such as oil level, oil temperature, and bearing temperature, a multi-parameter fusion diagnostic model is constructed to achieve quantitative identification of faults such as pump outlet check valve jamming or incomplete closure in the bearing oil system, thereby improving the accuracy of fault diagnosis.

[0022] 2. This invention analyzes oil temperature and bearing temperature characteristics when the unit is running, and analyzes oil level change characteristics when the unit is shut down. By combining the characteristics of the running and shut-down states, it identifies check valve backflow faults, effectively avoiding misjudgment problems caused by relying on a single temperature parameter.

[0023] 3. This invention verifies auxiliary measuring points such as cooling water flow rate, cooling water pressure, and cooler outlet temperature to eliminate temperature rise interference caused by abnormalities in the cooling system, thereby improving the accuracy and reliability of check valve fault identification.

[0024] 4. This invention can achieve fault diagnosis by relying only on conventional measuring point data in the existing monitoring system of hydropower plants, without the need to add additional testing equipment, resulting in low implementation costs and easy application in existing unit monitoring systems. Attached Figure Description

[0025] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0026] Figure 1 This is the overall flowchart of the present invention, which is a fault diagnosis method for the outlet check valve of the bearing oil pump of a hydroelectric generator based on the backflow anomaly index.

[0027] Figure 2 This is a detailed flowchart of the construction and determination of the reflux anomaly index in this invention. Detailed Implementation

[0028] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.

[0029] It should be noted that the threshold conditions, time window length, effective data ratio, deviation range, weight coefficient, statistical boundary and model parameters involved in this embodiment are all reference parameters or preferred parameters given for the convenience of explaining the method of the present invention, and do not constitute a limitation on the scope of protection of the present invention.

[0030] The mechanism of this invention lies in constructing a diagnostic method based on the abnormal backflow fault mechanism of check valves, which combines the characteristics of unit operating and shutdown states. Fault identification is achieved through cooling-side consistency verification, benchmark model construction, degradation degree calculation, and backflow anomaly index fusion. For different unit types, measurement point conditions, sampling cycles, and operating conditions, the above parameters can be adaptively adjusted, equivalently replaced, or optimized by those skilled in the art based on historical data, statistical laws, or model training results, without departing from the spirit and scope of protection of this invention.

[0031] This invention can be used for both historical case matching analysis and real-time matching analysis after a new warning is issued. The basic idea is as follows: extract oil temperature and bearing temperature characteristics when the unit is running, extract oil level characteristics when the unit is shut down, and, excluding the influence of abnormal factors on the cooling side, establish a normal state range through a benchmark model. Then, map the degree of deviation of characteristics during the evaluation stage to a degree of deterioration, and finally fuse them to form a backflow anomaly index to determine whether there is a jamming or improper closure fault in the bearing oil pump outlet check valve.

[0032] Example 1: A fault diagnosis method for the outlet check valve of a hydroelectric generator bearing oil pump based on a backflow anomaly index includes: I. Data Acquisition and Analysis Window Construction In this embodiment, minute-level operating data of the unit is collected, and the sampling interval is preferably 1 minute, but it is not limited to this. Second-level, 2-minute-level or other suitable sampling intervals can also be used depending on the configuration of the monitoring system.

[0033] The collected data includes at least: unit active power, unit speed, guide vane opening, oil level in the oil tank, oil temperature in the oil tank, bearing temperature, cooling water main flow rate, cooling water main pressure, cooler outlet water temperature, and oil-water mixing measurement point data.

[0034] To distinguish between normal operating conditions and conditions to be evaluated, this invention constructs a baseline data window and an evaluation data window. The baseline data window is used to characterize the features of the equipment's normal operating condition, preferably taking the warning time. The evaluation data window is used for fault matching and identification. In historical case analysis, it can take data from a continuous period after the warning time, while in real-time matching analysis, it can take data from the warning time to the current analysis time. The specific length of the evaluation data window can be adjusted according to the on-site business needs, the speed of fault development, and data integrity, preferably not exceeding 7 days, but not limited to this.

[0035] II. Data Preprocessing and Start-up / Shutdown Division The raw minute-level data is preprocessed. Preferably, missing values ​​are filled using a forward-filling method, with the filling time preferably not exceeding 5 minutes; records with remaining null values ​​are removed before status identification and feature extraction.

[0036] The start-up and shutdown status of the generator unit is determined jointly by three indicators: active power, generator speed, and guide vane opening. For any time t, the status discrimination variables are constructed as follows: ; ; ; in, The active power of the unit. For unit speed, For guide vane opening; , , These are the corresponding thresholds, with the preferred value being... Rated power Rated speed, Guide vane opening.

[0037] Define the overall voting result as: ; when When, it is determined to be in running state; when When the machine is in a stopped state, it is determined to be in a stopped state. When the time limit is reached, it is determined to be an unknown state. Unknown states are filled with previous valid states; short stop segments with a duration of no more than 5 minutes and both the preceding and following states are in running state are corrected to running state; invalid state segments with too short a duration are filtered out to obtain stable running and stop segment division results.

[0038] III. Consistency Verification of Auxiliary Measurement Points To eliminate non-target temperature rises caused by cooling system malfunctions, medium deterioration, or abnormal measurement points, the consistency of cooling-side operating conditions within the evaluation data window is verified before calculating the reflux anomaly index.

[0039] Preferably, the following determination is made only within the running status segment of the evaluation data window: 1. The maximum value of oil-water mixture satisfies: ; in, This indicates the maximum value of the oil-water mixture within the operating status segment of the evaluation data window. Indicates the permissible threshold for oil-water mixing; 2. The deviation of the cooling water main flow rate from the median of the evaluation window relative to the reference window satisfies: ; in, To assess the median flow rate of the main cooling water pipe inside the window, The median flow rate of the main cooling water pipe within the reference window. This is the flow deviation threshold; 3. The deviation of the cooling water main pressure from the median of the evaluation window relative to the reference window satisfies the following: ; in, To assess the median pressure in the main cooling water pipe inside the window, The median pressure in the main cooling water pipe within the reference window. This is the pressure deviation threshold; 4. The median difference in cooler outlet water temperature satisfies: ; in, To assess the median outlet water temperature of the in-window cooler, The median outlet water temperature of the reference in-window cooler. This is the threshold value for the outlet water temperature difference. In this example, we take... It can be adaptively adjusted according to the type of unit, the accuracy of the measuring points, historical statistical patterns, model training results, or on-site operating experience.

[0040] When the above conditions are met simultaneously, the cooling side is determined to be stable, and the subsequent fault characteristic calculation is performed; otherwise, the cooling side is determined to have a suspected abnormality, and the suspected abnormality of the cooling side is output or the matching confidence is reduced.

[0041] IV. Feature Extraction After passing the cooling-side consistency check, the baseline data window and the evaluation data window are divided into fixed-length time windows, and representative features are extracted within each window. The preferred time window length is a fixed minute-level window, such as a 10-minute window, but 5-minute, 15-minute, 30-minute, or other suitable lengths can be selected based on the field application requirements.

[0042] For the shutdown state window, extract the oil level characteristic value. Preferably, the maximum oil level within the window can be used as the oil level characteristic of that window to reflect the oil level rise trend caused by oil backfilling during shutdown.

[0043] For the operating status window, extract the oil temperature characteristic value. Preferably, the maximum oil temperature in the oil tank within the window can be used as the oil temperature characteristic of that window to reflect the abnormal rise in oil temperature caused by hot oil reflux during operation.

[0044] For the tile temperature characteristics, if multiple tile temperature measurement points are set up, the multiple tile temperature measurement points can be aggregated at the minute level to form a representative tile temperature sequence per minute; then, the representative sequence can be statistically analyzed within a time window to obtain the tile temperature window characteristics. The aggregation method and the statistical method within the window can use the maximum value, median, mean, quantile value or other robust statistical measures, which can be selected according to the number of measurement points, measurement stability and field experience.

[0045] To ensure the validity of window features, only time windows within which the amount of valid data meets preset requirements are retained. The required amount of valid data can be expressed as a certain proportion not less than the total length of the window, and this proportion can be set according to the sampling period and data integrity requirements.

[0046] V. Construction of the Benchmark Range This invention uses a machine learning normal sample model to adaptively establish normal ranges and abnormal boundaries based on the statistical distribution of each feature in the benchmark data window.

[0047] Example method: Constructing a baseline range based on a Gaussian statistical model For any index ,in Let the feature sample set corresponding to the baseline window be: ; Before fitting, it is preferable to perform robust extremum removal on the sample set. The median absolute deviation method is preferred for calculation: ; ; Will satisfy Samples that are considered abnormal baseline samples are removed. The preferred value is 3.

[0048] After removing extreme samples, the Gaussian distribution parameters of the remaining samples are estimated to obtain the mean. and standard deviation : ; ; When the sample size is small or when there is a higher requirement to combat outliers, a robust estimation method can be used, replacing the mean with the median. Alternative standard deviation.

[0049] Since the fault described in this invention is an abnormal increase in oil level, oil temperature, and bearing temperature, a single-sided upper tail reference range is used. For any index... Construct a normal reference upper boundary With significant anomaly boundary : ; ; in, and It is a positive parameter, and Preferably, Take a value between 1.5 and 2.0. Take a value between 3.0 and 4.0. Therefore, to Corresponding to the normal fluctuation range, to The corresponding degradation transition range exceeds Corresponding to a significantly abnormal interval.

[0050] When the benchmark samples exhibit mixed characteristics across multiple operating conditions, a Gaussian mixture model can be used to model them, describing the multi-peaked distribution characteristics corresponding to different operating sub-conditions under normal conditions. Alternatively, machine learning methods can be employed to establish a normal-state benchmark model. These machine learning methods may include single-class support vector machines, isolated forests, autoencoders, Gaussian process regression models, or other models suitable for learning the boundaries, distributions, or reconstruction errors of normal samples. After training the normal-state model using benchmark window samples, it can be further used to calculate the degree of deviation of feature samples within the evaluation window from the normal pattern.

[0051] The aforementioned statistical or machine learning models are all used to achieve the same goal: to adaptively establish the normal state range or normal state boundary based on benchmark samples, so as to reduce the problem that fixed empirical thresholds are not adaptable to different units, different measuring points and different operating conditions.

[0052] VI. Deterioration Calculation After obtaining the baseline model, the degradation degree is calculated for the feature values ​​in the evaluation data window. The degradation degree refers to an index that maps feature values ​​of different dimensions and fluctuation levels, such as oil level, oil temperature, and bearing temperature, to the same numerical range, representing the degree of abnormality of the current window relative to the normal state. Preferably, the degradation degree can be normalized to the range of 0 to 1, where a larger value indicates a more significant deviation from the normal state.

[0053] When using a Gaussian statistical model to establish a benchmark model, the standardized deviation of the evaluation eigenvalues ​​relative to the center position parameter of the benchmark sample can be calculated first. Then, the degradation degree can be calculated based on the positional relationship between the standardized deviation and the reference boundary and the anomaly boundary. Specifically, when the current eigenvalue is within the normal fluctuation range, the degradation degree takes a low value or zero; when the current eigenvalue is between the normal reference boundary and the significant anomaly boundary, the degradation degree increases monotonically with the degree of deviation; when the current eigenvalue exceeds the significant anomaly boundary, the degradation degree approaches 1 or takes 1. This mapping can be implemented using linear functions, nonlinear functions, probability cumulative functions, exponential functions, sigmoid functions, or other monotonic mapping functions.

[0054] The standardized deviation can be converted into anomaly probability using the Gaussian cumulative distribution function, and the degree of degradation can then be determined based on the magnitude of the anomaly probability. Compared with simple fixed threshold discrimination, this method can better adapt to the variability of field measurement points and the distribution characteristics of normal samples.

[0055] When using a Gaussian mixture model, the tail exceedance probability of the evaluation eigenvalue in a normal mixture distribution can be calculated, and this probability can be mapped to the degree of degradation. The smaller the exceedance probability, the higher the degree of degradation.

[0056] When using a machine learning model, the degree of degradation can be calculated based on the anomaly score, reconstruction error, boundary distance, or probability score output by the model, and then normalized to map it to a uniform range. The low-risk and high-risk boundaries of the anomaly score can be obtained from the distribution of benchmark window sample scores, adaptive training results, or analysis of historical failure samples.

[0057] Using the methods described above, the oil level degradation degree, oil temperature degradation degree, and bearing temperature degradation degree are obtained respectively. It should be noted that the boundary parameters, probability thresholds, scoring thresholds, and normalization methods used in the degradation degree calculation are all reference settings and can be adjusted according to the actual application requirements on site.

[0058] In this embodiment, a degradation degree function based on the cumulative probability of a Gaussian distribution is preferably used.

[0059] For any window in the evaluation window eigenvalues The standardized deviation is defined as: ; in, To prevent the extremely small positive number from causing instability in the denominator due to an excessively small standard deviation, it is preferable to take... to .

[0060] set up If the cumulative distribution function is a standard normal distribution, then the index The degree of degradation is defined as: ; in, , For normal reference boundary coefficients, This represents the boundary coefficient for significant anomalies. When the current eigenvalue is within the normal fluctuation range, the degradation degree is 0; when the current eigenvalue enters the transition range above the normal fluctuation range, the degradation degree increases monotonically with the deviation probability; when the current eigenvalue enters the significant anomaly range, the degradation degree is 1.

[0061] VII. Calculation of Reflux Anomaly Index When using a single-index degradation model, the oil level degradation degree is obtained respectively. Oil temperature deterioration and temperature degradation .

[0062] Since the available features of the running window and the shutdown window are different, this embodiment preferably performs normalized weighted fusion on the available features within the current window. Let the window... The available set of indicators is The formula for calculating the backflow anomaly index (BAI) is: ; In the formula This refers to the degree of oil level deterioration. This refers to the degree of oil temperature degradation. The degree of temperature degradation; , , Let be the weighting coefficient, and satisfy: . Representing windows respectively Are the internal oil level, oil temperature, and bearing temperature characteristics available?

[0063] When window When running the window, ; When window When the shutdown window is open, .

[0064] In this example, take , , .

[0065] VIII. Matching Degree and Fault Diagnosis The backflow anomaly indices corresponding to each effective time window within the assessment data window are constructed into a time series, and the overall matching degree is calculated based on this series. Preferably, a high quantile statistic can be used to characterize the fault matching level within the assessment phase, such as the 90th quantile value as the matching degree, but it is not limited to this; the 95th quantile value, the maximum value, the weighted average value, or other suitable statistics can also be used.

[0066] When the matching degree reaches or exceeds the preset diagnostic threshold, the warning is determined to match a fault of jammed or improperly closed bearing oil pump outlet check valve; when the matching degree is lower than the threshold, it is determined to be a mismatch or weak match. The diagnostic threshold can be determined based on historical normal samples, fault samples, false alarm rate constraints, recall rate requirements, or model training results, and can be adjusted according to the actual situation of different units.

[0067] IX. Implementation Results When using the method of this invention, under the premise that the overall cooling side conditions are stable, if the bearing oil pump outlet check valve is stuck or not closing tightly, the unit will typically exhibit an abnormal increase in oil temperature in the oil sump and bearing temperature during operation, and an abnormal increase in oil level in the oil sump during shutdown. By extracting the joint features of the operating and shutdown states, and establishing a backflow anomaly index by combining a benchmark model and a degradation degree mapping, the abnormal backflow fault of the check valve can be identified.

[0068] This invention does not rely on new sensors; it can achieve diagnosis based solely on existing measurement data of the unit. Furthermore, since the baseline range and degradation degree can be adaptively constructed through statistical or machine learning models, it has good adaptability, scalability, and engineering application value.

[0069] Example 2: Fault Diagnosis System Based on the Method of the Invention This embodiment provides a fault diagnosis system for the outlet check valve of a hydroelectric generator bearing oil pump based on a backflow anomaly index, including: Data Acquisition Module: Connects to the existing monitoring system of the hydropower plant (such as SCADA system) to collect real-time operating data of the unit and bearing oil system, including unit active power, unit speed, guide vane opening, oil level in the oil sump, oil temperature in the oil sump, bearing temperature, cooling water flow rate, cooling water pressure, and cooler outlet water temperature. The data sampling interval is configurable, with a default setting of 1 minute.

[0070] Preprocessing and Status Identification Module: This module preprocesses the collected raw data, including missing value imputation (forward imputation, maximum imputation time 5 minutes), outlier removal (based on the 3σ criterion), and time series data processing. Then, based on three indicators—unit active power, unit speed, and guide vane opening—the module jointly determines the unit's operating status, dividing the data into operating status data and shutdown status data.

[0071] Cooling system verification module: This module uses data from cooling water flow rate, cooling water pressure, cooler outlet water temperature, and oil-water mixing measurement points to verify the consistency of the cooling system's operating status. If all verification conditions are met, the cooling system is considered to be operating normally; otherwise, a cooling system malfunction alarm is output, and the fault diagnosis process is paused.

[0072] Feature extraction module: Divides the preprocessed data into fixed-length time windows (default 10 minutes). Within the running status window, it extracts oil temperature and bearing temperature features; within the stopped status window, it extracts oil level features. The feature extraction method is configurable; the default is to use the maximum value within the window as the feature value.

[0073] The baseline model building module establishes a normal-state baseline model based on feature data within the baseline data window (default 72 hours before the warning). It supports multiple model types, including Gaussian distribution models, Gaussian mixture models, single-class support vector machines, isolated forests, autoencoders, and Gaussian process regression models. Users can choose the appropriate model type based on their specific needs.

[0074] Deterioration Calculation Module: Based on the deviation of characteristic data within the evaluation data window from the normal baseline range, calculates the oil level deterioration, oil temperature deterioration, and bearing temperature deterioration. The deterioration degree is normalized to the range of 0 to 1, with larger values ​​indicating a more significant deviation from the normal state.

[0075] The reflux anomaly index calculation module: Based on different window types, it normalizes and weights the available degradation indicators to construct the reflux anomaly index (BAI). The weighting coefficients are configurable, with default settings of 0.40 for oil level, 0.35 for oil temperature, and 0.25 for bearing temperature.

[0076] Fault determination module: Constructs a time series of reflux anomaly indices corresponding to each valid time window within the evaluation data window, and calculates the 90th percentile value of this series as the overall matching degree. When the matching degree reaches or exceeds the preset diagnostic threshold (default 0.7), an alarm is output indicating that the check valve is stuck or not closing properly; otherwise, the system is considered normal.

[0077] Human-Machine Interface: Provides a visual interface to display the unit's operating status, real-time data from various measuring points, trends in the return flow anomaly index, and fault diagnosis results. Supports historical data query, fault case management, and system parameter configuration functions.

[0078] This system can be integrated into the existing equipment condition monitoring system of hydropower plants to achieve online real-time diagnosis of faults in the outlet check valve of bearing oil pumps. No new sensor equipment is required, and it has the advantages of low implementation cost, high diagnostic accuracy, and easy promotion and application.

[0079] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of this invention. It should be understood that the above descriptions are merely specific embodiments of this invention and are not intended to limit the scope of protection of this invention. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method for diagnosing the fault of the outlet check valve of a hydroelectric generator bearing oil pump based on a backflow anomaly index, characterized in that, Includes the following steps: S1: Collect unit operating condition data and bearing oil system operating data; S2: Preprocess the collected data and identify the unit's operating status based on the preprocessed data, dividing the data into operating status data and shutdown status data; S3: Use relevant measurement data of the cooling water system to verify the operating status of the cooling system and eliminate abnormal factors of the cooling system; S4: Extract oil temperature and bearing temperature characteristics while the machine is running; extract oil level characteristics while the machine is stopped. S5: Establish a normal state benchmark model based on the feature data within the benchmark data window, and determine the normal benchmark range corresponding to each feature; S6: Calculate the oil level deterioration, oil temperature deterioration, and bearing temperature deterioration based on the degree of deviation of the characteristic data within the evaluation data window from the normal reference range; S7: Based on the oil level deterioration, oil temperature deterioration, and bearing temperature deterioration, and combined with the window type, a normalized weighted fusion is performed to construct the reflux anomaly index BAI; S8: Determine whether the bearing oil pump outlet check valve is stuck or not closing properly based on the backflow anomaly index BAI.

2. The method for diagnosing the fault of the outlet check valve of a hydroelectric generator bearing oil pump based on the backflow anomaly index as described in claim 1, characterized in that, The data collected in step S1 includes the unit's active power, unit speed, guide vane opening, oil level in the oil tank, oil temperature in the oil tank, bearing temperature, cooling water flow rate, cooling water pressure, and cooler outlet water temperature.

3. The method for diagnosing the fault of the outlet check valve of a hydroelectric generator bearing oil pump based on the backflow anomaly index as described in claim 1, characterized in that, Step S2, data preprocessing, includes missing value imputation, outlier removal, and time series data processing. Unit operating status identification is determined jointly by three indicators: unit active power, unit speed, and guide vane opening. Specifically: Construct state discriminants separately: ； ； ； in, The active power of the unit. For unit speed, For guide vane opening; , , These are the corresponding thresholds; Define the overall voting result as: ； when When, it is determined to be in running state; when When the machine is in a stopped state, it is determined to be in a stopped state. When the time is right, it is determined to be an unknown state.

4. The method for diagnosing the fault of the outlet check valve of a hydroelectric generator bearing oil pump based on the backflow anomaly index as described in claim 1, characterized in that, The cooling system operating status verification in step S3 includes the following conditions: (1) The maximum value of oil-water mixture satisfies: ； in, This indicates the maximum value of the oil-water mixture within the operating status segment of the evaluation data window. Indicates the permissible threshold for oil-water mixing; (2) The deviation of the cooling water main flow rate from the median of the evaluation window relative to the reference window satisfies: ； in, To assess the median flow rate of the main cooling water pipe inside the window, The median flow rate of the main cooling water pipe within the reference window. This is the flow deviation threshold; (3) The deviation of the cooling water main pressure from the median of the evaluation window relative to the reference window satisfies: ； in, To assess the median pressure in the main cooling water pipe inside the window, The median pressure in the main cooling water pipe within the reference window. This is the pressure deviation threshold; (4) The median difference in cooler outlet water temperature satisfies: ； in, To assess the median outlet water temperature of the in-window cooler, The median outlet water temperature of the reference in-window cooler. The threshold for the outlet water temperature difference; When the above conditions are met simultaneously, the cooling side is determined to be stable, and the subsequent fault characteristic calculation is performed; otherwise, the cooling side is determined to have a suspected abnormality, and the suspected abnormality of the cooling side is output or the matching confidence is reduced.

5. The method for diagnosing the fault of the outlet check valve of a hydroelectric generator bearing oil pump based on the backflow anomaly index as described in claim 1, characterized in that, In step S5, the normal state baseline model is constructed using a statistical distribution model or a machine learning model; the statistical distribution model includes a Gaussian distribution model or a Gaussian mixture distribution model; the machine learning model includes at least one of a single-class support vector machine, an isolated forest, an autoencoder, or a Gaussian process regression model.

6. The method for diagnosing the fault of the outlet check valve of a hydroelectric generator bearing oil pump based on the backflow anomaly index as described in claim 5, characterized in that, When using a Gaussian distribution model to construct a normal state baseline model, the degradation degree calculation in step S6 is as follows: For any window in the evaluation window eigenvalues The standardized deviation is defined as: ； in, Features within the reference window k The mean, Features within the reference window k standard deviation To prevent the denominator from becoming unstable due to excessively small standard deviations; set up If the cumulative distribution function is a standard normal distribution, then the index The degree of degradation is defined as: ； in, , For normal reference boundary coefficients, This represents the boundary coefficient for significant anomalies.

7. The method for diagnosing the fault of the outlet check valve of a hydroelectric generator bearing oil pump based on the backflow anomaly index as described in claim 1, characterized in that, The formula for calculating the reflux anomaly index BAI in step S7 is as follows: ； in, This refers to the degree of oil level deterioration. This refers to the degree of oil temperature degradation. The degree of temperature degradation; , , Let be the weighting coefficient, and satisfy: ; Representing windows respectively Are the internal oil level, oil temperature, and bearing temperature characteristics available? When window When running the window, ； When window When the shutdown window is open, 。 8. The method for diagnosing the fault of the outlet check valve of the bearing oil pump of a hydroelectric generator based on the backflow anomaly index as described in claim 1, characterized in that, Step S8 specifically involves: constructing a time series from the backflow anomaly index corresponding to each effective time window within the evaluation data window, and calculating the high quantile statistic of the series as the overall matching degree; when the matching degree reaches or exceeds the preset diagnostic threshold, it is determined that the bearing oil pump outlet check valve has a jamming or improper closing fault.

9. A fault diagnosis system for the outlet check valve of a hydroelectric generator bearing oil pump based on a backflow anomaly index, characterized in that, The system is implemented based on the method described in any one of claims 1-8, comprising: The data acquisition module is used to collect unit operating condition data and bearing oil system operating data; The preprocessing and status identification module is used to preprocess the collected data and identify the unit's operating status based on the preprocessed data, dividing the data into operating status data and shutdown status data. The cooling system verification module is used to verify the operating status of the cooling system using relevant measurement point data of the cooling water system and to eliminate abnormal factors in the cooling system. The feature extraction module is used to extract oil temperature and bearing temperature features when the machine is running, and to extract oil level features when the machine is stopped. The benchmark model construction module is used to build a normal state benchmark model based on the feature data within the benchmark data window and to determine the normal benchmark range corresponding to each feature. The degradation calculation module is used to calculate the oil level degradation, oil temperature degradation, and bearing temperature degradation based on the degree of deviation of the characteristic data in the evaluation data window from the normal reference range. The reflux anomaly index calculation module is used to construct the reflux anomaly index BAI by performing normalized weighted fusion based on the oil level deterioration, oil temperature deterioration, and bearing temperature deterioration, combined with the window type. The fault determination module is used to determine whether there is a jamming or improper closing fault in the bearing oil pump outlet check valve based on the backflow anomaly index (BAI).

10. The fault diagnosis system for the outlet check valve of the hydroelectric generator bearing oil pump based on the backflow anomaly index according to claim 9, characterized in that, The data collected by the data acquisition module includes the unit's active power, unit speed, guide vane opening, oil level in the oil tank, oil temperature in the oil tank, bearing temperature, cooling water flow rate, cooling water pressure, and cooler outlet water temperature.

11. The fault diagnosis system for the outlet check valve of a hydroelectric generator bearing oil pump based on the backflow anomaly index according to claim 9, characterized in that, In the preprocessing and status identification module, data preprocessing includes missing value filling, outlier removal, and time series data processing; unit operating status identification is determined by a combination of three indicators: unit active power, unit speed, and guide vane opening.

12. The fault diagnosis system for the outlet check valve of the hydroelectric generator bearing oil pump based on the backflow anomaly index according to claim 9, characterized in that, The benchmark model building module constructs a normal state benchmark model through a statistical distribution model or a machine learning model; the statistical distribution model includes a Gaussian distribution model or a Gaussian mixture distribution model; the machine learning model includes at least one of a single-class support vector machine, an isolated forest, an autoencoder, or a Gaussian process regression model.

Images