A seepage probability prediction model training and reservoir seepage point positioning method and device

Abstract

The present application relates to the technical field of safety monitoring, and discloses a seepage probability prediction model training method, a reservoir basin seepage point positioning method and device, the present application simulates the readings of each osmometer when seepage occurs in different regions of the reservoir basin anti-seepage region by using the finite element seepage model of the reservoir basin anti-seepage region; the readings of each osmometer in the sample data are taken as input features, the seepage probabilities of different regions in the sample data are taken as output labels, the model is trained, and a seepage probability prediction model is obtained, the seepage probability prediction model is used to determine the seepage probabilities of different regions according to the readings of each osmometer in the reservoir basin anti-seepage region, a mapping relationship between osmometer readings and seepage probabilities of each region is constructed by finite element simulation, and the seepage probability prediction model is trained, the present scheme realizes the positioning of seepage regions based on the existing osmometer network, reduces the seepage monitoring cost, and improves the positioning accuracy and early warning timeliness.

Description

Technical Field

[0001] This invention relates to the field of safety monitoring technology, specifically to a method and device for training a leakage probability prediction model and locating leakage points in reservoirs and basins. Background Technology

[0002] Currently, leakage monitoring in pumped storage power station reservoirs mainly relies on piezometer point monitoring, geophysical exploration, tracer methods, and distributed fiber optic thermometry. Among these, piezometer point monitoring, as the most commonly used method, can only qualitatively determine leakage based on changes in water head, and cannot pinpoint the specific location of the leak. Geophysical exploration, while able to delineate anomaly zones, is costly and cannot provide real-time online early warning. The tracer method not only has a delayed response but also poses environmental pollution risks. While distributed fiber optic thermometry offers high accuracy, it requires pre-laying optical fibers during construction, making it impossible to retrofit existing power stations, and the overall construction and operation costs are high. Summary of the Invention

[0003] This invention provides a training method for a leakage probability prediction model to address the problems in existing pumped storage power station reservoir leakage monitoring technologies, such as the inability of some methods to accurately locate the leakage location and the excessively high monitoring costs.

[0004] In a first aspect, the present invention provides a method for training a leakage probability prediction model, comprising: The finite element seepage model of the reservoir basin seepage prevention area was used to simulate the seepage pressure change data and response time delay data of each piezometer when seepage occurred in different areas of the reservoir basin seepage prevention area. Multiple sets of sample data were obtained. The sample data included reservoir water level data, seepage pressure change data of each piezometer, response time delay data, and the leakage probability of different areas. Different piezometers were set at different locations in the reservoir basin seepage prevention area. The reservoir water level data, pressure change data and response time delay data of each piezometer in the sample data are used as input features, and the leakage probability of different areas in the sample data is used as output labels. The leakage probability prediction model to be trained is obtained. The leakage probability prediction model is used to determine the leakage probability of different areas based on the pressure change data, response time delay data and reservoir water level data of each piezometer in the reservoir seepage prevention area.

[0005] By simulating the changes in seepage pressure and response time delay data when seepage occurs in different areas using a finite element seepage model, a sample library containing water level, seepage pressure response characteristics, and seepage probability is constructed. This library is then used to train a seepage probability prediction model, enabling the prediction and location of seepage probabilities. In particular, by introducing response time delay data, the model's ability to determine the spatial location of seepage sources is enhanced, improving the location accuracy. Furthermore, the data required by this solution is entirely based on the existing piezometer network, requiring no additional hardware investment and reducing monitoring costs.

[0006] In one optional implementation, a finite element seepage model of the reservoir basin's seepage prevention zone is used to simulate the seepage pressure changes and response time delays of various piezometers when leakage occurs in different areas within the reservoir basin's seepage prevention zone, obtaining multiple sets of sample data, including: Determine the baseline leakage intensity; The pressure change data and response time delay data of each piezometer were simulated by the finite element seepage model when the reference leakage intensity occurred in each region, and sample data of the reference leakage intensity occurred in different regions were obtained. Based on the linear relationship between leakage intensity and seepage pressure change data, the seepage pressure change data in each sample data is linearly scaled to obtain multiple sets of sample data corresponding to different leakage intensities.

[0007] By first determining the baseline leakage intensity, the finite element flow model is used to simulate the pressure change and response time delay data of each region under the baseline leakage intensity to obtain baseline samples. Then, the pressure change is linearly scaled using the linear relationship between leakage intensity and pressure change, thereby quickly generating multiple sets of sample data under different leakage intensities. This process only uses the finite element flow model for one round of simulation, which can derive a large number of samples based on the known linear relationship, reducing the time and computational cost of sample generation, while ensuring the accuracy of the samples.

[0008] In one alternative implementation, determining the baseline leakage intensity includes: The initial baseline leakage intensity is determined based on the design standard for the total leakage of the reservoir basin. The finite element flow model was used to obtain the seepage pressure change data of each piezometer under different leakage intensities, and the relationship between leakage intensity and seepage pressure change data was determined. When there is a linear relationship between leakage intensity and seepage pressure change data, the range of values ​​for leakage intensity is selected to determine the range of values ​​for benchmark leakage intensity. The finite element flow model was used to obtain the seepage pressure change data of each piezometer under the calculated intensity, and the minimum seepage pressure change data of each piezometer under the calculated intensity was determined. The initial reference leakage strength is adjusted based on the minimum seepage pressure change data, instrument accuracy, and the range of reference leakage strength values ​​to obtain the reference leakage strength.

[0009] The leakage intensity and pressure change were obtained through finite element simulation, and the linear relationship between the leakage intensity and pressure change was obtained to determine the range of the benchmark leakage intensity. Then, the minimum pressure change under the trial intensity and the instrument accuracy were checked and adjusted to determine the optimal benchmark leakage intensity. This ensured that the benchmark leakage intensity was within the linear range, which ensured the accuracy of the subsequent scaling to generate more samples, and also ensured that the signal intensity generated by all piezometers was greater than the instrument accuracy, thus improving the sample quality.

[0010] In one alternative implementation, the leakage probability is obtained by performing a two-dimensional Gaussian distribution on the distance between each region and the actual leakage point.

[0011] In one alternative implementation, the method further includes: The accuracy of the location is determined by the proportion of the area where the maximum probability predicted by the leakage probability prediction model is located to the actual leakage area. The average positioning error is determined based on the Euclidean distance between the centroid of the probability distribution predicted by the leakage probability prediction model and the actual location. The entropy value of the predicted probability distribution is obtained by calculating the uncertainty of the prediction results based on the leakage probability prediction model. The model performance is evaluated based on the positioning accuracy, average positioning error, and entropy value of the predicted probability distribution to obtain the model performance evaluation results.

[0012] By comprehensively evaluating the leakage probability prediction model from multiple dimensions using the average positioning error, positioning accuracy, and entropy value of the predicted probability distribution, the accuracy, deviation, and confidence of the model's prediction results were determined, providing an evaluation basis for model optimization, selection, and practical application.

[0013] Secondly, the present invention provides a method for locating leakage points in a reservoir, comprising: Acquire the current reservoir water level data, seepage pressure change data of each piezometer in the reservoir basin seepage prevention area, and response time delay data; The current reservoir water level data, the seepage pressure change data of each piezometer, and the response time delay data are input into the pre-constructed leakage probability prediction model for prediction, and the leakage probability prediction results of different areas of the reservoir basin seepage prevention area are obtained.

[0014] The method inputs real-time monitored reservoir water level data, seepage pressure change data, and response time delay data into the leakage probability prediction model, and outputs the leakage probability of each area. The data acquisition of this method relies entirely on the existing piezometer network, without the need for additional hardware investment. This avoids the problems of high cost or difficulty in installing data acquisition equipment for geophysical exploration, fiber optics, and other methods. The introduction of response time delay data enhances the model's sensitivity to leakage location and improves the positioning accuracy.

[0015] In one alternative implementation, the method further includes: Interpolation calculations are performed on the leakage probability prediction results of different areas to obtain a continuous probability field; A probability heatmap of leakage locations is generated based on a continuous probability field.

[0016] By interpolating the leakage probability output by the model, a continuous probability field is generated and then visualized in the form of a heat map. The numerical results are transformed into a readable spatial risk distribution map, which intuitively shows the leakage prediction results, enabling maintenance personnel to quickly and accurately identify leakage risks and their extent.

[0017] Thirdly, the present invention provides a training device for a leakage probability prediction model, comprising: The sample acquisition module is used to simulate the seepage pressure change data and response time delay data of each piezometer when leakage occurs in different areas of the reservoir seepage prevention area using the finite element seepage model of the reservoir seepage prevention area. The sample data includes reservoir water level data, seepage pressure change data of each piezometer, response time delay data, and leakage probability of different areas. Different piezometers are set in different locations in the reservoir seepage prevention area. The model training module is used to train the leakage probability prediction model to be trained by taking the reservoir water level data, the seepage pressure change data and response time delay data of each piezometer in the sample data as input features and the leakage probability of different areas in the sample data as output labels. The leakage probability prediction model is used to determine the leakage probability of different areas based on the seepage pressure change data, response time delay data and reservoir water level data of each piezometer in the reservoir seepage prevention area.

[0018] Fourthly, the present invention provides a reservoir leakage point locating device, comprising: The data acquisition module is used to acquire the current reservoir water level data, the seepage pressure change data and response time delay data of each piezometer in the reservoir basin seepage prevention area; The result prediction module is used to input the current reservoir water level data, the seepage pressure change data of each piezometer, and the response time delay data into the pre-built leakage probability prediction model for prediction, and obtain the leakage probability prediction results of different areas of the reservoir basin seepage prevention area.

[0019] Fifthly, the present invention provides an electronic device, comprising: a memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the computer instructions to perform a leakage probability prediction model training method or a reservoir leakage point location method according to the first aspect or any corresponding embodiment described above.

[0020] In a sixth aspect, the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to execute a leakage probability prediction model training method or a reservoir leakage point location method according to the first aspect or any corresponding embodiment described above.

[0021] In a seventh aspect, the present invention provides a computer program product, including computer instructions, which are used to cause a computer to execute a leakage probability prediction model training method or a reservoir leakage point location method according to the first aspect or any corresponding embodiment described above. Attached Figure Description

[0022] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0023] Figure 1 This is a schematic diagram of an application scenario according to an embodiment of the present invention; Figure 2 This is a flowchart illustrating a method for training a leakage probability prediction model according to an embodiment of the present invention. Figure 3 This is a flowchart illustrating a method for locating leakage points in a reservoir according to an embodiment of the present invention. Figure 4 This is a structural block diagram of a leakage probability prediction model training device according to an embodiment of the present invention; Figure 5 This is a structural block diagram of a reservoir leakage point locating device according to an embodiment of the present invention; Figure 6 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0024] 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, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] It is understood that before using the technical solutions disclosed in the various embodiments of the present invention, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in the present invention and their authorization should be obtained in accordance with relevant laws and regulations through appropriate means.

[0026] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0027] As an optional application scenario of this invention, such as Figure 1 As shown, this leakage probability prediction model training system may include at least one terminal device and at least one server. Figure 1 The system is illustrated in the example, which includes a computer 101, a mobile terminal 102, and a server 103, and the terminal devices such as the computer 101 and the mobile terminal 102 are connected to the server 103 through a network 110.

[0028] Specifically, the terminal device can be a smartphone, tablet, laptop, PDA, desktop computer, game console, smart TV, smart wearable device, in-vehicle terminal, VR (Virtual Reality) device, AR (Augmented Reality) device, etc. Server 103 can be a standalone physical server, a server cluster, a distributed system, or a cloud server providing cloud services. Network 110 can be a wired or wireless network, examples of which include, but are not limited to, the Internet, corporate intranet, local area network, wide area network, mobile communication network, and combinations thereof.

[0029] This invention provides a method for training a leakage probability prediction model. By constructing a mapping relationship between osmotic pressure readings and leakage probabilities in each region through finite element simulation, a leakage probability prediction model is trained. The leakage probability prediction model can locate leakage areas based on existing osmotic pressure gauge networks, thereby reducing leakage monitoring costs and improving positioning accuracy and early warning timeliness.

[0030] According to an embodiment of the present invention, a method for training a leakage probability prediction model is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0031] This embodiment provides a method for training a leakage probability prediction model. Figure 2 This is a flowchart of a leakage probability prediction model training method according to an embodiment of the present invention, such as... Figure 2 As shown, the process includes the following steps: Step S201: Using the finite element seepage model of the reservoir basin seepage prevention area, simulate the seepage pressure change data and response time delay data of each piezometer when leakage occurs in different areas of the reservoir basin seepage prevention area, and obtain multiple sets of sample data. The sample data includes reservoir water level data, seepage pressure change data of each piezometer, response time delay data, and leakage probability of different areas. Different piezometers are set at different locations in the reservoir basin seepage prevention area.

[0032] The finite element method (FEM) seepage model is an offline simulation tool. It's a numerical calculation model based on the FEM method used to simulate the distribution and changes of the seepage field within a reservoir's seepage prevention area. In this scheme, it's used to simulate the seepage pressure changes and response delays of each piezometer during seepage, based on reservoir structural parameters, reservoir water level conditions, and preset seepage conditions, thus obtaining the training samples needed for model training. A piezometer is a sensor used to measure water pressure, i.e., osmotic pressure. It is typically buried in the reservoir's seepage prevention area to monitor changes in water head or pore water pressure at that location in real time. In this scheme, it serves as a data acquisition terminal, collecting data required by the seepage probability prediction model. Seepage pressure change data refers to the difference in water pressure at the same piezometer before and after a seepage event, reflecting the degree of impact of the seepage event on the water pressure at the current location. Response delay data refers to the time interval from the occurrence of a seepage event to the time it takes for the piezometer readings to reach a specific characteristic threshold, reflecting the spatial distance between the seepage source and the monitoring point, as well as the characteristics of seepage propagation. Reservoir water level data refers to the water level elevation of a pumped storage power station, reflecting changes in the water storage within the reservoir basin and influencing the distribution of the seepage field. Leakage probability refers to a label value directly calculated based on pre-defined leakage locations, used as a standard value for model training.

[0033] For example, response time delay data can be expressed as the time required from the occurrence of leakage to the point where the piezometer reading changes by 50% of its peak value. As an example, the change in piezometer pressure can be expressed as: ΔP kij (H)=[ΔP1,ΔP2,...,ΔP K ], where k is the piezometer number, i and j are coordinate indices, and ΔP kij Let τ be the osmotic pressure change data for the k-th piezometer. The response time delay data can be represented as τ. kij (H)=[τ1,τ2,...,τ K ], where k is the piezometer number, i and j are coordinate indices, and τ kij The response time delay data for the kth piezometer is given.

[0034] For example, the location of the piezometers can be determined by dividing the reservoir seepage prevention area into I×J rectangular areas according to planar coordinates, and placing the piezometers at the center point of each sub-area. The center point of each sub-area is denoted as the potential leakage source S. ij Where i=1~I, j=1~J, using K piezometers A already buried in the reservoir basin. k A monitoring network is formed, where k = 1 to K.

[0035] For example, the response time delay data can be the time required from the occurrence of leakage to the change in the piezometer reading reaching 50% of its peak value.

[0036] Step S202: The reservoir water level data, the seepage pressure change data and response time delay data of each piezometer in the sample data are used as input features, and the leakage probability of different areas in the sample data is used as output labels to train the leakage probability prediction model to be trained, so as to obtain the leakage probability prediction model. The leakage probability prediction model is used to determine the leakage probability of different areas based on the seepage pressure change data, response time delay data and reservoir water level data of each piezometer in the reservoir seepage prevention area.

[0037] Output labels refer to the true values ​​provided for the model output during model training, serving as a benchmark for measuring prediction accuracy.

[0038] This process inputs the features into the leakage probability prediction model to be trained. The model outputs the predicted probability of each region. The error between the predicted value and the true label is calculated through the loss function. Then, the backpropagation algorithm is used to propagate the error layer by layer and adjust the internal parameters of the model so that the predicted value continuously approaches the true label. After multiple rounds of iterative optimization, a model that can be used for real-time leakage probability prediction is obtained.

[0039] For example, leakage probability prediction models can employ deep neural networks (DNNs), convolutional neural networks (CNNs), or long short-term memory networks (LSTMs).

[0040] As an example, if a CNN model is used, the neural network structure can be designed as follows: Input layer: The number of neurons is equal to the dimension of the input features, including the seepage change data and response time delay data of each piezometer, as well as the reservoir water level data, i.e., 2K+1.

[0041] Hidden layers: Set up multiple fully connected layers, with a typical configuration as follows: Output layer: The number of neurons equals the total number of sub-regions N = I × J, and the activation function is the softmax function. Each neuron in the output layer corresponds to a sub-region, and its original output value is converted into a probability value through the softmax function, forming a probability distribution across the entire basin. Let the sub-region S be... ijThe probability value of the corresponding output neuron after Softmax calculation is The sum of the probabilities of all regions is 1. The structure of the hidden layers can be: First hidden layer: 256 neurons, ReLU activation function + regularization method (Dropout(0.2)); Second hidden layer: 512 neurons, ReLU activation function + regularization method (Dropout(0.2)); Third hidden layer: 256 neurons, ReLU activation function + regularization method (Dropout(0.2)).

[0042] For example, the loss function can be a cross-entropy loss model, which measures the difference between the predicted probability distribution and the true label, and the loss value can be calculated according to the following formula:

[0043] in, This is the loss value. The output label represents the true leakage probability of the j-th region in the i-th sample. is the model prediction value, representing the predicted leakage probability of the j-th region in the i-th sample.

[0044] This embodiment provides a method for training a leakage probability prediction model. It uses a finite element flow model to simulate the changes in seepage pressure and response time delay data when leakage occurs in different areas. A sample library containing water level, seepage pressure response characteristics, and leakage probability is constructed, and this library is used to train the leakage probability prediction model. This enables the prediction and location of leakage probabilities. By introducing response time delay data, the model's ability to determine the spatial location of leakage sources is enhanced, improving location accuracy. Furthermore, the data required by this solution is entirely based on the existing piezometer network, requiring no additional hardware investment and reducing monitoring costs.

[0045] In an optional embodiment, step S201 involves using a finite element seepage model of the reservoir basin seepage prevention area to simulate the seepage pressure change data and response time delay data of each piezometer when leakage occurs in different areas of the reservoir basin seepage prevention area, obtaining multiple sets of sample data, specifically including: Step a1: Determine the baseline leakage intensity.

[0046] The benchmark leakage intensity refers to the unit leakage intensity used as a reference standard in finite element simulation. It is used to uniformly measure the degree of leakage under different leakage conditions, making the sample data comparable. By defining the benchmark leakage intensity, a data foundation is provided for subsequent scaling through linear relationships to generate more samples.

[0047] Step a2: Using the finite element seepage model, simulate the seepage pressure change data and response time delay data of each piezometer when the reference seepage intensity occurs in each region, and obtain sample data when the reference seepage intensity occurs in different regions.

[0048] In finite element simulations, using a unified baseline leakage intensity as a standard, the simulation iterates through the pressure change data and response time delay data of each piezometer when leakage occurs in each potential leakage sub-region, thereby generating standardized sample data that differs only in leakage location. This process ensures that the differences between samples originate only from the leakage location, allowing model training to focus more on learning the mapping relationship between location and input features.

[0049] Step a3: Based on the linear relationship between leakage intensity and seepage pressure change data, linearly scale the seepage pressure change data in each sample data to obtain multiple sets of sample data corresponding to different leakage intensities.

[0050] Since the leakage intensity and the change in seepage pressure are linearly related within a certain range, the data on each change in seepage pressure can be linearly scaled according to the multiple relationship between the benchmark leakage intensity and other leakage intensities, thereby quickly expanding the sample data.

[0051] For example, the leakage intensity can be 0.5Q0, 2Q0 or 5Q0, where Q0 is the baseline leakage intensity.

[0052] This embodiment provides a method for training a leakage probability prediction model. By first determining the baseline leakage intensity, the method uses a finite element flow model to simulate the pressure change and response time delay data of each region under the baseline leakage intensity to obtain baseline samples. Then, the method uses the linear relationship between leakage intensity and pressure change to linearly scale the pressure change, thereby quickly generating multiple sets of sample data under different leakage intensities. This process only uses the finite element flow model for one round of simulation, which can derive a large number of samples based on the known linear relationship, reducing the time and computational cost of sample generation while ensuring the accuracy of the samples.

[0053] In an optional embodiment, step a1, determining the baseline leakage intensity, specifically includes: Step b1: Confirm the initial baseline leakage intensity based on the design standard for total leakage of the reservoir basin.

[0054] The design standard for total leakage of the reservoir basin refers to the maximum allowable total leakage limit under normal operating conditions specified in the design of the seepage prevention system of the reservoir basin of a pumped storage power station. It is an important design indicator for measuring the seepage prevention effect and controlling leakage risk.

[0055] For example, the design standard for the total leakage of the reservoir basin of a pumped storage power station is usually in the range of tens of liters per second, and the flow rate of local concentrated leakage points is mostly between 0.1 and 10 L / s. Therefore, the initial benchmark leakage intensity can be Q0=1L / s or Q0=0.1L / s as the benchmark value.

[0056] Step b2: Use the finite element flow model to obtain the seepage pressure change data of each piezometer under different leakage intensities, and determine the relationship between leakage intensity and seepage pressure change data.

[0057] Through finite element simulation, a mathematical relationship between leakage intensity and seepage pressure change was established, verifying the existence of a linear relationship between leakage intensity and seepage pressure change, determining the linear interval and scaling factor, and providing a foundation for subsequently determining the interval of benchmark leakage intensity and rapidly generating diverse samples based on linear scaling.

[0058] For example, different leakage intensities can be 0.01L / s, 0.1L / s, 1L / s, 10L / s, or 100L / s, and there is no limitation here.

[0059] Step b3: Select the range of values ​​for leakage intensity when there is a linear relationship between leakage intensity and seepage pressure change data, and determine the range of values ​​for benchmark leakage intensity.

[0060] Under normal leakage conditions, there is a linear relationship between leakage intensity and seepage pressure change. However, when leakage is excessive, it may cause damage such as piping and underground cavities, leading to abrupt changes in seepage pressure change. In this case, the relationship between leakage intensity and seepage pressure change becomes non-linear. In this embodiment of the invention, after determining the baseline leakage intensity, it is necessary to use the linear relationship between the baseline intensity and seepage pressure change to linearly scale the sample data corresponding to the baseline leakage intensity to obtain more samples that conform to physical constraints. Therefore, when determining the baseline leakage intensity, it is necessary to first determine the range of values ​​corresponding to the leakage intensity when there is a linear relationship between leakage intensity and seepage pressure change, and then further determine the baseline leakage intensity from this range.

[0061] For example, by plotting the relationship curve between leakage intensity and seepage pressure change data, if there is a standard linear relationship between leakage intensity and seepage pressure change data in the range of 0.1~10L / s, but a significant nonlinear relationship appears at 100L / s, then the range of 0.5~1L / s can be selected as the range of values ​​for the benchmark leakage intensity, without any restrictions.

[0062] Step b4: Use the finite element seepage model to obtain the seepage pressure change data of each piezometer under the calculated intensity, and determine the minimum seepage pressure change data of each piezometer under the calculated intensity.

[0063] Trial strength refers to the temporary leakage strength value used to simulate and obtain the minimum seepage pressure change.

[0064] By finding the minimum pressure change across all leakage locations using all piezometers under the calculated intensity, the system's ability to capture signals from the least sensitive areas was determined. This minimum pressure change provided a quantitative basis for subsequent adjustments to the baseline leakage intensity, ensuring that the final selected baseline leakage intensity could generate a sufficiently strong leakage signal that could be effectively detected by all piezometers, thereby guaranteeing the quality of the sample data and the training effectiveness of the model.

[0065] For example, the value of the calculated strength can be Q. trial =0.01L / s.

[0066] Step b5: Adjust the initial reference leakage intensity based on the minimum seepage pressure change data, instrument accuracy, and the range of reference leakage intensity values ​​to obtain the reference leakage intensity.

[0067] The minimum pressure change data refers to the pressure change data with the smallest pressure change among all piezometers obtained through trial strength simulation. Instrument accuracy refers to the parameter calibration of the minimum pressure change that the piezometer can reliably detect. It is usually determined by the technical parameters of the equipment and reflects the resolution and sensitivity of the piezometer. Therefore, the minimum pressure change data needs to be greater than the instrument accuracy in order to be captured by the piezometer and avoid being drowned out by measurement noise.

[0068] This process adjusts the initial reference leakage intensity by taking into account the minimum osmotic pressure change, instrument accuracy, and linear range, and determines a reference leakage intensity value that ensures effective detection by all piezometers while remaining within the linear range.

[0069] For example, the data on the change in seepage pressure caused by the benchmark leakage intensity is at least 5 to 10 times the instrument accuracy to meet the signal-to-noise ratio requirements.

[0070] As an example, the baseline leakage intensity can be selected using a weighted scoring method based on the following formula:

[0071] in, Candidate leakage strength The overall score, Candidate leakage strength The linearity score, Candidate leakage strength The signal-to-noise ratio score, For linearity weights, For signal-to-noise ratio weights. As an example, linearity weights. The value can be 0.6, and the signal-to-noise ratio weight is... The value can be 0.4, and there is no restriction here.

[0072] When selecting the benchmark leakage intensity, the candidate leakage intensity with the highest comprehensive score is selected. As a baseline leakage intensity, if multiple candidate leakage intensity scores are close, the smaller value is selected to keep the system within the linear range. For example, a difference of <0.05 can be considered as multiple candidate leakage intensity scores being close.

[0073] Among them, it can be based on or Calculate the linearity score, where, Candidate leakage strength The linearity score, Candidate leakage strength The coefficient of determination for linear fit, This refers to the relative error between the actual response value and the linear prediction value. The coefficient of determination for linear fitting can be determined based on the linear relationship between the change in seepage pressure and the seepage intensity under different candidate seepage intensities.

[0074] The signal-to-noise ratio score can be calculated using the following formula:

[0075] in, Candidate leakage strength The signal-to-noise ratio score, Minimum theoretical variation of piezometer and instrument resolution The ratio, This represents the target signal-to-noise ratio. As an example, the target signal-to-noise ratio can be 5 or 10; there is no restriction here. It can be determined based on... Determine the ratio of the minimum theoretical change in the piezometer to the instrument resolution, where This is the minimum theoretical change value of the piezometer.

[0076] This embodiment provides a method for training a leakage probability prediction model. It obtains the leakage intensity and pressure change through finite element simulation, and acquires the linear relationship between them to determine the range of benchmark leakage intensity. Then, it verifies and adjusts the benchmark based on the minimum pressure change under the calculated intensity and the instrument accuracy, thereby determining the optimal benchmark leakage intensity. This ensures that the benchmark leakage intensity is within the linear range, guaranteeing the accuracy of subsequent scaling to generate more samples, and ensuring that the signal strength generated by all piezometers is greater than the instrument accuracy, thus improving sample quality.

[0077] In an optional embodiment, when performing step S201, the finite element seepage model of the reservoir basin seepage prevention area is used to simulate the seepage pressure change data and response time delay data of each piezometer when leakage occurs in different areas of the reservoir basin seepage prevention area, and multiple sets of sample data are obtained, the leakage probability in the sample data is obtained by performing a two-dimensional Gaussian distribution based on the distance between each area and the actual leakage point.

[0078] The distance between each region and the actual leakage point refers to the spatial distance between the center point of each sub-region after the reservoir's seepage prevention area is divided into multiple sub-regions and the preset actual leakage point in the simulation. As an example, the distance between each region and the actual leakage point can be Euclidean distance.

[0079] For example, the leakage probability can be calculated based on the pressure change data and response time lag data using the following formula:

[0080] in, This represents the probability of leakage. The distance between the area and the actual leak point. The standard deviation is denoted as .

[0081] In an optional embodiment, the method for training a leakage probability prediction model further includes: Step c1: Determine the location accuracy based on the consistency ratio between the area where the maximum probability predicted by the leakage probability prediction model is located and the actual leakage area.

[0082] The region with the highest probability refers to the region with the highest probability of leakage among all regions output by the leakage probability prediction model. It represents the location where the model judges leakage to be most likely. The consistency ratio refers to the percentage of samples whose predicted results match the actual situation out of the total number of test samples, and is used to measure the accuracy of the model's predictions.

[0083] Step c2: Determine the average positioning error based on the Euclidean distance between the centroid of the probability distribution predicted by the leakage probability prediction model and the actual location.

[0084] The centroid of the probability distribution refers to the spatial location obtained by weighting the leakage probabilities of each region output by the model as weights. Physically, it represents the expected location where the leakage source is most likely to occur. The average location error is the average Euclidean distance between the centroid of the probability distribution predicted by the model and the actual leakage point over all test samples. It is used to measure the average deviation between the model's predicted location and the actual location.

[0085] Step c3: Calculate the entropy value of the predicted probability distribution based on the uncertainty of the prediction results from the leakage probability prediction model.

[0086] Uncertainty in the outcome refers to the confidence level of the predicted result, reflected in the concentration or dispersion of the model's output probability distribution. High uncertainty indicates that the model's judgment of the leakage area is ambiguous, manifested as a dispersed probability distribution; low uncertainty indicates that the model's judgment of the leakage area is clear, manifested as a concentrated probability distribution. The entropy value of the predicted probability distribution is used to quantify the uncertainty or confidence level of the prediction result.

[0087] Step c: Evaluate the model performance based on the positioning accuracy, average positioning error, and entropy value of the predicted probability distribution to obtain the model performance evaluation results.

[0088] The model performance evaluation result refers to the conclusion on the reliability of the model after a comprehensive evaluation of the leakage probability prediction model that has been trained, which is used to determine whether the model meets the application requirements.

[0089] This embodiment provides a method for training a leakage probability prediction model. By comprehensively evaluating the leakage probability prediction model from multiple dimensions using the average positioning error, positioning accuracy, and entropy value of the predicted probability distribution, the accuracy, deviation, and confidence of the model's prediction results are determined, providing an evaluation basis for model optimization, screening, and practical application.

[0090] This embodiment provides a method for locating leakage points in a reservoir basin. Figure 3 This is a flowchart illustrating a method for locating leakage points in a reservoir according to an embodiment of the present invention. Figure 3 As shown, the process includes the following steps: Step S301: Obtain the current reservoir water level data, the seepage pressure change data and response time delay data of each piezometer in the reservoir basin seepage prevention area.

[0091] During application, real-time seepage pressure change data and response time delay data are obtained from various piezometers deployed in the reservoir seepage prevention area. The data types are the same as in step S201, and will not be repeated here.

[0092] Step S302: Input the current reservoir water level data, the seepage pressure change data of each piezometer, and the response time delay data into the pre-constructed leakage probability prediction model for prediction, and obtain the leakage probability prediction results of different areas of the reservoir basin seepage prevention area.

[0093] Real-time monitored reservoir water level data, seepage pressure change data, and response time delay data are input into a trained leakage probability prediction model to obtain the leakage probability of each area. The output is a quantified probability distribution, which intuitively reflects the probability of leakage in different areas, providing maintenance personnel with location results.

[0094] This embodiment provides an application method for a leakage probability prediction model. The model inputs real-time monitored reservoir water level data, seepage pressure change data, and response time delay data, and outputs the leakage probability of each area. The data acquisition method relies entirely on the existing piezometer network, without the need for additional hardware investment. This avoids the problems of high cost or difficulty in installing data acquisition equipment for geophysical exploration, fiber optics, and other methods. The introduction of response time delay data enhances the model's sensitivity to leakage location and improves positioning accuracy.

[0095] In an optional embodiment, after obtaining the leakage probability prediction results for different regions by performing step S303, the application method of the leakage probability prediction model provided in this embodiment of the invention further includes: Step d1 involves interpolating the leakage probability prediction results for different regions to obtain a continuous probability field.

[0096] A continuous probability field refers to a function that calculates discrete leakage probabilities as a continuous distribution function containing probability values ​​at all locations within the reservoir's seepage prevention area through interpolation, reflecting the continuous spatial variation trend of leakage risk.

[0097] For example, the interpolation operation can be performed using bilinear interpolation or spline interpolation.

[0098] Step d2: Generate a probability heatmap of the leakage location based on the continuous probability field.

[0099] A leak location probability heatmap is a visual image generated by mapping the probability values ​​of a continuous probability field to different colors. Different colors represent different leak probabilities.

[0100] This embodiment provides an application method for a leakage probability prediction model. By interpolating the leakage probability output by the model, a continuous probability field is generated and then visualized in the form of a heat map. The numerical results are transformed into a readable spatial risk distribution map, which intuitively displays the leakage prediction results, enabling maintenance personnel to quickly and accurately identify leakage risks and their extent.

[0101] This embodiment also provides a leakage probability prediction model training device, which is used to implement the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can be a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0102] This embodiment provides a training device for a leakage probability prediction model, such as... Figure 4 As shown, it includes: The sample acquisition module 401 is used to simulate the seepage pressure change data and response time delay data of each piezometer when leakage occurs in different areas of the reservoir basin seepage prevention area using the finite element seepage model of the reservoir basin seepage prevention area, and obtain multiple sets of sample data. The sample data includes reservoir water level data, seepage pressure change data of each piezometer, response time delay data, and leakage probability of different areas. Different piezometers are set at different locations in the reservoir basin seepage prevention area. The model training module 402 is used to train the leakage probability prediction model to be trained by taking the reservoir water level data, the seepage pressure change data and response time delay data of each piezometer in the sample data as input features and the leakage probability of different areas in the sample data as output labels. The leakage probability prediction model is used to determine the leakage probability of different areas based on the seepage pressure change data, response time delay data and reservoir water level data of each piezometer in the reservoir seepage prevention area.

[0103] The leakage probability prediction model training device provided in this embodiment of the invention can execute a leakage probability prediction model training method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method. Further functional descriptions of the above modules and units are the same as in the corresponding embodiments described above, and will not be repeated here.

[0104] This embodiment also provides a reservoir leakage point locating device, which is used to implement the above embodiments and preferred embodiments. Details that have been described will not be repeated here.

[0105] This embodiment provides a device for locating leakage points in a reservoir, such as... Figure 5 As shown, it includes: Data acquisition module 501 is used to acquire the current reservoir water level data, seepage pressure change data and response time delay data of each piezometer in the reservoir basin seepage prevention area; The result prediction module 502 is used to input the current reservoir water level data, the seepage pressure change data of each piezometer and the response time delay data into the pre-built leakage probability prediction model for prediction, so as to obtain the leakage probability prediction results of different areas of the reservoir basin seepage prevention area.

[0106] The leakage probability prediction model training device provided in this embodiment of the invention can execute a leakage probability prediction model training method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method. Further functional descriptions of the above modules and units are the same as in the corresponding embodiments described above, and will not be repeated here.

[0107] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention.

[0108] The following is a detailed reference. Figure 6This diagram illustrates a suitable structural design for implementing an electronic device according to embodiments of the present invention. The electronic device may include a processor (e.g., a central processing unit, graphics processor, etc.) 601, which can perform various appropriate actions and processes based on a program stored in read-only memory (ROM) 602 or a program loaded from memory 608 into random access memory (RAM) 603. RAM 603 also stores various programs and data required for the operation of the electronic device. The processor 601, ROM 602, and RAM 603 are interconnected via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0109] Typically, the following devices can be connected to I / O interface 605: input devices 606 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 607 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; memory devices 608 including, for example, magnetic tapes, hard disks, etc.; and communication devices 609. Communication device 609 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 6 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown, and more or fewer devices may be implemented or have instead.

[0110] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 609, or installed from a memory 608, or installed from a ROM 602. When the computer program is executed by the processor 601, it performs the functions defined in the leakage probability prediction model training method of the embodiments of the present invention.

[0111] Figure 6 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0112] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as recordable on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and subsequently stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, a leakage probability prediction model training method shown in the above embodiments is implemented.

[0113] A portion of this invention can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to the invention through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0114] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method for training a leakage probability prediction model, characterized in that, The method includes: The finite element seepage model of the reservoir basin seepage prevention area was used to simulate the seepage pressure change data and response time delay data of each piezometer when leakage occurred in different areas of the reservoir basin seepage prevention area. Multiple sets of sample data were obtained. The sample data included reservoir water level data, seepage pressure change data of each piezometer, response time delay data, and leakage probability of different areas. Different piezometers were set at different locations in the reservoir basin seepage prevention area. The reservoir water level data, the seepage pressure change data and the response time delay data of each piezometer in the sample data are used as input features, and the leakage probability of different areas in the sample data is used as output labels to train the leakage probability prediction model to obtain the leakage probability prediction model. The leakage probability prediction model is used to determine the leakage probability of different areas based on the seepage pressure change data, the response time delay data and the reservoir water level data of each piezometer in the reservoir seepage prevention area.

2. The method according to claim 1, characterized in that, The finite element seepage model of the reservoir basin seepage prevention area is used to simulate the seepage pressure change data of each piezometer and the response time delay data when seepage occurs in different areas of the reservoir basin seepage prevention area, to obtain multiple sets of sample data, including: Determine the baseline leakage intensity; The pressure change data and response time delay data of each piezometer were simulated by using the finite element seepage model when the reference leakage intensity occurred in each region, so as to obtain sample data when the reference leakage intensity occurred in different regions. Based on the linear relationship between leakage intensity and the seepage pressure change data, the seepage pressure change data in each sample data is linearly scaled to obtain multiple sets of sample data corresponding to different leakage intensities.

3. The method according to claim 2, characterized in that, The determination of the benchmark leakage intensity includes: The initial baseline leakage intensity is determined based on the design standard for the total leakage of the reservoir basin. The finite element seepage model was used to obtain the seepage pressure change data of each piezometer under different leakage intensities, and the relationship between leakage intensity and seepage pressure change data was determined. When there is a linear relationship between leakage intensity and seepage pressure change data, the range of values ​​for leakage intensity is selected to determine the range of values ​​for benchmark leakage intensity. The finite element seepage model was used to obtain the seepage pressure change data of each piezometer under the calculated intensity, and the minimum seepage pressure change data of each piezometer under the calculated intensity was determined. The initial reference leakage intensity is adjusted based on the minimum osmotic pressure change data, instrument accuracy, and the range of the reference leakage intensity to obtain the reference leakage intensity.

4. The method according to claim 1, characterized in that, The leakage probability is obtained by performing a two-dimensional Gaussian distribution on the distance between each region and the actual leakage point.

5. The method according to claim 1, characterized in that, The method further includes: The location accuracy is determined based on the consistency ratio between the area where the maximum probability predicted by the leakage probability prediction model is located and the actual leakage area. The average positioning error is determined based on the Euclidean distance between the centroid of the probability distribution predicted by the leakage probability prediction model and the actual location. The entropy value of the predicted probability distribution is obtained by calculating the uncertainty of the prediction results of the leakage probability prediction model. The model performance is evaluated based on the positioning accuracy, the average positioning error, and the entropy value of the predicted probability distribution to obtain the model performance evaluation results.

6. A method for locating leakage points in a reservoir basin, characterized in that, The method includes: Acquire the current reservoir water level data, seepage pressure change data and response time delay data of each piezometer in the reservoir basin seepage prevention area; The current reservoir water level data, the seepage pressure change data of each piezometer, and the response time delay data are input into a pre-constructed leakage probability prediction model for prediction, so as to obtain the leakage probability prediction results of different areas of the reservoir basin seepage prevention zone.

7. The method according to claim 6, characterized in that, The method further includes: Interpolation calculations are performed on the leakage probability prediction results of the different regions to obtain a continuous probability field; A probability heatmap of leakage locations is generated based on the continuous probability field.

8. A training device for a leakage probability prediction model, characterized in that, The device includes: The sample acquisition module is used to simulate the seepage pressure change data and response time delay data of each piezometer when leakage occurs in different areas of the reservoir basin seepage prevention area using the finite element seepage model of the reservoir basin seepage prevention area, and obtain multiple sets of sample data. The sample data includes reservoir water level data, seepage pressure change data of each piezometer, response time delay data, and leakage probability of different areas. Different piezometers are set at different locations in the reservoir basin seepage prevention area. The model training module is used to train the leakage probability prediction model to be trained by taking the reservoir water level data, the seepage pressure change data and the response time delay data of each piezometer in the sample data as input features, and the leakage probability of different areas in the sample data as output labels. The leakage probability prediction model is used to determine the leakage probability of different areas based on the seepage pressure change data, the response time delay data and the reservoir water level data of each piezometer in the reservoir seepage prevention area.

9. A training device for a leakage probability prediction model, characterized in that, The device includes: The data acquisition module is used to acquire the current reservoir water level data, the seepage pressure change data of each piezometer in the reservoir basin seepage prevention area, and the response time delay data. The result prediction module is used to input the current reservoir water level data, the seepage pressure change data of each piezometer, and the response time delay data into a pre-constructed leakage probability prediction model for prediction, so as to obtain the leakage probability prediction results of different areas of the reservoir basin seepage prevention area.

10. An electronic device, characterized in that, include: The system includes a memory and a processor, which are interconnected. The memory stores computer instructions, and the processor executes the computer instructions to perform a leakage probability prediction model training method according to any one of claims 1 to 5, or to perform a reservoir leakage point location method according to claim 6 or 7.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to execute a leakage probability prediction model training method according to any one of claims 1 to 5, or to execute a reservoir leakage point location method according to claim 6 or 7.

12. A computer program product, characterized in that, The computer instructions include computer commands for causing the computer to execute a leakage probability prediction model training method according to any one of claims 1 to 5, or to execute a reservoir leakage point location method according to claim 6 or 7.

Images