An open-air seepage flow monitoring system behind a dam and a measured value attribution analysis method

Abstract

The present invention discloses a post-dam open seepage flow monitoring system and a measured value attribution analysis method: determining the seepage flow measured value influence area and influencing factors based on the measuring point position, topography and precipitation characteristics; determining the corresponding monitoring items, measuring points and instruments based on the spatio-temporal distribution of sensitive factors within the influence area; dividing the measured data sequence of seepage flow into three types, namely only affected by reservoir water seepage, significantly affected by rainfall, and uncertain, according to the determined lag time by means of a corresponding mathematical model; respectively constructing rainfall-reservoir water coupled seepage and reservoir water single seepage data assimilation models according to the measured seepage flow data during the period significantly affected by rainfall and the period only affected by reservoir water seepage; respectively establishing lightweight temporal convolutional models for seepage flow affected by rainfall and seepage flow only affected by reservoir water seepage according to the assimilated samples, and separating the rainfall influence component and the reservoir water seepage component according to the difference between the two. It realizes the accurate separation of seepage components under rainfall conditions and provides support for the seepage safety analysis of the dam.

Description

Technical Field

[0001] The present invention relates to a dam-back open seepage flow monitoring system and a measured value attribution analysis method, belonging to the technical fields of dam safety monitoring and data analysis. Background Art

[0002] Seepage flow monitoring is an essential monitoring item for the safety monitoring of large and medium-sized reservoir dams, and is also a primary monitoring item emphasized for the safety monitoring of small reservoir dams and dikes. Its measured value is of great significance for analyzing the anti-sliding stability of dams, especially the seepage safety of earth-rock dams. However, it is found in actual projects that almost all the dam-back leakage amounts, especially the total seepage flow monitoring facilities (such as the weir behind the dam), are set outdoors, thus being easily affected by rainfall factors. Long-term rainfall or heavy rainstorms are adverse working conditions for dams, which are likely to cause dam failure events such as uneven settlement deformation, dam surface cracking, seepage failure, and even landslides. Therefore, under adverse working conditions, it is more desirable that the measured seepage flow monitoring data can reflect the internal seepage condition of the dam. However, due to the influence of rainfall, problems such as the difficulty in separating the main and guest waters in the measured leakage amount, the uncertainty of the amount of precipitation component in the measured data, and the difficulty in judging the seepage safety state exist. Summary of the Invention

[0003] The purpose of the present invention is to overcome the deficiencies in the prior art and provide a dam-back open seepage flow monitoring system and a measured value attribution analysis method to solve the problem of separating and extracting the reservoir water seepage component and the rainfall influence component in the total seepage flow under rainfall or non-rainfall conditions.

[0004] To achieve the above purpose, the present invention is implemented by the following technical solutions:

[0005] In the first aspect, the present invention provides a measured value attribution analysis method for a dam-back open seepage flow monitoring system, including:

[0006] Determining the influence area and influencing factors of the seepage flow measured value based on the measuring point position, topography, and precipitation characteristics;

[0007] Based on the spatial distribution of various sensitive factors affecting the seepage flow within the influence area, determining the corresponding monitoring items and arranging the measuring points by using the representative principle;

[0008] Using the time process numerical simulation method to determine the lag time of the reservoir water seepage and rainfall influence according to the mathematical model corresponding to the non-reservoir water seepage flow in the monitoring facility caused by the reservoir water seepage and precipitation, and dividing the measured data sequence of the seepage flow into three types: the period only affected by the reservoir water seepage, the period significantly affected by rainfall, and the uncertain period;

[0009] Constructing rainfall influence and reservoir water seepage data assimilation models respectively according to the measured seepage flow data in the period significantly affected by rainfall and the period not affected by rainfall;

[0010] Based on the assimilation results, new samples are obtained to establish lightweight temporal convolutional models that are only affected by reservoir water seepage and affected by the coupling of rainfall and reservoir water seepage, with the measured seepage flow rate as the output. The rainfall influence component and the reservoir water seepage component are separated according to the difference between the two.

[0011] Optionally, the rainfall characteristics are determined based on the ground watershed in combination with the hydrogeological conditions.

[0012] Optionally, the determination of the influencing factors affecting the seepage flow rate within the influence area is carried out by using a distributed hydrological model in combination with factor analysis of the shallow water equation under the combined action of rainfall and seepage, combined with the variation range of the above factors, so as to determine the rainfall-reservoir water seepage influencing factors that have an impact on the measured values of the seepage flow rate monitoring facilities and change with time, and determine the monitoring items, their spatially representative measurement points, and the time sampling rate according to the spatial distribution and time variation of these changing factors.

[0013] Optionally, the determination of the monitoring items and measurement points based on the spatio-temporal distribution of sensitive factors in the influence area includes:

[0014] Extract the changing factors of the rainfall influence factors affecting the entry into the seepage flow rate monitoring facilities from the influencing factors, including rainfall amount, slope, roughness, crack distribution, vegetation cover, soil temperature, air temperature, and soil moisture content;

[0015] Improve or supplement on-site monitoring items such as rainfall amount, slope, roughness, cracks, vegetation height, soil temperature, air temperature, and soil moisture content, and use the corresponding monitoring instruments and their data acquisition and control devices to form a seepage flow rate monitoring system; the monitoring instruments include a combination of fixed-installed buried sensors, inspection robot dogs, and unmanned aerial vehicle remote sensing. Among them, the fixed-installed buried sensors include fixed-position potential image sensors, tipping bucket rain gauges, radar rain gauges, disdrometers, TDRs, and soil moisture sensors; the inspection robot dogs and unmanned aerial vehicles carry lidar, hyperspectral cameras, and infrared cameras, etc., to monitor the slope, roughness, vegetation, etc.

[0016] If the influence area is relatively large, multiple measurement points should be set for the same monitoring item according to whether the distribution of each element within the influence area is uniform and the representativeness of the measurement points, and the sampling frequency of each measurement point should be determined according to the time variation of each monitoring element based on the Shannon sampling theorem or the compressive sensing theory.

[0017] Optionally, the determination of the lag time according to the corresponding mathematical model includes:

[0018] Establish an unsteady seepage model under the action of reservoir water, select a three-dimensional spatial range according to the seepage influence area, and calculate the lag influence time of the reservoir seepage to the seepage flow rate measurement point monitoring facility under the conditions of extreme variable amplitude of reservoir water and different elevation change combinations;

[0019] A mathematical model for surface runoff generation and subsurface seepage runoff under rainfall conditions is established. The model consists of three parts: the surface model, the subsurface model, and the infiltration model. For complex regions, the surface model uses direct numerical simulation based on the Navier-Stokes equations, while for general regions, it uses the two-dimensional shallow water equations. The subsurface part is described by the Richards equation. The infiltration equation is determined based on rainfall type, vegetation and soil type, soil saturation, and rainfall intensity factors, and numerical calculations are used to obtain the lag time of rainfall runoff generation and seepage to the seepage flow measurement points in the affected area.

[0020] Optionally, the division of the historical measured data series of seepage flow into three types: the period only affected by reservoir seepage, the period significantly affected by rainfall, and the uncertain period includes:

[0021] Based on the measured rainfall process in the affected area; the above model is used to calculate the influence duration of rainfall on the measured seepage flow value according to the relevant on-site parameters. Considering the influence of model calculation error factors, the sum of 1 / 3 of the duration at the junction of the two periods affected by rainfall and not affected by rainfall at all is defined as the uncertain duration of rainfall influence, so as to divide the measured seepage flow period into three types: the period only affected by reservoir seepage, the period significantly affected by precipitation, and the uncertain period.

[0022] Optionally, the establishment of a lightweight temporal convolutional model affected by rainfall based on the samples of the assimilation results includes:

[0023] Perform four-dimensional assimilation on the mathematical model of surface runoff generation and subsurface seepage runoff under rainfall conditions and the measured data of the measured seepage flow that is significantly affected by precipitation during the same period;

[0024] Use the assimilated model to generate the total seepage flow, and at the same time divide the sample into a training sample and a test sample;

[0025] Combine the measured data of the influencing factors affecting the measured seepage flow value and the above samples to establish a lightweight time series convolutional neural network L-TCN model. The influencing factors and their changes are used as the input of the model, and the total seepage flow is used as the output of the L-TCN;

[0026] After model training and testing, select a qualified model as the separation model of the total seepage flow.

[0027] Optionally, the establishment of a lightweight temporal convolutional model only affected by reservoir seepage based on the samples of the assimilation results includes:

[0028] Perform four-dimensional assimilation on the unsaturated-saturated dynamic unstable seepage equation of reservoir water only affected by reservoir seepage and the measured data of the measured seepage flow during the same period;

[0029] Use the assimilated model to predict and generate a reservoir water seepage flow sample, and divide the sample into a training sample and a test sample;

[0030] Factors affecting the seepage flow rate of reservoir water, such as water level, air temperature, time, etc., are used as the input of the lightweight time series convolutional neural network L-TCN model, and the seepage flow rate of reservoir water is used as the output of the L-TCN model. After model training and verification, a qualified model is selected as the separation model of the seepage flow rate of reservoir water.

[0031] Optionally, the separating the rainfall influence component and the reservoir water seepage component according to the difference between the two includes:

[0032] Predicting the total seepage flow rate and the reservoir water seepage component respectively through the lightweight time convolutional model affected by rainfall and only affected by reservoir water seepage according to the measured input, and obtaining the rainfall influence component by subtracting the reservoir water seepage component from the total seepage flow rate, so as to realize the separation of the total measured seepage flow rate under rainfall conditions and non-precipitation conditions, that is, attribution analysis.

[0033] In a second aspect, the present invention provides a post-dam open-air seepage flow rate monitoring system adapted to the above-mentioned measured value attribution analysis method, which is characterized in that it includes seepage flow rate monitoring facilities, and necessary monitoring items and monitoring facilities are added according to the composition and influencing factors of the measured seepage flow rate data, for realizing the all-element, full-range, all-weather, and all-scenario information perception equipment of the post-dam seepage flow rate and its main influencing factors.

[0034] Compared with the prior art, the beneficial effects achieved by the present invention:

[0035] A post-dam open-air seepage flow rate monitoring system and a measured value attribution analysis method provided by the present invention add the monitoring of relevant factors for seepage flow rate attribution analysis, and construct a new monitoring system according to the changing factors affecting seepage flow and runoff generation and concentration. Based on the monitoring results, the rainfall influence, seepage lag time and influence duration are obtained based on the corresponding mathematical model, and the measured data are classified and processed based on the lag time and duration. Based on the monitoring data of the period affected by rainfall and the period only affected by reservoir water, the separation models of the rainfall seepage component and the reservoir water seepage component are respectively established by combining the four-dimensional assimilation and the lightweight convolutional neural network model, and the rainfall influence component and the reservoir water seepage component are separated according to the difference between the above models; the present invention can solve the problem of separating the measured value components of the post-dam open-air measured seepage flow rate under rainfall influence, and provide a basis for the dam stability analysis and seepage safety analysis. Description of the Drawings

[0036] Figure 1 It is a flow chart of a measured value attribution analysis method for a post-dam open-air seepage flow rate monitoring system provided by an embodiment of the present invention. Detailed Embodiment

[0037] The present invention will be further described below with reference to the drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention, and cannot be used to limit the protection scope of the present invention.

[0038] Example 1:

[0039] As Figure 1 shown, the present invention provides a method for attributing measured values of a post-dam open-air seepage flow monitoring system, including the following steps:

[0040] 1. Determine the influence area and influencing factors of the seepage flow measured value based on the measuring point location, topography and geomorphology, and precipitation characteristics;

[0041] The rainfall characteristics are determined according to the ground watershed combined with the hydrogeological conditions.

[0042] The determination of the influencing factors affecting the seepage flow within the influence area is carried out by using a distributed hydrological model or a shallow groundwater equation on the ground, a rainfall infiltration model, and a factor analysis of the groundwater flow model in combination with the change speed and amplitude of the corresponding influencing factors. The rainfall influencing factors, spatial distribution, and temporal variation characteristics affecting the seepage flow monitoring facility are determined from the above models, thereby establishing a corresponding monitoring system.

[0043] 2. Determine the monitoring items and measuring point layout based on the spatio-temporal distribution of the sensitive factors in the influence area;

[0044] Extract the influencing factors affecting the rainfall entering the seepage flow monitoring facility from the influencing factors, including rainfall amount, slope, roughness, ground cracks, vegetation cover, ground temperature, air temperature, and soil moisture content;

[0045] The monitoring system includes monitoring items such as rainfall, slope, roughness, vegetation height monitoring, ground cracks, soil moisture content, and air temperature monitoring, and constructs a seepage flow monitoring system; the monitoring system adopts a combination of fixed installation and buried sensors, patrol robot dogs, and unmanned aerial vehicle remote sensing. The fixed installation and buried sensors include fixed-position potential image sensors, tipping bucket rain gauges, radar rain gauges, disdrometers, TDRs, and soil moisture sensors; the patrol robot dogs and unmanned aerial vehicles carry lidar, hyperspectral cameras, and infrared cameras, etc., to monitor the slope, roughness, vegetation, etc.

[0046] If the influence area is relatively large, multiple measuring points should be set for the same monitoring item according to whether the distribution of each element in the influence area is uniform and the representativeness of the measuring points, and the sampling frequency of each measuring point is determined according to the time variation of each monitoring element according to the Shannon sampling theorem or the compressed sensing theory.

[0047] 3. Divide the historical measured data sequence of the seepage flow into three types: the period only affected by the reservoir water seepage, the period significantly affected by rainfall, and the uncertain period according to the lag time determined by the corresponding mathematical model;

[0048] 3.1. Determining the lag time according to the corresponding mathematical model includes:

[0049] Establish an unsteady seepage model under the action of reservoir water, including:

[0050] Based on the water balance equation and seepage law (Darcy's law is adopted when the Reynolds number is between 1 and 10. For cases where Darcy's seepage conditions are not satisfied, the corresponding velocity-pressure gradient function relationship is used to replace Darcy's law. For example, for rock seepage, according to the starting pressure gradient formula:

[0051]

[0052] In the formula, λ is the starting pressure gradient, k is the permeability coefficient, p is the pressure, and μ is the viscosity coefficient; for unsaturated soil seepage, the Books-Corey, Gardner, or Van Genuchten-Muale model is adopted; for high-speed seepage, the Forchheimer formula is used.) etc., establish an unsaturated-saturated dynamic unsteady seepage model of reservoir water under the action of reservoir water;

[0053] Select a three-dimensional spatial range according to the seepage influence area, and calculate the seepage lag time of the reservoir under extreme amplitude and elevation conditions of reservoir water;

[0054] Establish a mathematical model of surface runoff generation and underground seepage runoff under rainfall conditions. For the surface model in complex areas, an approximate form based on the Navier-Stokes equation is adopted, and for general areas, the two-dimensional shallow water equation is used; the underground part is described by the Richard equation; the infiltration equation is determined according to rainfall patterns, vegetation and soil types, soil saturation, and rainfall intensity factors, and the influence lag time and duration of rainfall infiltration volume are calculated.

[0055] Neglect the wind stress and Coriolis force terms and simplify the two-dimensional shallow water equation of overland flow to:

[0056]

[0057] In the formula: t is time, x and y are spatial coordinates; G and H are flux vectors in the x and y directions respectively, and S is the source term.

[0058]

[0059] In the formula: h is the water depth, u and v are the average velocity components in the x and y directions respectively, r is the rainfall intensity, f is the infiltration intensity, g is the acceleration due to gravity, S ox and S oy represent the bottom slopes of the water bottom in the x and y directions, and S fx and S fy are the frictional bottom slopes in the x and y directions respectively.

[0060]

[0061]

[0062] In the formula: z is the bottom elevation, and n is the Manning coefficient.

[0063] 3.2. Divide the measured data series of seepage flow into three types: the period only affected by reservoir water seepage, the period significantly affected by rainfall, and the uncertain period, including:

[0064] Calculate the lag time and its process of the seepage flow affected by rainfall and the seepage flow only affected by reservoir water seepage respectively according to the rainfall process and the process of reservoir water rise and fall; considering the influence of model calculation error factors, define the sum of 1 / 3 of the duration at the junction of the two periods as the uncertain duration of rainfall influence, so as to divide the measured seepage flow period into three types: the period only affected by reservoir water seepage, the period significantly affected by precipitation, and the uncertain period.

[0065] 4. Obtain the measured seepage flow data of the period significantly affected by rainfall and the period only affected by seepage respectively, and construct rainfall-seepage and reservoir-water-seepage data assimilation models;

[0066] 5. Establish lightweight temporal convolutional models for seepage flow affected by rainfall and seepage flow only affected by reservoir water according to the samples obtained from the assimilation results, and separate the rainfall-seepage component and the reservoir-water influence component according to the difference between the two.

[0067] 5.1. Establish a lightweight temporal convolutional model for seepage flow affected by rainfall according to the samples obtained from the assimilation results, including:

[0068] Perform four-dimensional assimilation on the coupled mathematical model of overland runoff and subsurface seepage runoff under rainfall conditions and the measured seepage flow data significantly affected by precipitation during the same period;

[0069] Divide the assimilated samples into training samples and test samples;

[0070] Use the measured data of the factors affecting the change of seepage flow as the input of the lightweight time series convolutional neural network L-TCN model, and use the total seepage flow as the output of the L-TCN;

[0071] After model training and testing, select a qualified model as the separation model of the total seepage flow.

[0072] 5.2. Establish a lightweight temporal convolutional model for seepage flow only affected by reservoir water according to the samples obtained from the assimilation results, including:

[0073] Perform four-dimensional assimilation on the unsaturated-saturated dynamic unstable seepage equation of reservoir water only affected by reservoir water seepage during the period and the measured data during the same period;

[0074] Generate reservoir water seepage flow samples from the assimilated model, and divide the samples into training samples and test samples;

[0075] Factors affecting the seepage flow rate of reservoir water, such as water level, air temperature, time, etc., are used as the input of the lightweight time series convolutional neural network L-TCN model, and the seepage flow rate of reservoir water is used as the output of the L-TCN model. After model training and verification, a qualified model is selected as the separation model of the seepage flow rate of reservoir water.

[0076] The present invention adopts a weak-constraint 4D variational assimilation mode, and the mode state equation with weak-constraint of the error forcing control variable is:

[0077] X i =M i-1,i (X i-1 )+η i

[0078] Among them, X i represents the mode state variable vector at the i-th moment, M i-1,i represents the non-linear operator for integrating the mode state variable from time t i-1 to time t i η i represents the mode error at time t i , which is a vector with the same dimension as the state variable X i .

[0079] At this time, the weak-constraint 4dvar objective function of the mode error control variable is:

[0080]

[0081] Among them, <·,·> and (·,·) respectively represent the inner products in the spaces of R n and R m , N represents the length of the assimilation interval, H i represents the non-linear observation operator at time t i , X b is the background field of the state variable at the initial time, Y i represents the observation data at time t i , R i and Q i respectively represent the observation error and the mode error covariance matrix at time t i , and B represents the background field error covariance matrix. It can be seen that, compared with the traditional 4dvar cost function, the weak-constraint 4D variational assimilation objective function of the mode error control variable adds a mode error term

[0082] 5.3. Separating the rainfall seepage component and the reservoir water influence seepage component according to the difference between the two includes:

[0083] According to the measured input, the total seepage flow rate, the reservoir water seepage component are predicted respectively through a lightweight temporal convolutional model affected by rainfall and not affected by rainfall, and the rainfall-affected seepage component obtained by subtracting the reservoir water seepage component from the total seepage flow rate, so as to realize the separation of the total measured seepage flow rate under rainfall conditions and non-precipitation conditions, that is, attribution analysis.

[0084] Embodiment 2:

[0085] The embodiment of the present invention provides a post-dam open-air seepage flow rate monitoring system adapted to the measured value attribution analysis method of Embodiment 1, including seepage flow rate monitoring facilities, adding necessary monitoring items and monitoring facilities according to the composition and influencing factors of the measured seepage flow rate data, for realizing the all-element, full-range, all-weather, and full-scenario information perception of the post-dam seepage flow rate and its main influencing factors.

[0086] The present invention first adds necessary monitoring items and monitoring facilities according to the composition and influencing factors of the measured seepage flow rate data, so as to realize the all-element, full-range, all-weather, and full-scenario information perception of the post-dam seepage flow rate and its main influencing factors. In addition to the relevant parameters such as the upstream reservoir water level, downstream reservoir water level, dam body permeability coefficient and zoning that affect the distribution and evolution of the dam body seepage field, the factors affecting the post-dam seepage flow rate also include the relevant factors affecting rainfall runoff generation and concentration, such as the topography, slope, roughness, vegetation, soil type, slope, rainfall amount, soil moisture content, soil temperature and air temperature in the affected area. To realize the attribution analysis of the post-dam seepage flow rate, it is necessary to make necessary improvements and supplements to the changing factors affecting seepage and runoff generation and concentration, so as to realize the improvement of the existing monitoring facilities (generally just a water measuring weir and a weir head monitoring device); the data collected by the monitoring facilities are processed through a runoff-seepage coupling mathematical model and an unsteady seepage model, and separation models for rainfall seepage components and reservoir water seepage components are respectively constructed, so as to realize the separation of rainfall seepage components and reservoir water influence components.

[0087] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0088] This application is described with reference to the flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, as well as the combination of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to the processors of general-purpose computers, special-purpose computers, embedded processors, or other programmable data processing devices to generate a machine, such that the instructions executed by the processors of the computer or other programmable data processing devices produce means for implementing the functions specified in the Figure 1 one or more flows and / or blocks Figure 1 means for implementing the functions specified in one or more blocks or a plurality of blocks.

[0089] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory produce a manufactured article including instruction means that implement the functions specified in the Figure 1 one or more flows and / or blocks Figure 1 means for implementing the functions specified in one or more blocks or a plurality of blocks.

[0090] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in the Figure 1 one or more flows and / or blocks Figure 1 means for implementing the functions specified in one or more blocks or a plurality of blocks.

[0091] The above is only the preferred embodiment of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the technical principle of the present invention, several improvements and modifications can be made, and these improvements and modifications should also be regarded as the protection scope of the present invention.

Claims

1. A method for attributing measured values of an open seepage flow monitoring system behind a dam, characterized in that, Including: Determine the influence area and influencing factors of the measured value of seepage flow based on the measuring point location, topography and rainfall characteristics; Based on the spatial distribution of various sensitive factors affecting seepage flow within the influence area, determine the corresponding monitoring items and arrange measuring points by adopting the representative principle; According to the mathematical model corresponding to the non-reservoir water seepage flow in the monitoring facilities caused by reservoir water seepage and precipitation, use the time process numerical simulation method to determine the lag time affected by reservoir water seepage and rainfall, and divide the measured data sequence of seepage flow into three types: the period only affected by reservoir water seepage, the period significantly affected by rainfall, and the uncertain period; Construct rainfall influence and reservoir water seepage data assimilation models respectively according to the measured seepage flow data in the period significantly affected by rainfall and the period not affected by rainfall; According to the new samples obtained from the assimilation results, establish lightweight temporal convolutional models with the measured seepage flow as the output, which are only affected by reservoir water seepage and affected by the coupling of rainfall and reservoir water seepage respectively. Separate the rainfall influence component and the reservoir water seepage component according to the difference between the two; Among them, establishing a lightweight temporal convolutional model affected by rainfall according to the samples of the assimilation results includes: Perform four-dimensional assimilation on the mathematical models of overland runoff generation and underground seepage runoff under rainfall conditions and the measured seepage flow data significantly affected by precipitation during the same period; Use the assimilated model to generate the total seepage flow, and at the same time divide the sample into a training sample and a test sample; Combine the measured data of the influencing factors affecting the measured value of seepage flow and the above samples to establish a lightweight time series convolutional neural network L-TCN model. The influencing factors and their changes are used as the input of the model, and the total seepage flow is used as the output of the L-TCN; After model training and testing, select a qualified model as the separation model of the total seepage flow.

2. The value attribution analysis method of the measured value of an open seepage flow monitoring system behind a dam according to claim 1, characterized in that The rainfall characteristics are determined according to the ground watershed combined with the hydrogeological conditions.

3. The method for attributing measured values of an open seepage flow monitoring system behind a dam according to claim 1, characterized in that, The determination of the influencing factors affecting seepage flow within the influence area adopts a distributed hydrological model combined with factor analysis of the shallow water equation under the combined action of rainfall and seepage, combined with the change speed and amplitude of the above factors, so as to determine the rainfall-reservoir water seepage influencing factors that have an impact on the measured value of the seepage flow monitoring facility and change with time, and determine the monitoring items, their spatial representative measuring points and time sampling rates according to the spatial distribution and time change of these changing factors.

4. The value attribution analysis method of the measured value of an open seepage flow monitoring system behind a dam according to claim 3, characterized in that, The determination of monitoring items and measuring points based on the spatio-temporal distribution of sensitive factors in the influence area includes: Extract the changing factors affecting the rainfall influencing factors entering the seepage flow monitoring facility from the influencing factors, including rainfall amount, slope, roughness, crack distribution, vegetation cover, soil temperature, air temperature, soil moisture content; The monitoring items include rainfall, slope, roughness, cracks, vegetation height, soil temperature, air temperature and soil moisture content. The corresponding monitoring instruments and their data acquisition control devices are used to form a seepage monitoring system; the monitoring instruments include a combination of fixed installed buried sensors, patrol robot dogs and drone remote sensing; the fixed installed buried sensors include fixed position potential image sensors, tipping bucket rain gauges, radar rain gauges, raindrop spectrometers, TDR and soil moisture sensors; patrol robot dogs and drones carry laser radars, hyperspectral cameras and infrared cameras to monitor slopes, roughness and vegetation; If the affected area is relatively large, multiple measuring points should be set for the same monitoring project based on whether the distribution of various elements in the affected area is uniform and the representativeness of the measuring points. The sampling frequency of each measuring point should be determined according to the Shannon sampling theorem or compressed sensing theory based on the time changes of each monitored element.

5. The method for attributing measured values of a post-dam open seepage flow monitoring system according to claim 1, characterized in that, Determining the lag time according to the corresponding mathematical model includes: Establish an unsteady seepage model under the action of reservoir water, select a three-dimensional spatial range according to the seepage influence area, and calculate the lag effect time from reservoir seepage to the seepage flow measurement point monitoring facility under the conditions of extreme reservoir water fluctuation and different elevation change combinations; A mathematical model of surface runoff and underground seepage runoff under rainfall conditions is established, and the model includes three parts: surface, underground and infiltration models. The surface model adopts indirect numerical simulation based on the Navier-Stokes equation for complex areas, and adopts a two-dimensional shallow water equation for general areas. The underground part is described by the Richard equation. The infiltration equation is determined according to the rainfall type, vegetation soil type, soil saturation and rainfall intensity factors, and numerical calculation is used to obtain the lag influence time of rainfall runoff and seepage to the monitoring facilities of the seepage flow measurement point in the affected area.

6. The method for attributing measured values of an open seepage flow monitoring system behind a dam according to claim 1, characterized in that The historical measured data series of seepage volume are divided into three types: the period only affected by reservoir water seepage, the period obviously affected by rainfall, and the uncertain period, including: According to the measured rainfall process in the affected area; the above model is used to calculate the duration of rainfall affecting the seepage measurement value according to the relevant parameters on site; taking into account the influence of model calculation error factors, the sum of 1 / 3 of the duration at the junction of the two periods affected by rainfall and the period not affected by rainfall is defined as the uncertain duration of rainfall influence, thereby dividing the measured seepage time period into three types: the period that is obviously affected only by reservoir water seepage, the period that is obviously affected by precipitation, and the uncertain period.

7. The method for attributing measured values of an open seepage flow monitoring system behind a dam according to claim 1, characterized in that, The lightweight time convolution model based on the sample of the assimilation result and only affected by the reservoir water seepage is established, including: The dynamic unsteady seepage equation of unsaturated-saturated reservoir water during the period that is obviously not affected by rainfall is assimilated in four dimensions with the measured seepage data of the same period; The assimilated model is used to predict and generate reservoir water seepage flow samples, which are divided into training samples and test samples; The factors affecting reservoir water seepage, water level, air temperature and time are taken as the input of the lightweight time series convolutional neural network L-TCN model, and the reservoir water seepage is taken as the output of the L-TCN model. After model training and testing, a qualified model is selected as the separation model of reservoir water seepage.

8. The value attribution analysis method of the measured value of an open seepage flow monitoring system behind a dam according to claim 1, characterized in that The separation of the rainfall influence component and the reservoir water seepage component based on the difference between the two includes: Calculating the total seepage flow and the reservoir water seepage component respectively through the lightweight temporal convolutional models affected by rainfall and only affected by reservoir water seepage according to the measured inputs, and obtaining the rainfall influence component by subtracting the reservoir water seepage component from the total seepage flow, so as to realize the separation of the total measured seepage flow under rainfall conditions and non-precipitation conditions, that is, attribution analysis.

9. A post-dam open-air seepage flow monitoring system adapted to the measured value attribution analysis method according to any one of claims 1-8, characterized in that, It includes seepage flow monitoring facilities, monitoring items and monitoring facilities added according to the measured data components and influencing factors of seepage flow, and is an information perception device for realizing the all-element, full-range, all-weather and full-scenario information of the seepage flow behind the dam and its main influencing factors.

Images