A method, device and system for detecting early leakage of a reservoir dam body

Abstract

The application relates to the technical field of dam body leakage detection, in particular to a reservoir dam body early leakage detection method, equipment and system, the method comprising the following steps: obtaining synchronous aligned reservoir water level and pore water pressure, extracting pressure response time lag and pressure response amplitude ratio under a dominant frequency by using frequency domain analysis, constructing a hydraulic retardation modulus representing medium properties, calculating the shortest seepage path length from each measuring point to the reservoir water boundary according to a dam geometric structure model and constructing a spatial adjacency distance matrix, performing linear trend fitting on the pressure response time lag and the seepage path length to generate data reliability weights, calculating the distribution recovery cost required for transforming the measured retardation distribution vector into a reference path distribution vector, and performing trend monitoring on the distribution recovery cost at continuous multiple time points to determine abnormalities. The application aims to extract the medium property change characteristics caused by early small reservoir dam body leakage, and realize spatial positioning of the leakage source position.

Description

Technical Field

[0001] This application relates to the field of dam leakage detection technology, specifically to a method, equipment, and system for detecting early leakage in reservoir dams. Background Technology

[0002] During operation, pumped storage power stations and peak-shaving reservoirs experience frequent and significant periodic rises and falls in water level. These high-frequency water level fluctuations induce a significant nonlinear seepage response within the dam body, resulting in inconsistent pore water pressure values ​​at the same water level during the rising and falling periods, forming a complex hysteresis loop effect.

[0003] Existing monitoring methods are mostly based on static regression models of "water level-seepage pressure." This method struggles to eliminate nonlinear interference caused by dynamic lag and is prone to false alarms during rapid water level fluctuations. Furthermore, for newly constructed dams or monitoring sections with newly installed sensors, the lack of sufficient long-term historical data makes it impossible to construct high-precision statistical models, leading to a cold start challenge for the monitoring system. More critically, the seepage pressure changes caused by early, minute leaks or initial piping are extremely weak and often masked by large water level fluctuations. Traditional single-point threshold monitoring methods struggle to extract these subtle structural damage features from strong background noise and are also inadequate for spatially locating the leak source in the absence of historical benchmarks. Summary of the Invention

[0004] In view of the above, it is necessary to provide a method, equipment and system for detecting early leakage of reservoir dams to solve the above problems.

[0005] According to one aspect of this application, a method for detecting early leakage in a reservoir dam body is provided, the method comprising: The reservoir water level and pore water pressure at each measuring point on the reservoir dam body at each time within a preset sliding window are collected. The amplitude attenuation and time lag of reservoir water level and pore water pressure at the dominant frequency are extracted by using discrete Fourier transform. The pressure response amplitude ratio and pressure response time lag of each measuring point are obtained, and the local hydraulic retardation modulus characterizing the water-blocking capacity of the medium is calculated. Based on the three-dimensional geometric model of the reservoir dam and the equation of the reservoir water boundary surface, the seepage path length of each measuring point is obtained. Combined with the pressure response time lag, a linear model for measuring the overall diffusion performance of the dam body is obtained. The data reliability weight of each measuring point is calculated using the residual of each measuring point on the linear model. Based on the local hydraulic resistance modulus and data reliability weights, the measured resistance distribution is normalized, the seepage path length is normalized to the reference path distribution, and the linear programming problem that minimizes the cumulative transportation cost is solved. The minimum cumulative transportation cost is defined as the distribution restoration cost. Based on a preset historical reference time window, a dynamic threshold is established to monitor the distribution of restoration costs and locate the leakage source.

[0006] Preferably, the specific steps for obtaining the pressure response amplitude ratio and pressure response time lag at each measuring point are as follows: The complex spectral component of the pore water pressure at the dominant frequency within the preset sliding window at each measuring point is denoted as the first spectral component; the complex spectral component of the reservoir water level sequence at the dominant frequency within the preset sliding window is denoted as the second spectral component. Obtain the variance of the reservoir water level sequence within a preset sliding window. If the variance is greater than or equal to a preset water level dead zone threshold, use the ratio of the magnitude of the first spectral component to the magnitude of the second spectral component as the pressure response amplitude ratio at each measuring point; otherwise, do not analyze the current preset sliding window. Cross-correlation is performed on the reservoir water level sequence within the preset sliding window and the pressure response sequence at each measuring point. The offset corresponding to the maximum value of the cross-correlation function sequence is used as the pressure response time lag at each measuring point.

[0007] Preferably, the calculation of the local hydraulic resistance modulus characterizing the water-blocking capacity of the medium specifically includes: For each measuring point, the negative correlation mapping result of the product of the pressure response amplitude ratio and the preset value is positively fused with the pressure response time lag to obtain the local hydraulic resistance modulus.

[0008] Preferably, the seepage path length of each measuring point is specifically the shortest Euclidean distance from the spatial coordinates of each measuring point to the reservoir water boundary.

[0009] Preferably, the linear model obtained for measuring the overall diffusion performance of the dam body is specifically as follows: A linear regression model is obtained by using the seepage path length of all measuring points at the current moment as the independent variable and the pressure response time lag as the dependent variable.

[0010] Preferably, the process of calculating the data reliability weight of each measuring point using the residuals of each measuring point on the linear model is as follows: Based on the fitted linear model, the propagation residual at each measurement point is calculated, and the propagation residual at the nth measurement point is denoted as... The median absolute deviation of the propagation residuals at all measurement points is denoted as . ; Let the data reliability weight of the nth measurement point be denoted as... The specific formula is as follows: In the formula, exp() represents an exponential function with the natural constant as the base. It represents the smallest positive number that is preset.

[0011] Preferably, the linear programming problem of minimizing cumulative transportation cost is defined as the distribution restoration cost, specifically: Let the measured retardation distribution at the nth measuring point be denoted as... Let the reference path distribution vector of the m-th measuring point be denoted as Based on the distance between the measured position of the nth measuring point and the reference position of the mth measuring point, obtain the elements of the spatial adjacency distance matrix, denoted as... ; Define the optimization objective function as the distributed restoration cost. : In the formula, denoted by ; min represents the minimum value function; Q represents the virtual mass transfer matrix; N represents the total number of measurement points; This represents the element in the nth row and mth column of the virtual mass transfer matrix; n represents the measured position index of the measurement point; m represents the reference position index of the measurement point. The optimization model needs to satisfy the following mass conservation constraints: for any ,have For any ,have For any ,have ; The distributed restoration cost is obtained by using the network simplex method. And the virtual mass transfer matrix.

[0012] Preferably, the specific process of establishing a dynamic threshold monitoring distribution restoration cost based on a preset historical reference time window and locating the leakage source is as follows: The dynamic alarm threshold is obtained by using the 3σ principle based on the distribution restoration cost of all times within a preset historical reference time window; if the distribution restoration cost of the current time exceeds the corresponding dynamic alarm threshold for a preset number of consecutive analysis cycles, it is determined that there is a structural leakage anomaly inside the dam body. Using the virtual mass transfer matrix and the spatial adjacency distance matrix, the stagnation deficit index at each reference location is obtained; the reference location corresponding to the maximum value of the stagnation deficit index is taken as the source center of leakage.

[0013] According to another aspect of this application, a detection system for early leakage of a reservoir dam is provided, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of the method as described in any of the preceding claims.

[0014] According to another aspect of this application, a detection device for early leakage of a reservoir dam is provided, the device storing a computer program that, when executed by a processor, implements the method described in any of the preceding claims.

[0015] This application has at least the following beneficial effects: This application first eliminates nonlinear water level interference by decoupling through frequency domain features, and constructs a hydraulic retardation modulus by using the ratio of pressure response time lag to amplitude ratio. This mechanism isolates the nonlinear influence of reservoir water level fluctuation amplitude on seepage pressure response, allowing changes in media properties caused by early minor leakage to be independently extracted from strong background water level noise.

[0016] Furthermore, single-frame monitoring without historical samples is achieved using optimal transmission topology analysis. By constructing a constant reference path distribution as a theoretical benchmark, leakage detection is transformed into a problem of verifying the consistency between the measured distribution and the theoretical benchmark topology. Since this benchmark does not change over time and does not depend on historical samples, this application only requires a single-frame spatial snapshot to identify structural anomalies, solving the cold-start monitoring problem in new projects or new sensor deployment phases. In addition, by analyzing the virtual mass transfer matrix to calculate the hindrance deficit index, the complex parameter inversion problem is transformed into a spatial transmission cost analysis problem, directly locating the leakage source without prior knowledge of geological parameters. Attached Figure Description

[0017] Figure 1 A flowchart illustrating the steps of a method for detecting early leakage in a reservoir dam, as provided in this application. Detailed Implementation

[0018] Please see Figure 1 The diagram illustrates a flowchart of a method for detecting early leakage in a reservoir dam, according to an embodiment of this application. The method includes the following steps: Step 1: Collect the reservoir water level and pore water pressure at each measuring point on the reservoir dam body at each time within the preset sliding window.

[0019] Since reservoir water level changes are a continuous dynamic process, and sensors distributed at different locations on the dam may have slight deviations in their sampling clocks, directly using instantaneous readings at a single moment cannot capture the hysteresis effect of pore water pressure relative to reservoir water level. Therefore, this application establishes a unified time reference and slices the continuous time stream into independent analysis units.

[0020] Firstly, by deploying within the dam body A pore water pressure sensor and a water level gauge in the reservoir area are used at a fixed sampling frequency. (In this embodiment, data is collected every 5 minutes.) Raw data is continuously collected. Next, a time span is defined as... The sliding time window (72 hours in this embodiment, to cover at least one complete water level rise and fall cycle).

[0021] At each discrete analysis time k, perform the following operation: extract the current time. The previous time length was The historical data segment is used; for the water level data and pressure data of each sensor within this segment, a linear interpolation algorithm is used to map all data points to a unified timestamp, eliminating the time deviation caused by hardware sampling; an analysis dataset for the current moment is generated, which contains a standard water level fluctuation sequence. as well as The corresponding pore pressure response sequence ,in, For sensor indexing, .

[0022] All subsequent calculations process only the data within the current analysis frame, without relying on future data input, thus ensuring real-time monitoring.

[0023] Step 2: Use Discrete Fourier Transform to extract the amplitude attenuation and time lag of reservoir water level and pore water pressure at the dominant frequency, obtain the pressure response amplitude ratio and pressure response time lag at each measuring point, and calculate the local hydraulic retardation modulus characterizing the water-blocking capacity of the medium.

[0024] Since the magnitude of pore water pressure is affected by both the "reservoir water level" and the "medium permeability", in order to eliminate the interference of water level fluctuations and obtain the properties of the medium alone, it is necessary to perform frequency domain analysis using the low-pass filtering characteristics of the dam medium on pressure waves.

[0025] For the data within the current analysis frame, the variance of the water level fluctuation sequence H(t) is calculated. If the variance is less than the preset water level dead zone threshold (0.01 in this embodiment), the current analysis frame is skipped to avoid errors in the subsequent denominator. A value of zero causes computational crashes; otherwise, for the water level fluctuation sequence... Power spectral density analysis was performed to identify the dominant frequencies of energy concentration. This dominant frequency corresponds to the daily regulation cycle of the pumped storage power station.

[0026] Subsequently, the Discrete Fourier Transform (DFT) algorithm was used to extract the dominant frequencies of each sequence. The spectral components at that location. For each sensor index. First, calculate the pressure response amplitude ratio at the corresponding measuring point. This characterizes the degree of attenuation of the pressure wave during its propagation through the medium. The calculation formula is as follows: in: This indicates taking the modulus of a complex number. This represents the pressure response sequence at the nth measuring point. In dominant frequency The complex spectral component at that point is denoted simply as the first spectral component. This indicates that the water level fluctuation sequence H is at the dominant frequency. The complex spectral component at that point is denoted as the second spectral component.

[0027] Next, the pressure response time lag at the nth measuring point is calculated. This index characterizes the time delay characteristic of pressure waves propagating from the infiltration boundary to the measuring point. The calculation process is as follows: The reservoir water level sequence within the current analysis frame Pressure response sequence at the nth measuring point Perform cross-correlation operations to find a sequence of cross-correlation functions. The offset corresponding to the maximum value This is defined as pressure response time hysteresis. This method directly obtains the true physical delay through global energy matching, and is not limited by the signal period.

[0028] After obtaining the frequency domain response characteristics of each measuring point, a comprehensive index needs to be constructed to quantify the medium's ability to impede water flow. In hydraulic mechanisms, a healthy dense medium should exhibit "significant amplitude attenuation (low...)". ")" and "significant time delay (high)" Conversely, if there is a well-connected leakage channel, it will manifest as "small amplitude attenuation (high amplitude)". "and "short time delay (low)" )".

[0029] Based on the above mechanism, define and calculate the first... Hydraulic retardation modulus at each measuring point The specific formula is as follows: In the formula, , The preset water level standard reference value and pressure standard reference value (in this embodiment, the upper limit of the range of the water level gauge and pressure sensor) are used. It is a very small positive regularization constant (in this embodiment, it takes the value of ). ), This represents the negative correlation mapping result between the pressure response amplitude ratio and the product of the preset value, used to prevent numerical instability when the denominator is zero. This modulus A higher value indicates a stronger impediment effect of the medium on pressure waves and better seepage prevention performance; a lower value indicates a more permeable medium and a risk of leakage. It should be noted that the hydraulic impediment modulus is an equivalent physical parameter characterizing the degree of resistance of the medium to the propagation of water pressure disturbances, and its value is equal to the ratio of phase lag time to amplitude transmission efficiency.

[0030] After calculating the hydraulic retardation modulus of all measuring points, they are assembled into the hydraulic retardation modulus field of the dam body at the current moment. This field manifests as a set of feature vectors containing spatial location information: in, For the first The three-dimensional spatial coordinates of each measuring point in the dam coordinate system. This hydraulic resistance modulus field. This will serve as the basis for input data in subsequent steps of topological geometry analysis.

[0031] Step 3: Based on the three-dimensional geometric model of the reservoir dam and the equation of the reservoir water boundary surface, obtain the seepage path length of each measuring point, and combine it with the pressure response time lag to obtain a linear model for measuring the overall diffusion performance of the dam body. Calculate the data reliability weight of each measuring point using the residual of each measuring point on the linear model.

[0032] To address the issue of monitoring systems failing to start due to a lack of historical data, this application utilizes the inherent geometric information of the dam to construct a constant reference baseline. This process is typically performed once during the initial deployment phase, and the generated parameters will serve as static constants for subsequent steps.

[0033] First, read the as-built 3D geometric model of the dam and the equations of the reservoir water boundary surface. For each sensor index , The spatial coordinates of the nth measuring point to the reservoir boundary The shortest Euclidean distance is defined as the seepage path length at the corresponding measuring point. This parameter characterizes the theoretical physical distance required for water to reach the measuring point under the assumption of an ideal homogeneous medium.

[0034] Simultaneously, calculate the spatial adjacency relationships within the sensor network and construct... 3D spatial adjacency distance matrix The first in the matrix Line 1 Column elements Defined as from the measured position To the reference position Euclidean distance: In the formula, This represents the measured position of the nth measuring point. This indicates the reference position of the m-th measuring point; Represents the L2 norm. Spatial adjacency distance matrix. The unit cost of transferring virtual mass between different spatial nodes in subsequent calculations is defined.

[0035] In the early stages of leakage, the pressure wave propagation speed in the leakage channel area is abnormally accelerated, resulting in a lag in the measured pressure response time. Significantly smaller than the length of its seepage path The determined theoretical value. In order to identify such anomalies that violate the laws of physical diffusion, this application performs robust statistical analysis on the data within the current analysis frame.

[0036] Obtain the pressure response time lag at each measuring point at the current moment. With seepage path length ; Use the Random Sample Consensus (RANSAC) algorithm on the point set Linear regression fitting is performed. The specific execution process of this algorithm is as follows: multiple random subsets of data are sampled to fit a linear model. The model is then calculated to determine the degree of fit of the remaining data points to the model. Finally, the model parameter with the most interior points is selected as the propagation reference (slope) of the dam's background medium at the current moment. and intercept This process is based entirely on the spatial statistical characteristics of the current moment and does not rely on historical benchmarks, thus adapting to the overall trend drift caused by changes in water level elevation.

[0037] Based on the fitted background model, the propagation residuals at each measuring point are calculated. .like A significantly positive value indicates a phase lead anomaly at that point; the median absolute deviation of the residuals at all measurement points is calculated. The reliability weights of the data at each measuring point are calculated using a Gaussian symmetric decay function. : In the formula, exp() represents an exponential function with the natural constant as the base. This represents a preset, extremely small positive number, which is set to 0.01 in this embodiment to prevent data from being perfectly fitted. Setting it to 0 will cause a division by zero error in the formula.

[0038] The coefficient of the absolute deviation of the median is selected from the normal distribution. Criterion. For measurement points that conform to the background propagation law, their residuals are close to 0, and the weights... Approaching 1; for leakage points that deviate significantly from the pattern, the residuals increase significantly, and the weights... It is nonlinearly suppressed to approach 0.

[0039] Step 4: Based on the local hydraulic resistance modulus and data reliability weights, normalize the measured resistance distribution, normalize the seepage path length to the reference path distribution, solve the linear programming problem that minimizes the cumulative transportation cost, and define the obtained minimum cumulative transportation cost as the distribution restoration cost.

[0040] In order to mathematically quantify the difference between the "measured state" and the "ideal state", this application transforms the physical parameters into a normalized probability distribution vector.

[0041] First, construct the measured retardation distribution vector. , define the first Measured effective resistance mass at each measuring point Through hydraulic retardation modulus Weighting based on data reliability The product is obtained, that is... Physically, it represents the equivalent retardation time after reliability weighting. When leakage occurs in a certain area, its hydraulic retardation modulus... Decrease and weight The suppression resulted in a significant "collapse" of the effective quality in the region. After normalization, we get The Each component Specifically: Secondly, construct the reference path distribution vector. Utilizing the seepage path length As the first Theoretical geometric mass of each reference node The system for After normalization, we get The Each component Specifically: This distribution represents the relative share of obstruction that each spatial node should bear under ideal conditions of homogeneous medium.

[0042] When leakage occurs in a localized area, it causes the measured resistance distribution vector to... When local collapse occurs, in order to mathematically transform it into a uniform reference path distribution vector... Mass must be "transferred" from other areas with abundant mass to fill the deficit in the subsidence area. The cumulative spatial cost of this "transfer" process directly reflects the degree of topological distortion caused by leakage.

[0043] A discrete optimal transport model is constructed to solve a linear programming problem that minimizes the cumulative transport cost. This application defines the objective function as the distributed restoration cost. : In the formula, The cost of distribution restoration represents the minimum total "spatial work" required to forcibly restore the distorted measured obstruction distribution to an ideal healthy distribution, serving as the core quantitative indicator of the degree of leakage distortion across the entire field; min represents the minimum value function; Q represents the virtual mass transfer matrix; and N represents the total number of measurement points. The element in the nth row and mth column of the virtual mass transfer matrix represents the position of the element at the measured location in order to eliminate distribution differences in the numerical algorithm. The blocked portion is virtually moved to the reference position. The relative hindering mass share.

[0044] This optimization calculation must satisfy the following mass conservation constraints: for any measured end index ,have Ensure from location The total mass removed is equal to the measured weight at that location; for any reference end index ,have Ensure the inflow location The total mass is equal to the theoretical weight at that location; for any ,have .

[0045] Solve the above linear programming problem using the network simplex method and output the distribution restoration cost in scalar form. as well as dimensional virtual mass transfer matrix The cost This will serve as the core indicator for determining whether leakage exists in subsequent steps.

[0046] Step 5: Establish a dynamic threshold monitoring distribution restoration cost based on a preset historical reference time window, and locate the leakage source.

[0047] Using the calculated distribution restoration cost As the sole indicator of the overall structural health, under normal operating conditions, this cost will only exhibit a slow, smooth baseline drift due to the influence of seasonal temperature and water level. However, when structural leakage occurs, the topological consistency of the retardation distribution is disrupted, leading to a significant step or exponential increase in this cost.

[0048] This application employs a sliding window statistical process control method for monitoring, setting a historical reference time window. (In this embodiment, the past 30 days are used to cover the monthly environmental change cycle.) At each analysis point... Read the history within this window For a sequence, calculate its mean. and standard deviation Set dynamic alarm thresholds. ,in, This is the confidence coefficient (in this example, it is set to 3, corresponding to a 99.7% confidence interval). The current time... With threshold Compare. If continuous One analysis cycle (in this embodiment) Exceeding the threshold The system determined that there was structural leakage inside the dam and triggered a Level 1 alarm signal.

[0049] After the alarm is triggered, the output virtual mass transfer matrix is ​​used. and the preset spatial adjacency distance matrix To perform spatial positioning. Its physical meaning lies in: regarding the theoretical position... If leakage occurs in the corresponding actual area, the measured resistive mass of that area will approach zero (mass collapse). This is to satisfy the theoretical distribution. Due to the conservation requirement, the optimal transmission algorithm is forced to use other measured locations at greater distances. Virtual transport quality to this location. Due to spatial distance. The large volume of long-distance transport leads to extremely high transfer costs at this node.

[0050] Calculate each reference position The stagnation deficit index The calculation formula is: The index The location was directly quantified. To maintain its theoretical geometric properties, the total cost dependent on external inputs is required. This involves traversing all reference location indices. Search The index corresponding to the global maximum value This application extracts the pre-stored spatial coordinates corresponding to the index. This location is marked as the center of the leak origin. The output includes the leak alarm status and the coordinates of the leak source. The test results are used by engineers to carry out targeted grouting or seepage prevention treatments.

[0051] It should be noted that the resistance modulus of the leakage area and reliability weight This dual reduction manifests as mass gaps and voids in the measured distribution. To satisfy the geometric conservation requirements of the theoretical distribution, the optimal transport algorithm must inevitably draw virtual mass from the non-leaking zone over a long distance to fill these voids. Therefore, the cumulative transport cost... The highest reference position is the area with the most severe quality deficit and the greatest need for external replenishment, corresponding to the physical core of the actual leakage.

[0052] Based on the same concept as the method embodiments of this application, a detection system for early leakage of a reservoir dam is proposed, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of the method as described in any of the above.

[0053] Based on the same concept as the method embodiments of this application, a detection device for early leakage of reservoir dam body is proposed. The device stores a computer program, which, when executed by a processor, implements the method as described in any of the above.

Claims

1. A method for detecting early leakage in a reservoir dam, characterized in that, The method includes the following steps: The reservoir water level and pore water pressure at each measuring point on the reservoir dam body at each time within a preset sliding window are collected. The amplitude attenuation and time lag of reservoir water level and pore water pressure at the dominant frequency are extracted by using discrete Fourier transform. The pressure response amplitude ratio and pressure response time lag of each measuring point are obtained, and the local hydraulic retardation modulus characterizing the water-blocking capacity of the medium is calculated. Based on the three-dimensional geometric model of the reservoir dam and the equation of the reservoir water boundary surface, the seepage path length of each measuring point is obtained. Combined with the pressure response time lag, a linear model for measuring the overall diffusion performance of the dam is obtained. The data reliability weight of each measuring point is calculated using the residual of each measuring point on the linear model. Based on the local hydraulic resistance modulus and data reliability weights, the measured resistance distribution is normalized, the seepage path length is normalized to the reference path distribution, and the linear programming problem that minimizes the cumulative transportation cost is solved. The minimum cumulative transportation cost is defined as the distribution restoration cost. Based on a preset historical reference time window, a dynamic threshold is established to monitor the distribution of restoration costs and locate the leakage source. The local hydraulic retardation modulus, which characterizes the water-blocking capacity of the medium, is specifically calculated as follows: For each measuring point, the negative correlation mapping result of the product of the pressure response amplitude ratio and the preset value is positively fused with the pressure response time lag to obtain the local hydraulic resistance modulus.

2. The method for detecting early leakage of a reservoir dam as described in claim 1, characterized in that, The specific steps for obtaining the pressure response amplitude ratio and pressure response time lag at each measuring point are as follows: The complex spectral component of the pore water pressure at the dominant frequency within the preset sliding window at each measuring point is denoted as the first spectral component; the complex spectral component of the reservoir water level sequence at the dominant frequency within the preset sliding window is denoted as the second spectral component. Obtain the variance of the reservoir water level sequence within a preset sliding window. If the variance is greater than or equal to a preset water level dead zone threshold, use the ratio of the magnitude of the first spectral component to the magnitude of the second spectral component as the pressure response amplitude ratio at each measuring point; otherwise, do not analyze the current preset sliding window. Cross-correlation is performed on the reservoir water level sequence within the preset sliding window and the pressure response sequence at each measuring point. The offset corresponding to the maximum value of the cross-correlation function sequence is used as the pressure response time lag at each measuring point.

3. The method for detecting early leakage of a reservoir dam as described in claim 1, characterized in that, The seepage path length of each measuring point is specifically the shortest Euclidean distance from the spatial coordinates of each measuring point to the reservoir water boundary.

4. The method for detecting early leakage of a reservoir dam as described in claim 1, characterized in that, The obtained linear model for measuring the overall diffusion performance of the dam body is as follows: A linear regression model is obtained by using the seepage path length of all measuring points at the current moment as the independent variable and the pressure response time lag as the dependent variable.

5. The method for detecting early leakage of a reservoir dam as described in claim 1, characterized in that, The process of calculating the data reliability weight of each measurement point using the residuals of each measurement point on the linear model is as follows: Based on the fitted linear model, the propagation residual at each measurement point is calculated, and the propagation residual at the nth measurement point is denoted as... The median absolute deviation of the propagation residuals at all measurement points is denoted as . ; Let the data reliability weight of the nth measurement point be denoted as... The specific formula is as follows: In the formula, exp() represents an exponential function with the natural constant as the base. It represents the smallest positive number that is preset.

6. The method for detecting early leakage of a reservoir dam as described in claim 1, characterized in that, The linear programming problem of minimizing the cumulative transportation cost is defined as the distribution restoration cost, specifically: Let the measured retardation distribution at the nth measuring point be denoted as... Let the reference path distribution vector of the m-th measuring point be denoted as Based on the distance between the measured position of the nth measuring point and the reference position of the mth measuring point, obtain the elements of the spatial adjacency distance matrix, denoted as... ; Define the optimization objective function as the distributed restoration cost. : In the formula, denoted by ; min represents the minimum value function; Q represents the virtual mass transfer matrix; N represents the total number of measurement points; This represents the element in the nth row and mth column of the virtual mass transfer matrix; n represents the measured position index of the measurement point; m represents the reference position index of the measurement point. The optimization model needs to satisfy the following mass conservation constraints: for any ,have For any ,have For any ,have ; The distributed restoration cost is obtained by solving the network simplex method. And the virtual mass transfer matrix.

7. The method for detecting early leakage of a reservoir dam as described in claim 6, characterized in that, The specific process of establishing a dynamic threshold monitoring distribution restoration cost based on a preset historical reference time window and locating the leakage source is as follows: The dynamic alarm threshold is obtained by using the 3σ principle based on the distribution restoration cost of all times within a preset historical reference time window; if the distribution restoration cost of the current time exceeds the corresponding dynamic alarm threshold for a preset number of consecutive analysis cycles, it is determined that there is a structural leakage anomaly inside the dam body. Using the virtual mass transfer matrix and the spatial adjacency distance matrix, the stagnation deficit index at each reference location is obtained; the reference location corresponding to the maximum value of the stagnation deficit index is taken as the source center of leakage.

8. A detection system for early leakage of a reservoir dam, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1-7.

9. A detection device for early leakage of a reservoir dam, the device storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1-7.

Images