A gate water leakage detection model construction method applied to a hydropower station

Abstract

The present application relates to the technical field of water leakage detection of hydropower station gate, and relates to a gate water leakage detection model construction method applied to hydropower station, the steps of the method comprising: obtaining time series pressure data of historical volute section and historical tail water section based on the laid pressure detection module, respectively performing data preprocessing on the time series pressure data of the historical volute section and the historical tail water section to represent the time series data set of the volute section and the time series data set of the tail water section, respectively constructing a volute section fault detection model and a tail water section fault detection model by applying an adaptive filtering algorithm, taking the time series data set of the volute section as the input of the volute section fault detection model and taking the time series data set of the tail water section as the input of the tail water section fault detection model, respectively performing model training on the volute section fault detection model and the tail water section fault detection model, and combining an evaluation index algorithm to obtain the trained volute section fault detection model and the trained tail water section fault detection model with optimal values.

Description

Technical Field

[0001] This invention relates to the field of gate leakage detection technology in hydropower stations, and more specifically, to a method for constructing a gate leakage detection model for hydropower stations. Background Technology

[0002] Currently, determining whether leakage exists in the emergency gates and tailrace gates of hydropower stations typically requires a comprehensive assessment combining the drainage times of the tailrace vent valve, the vent valve, and the pressure drop in the vent valve's pressure pipeline. If the emergency gates or tailrace gates in the vent valve have a malfunction, leakage is likely to occur. At present, hydropower stations often rely on maintenance personnel repeatedly raising and lowering the gates based on experience until they reach the set level before drainage. This process is not only time-consuming but also lacks objective comparison and analysis of the hydropower station's pressure data due to subjective, experience-based judgment. This can easily lead to discrepancies between measured and actual leakage data, resulting in significant uncertainty and failing to meet operational requirements. Therefore, we have designed a gate leakage detection model construction method for hydropower stations to address the aforementioned technical problems. Summary of the Invention

[0003] The purpose of this invention is to provide a method for constructing a gate leakage detection model for hydropower stations, which solves the above-mentioned technical problems.

[0004] The embodiments of the present invention are achieved through the following technical solutions:

[0005] A method for constructing a gate leakage detection model for hydropower stations, characterized by the following steps:

[0006] A reliability probability table is established. Based on the deployed pressure detection modules, time-series pressure data of the historical spiral section and historical tailrace section are obtained. The time-series pressure data of the historical spiral section and historical tailrace section are preprocessed to form single-mode information of pressure of the historical spiral section and historical tailrace section. According to the reliability probability table and Bayesian inference model, the corresponding multi-mode fusion probability is obtained. The reliable time-series information of the historical spiral section and historical tailrace section is obtained based on the multi-mode fusion probability, which is represented as time-series dataset of spiral section and time-series dataset of tailrace section, and then proceeds to the next step.

[0007] An adaptive filtering algorithm was applied to construct fault detection models for the spiral section and the tailrace section, respectively. The time series dataset of the spiral section was used as the input of the spiral section fault detection model, and the time series dataset of the tailrace section was used as the input of the tailrace section fault detection model. The spiral section fault detection model and the tailrace section fault detection model were trained respectively. The evaluation index algorithm was set and the optimal value was used as the evaluation result to obtain the trained spiral section fault detection model and tailrace section fault detection model.

[0008] The trained spiral section fault detection model and tailrace section fault detection model are deployed to the corresponding areas, and the real-time monitoring data of the spiral section and tailrace section are input into the corresponding models to complete the gate leakage detection of the hydropower station.

[0009] Optionally, the deployed pressure detection modules include: a pressure monitor deployed at the inlet of the spiral casing, a pressure monitor deployed at the tail of the spiral casing, a pressure monitor deployed at the inlet of the tailwater pipe, and a pressure monitor deployed at the tail of the tailwater pipe.

[0010] Optionally, the historical time-series pressure data for the volute segment includes: pressure data at the volute inlet and the corresponding displacement per unit time, as well as pressure data at the volute tail and the corresponding displacement per unit time.

[0011] Optionally, the historical tailrace section time-series pressure data includes: the pressure data at the tailrace pipe inlet and the corresponding discharge per unit time, as well as the pressure data at the tailrace pipe end and the corresponding discharge per unit time.

[0012] Optionally, the unit time drainage volume corresponding to the pressure data of the volute inlet and tail is specifically obtained by calculating the flow area of ​​the preset volute vent valve and using Bernoulli's equation.

[0013] The unit time drainage volume corresponding to the pressure data of the tailpipe inlet and tail is specifically obtained by calculating the preset tailpipe volume using Bernoulli's equation.

[0014] The formula for calculating Bernoulli's equation is:

[0015] p+ρgh+(1 / 2)*ρv^2=C*.

[0016] Optionally, the time-series pressure data of the historical spiral section and the historical tailrace section are preprocessed separately, and the data preprocessing method is Kalman filtering.

[0017] Optionally, the evaluation index algorithm can be set to one of the following: goodness-of-fit R-squared evaluation index algorithm, mean absolute error (MAE) evaluation index algorithm, mean square error (MSE) evaluation index algorithm, and root mean square error (RMSE) evaluation index algorithm.

[0018] A gate leakage detection model construction system for hydropower stations, employing the aforementioned gate leakage detection model construction method for hydropower stations, includes: a historical data acquisition module, a data preprocessing module, a model construction module, a model training module, and a model evaluation module connected sequentially. The historical data acquisition module acquires time-series pressure data for the historical spiral section and the historical tailrace section. The data preprocessing module preprocesses the time-series pressure data for the historical spiral section and the historical tailrace section. The model construction module uses an adaptive filtering algorithm to construct fault detection models for the spiral section and the tailrace section, respectively. The model training module trains the fault detection models for the spiral section and the tailrace section. The model evaluation module combines a set evaluation index algorithm and uses the optimal value as the evaluation result to evaluate the fault detection models for the spiral section and the tailrace section, respectively.

[0019] An electronic device, comprising:

[0020] Memory, used to store computer programs;

[0021] A processor is used to execute the computer program to implement the steps of the above-described method for constructing a gate leakage detection model for hydropower stations.

[0022] A readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described method for constructing a gate leakage detection model for hydropower stations.

[0023] The technical solutions of the embodiments of the present invention have at least the following advantages and beneficial effects:

[0024] This embodiment uses a model based on Bernoulli's principle to detect faults in the spiral section gate and tailrace section gate. This model can scientifically calculate gate leakage, replace manual timing to measure guide vane leakage, effectively monitor the gate closure status during unit maintenance, and solve problems such as whether the gate is properly closed and delays in maintenance schedules.

[0025] Furthermore, this embodiment only requires the use of existing data acquisition methods, without the need for additional monitoring devices, and does not affect the normal operation of the unit, resulting in extremely low cost. This diagnostic method aims to efficiently identify gate leakage faults caused by unsuccessful gate opening in the spiral casing section and tailrace section, thereby improving maintenance efficiency, shortening maintenance cycles, and saving maintenance costs. Attached Figure Description

[0026] Figure 1 A flowchart illustrating a method for constructing a gate leakage detection model for hydropower stations, provided by the present invention.

[0027] Figure 2 A schematic diagram of the volute pressure drop curve provided by the present invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0029] like Figure 1 , Figure 2 As shown, the present invention provides one embodiment: a method for constructing a gate leakage detection model for hydropower stations, characterized in that the method includes the following steps:

[0030] A stable probability table is established. Based on the deployed pressure detection modules, time-series pressure data of the historical spiral section and historical tailrace section are obtained. The time-series pressure data of the historical spiral section and historical tailrace section are preprocessed to form single-mode information of pressure of the historical spiral section and historical tailrace section. Based on the stable probability table and Bayesian inference model, the corresponding multi-modal fusion probability is obtained. Based on the multi-modal fusion probability, the stable time-series information of the historical spiral section and historical tailrace section is obtained to represent the time-series dataset of the spiral section and the time-series dataset of the tailrace section, and then proceed to the next step.

[0031] An adaptive filtering algorithm was applied to construct fault detection models for the spiral section and the tailrace section, respectively. The time series dataset of the spiral section was used as the input of the spiral section fault detection model, and the time series dataset of the tailrace section was used as the input of the tailrace section fault detection model. The spiral section fault detection model and the tailrace section fault detection model were trained respectively. The evaluation index algorithm was set and the optimal value was used as the evaluation result to obtain the trained spiral section fault detection model and tailrace section fault detection model.

[0032] The trained spiral section fault detection model and tailrace section fault detection model are deployed to the corresponding areas, and the real-time monitoring data of the spiral section and tailrace section are input into the corresponding models to complete the gate leakage detection of the hydropower station.

[0033] There are m historical single-mode pressure data points for the spiral section, each corresponding to k stable probability tables. There are n historical single-mode pressure data points for the tailrace section, each corresponding to j stable probability tables. The reliability probability table Efg stores the reliability probability of the detection data of the f-th gate representing the g-th single-mode information. Therefore, by using the reliability probability table Efg, the reliability probability P(M) of the F-th gate representing that measurement value can be retrieved based on the detection data of the g-th single-mode information. i |N j Since the detection data is relatively independent and difficult to interact with, in order to avoid false detection and missed detection, objectively acquired operational data is used, and multimodal fusion is adopted to reconstruct and integrate the data. This can break down the barriers between multi-dimensional detection data and effectively obtain accurate detection data.

[0034] In one application of this embodiment, pressure monitors installed at the inlet of the spiral casing and at the tail of the spiral casing collect the corresponding time-series pressure data when the emergency valve of the spiral casing is closed over the years; pressure monitors installed at the inlet of the tailrace pipe and at the tail of the tailrace pipe collect the corresponding time-series pressure data when the tailrace section is closed over the years; and the flow area of ​​the spiral casing vent valve and the volume of the unit's tailrace pipe are obtained from the turbine data.

[0035] The time-series pressure data (the four pressure data types mentioned above and their corresponding unit-time drainage data) obtained from collection and calculation are preprocessed. Kalman filtering is used to smooth the raw data and reduce noise, thereby improving the accuracy of model predictions.

[0036] In this embodiment, the deployed pressure detection module includes: a pressure monitor deployed at the inlet of the volute, a pressure monitor deployed at the tail of the volute, a pressure monitor deployed at the inlet of the tailwater pipe, and a pressure monitor deployed at the tail of the tailwater pipe.

[0037] In one application of this embodiment, the preprocessed pressure data of the spiral casing section and its corresponding discharge per unit time are used as features of the time-series dataset of the spiral casing section to construct a fault detection model for the spiral casing section; similarly, the preprocessed pressure data of the tailrace section and its corresponding discharge per unit time are used as features of the time-series dataset of the tailrace section to construct a fault detection model for the tailrace section. Furthermore, records from historical maintenance logs showing that maintenance work can begin with a single gate closure are labeled "normal data," while records requiring multiple gate closures to begin maintenance are labeled "fault data."

[0038] The data with the above-mentioned characteristics, namely normal and fault data labels, are input into the corresponding fault detection model. The model is trained using an adaptive filtering algorithm (Least Mean Square) to obtain a model that can diagnose faults based on pressure data collected in a short period of time during new maintenance work. In the future, when the valves of the spiral section and tailrace section are closed during maintenance, the model can quickly diagnose whether the closure was successful.

[0039] This embodiment utilizes existing data acquisition methods without requiring additional monitors, thus not affecting normal unit operation and at extremely low cost. This diagnostic method aims to efficiently identify gate leakage faults caused by unsuccessful gate opening in the spiral casing or tailrace section, thereby improving maintenance efficiency, shortening maintenance cycles, and saving maintenance costs.

[0040] In this embodiment, the time-series pressure data of the historical volute segment includes: the pressure data at the volute inlet and the corresponding displacement per unit time, as well as the pressure data at the volute tail and the corresponding displacement per unit time.

[0041] Furthermore, the fault detection model described in this embodiment is an autoregressive model. An autoregressive model (AR) is a machine learning model used for regression. This model is mainly used to process univariate time series data. Any variable in the model can be represented by an equation composed of its own lagged value, other related variables, constant terms, and error terms, and its future value can be predicted through iteration.

[0042] In this embodiment, actual measurement data is used to construct a fault detection model for the spiral section and tailrace section based on the autoregressive method. It uses minimal theoretical assumptions, takes the statistical characteristics of time series as the starting point, requires little data, and can use its own variable series for prediction.

[0043] This embodiment takes into account that during each maintenance period, the amount of data and features available for fault diagnosis is limited, making it unsuitable for machine learning methods that rely on large amounts of data. Therefore, an autoregressive method based on machine learning is adopted.

[0044] In this embodiment, time-series data is a series of values ​​with the same statistical indicators arranged in chronological order of occurrence. If a variable can be observed over time, and past data contains information about future changes in that variable, then a function of the past observation data can be used to predict the future value of that variable. Therefore, the time-series pressure data in the fault detection model in this embodiment is as described above, such as the historical pressure changes of the volute and tailrace sections. Furthermore, the past pressure data contains information about future pressure changes, so historical pressure data can be used to predict future data.

[0045] Based on the aforementioned time-series data, the adaptive filtering method in this embodiment uses a weighted average of historical observations in the time series for prediction. It seeks a "best" set of weights by first calculating a predicted value using a given set of weights, then calculating the prediction error, and finally adjusting the weights based on the error to reduce it. This process is repeated until a "best" set of weights is found, minimizing the error. This weight adjustment process is very similar to the transmission noise filtering process in communication engineering.

[0046] Therefore, the specific steps for model construction in this embodiment are as follows:

[0047] (1) Calculate the displacement according to Bernoulli's equation and preprocess the data using the Kalman filter method.

[0048] (2) Model building: Supervised learning algorithm (autoregressive method) and adaptive filtering method are used for modeling. The processed data is input into the model algorithm for training and validation, the generalization ability of the model is evaluated, and the results are predicted on the test data.

[0049] (3) Establish a model result evaluation system, including but not limited to the goodness-of-fit R-squared evaluation index, the mean absolute error (MAE) evaluation index algorithm, the mean square error (MSE) evaluation index algorithm, and the root mean square error (RMSE) evaluation index algorithm, to evaluate the prediction results of the model algorithm in order to adjust the optimal prediction model.

[0050] In this embodiment, the time-series pressure data of the historical tailrace section includes: the pressure data at the tailrace pipe inlet and the corresponding discharge volume per unit time, as well as the pressure data at the tailrace pipe end and the corresponding discharge volume per unit time.

[0051] In this embodiment, the unit time drainage volume corresponding to the pressure data of the volute inlet and tail is specifically obtained by calculating the flow area of ​​the preset volute vent valve and using Bernoulli's equation.

[0052] The unit time drainage volume corresponding to the pressure data of the tailpipe inlet and tail is specifically obtained by calculating the preset tailpipe volume using Bernoulli's equation.

[0053] The formula for calculating Bernoulli's equation is:

[0054] p+ρgh+(1 / 2)*ρv^2=C*.

[0055] This formula involves the principle of energy balance, specifically that pressure energy + potential energy + kinetic energy are a constant, the volume of the volute is constant (fixed water volume), after determining the opening of the vent valve, the drainage volume is calculated, and the leakage volume can be obtained by subtracting the fixed water volume from the drainage volume.

[0056] In this embodiment, the time-series pressure data of the historical spiral section and the historical tailrace section are preprocessed, and the data preprocessing method is specifically the Kalman filtering method.

[0057] In this embodiment, the evaluation index algorithm is specifically selected from one of the following: goodness-of-fit R-squared evaluation index algorithm, mean absolute error (MAE) evaluation index algorithm, mean square error (MSE) evaluation index algorithm, and root mean square error (RMSE) evaluation index algorithm.

[0058] A gate leakage detection model construction system for hydropower stations, employing the aforementioned gate leakage detection model construction method for hydropower stations, includes: a historical data acquisition module, a data preprocessing module, a model construction module, a model training module, and a model evaluation module connected sequentially. The historical data acquisition module acquires time-series pressure data for the historical spiral section and the historical tailrace section. The data preprocessing module preprocesses the time-series pressure data for the historical spiral section and the historical tailrace section. The model construction module uses an adaptive filtering algorithm to construct fault detection models for the spiral section and the tailrace section, respectively. The model training module trains the fault detection models for the spiral section and the tailrace section. The model evaluation module combines a set evaluation index algorithm and uses the optimal value as the evaluation result to evaluate the fault detection models for the spiral section and the tailrace section, respectively.

[0059] An electronic device, comprising:

[0060] Memory, used to store computer programs;

[0061] A processor is used to execute the computer program to implement the steps of the above-described method for constructing a gate leakage detection model for hydropower stations.

[0062] A readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described method for constructing a gate leakage detection model for hydropower stations.

[0063] In summary, this embodiment establishes a fault detection model for the volute section gate and a fault detection model for the tailrace section gate based on Bernoulli's principle. This model can scientifically calculate gate leakage, replace manual timing methods for measuring guide vane leakage, effectively monitor the gate's maintenance status, and solve problems such as whether the gate is properly closed and delays in maintenance schedules.

[0064] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for constructing a gate leakage detection model for hydropower stations, characterized in that, The steps of this method include: A reliability probability table is established. Based on the deployed pressure detection modules, time-series pressure data of the historical spiral section and historical tailrace section are obtained. The time-series pressure data of the historical spiral section and historical tailrace section are preprocessed to form single-mode information of pressure of the historical spiral section and historical tailrace section. According to the reliability probability table and Bayesian inference model, the corresponding multi-mode fusion probability is obtained. The reliable time-series information of the historical spiral section and historical tailrace section is obtained based on the multi-mode fusion probability, which is represented as time-series dataset of spiral section and time-series dataset of tailrace section, and then proceeds to the next step. An adaptive filtering algorithm was applied to construct fault detection models for the spiral section and the tailrace section, respectively. The time series dataset of the spiral section was used as the input of the spiral section fault detection model, and the time series dataset of the tailrace section was used as the input of the tailrace section fault detection model. The spiral section fault detection model and the tailrace section fault detection model were trained respectively. The evaluation index algorithm was set and the optimal value was used as the evaluation result to obtain the trained spiral section fault detection model and tailrace section fault detection model. The trained spiral casing section fault detection model and tailrace section fault detection model are deployed to the corresponding areas, and the real-time monitoring data of the spiral casing section and tailrace section are input into the corresponding models to complete the gate leakage detection of the hydropower station. The historical time-series pressure data for the volute section includes: pressure data at the volute inlet and the corresponding displacement per unit time, as well as pressure data at the volute tail and the corresponding displacement per unit time. The historical tailrace pressure data includes: the pressure data at the tailrace inlet and the corresponding discharge per unit time, as well as the pressure data at the tailrace and the corresponding discharge per unit time.

2. The method for constructing a gate leakage detection model for hydropower stations according to claim 1, characterized in that, The deployed pressure detection modules include: a pressure monitor installed at the inlet of the spiral casing, a pressure monitor installed at the tail of the spiral casing, a pressure monitor installed at the inlet of the tailwater pipe, and a pressure monitor installed at the tail of the tailwater pipe.

3. The method for constructing a gate leakage detection model for hydropower stations according to claim 2, characterized in that, The unit time drainage volume corresponding to the pressure data of the volute inlet and tail mentioned above is specifically obtained by calculating the flow area of ​​the preset volute vent valve and using Bernoulli's equation. The unit time drainage volume corresponding to the pressure data of the tailpipe inlet and tail is specifically obtained by calculating the preset tailpipe volume using Bernoulli's equation. The formula for calculating Bernoulli's equation is: p+ρgh+(1 / 2) pv^2=C ; This formula involves the principle of energy balance, where pressure energy + potential energy + kinetic energy are constants. The volume of the volute is constant, which means the water volume is fixed. After determining the opening of the vent valve, the drainage volume is calculated. The leakage volume can be obtained by subtracting the fixed water volume from the drainage volume.

4. The method for constructing a gate leakage detection model for hydropower stations according to claim 3, characterized in that, The time-series pressure data of the historical spiral section and the historical tailrace section were preprocessed separately, and the specific data preprocessing method was Kalman filtering.

5. The method for constructing a gate leakage detection model for hydropower stations according to claim 4, characterized in that, The evaluation index algorithm is specifically selected from one of the following: goodness-of-fit R-squared evaluation index algorithm, mean absolute error (MAE) evaluation index algorithm, mean square error (MSE) evaluation index algorithm, and root mean square error (RMSE) evaluation index algorithm.

6. A system for constructing a gate leakage detection model for hydropower stations, characterized in that, The method for constructing a gate leakage detection model for a hydropower station according to any one of claims 1-5 includes: a historical data acquisition module, a data preprocessing module, a model construction module, a model training module, and a model evaluation module connected in sequence; wherein, the historical data acquisition module is used to acquire the time-series pressure data of the historical spiral section and the historical tailrace section; the data preprocessing module is used to preprocess the time-series pressure data of the historical spiral section and the historical tailrace section; the model construction module uses an adaptive filtering algorithm to construct a fault detection model for the spiral section and a fault detection model for the tailrace section respectively; the model training module is used to train the fault detection model for the spiral section and the fault detection model for the tailrace section respectively; and the model evaluation module is used to evaluate the fault detection model for the spiral section and the fault detection model for the tailrace section respectively by combining a set evaluation index algorithm and using the optimal value as the evaluation result. The historical time-series pressure data for the volute section includes: pressure data at the volute inlet and the corresponding displacement per unit time, as well as pressure data at the volute tail and the corresponding displacement per unit time. The historical tailrace pressure data includes: the pressure data at the tailrace inlet and the corresponding discharge per unit time, as well as the pressure data at the tailrace and the corresponding discharge per unit time.

7. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to execute the computer program to implement the steps of the gate leakage detection model construction method for hydropower stations as described in any one of claims 1-5.

8. A readable storage medium, characterized in that, The readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the gate leakage detection model construction method for hydropower stations as described in any one of claims 1-5.

Images