Flood runoff prediction method for valley tailings pond based on dynamic correction of underlying surface parameters

Abstract

The application discloses a valley type tailing pond flood runoff prediction method based on dynamic correction of underlying surface parameters, comprising the following steps: S1, obtaining multi-source data of the valley type tailing pond for pretreatment, and constructing a basic database; S2, obtaining underlying surface parameters of the valley type tailing pond, and determining initial value ranges of the parameters; S3, establishing a dynamic correction model of the underlying surface parameters, setting differentiated runoff and confluence parameters for different regions, and performing dynamic correction; S4, constructing a coupled hydrology-hydrodynamic model, simulating a runoff formation process and time-varying characteristics, and outputting prediction values of runoff flow, flood peak time and flood peak water level, and verifying and calibrating the prediction values by comparing measured values of historical flood inundation areas of satellite remote sensing images; and S5, inputting real-time data into the hydrology-hydrodynamic model, and predicting a flood runoff inundation range, inundation time and inundation degree based on a geographic spatial analysis method. The application realizes the improvement of timeliness and accuracy of runoff prediction in small watershed regions of the valley type tailing pond and the like without runoff data.

Description

Technical Field

[0001] This invention relates to the field of runoff prediction technology for valley-type tailings ponds, and more specifically to a method for predicting flood runoff from valley-type tailings ponds based on dynamic correction of underlying surface parameters. Background Technology

[0002] Tailings dams are artificial dams formed by the accumulation of waste generated during mining operations. Due to their unique geographical location and accumulation method, they pose a certain threat to the surrounding environment and water resource safety. Most of the existing tailings dams in Jiangxi Province are valley-type tailings dams, characterized by complex underlying surface conditions and extremely short runoff durations within very small watersheds. However, the unique topographical conditions of valley-type tailings dams present a severe flood threat under complex rainfall conditions. Once flooded, tailings dams may experience overtopping, dam failure, and other disasters, leading not only to damage to the dam's facilities but also to tailings leakage, causing catastrophic consequences for the surrounding ecological environment and the safety of residents' lives and property.

[0003] Traditional flood runoff prediction techniques for valley-type tailings dams often rely on simple hydrological models and empirical formulas. These methods typically consider only single meteorological or topographical factors, failing to adequately address the complexity of underlying surface conditions. For instance, some early prediction methods relied solely on linear extrapolation based on rainfall and historical flood data, neglecting the influence of vegetation and soil characteristics on flood generation and runoff processes. While this traditional approach is computationally simple and requires minimal data, allowing for a rough assessment of flood trends to some extent, its prediction accuracy is extremely low. Because it doesn't comprehensively consider the complex geographical environment and the tailings dam's own operational status, it often fails to accurately predict runoff volume, peak flood time, and flood inundation extent when dealing with valley-type tailings dams where underlying surface factors such as slope, aspect, soil, and vegetation cover vary significantly, thus failing to meet practical flood control and disaster reduction needs.

[0004] With the continuous advancement of science and technology, existing technologies have made some progress in flood runoff prediction for large-scale watersheds. However, current research still has weaknesses and shortcomings in predicting flood runoff in small-scale watersheds within tailings dam areas. Some studies utilize abundant spatiotemporal rainfall and runoff data from meteorological and hydrological monitoring stations to construct hydrological models for flood runoff prediction, but existing technologies still face many key challenges in predicting flood runoff from tailings dams. On the one hand, fixed parameter values ​​are typically used in model parameter settings. This approach fails to consider the dynamic impact of the spatiotemporal heterogeneity of tailings dam operation and vegetation growth on underlying surface parameters, resulting in models that cannot accurately reflect the actual situation. On the other hand, tailings ponds are mostly located in valleys and hilly areas with high flood peaks and large volumes, and steep rises and falls. These areas are often small watershed-scale areas without runoff data, so it is impossible to directly use traditional large watershed-scale hydrological or hydraulic models to predict flood runoff. It is necessary to make full use of distributed data such as WeChat remote sensing data, digital elevation and high precision, and use geographic information systems (GIS) and remote sensing technology to invert the various underlying surface parameters of tailings ponds and the hydrological response runoff at the small watershed scale. Then, the flood runoff of tailings ponds can be predicted by establishing hydrological and hydrodynamic models.

[0005] Therefore, the field of flood runoff prediction for valley-type tailings ponds still faces many challenges. Traditional technical solutions are too simplistic and cannot cope with complex realities; existing technologies rely too heavily on large-scale watersheds with mature runoff monitoring stations providing abundant historical data for flood runoff prediction, which is insufficient for accurate and dynamic flood runoff prediction in areas like tailings ponds where historical runoff data is unavailable. To address these issues, this invention proposes a flood runoff prediction method for valley-type tailings ponds based on dynamic correction of underlying surface parameters. Summary of the Invention

[0006] Based on the above, this application presents a method for predicting flood runoff from valley-type tailings dams based on dynamic correction of underlying surface parameters, specifically including:

[0007] S1. Acquire topographic and geomorphological data, vegetation and soil types, land use data, meteorological and hydrological data, tailings dam parameters, and historical flood data of valley-type tailings dams. Preprocess the collected multi-source data to construct a basic database.

[0008] S2. Obtain the underlying surface parameters of the valley-type tailings dam, including vegetation interception parameters, vegetation cover index, soil type, soil saturated hydraulic conductivity, impermeable area ratio, and determine the initial value range of each underlying surface parameter.

[0009] S3. By considering vegetation growth factors, soil moisture, and the operation of tailings ponds, differentiated runoff and runoff parameters are set for different regions. A dynamic correction model for underlying surface parameters is established, and the preliminary underlying surface parameters are dynamically corrected based on a dynamic update mechanism of real-time remote sensing data.

[0010] S4. Construct a coupled hydrological-hydrodynamic model, simulate the watershed runoff formation process and time-varying characteristics through multi-risk flood scenarios, and output predicted values ​​of runoff flow, peak flood time, and peak flood level. Verify and calibrate the hydrological-hydrodynamic model by comparing it with measured values ​​of floodplains in historical flood events from satellite remote sensing images.

[0011] S5. Input real-time meteorological and hydrological data, topography, land use and vegetation coverage data into the hydrological-hydrodynamic model. Through geospatial analysis methods, based on the predicted runoff, flood peak time and flood peak water level, predict the inundation range, inundation time and inundation degree of flood runoff, and realize the prediction of flood runoff of valley-type tailings dams.

[0012] Preferably, in step S1, the collected multi-source data is preprocessed. A multi-source data association feature mining filtering algorithm is used to analyze the correlation between topography, vegetation and soil type, land use, meteorology and hydrology, tailings dam parameters and historical flood data, and to filter out feature data that has an impact on flood risk prediction. The multi-source data fusion mechanism of graph neural network is used to map different types of data into a unified feature space, and the graph structure is used to capture the relationship between feature data and perform fusion. The fused data is standardized to give it a unified dimension and scale. The data that has been filtered, fused and standardized is used to build a basic database according to the database architecture and indexing method.

[0013] Preferably, the dynamic correction model for the underlying surface parameters is established in step S3, specifically as follows:

[0014] Obtain comprehensive indicators of tailings dam operation status This includes tailings dam water level, tailings dam dry beach length, and flood discharge capacity of flood control structures; comprehensive vegetation growth indicators were obtained. This includes the rate of change in vegetation cover and the amount of vegetation height growth; a dynamic correction model for underlying surface parameters is constructed using a multilayer sensing mechanism, with the model input being... The hidden layer uses the LeakyReLU activation function, with the following formula: ,in For the first The weight matrix of the layer, For bias vectors, For the first The output of the hidden layer after processing by the activation function is expressed by the following formula: ,in , The input values ​​for the activation function. The initial input vector is used; the output layer uses a linear activation function to output the correction amount of the underlying surface parameters. The formula is: , The total number of network layers. This is the output layer weight matrix. This is the output layer bias vector. For the first The output of the layer is determined by the correction amount of the underlying surface parameters. And original underlay parameters The constructed dynamic correction model for the underlying surface parameters is given by the following formula: ,in To obtain the corrected underlying surface parameters, the correction amount is calculated by continuously inputting real-time tailings dam operation status and vegetation growth data, and the underlying surface parameters are dynamically corrected.

[0015] Preferably, when dynamically correcting the initially determined underlying surface parameters using the established dynamic correction model for underlying surface parameters, the specific steps are as follows:

[0016] Using a Geographic Information System (GIS), the valley-type tailings dam basin is divided into multiple sub-regions with different topographic, vegetation, and soil characteristics. For each sub-region, a local parameter adjustment model based on fuzzy inference is constructed. The sub-region's topographic slope, aspect, and vegetation cover are used as input variables. Membership functions are used to determine the membership degrees of each input variable in different fuzzy sets. Fuzzy inference is then performed on the input variables to obtain the adjustment coefficients for the runoff generation and confluence parameters of each sub-region. For the initially determined underlying surface parameters, the global correction amount obtained from the established underlying surface parameter dynamic correction model is combined with the local adjustment coefficients obtained from the adjustment coefficients of each sub-region to correct the initially determined underlying surface parameters. The formula is as follows: ,in For the first The underlying surface parameters at the first The corrected values ​​for each sub-region For the underlying surface parameters in the first... The initial values ​​determined for each sub-region The first obtained by the dynamic correction model of the underlying surface parameters Global correction amount for each underlying surface parameter For the first The local adjustment coefficients of each sub-region form the settings for differentiated runoff generation parameters in different regions, and dynamically correct the initially determined underlying surface parameters.

[0017] Preferably, a local parameter adjustment model based on fuzzy inference is constructed for each sub-region, specifically as follows:

[0018] Determine the slope of the sub-region Slope aspect Vegetation coverage The feature vectors are defined, multiple fuzzy sets are set, and Gaussian membership functions are used to define them. Calculate the membership degree of each input variable in different fuzzy sets, where For input variable values, The membership function center value, To determine the standard deviation, a fuzzy rule base is constructed. A fuzzy synthesis inference algorithm is then used to infer and calculate the fuzzy rules, yielding the adjustment coefficients for the flow generation and merging parameters of each sub-region. The calculation formula is: ,in To determine the number of fuzzy rules, For the first The weight of a fuzzy rule, For the first The comprehensive membership degree of the input variables in the fuzzy rules is used to construct a local parameter adjustment model.

[0019] Preferably, the hydrological-hydrodynamic model in S4 includes a distributed hydrological model and a flood propagation hydrodynamic model, which are coupled to construct the hydrological-hydrodynamic model. The distributed hydrological model simulates the runoff generation and confluence process of the watershed and outputs the flood hydrograph at the tailings dam inlet. The flood propagation hydrodynamic model simulates the evolution of floods in the tailings dam area based on the modified topography and inflow boundary, and calculates the inundation depth and flow velocity. The hydrological-hydrodynamic model obtains the runoff generation and confluence process of the watershed by simulating multiple risk scenarios, including local dam failure and vegetation degradation, and outputs simulated values. The simulated values ​​are compared with the measured values ​​to verify and calibrate the hydrological-hydrodynamic model through historical flood event inversion.

[0020] Preferably, the distributed hydrological model simulates the runoff generation and confluence process of the watershed and outputs the flood hydrograph at the tailings dam inlet, specifically as follows:

[0021] The watershed is divided into grids, with each grid cell... Corresponding underlying surface parameters Construct a flow generation model and utilize a deep residual shrinkage network. Meteorological and hydrological data, topographic data, and underlying surface parameters of the grid cells Perform feature extraction to obtain feature vectors The formula is: ,in For rainfall, For evaporation, Given the slope, runoff generation was calculated using an infiltration model, taking into account the impact of vegetation interception on infiltration, and the corrected infiltration rate was calculated. The formula is: ,in For the duration of rainfall, For grid cells The formula for the water accumulation delay time is: , For the vegetation interception correction factor, obtain the first... Flow rate of grid cells The formula is: ,in For grid cells The area is calculated by summing the flow rates of all grid cells flowing into the tailings dam inlet and subtracting the discharge flow rate from the combined drainage wells of the flood discharge system that flows into the main tunnel. The discharge flow rate of the main tunnel is calculated as pressurized flow. The formula is: , ,in It is a resistance-corrected flow coefficient. It is the acceleration due to gravity. For water flow section, The calculated head at the tunnel inlet. The tunnel has a longitudinal slope. The length of the tunnel. The height of the tunnel. For the Xie Cai coefficient, The drag coefficient is used to output the flood hydrograph at the tailings dam inlet, using the following formula: ,in The number of grid cells flowing into the tailings dam inlet. For the first The flow rate of each grid cell flowing into the tailings dam inlet. For the first The main tunnel discharge flow rate is associated with each grid cell flowing into the tailings dam inlet.

[0022] Preferably, the flood propagation hydrodynamic model is based on modified topography and inflow boundary to simulate the evolution of flood in the tailings dam area and calculate the inundation depth and flow velocity, specifically as follows:

[0023] In simulating flood evolution, a multi-scale spatiotemporal grid division technique is used to dynamically refine and coarsen the grid based on topographic changes and water flow velocity, thereby establishing the water flow motion equations: ,in Because of the water depth, , They are respectively , Flow velocity component in the direction, , The inflow and outflow rates of the grid cells are respectively used. Through iterative solutions, combined with corrected topography and inflow boundary conditions, the evolution of the flood in the tailings dam area is simulated; the current water depth is used as a reference. Subtract terrain elevation Calculate the flood depth The formula is: ; obtained by solving the equation , Flow velocity component in direction , To obtain the submerged flow velocity The formula is: .

[0024] Preferably, the hydrological-hydrodynamic model simulates multiple risk scenarios to obtain the runoff generation and confluence process of the watershed, outputs simulated values, and verifies and calibrates the hydrological-hydrodynamic model by comparing the simulated values ​​with measured values ​​through historical flood event inversion. Specifically:

[0025] The runoff generation and confluence processes of the watershed are obtained through a hydro-hydrodynamic model. The tailings dam inlet flood hydrograph output by the WRF-Hydro distributed hydrological model in the hydro-hydrodynamic model is used to simulate the inflow boundary of the flood propagation hydrodynamic model. The flood propagation hydrodynamic model simulates flood evolution and outputs simulated runoff. Peak flood time Flood peak water level Through inversion verification of historical flood events, measured runoff is obtained from historical flood events. Peak flood time Flood peak water level Construct an error evaluation function ,in , , The weights of each indicator are determined based on their importance to flood risk assessment. The comprehensive error assessment function is adjusted by modifying the parameters in the hydrological-hydrodynamic model. Minimum, enabling the verification and calibration of hydrological-hydrodynamic models.

[0026] Preferably, in step S5, the predicted runoff flow rate is used... Peak flood time Flood peak water level The extent and duration of flooding can be determined by reversing these parameters, thus establishing the degree of inundation.

[0027] Constructing terrain surfaces using digital elevation model data ,in Use geographical coordinates to determine the extent of flood inundation. The formula is: in, For point The cross-sectional area of ​​the water passage, To preset the water level difference, ,point Within the flood zone, At time, point Those not submerged were identified by coordinating data points to determine the extent of flooding; the timing of the flood peak was also considered. Determine the flooding time The formula is: ,in, For point The time of submersion, This represents the shortest water flow path distance from this point to the starting point of the flood peak. This is a correction factor for water flow velocity. The average water flow velocity is used; the degree of inundation is determined by the inundation range and inundation time, using the following formula: in, To determine the degree of flooding, To submerge the water level, and These are the highest and lowest water levels in history. and These represent the longest and shortest flooding times in history, respectively. This is the highest runoff volume in history. , and These are the corresponding weighting coefficients.

[0028] Compared with the prior art, the technical solution of this application has the following technical effects:

[0029] This invention constructs a basic database and uses a multi-source data association feature mining filtering algorithm to analyze the correlation relationships of various types of data and filter out the feature data that affect flood runoff prediction. It uses a graph neural network multi-source data fusion mechanism to map different types of data to a unified feature space and capture their relationships for fusion. The fused data is standardized and then a basic database is constructed according to the database architecture and indexing method. This solves the technical problems of fragmented data, lack of systematic integration, and difficulty in obtaining key underlying surface data that affect flood runoff prediction in traditional data processing methods.

[0030] This invention employs a dynamic correction model for underlying surface parameters. It acquires comprehensive indicators of tailings dam operation (tailings dam water level, reservoir storage capacity, tailings dam dry beach length, and flood discharge capacity of flood control structures) and comprehensive indicators of vegetation growth (vegetation coverage change rate and vegetation height growth). Using a multilayer perceptron, these indicators are used as input to construct a model. The hidden layer employs the LeakyReLU activation function, and the output layer uses a linear activation function to output the correction amount of the underlying surface parameters. This is then combined with the original underlying surface parameters to construct a dynamic correction model. By continuously inputting real-time data to calculate the correction amount, the parameters are dynamically corrected. This solves the technical problem in traditional prediction methods that ignore the impact of tailings dam operation and vegetation growth changes on underlying surface parameters, leading to model parameters that cannot adapt to changes in actual conditions.

[0031] This invention employs a technical solution for constructing a coupled hydrological-hydrodynamic model. The distributed hydrological model divides the watershed into grids and utilizes a deep residual shrinking network to extract meteorological, hydrological, topographic, and underlying surface parameter characteristics from the grid units. It then combines this with a Philips infiltration model considering vegetation interception to calculate runoff generation. The runoff generation from the grid units flowing into the tailings dam inlet is accumulated, and the flood discharge from the drainage system is subtracted to output the flood hydrograph. The flood propagation hydrodynamic model, based on modified underlying surface factors such as vegetation interception, soil saturated hydraulic conductivity, vegetation cover, and reservoir inflow boundary, uses multi-scale spatiotemporal grid division technology to establish a flow motion equation to simulate runoff evolution, calculating runoff volume, peak flood time, and inundation level. The model is validated and calibrated through inversion using simulations of multiple risk scenarios and historical flood events, along with measured values ​​from satellite remote sensing images of floodplains. Underlying surface parameters such as topographic slope, aspect, soil, and vegetation cover are dynamically corrected. Based on geospatial analysis methods, the model predicts the inundation range, inundation time, and inundation degree of flood runoff. This invention solves the technical problem of improving the accuracy of flood runoff prediction in small-scale watersheds of tailings ponds where historical spatiotemporal runoff data is unavailable. The above description is merely an overview of the technical solution of this application. To better understand the technical means of this application and to enable its implementation according to the description, and to make the above and other objects, features, and advantages of this application more apparent, the following detailed description, with reference to preferred embodiments and accompanying drawings, is provided.

[0032] The above and other objects, advantages and features of this application will become more apparent to those skilled in the art from the following detailed description of specific embodiments in conjunction with the accompanying drawings. Attached Figure Description

[0033] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In all drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0034] Figure 1 The flowchart shows a method for predicting flood runoff in valley-type tailings ponds based on dynamic correction of underlying surface parameters.

[0035] Figure 2 This is a land use type distribution map of a valley-type tailings dam watershed in Jiangxi Province.

[0036] Figure 3 A comparison diagram of the flood process lines at the inlet of a valley-type tailings dam in Jiangxi Province, showing the corrected and uncorrected groups.

[0037] Figure 4 The HEC-RAS model simulation of the flood evolution and water level change curve on June 27, 2024, for a valley-type tailings dam in Jiangxi Province;

[0038] Figure 5 Runoff flow curves for simulating flood evolution using the HEC-RAS model of a valley-type tailings dam in Jiangxi Province;

[0039] Figure 6 The curve showing the flood evolution and water level change of a tailings dam in a valley in Jiangxi Province during an extreme rainfall event (daily rainfall of 96.8 mm).

[0040] Figure 7 The runoff-discharge curve of the flood evolution during an extreme rainfall event at a tailings dam in a valley in Jiangxi Province (simulated by the HEC-RAS model).

[0041] Figure 8 This is a diagram showing the rainfall-runoff process of a valley-type tailings dam in Jiangxi Province from June 20 to July 5, 2024.

[0042] Figure 9 This is an overlay map showing the predicted flood inundation area of ​​a valley-type tailings dam in Jiangxi Province on June 27, 2024, and the actual floodplain measured by satellite. Detailed Implementation

[0043] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. In the following description, specific details such as specific configurations and components are provided merely to help fully understand the embodiments of this application. Therefore, those skilled in the art should understand that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this application. In addition, for clarity and brevity, descriptions of known functions and structures are omitted in the embodiments.

[0044] It should be understood that the phrase "an embodiment" or "this embodiment" throughout the specification means that a specific feature, structure, or characteristic related to the embodiment is included in at least one embodiment of this application. Therefore, "an embodiment" or "this embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments.

[0045] Furthermore, reference numerals and / or letters may be repeated in different examples within this application. Such repetition is for the purpose of simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or settings discussed.

[0046] In this article, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can mean: A exists alone, B exists alone, and A and B exist simultaneously. The term " / and" in this article describes another type of relationship between related objects, indicating that two relationships can exist. For example, A / and B can mean: A exists alone, and A and B exist alone. In addition, the character " / " in this article generally indicates that the related objects before and after it are in an "or" relationship.

[0047] In this article, the term "at least one" is merely a description of the relationship between related objects, indicating that there can be three relationships. For example, "at least one of A and B" can mean: A exists alone, A and B exist simultaneously, or B exists alone.

[0048] It should also be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion.

[0049] Example 1

[0050] This embodiment mainly describes a flood runoff prediction method for valley-type tailings dams based on dynamic correction of underlying surface parameters, such as... Figure 1 As shown, it includes:

[0051] S1. Acquire topographic and geomorphological data, vegetation and soil types, land use data, meteorological and hydrological data, tailings dam parameters, and historical flood data of valley-type tailings dams. Preprocess the collected multi-source data to construct a basic database.

[0052] S2. Obtain the underlying surface parameters of the valley-type tailings dam, including vegetation interception parameters, vegetation cover index, soil type, soil saturated hydraulic conductivity, impermeable area ratio, and determine the initial value range of each underlying surface parameter.

[0053] S3. By considering vegetation growth factors, soil moisture, and the operation of tailings ponds, differentiated runoff and runoff parameters are set for different regions. A dynamic correction model for underlying surface parameters is established, and the preliminary underlying surface parameters are dynamically corrected based on a dynamic update mechanism of real-time remote sensing data.

[0054] S4. Construct a coupled hydrological-hydrodynamic model, simulate the watershed runoff formation process and time-varying characteristics through multi-risk flood scenarios, and output predicted values ​​of runoff flow, peak flood time, and peak flood level. Verify and calibrate the hydrological-hydrodynamic model by comparing it with measured values ​​of floodplains in historical flood events from satellite remote sensing images.

[0055] S5. Input real-time meteorological and hydrological data, topography, land use and vegetation coverage data into the hydrological-hydrodynamic model. Through geospatial analysis methods, based on the predicted runoff, flood peak time and flood peak water level, predict the inundation range, inundation time and inundation degree of flood runoff, and realize the prediction of flood runoff of valley-type tailings dams.

[0056] Furthermore, in S1, the collected multi-source data is preprocessed. Through a multi-source data association feature mining filtering algorithm, the correlation between topography, vegetation and soil type, land use, meteorology and hydrology, tailings dam parameters and historical flood data is analyzed to screen out feature data that has an impact on flood risk prediction. Using the multi-source data fusion mechanism of graph neural network, different types of data are mapped to a unified feature space. The graph structure is used to capture the relationship between feature data and perform fusion. The fused data is standardized to give it a unified dimension and scale. The filtered, fused and standardized data are used to build a basic database according to the database architecture and indexing method.

[0057] Furthermore, a dynamic correction model for the underlying surface parameters is established in S3, specifically as follows:

[0058] Obtain comprehensive indicators of tailings dam operation status This includes tailings dam water level, tailings dam dry beach length, and flood discharge capacity of flood control structures; and obtaining comprehensive vegetation growth indicators. This includes the rate of change in vegetation cover and the amount of vegetation height growth; a dynamic correction model for underlying surface parameters is constructed using a multilayer sensing mechanism, with the model input being... The hidden layer uses the LeakyReLU activation function, with the following formula: ,in For the first The weight matrix of the layer, For bias vectors, For the first The output of the hidden layer after processing by the activation function is expressed by the following formula: ,in , The input values ​​for the activation function. The initial input vector is used; the output layer uses a linear activation function to output the correction amount of the underlying surface parameters. The formula is: , The total number of network layers. This is the output layer weight matrix. This is the output layer bias vector. For the first The output of the layer is determined by the correction amount of the underlying surface parameters. And original underlay parameters The constructed dynamic correction model for the underlying surface parameters is given by the following formula: ,in To obtain the corrected underlying surface parameters, the correction amount is calculated by continuously inputting real-time tailings dam operation status and vegetation growth data, and the underlying surface parameters are dynamically corrected.

[0059] Furthermore, when dynamically correcting the initially determined underlying surface parameters using the established dynamic correction model, the specific steps are as follows:

[0060] Using a Geographic Information System (GIS), the valley-type tailings dam basin is divided into multiple sub-regions with different topographic, vegetation, and soil characteristics. For each sub-region, a local parameter adjustment model based on fuzzy inference is constructed. The sub-region's topographic slope, aspect, and vegetation cover are used as input variables. Membership functions are used to determine the membership degrees of each input variable in different fuzzy sets. Fuzzy inference is then performed on the input variables to obtain the adjustment coefficients for the runoff generation and confluence parameters of each sub-region. For the initially determined underlying surface parameters, the global correction amount obtained from the established underlying surface parameter dynamic correction model is combined with the local adjustment coefficients obtained from the adjustment coefficients of each sub-region to correct the initially determined underlying surface parameters. The formula is as follows: ,in For the first The underlying surface parameters at the first The corrected values ​​for each sub-region For the underlying surface parameters in the first... The initial values ​​determined for each sub-region The first obtained by the dynamic correction model of the underlying surface parameters Global correction amount for each underlying surface parameter For the first The local adjustment coefficients of each sub-region form the settings for differentiated runoff generation parameters in different regions, and dynamically correct the initially determined underlying surface parameters.

[0061] Furthermore, for each sub-region, a local parameter adjustment model based on fuzzy inference is constructed, specifically as follows:

[0062] Determine the slope of the sub-region Slope aspect Vegetation coverage The feature vectors are defined, multiple fuzzy sets are set, and Gaussian membership functions are used to define them. Calculate the membership degree of each input variable in different fuzzy sets, where For input variable values, The membership function center value, To determine the standard deviation, a fuzzy rule base is constructed. A fuzzy synthesis inference algorithm is then used to infer and calculate the fuzzy rules, yielding the adjustment coefficients for the flow generation and merging parameters of each sub-region. The calculation formula is: ,in To determine the number of fuzzy rules, For the first The weight of a fuzzy rule, For the first The comprehensive membership degree of the input variables in the fuzzy rules is used to construct a local parameter adjustment model.

[0063] Furthermore, the hydro-hydrodynamic model in S4 includes a distributed hydrological model and a flood propagation hydrodynamic model, which are coupled to construct the hydro-hydrodynamic model. The distributed hydrological model simulates the runoff generation and confluence process of the watershed and outputs the flood hydrograph at the tailings dam inlet. The flood propagation hydrodynamic model, based on modified topography and inflow boundary, simulates the evolution of flood in the tailings dam area and calculates the inundation depth and flow velocity. The hydro-hydrodynamic model simulates multiple risk scenarios, including local dam failure and vegetation degradation, to obtain the runoff generation and confluence process of the watershed, output simulated values, and verifies and calibrates the hydro-hydrodynamic model by comparing simulated values ​​with measured values ​​through historical flood event inversion.

[0064] Furthermore, the distributed hydrological model simulates the runoff generation and confluence process of the watershed and outputs the flood hydrograph at the tailings dam inlet, specifically:

[0065] The watershed is divided into grids, with each grid cell... Corresponding underlying surface parameters Construct a flow generation model and utilize a deep residual shrinkage network. Meteorological and hydrological data, topographic data, and underlying surface parameters of the grid cells Perform feature extraction to obtain feature vectors The formula is: ,in For rainfall, For evaporation, Given the slope, runoff generation was calculated using an infiltration model, taking into account the impact of vegetation interception on infiltration, and the corrected infiltration rate was calculated. The formula is: ,in For the duration of rainfall, For grid cells The formula for the water accumulation delay time is: , For the vegetation interception correction factor, obtain the first... Flow rate of grid cells The formula is: ,in For grid cells The area is calculated by summing the flow rates of all grid cells flowing into the tailings dam inlet and subtracting the discharge flow rate from the combined drainage wells of the flood discharge system that flows into the main tunnel. The discharge flow rate of the main tunnel is calculated as pressurized flow. The formula is: , ,in It is a resistance-corrected flow coefficient. It is the acceleration due to gravity. For water flow section, The calculated head at the tunnel inlet. The tunnel has a longitudinal slope. The length of the tunnel. The height of the tunnel. For the Xie Cai coefficient, The drag coefficient is used to output the flood hydrograph at the tailings dam inlet, using the following formula: ,in The number of grid cells flowing into the tailings dam inlet. For the first The flow rate of each grid cell flowing into the tailings dam inlet. For the first The main tunnel discharge flow rate is associated with each grid cell flowing into the tailings dam inlet.

[0066] Furthermore, the flood propagation hydrodynamic model, based on modified topography and inflow boundary, simulates the evolution of floodwaters in the tailings dam area, calculating inundation depth and flow velocity, specifically:

[0067] In simulating flood evolution, a multi-scale spatiotemporal grid division technique is used to dynamically refine and coarsen the grid based on topographic changes and water flow velocity, thereby establishing the water flow motion equations: ,in Because of the water depth, , They are respectively , Flow velocity component in the direction, , The inflow and outflow rates of the grid cells are respectively used. Through iterative solutions, combined with corrected topography and inflow boundary conditions, the evolution of the flood in the tailings dam area is simulated; the current water depth is used as a reference. Subtract terrain elevation Calculate the flood depth The formula is: ; obtained by solving the equation , Flow velocity component in direction , To obtain the submerged flow velocity The formula is: .

[0068] Furthermore, the hydro-hydrodynamic model simulates multiple risk scenarios to obtain the runoff generation and confluence processes in the watershed, outputs simulated values, and verifies and calibrates the hydro-hydrodynamic model by comparing simulated and measured values ​​through historical flood event inversion. Specifically:

[0069] The runoff generation and confluence processes of the watershed are obtained through a hydro-hydrodynamic model. The tailings dam inlet flood hydrograph output by the WRF-Hydro distributed hydrological model in the hydro-hydrodynamic model is used to simulate the inflow boundary of the flood propagation hydrodynamic model. The flood propagation hydrodynamic model simulates flood evolution and outputs simulated runoff. Peak flood time Flood peak water level Through inversion verification of historical flood events, measured runoff is obtained from historical flood events. Peak flood time Flood peak water level Construct an error evaluation function ,in , , The weights of each indicator are determined based on their importance to flood risk assessment. The comprehensive error assessment function is adjusted by modifying the parameters in the hydrological-hydrodynamic model. Minimum, enabling the verification and calibration of hydrological-hydrodynamic models.

[0070] Furthermore, in S5, based on the predicted runoff flow... Peak flood time Flood peak water level The extent and duration of flooding can be determined by reversing these parameters, thus establishing the degree of inundation.

[0071] Constructing terrain surfaces using digital elevation model data ,in Use geographical coordinates to determine the extent of flood inundation. The formula is: in, For point The cross-sectional area of ​​the water passage, To preset the water level difference, ,point Within the flood zone, At time, point Those not submerged were identified by coordinating data points to determine the extent of flooding; the timing of the flood peak was also considered. Determine the flooding time The formula is: ,in, For point The time of submersion, This represents the shortest water flow path distance from this point to the starting point of the flood peak. This is a correction factor for water flow velocity. The average water flow velocity is used; the degree of inundation is determined by the inundation range and inundation time, using the following formula: in, To determine the degree of flooding, To submerge the water level, and These are the highest and lowest water levels in history. and These represent the longest and shortest flooding times in history, respectively. This is the highest runoff volume in history. , and These are the corresponding weighting coefficients.

[0072] This embodiment details a technical solution that addresses issues in traditional runoff prediction methods, such as incomplete data processing, static parameters, incomplete simulation, and inaccurate risk assessment, by constructing a basic database, a dynamic correction model for underlying surface parameters, a coupled hydrological-hydrodynamic model, and a risk assessment system. This solution enables accurate prediction of flood risk for valley-type tailings ponds, effectively improving the scientific nature of flood control decisions, reducing flood damage to tailings ponds and surrounding areas, and safeguarding people's lives and property as well as the stability of the ecological environment.

[0073] Based on Example 1, this example describes in detail the construction of the basic database in S1, specifically as follows:

[0074] The collected multi-source data were preprocessed, and the topography, vegetation and soil types, land use, meteorological and hydrological data, tailings dam parameters, and historical flood data were encoded and represented as vectors. Construct a correlation analysis matrix , For the number of data samples, For the first In the class of data, the first The original values ​​of each sample, For the first In the class of data, the first The original values ​​of each sample, For the first The sample mean of class data. For the first The sample mean of class data. Reflects the data and The degree of correlation between them, setting a threshold ,when At that time, the data was identified. and There is a strong correlation, so data related to flood runoff prediction are selected as feature data;

[0075] When using graph neural networks for data fusion, the selected feature data is constructed into a graph structure, where nodes represent data features and edges represent the relationships between features. Layer node characteristics are Through formula To carry out dissemination and updates, among which The graph structure is represented by an adjacency matrix. For degree matrix, For the first Layer weight matrix, For bias vectors, As an activation function, after multiple propagations, it maps different types of data to a unified feature space, thereby achieving data fusion;

[0076] The merged data is standardized using a formula. ,in This is the original data. The mean of the data. To determine the variance, a basic database is constructed. According to the pre-designed database architecture, the standardized data is stored in key-value pairs. The key is a unique identifier for the data, such as a combination of data type and timestamp. The value is the standardized feature data. The index is built based on the key attributes of the data, and the data is quickly stored in the corresponding location, thus completing the construction of the basic database.

[0077] This embodiment effectively integrates various complex data through a technical process of multi-source data association feature mining, graph neural network fusion, data standardization, and architecture-based indexing to build a basic database. It can accurately analyze the relationships between data, screen out key features, and avoid interference from redundant information. Unifying the units and scales eliminates the impact of data differences, making the data more comparable. The reasonable architecture and index design improves data storage and query efficiency, providing comprehensive, accurate, and efficient data support for subsequent model calculations and flood runoff prediction, significantly enhancing the reliability and accuracy of prediction.

[0078] Based on Embodiment 1, this embodiment describes in detail the acquisition of the initial value range of each parameter of the underlying surface in this application, specifically as follows:

[0079] For vegetation interception parameters, representative vegetation plots were selected around the tailings dam, and multiple observation units with different heights, densities, and vegetation types were set up. During a complete precipitation cycle, the amount of water intercepted by vegetation during each precipitation event was recorded, and the vegetation interception parameters were calculated. The formula is: The total number of precipitation events is set as follows: , For the first Vegetation interception during the next rainfall The biomass of vegetation. Leaf area index, Using empirical coefficients representing local vegetation characteristics, the initial range of values ​​for the vegetation interception parameter is determined as follows: ,in This is the vegetation interception adjustment value;

[0080] For the vegetation cover index, the tailings pond area is divided into multiple grid units, and the vegetation cover area within each grid is identified. and total grid area The vegetation cover index for each grid is calculated using the following formula: For all grids Statistical analysis was performed to obtain the mean. and standard deviation The initial range of values ​​for the vegetation cover index was determined as follows: ,in The vegetation cover index bias coefficient;

[0081] To determine the saturated hydraulic conductivity of the soil, multiple measurement points were set up at different locations within the tailings dam to obtain the original saturated hydraulic conductivity of the soil at each measurement point. Considering soil heterogeneity and measurement errors, a correction is made, and the formula is as follows: ,in This represents the average saturated hydraulic conductivity of the soil at all measurement points. The corrected saturated hydraulic conductivity of the soil. Correction parameters for soil saturated hydraulic conductivity; by combining The initial range of values ​​for the saturated hydraulic conductivity of the soil was determined to be: ,in This is the adjustment coefficient for soil saturated hydraulic conductivity;

[0082] To determine the proportion of impermeable areas, the types of impermeable surfaces within the tailings dam area are determined, and the area of ​​each impermeable surface type is calculated. The summation yields the total impermeable area. Calculate the ratio of impermeable area The formula is: , Given the total area of ​​the tailings dam area, the initial range for determining the proportion of impermeable area is as follows: ,in This is the ratio of impermeable area.

[0083] This embodiment determines the initial range of underlying surface parameters, improves the accuracy of runoff prediction for valley-type tailings ponds, integrates multi-source data, and fully considers the spatiotemporal variation characteristics of vegetation, soil, and land use. This provides a reliable foundation for subsequent dynamic correction of underlying surface parameters and model simulation, reduces prediction deviations caused by unreasonable parameter values, and enhances the adaptability of the runoff prediction model to complex environments and the credibility of prediction results.

[0084] Based on Example 1, this example details a comparison between the flood runoff prediction method for valley-type tailings ponds of this application and existing technologies, specifically:

[0085] like Figure 2As shown, a typical valley-type tailings dam in Jiangxi Province was selected as the research object. The catchment area of ​​the tailings dam is 0.719 km², and the terrain is characterized by a "two mountains and one valley". The highest elevation in the catchment is 326.5 m, the lowest is 112.3 m, and the relative elevation difference is 214.2 m. The main ditch is about 1.7 km long, and the average slope is 18.7°. The vegetation within the catchment area of ​​the tailings dam is mainly subtropical evergreen broad-leaved forest, accounting for 68.11% of the total vegetation coverage, followed by shrub forest (10.26%) and grassland (5.18%). Mining land accounts for 16.45%, and the rest is dry land and paddy fields. The soil type is red soil, accounting for 82.47%, which is characterized by heavy clay and poor permeability. The average measured soil saturated hydraulic conductivity is [value missing]. ;

[0086] The region has a subtropical monsoon climate with an average annual precipitation of 1356.42 mm, of which 57.3% occurs from June to August, with a maximum hourly rainfall of 48.6 mm. The average annual evaporation is 987.36 mm. During the experimental period (June-August 2024), the tailings dam water level fluctuated between 208 and 208.66 m, the dry beach length ranged from 95.01 to 165 m, and the drainage well's discharge capacity ranged from 0 to 7.47 m³ / s. Topographic data was obtained through a combination of satellite remote sensing (5 m resolution) and ground measurements. Land use status was obtained using drone aerial photography. Fifty soil samples were collected for physicochemical analysis. Hourly rainfall and evaporation data from a meteorological station 5 km from the tailings dam over the past 20 years, as well as measured data from eight typical flood events in the past five years, were also collected, as shown in Table 1.

[0087] Table 1. Measured data of flood events

[0088] Time of occurrence Rainfall (mm) Runoff flow rate (m³ / s) Discharge volume (m³ / s) Peak flood time (h) Peak flood level (m) Flooded area (km²) 2019.07.21 94.5 21.3 6.43 1.26 208.55 0.39 2020.06.15 94.7 21.4 6.89 1.25 208.57 0.40 2021.08.09 88.3 19.2 4.14 1.46 208.42 0.35 2022.07.03 92.9 20.8 5.43 1.34 208.48 0.38 2023.06.28 89.4 19.7 4.76 1.42 208.45 0.36

[0089] Through field measurements and data analysis, initial values ​​of underlying surface parameters were obtained: the vegetation interception parameter of subtropical evergreen broad-leaved forest was 2.35 mm, the average vegetation cover index was 0.65, and the soil saturated hydraulic conductivity was [missing value]. The impermeable area ratio is 12.38%. The initial value range of each parameter was determined through statistical analysis. For example, the vegetation cover index was centered at a mean of 0.65 and combined with a standard deviation of 0.08, the range was determined to be [0.57, 0.73].

[0090] A multilayer perceptron model was constructed, taking the tailings dam's operational status (water level, dry beach length, etc.) and vegetation growth indicators (coverage change rate, height growth) as inputs, and outputting corrections for underlying surface parameters. Taking vegetation interception parameters as an example, the model calculated a correction of 0.08 mm, resulting in a corrected parameter of 2.43 mm; the corrected soil saturated hydraulic conductivity was... The parameters are divided into a "corrected group" and an "uncorrected group": the corrected group uses the dynamically updated parameters, while the uncorrected group retains the initial parameters. A comparison of the parameters before and after correction is shown in Table 2.

[0091] Table 2 Comparison of Correction Parameters

[0092]

[0093] like Figure 3 As shown, the watershed was meshed using the WRF-Hydro model, with the corrected and uncorrected groups using the same mesh precision (100m × 100m). A typical rainfall event (daily rainfall of 45.63 mm) measured in August 2023 was selected for validation. The measured data showed that under this rainfall condition, the peak value of the flood hydrograph at the tailings dam inlet was 18.5 m³ / s, occurring at 1.62 h.

[0094] Based on this rainfall event, the corrected group calculated an infiltration rate of 0.0012 m³ / h and a single-grid flow rate of 0.035 m³ / s. After superimposing the flow rates of each grid along the confluence path, the peak value of the tailings dam inlet flood process line for the corrected group was 18.82 m³ / s, occurring at 1.55 h, with an error of only 1.72% compared to the measured peak value and a peak time error of 1.85%. The uncorrected group had an infiltration rate of 0.0010 m³ / h and a single-grid flow rate of 0.042 m³ / s; after superposition calculation, the peak value was 16.3 m³ / s, occurring at 2.14 h, which was 11.89% lower than the measured peak value and had a peak time lag of 32.09%. The comparison with the measured data shows that the accuracy of the corrected group in the flow generation calculation is significantly better than that of the uncorrected group.

[0095] like Figures 4-5 As shown, the evolution of flood in the tailings dam area was simulated using the HEC-RAS model. Taking the historical flood event of June 27, 2024 as an example, the rainfall of this event was 92.5 mm, the measured runoff was 20.6 m³ / s, the peak time was 1.36 h, and the peak water level was 208.46 m. ​​The simulation results of the corrected group were: runoff 21.21 m³ / s (error 3.0%), peak time 1.28 h (error -5.9%), and peak water level 208.53 m (error 0.033%); the simulation results of the uncorrected group were: runoff 18.85 m³ / s (error -8.49%), peak time 1.64 h (error 20.6%), and peak water level 208.23 m (error -0.096%).

[0096] To further verify the model's predictive ability under extreme weather conditions, such as Figures 6-7As shown, real-time meteorological data is input into the model to predict the flood risk of a certain extreme rainfall event (daily rainfall of 96.8 mm). Based on historical measured data of extreme rainfall events (the average measured runoff for this type of extreme rainfall event was 23.5 m³ / s, the average peak time was 1.18 h, the average peak water level was 208.71 m, the average inundation area was 0.45 m², the average inundation time was 3.1 h, and the average inundation severity was 0.55), the revised group predicted a runoff of 23.7 m³ / s, a peak time of 1.15 h, a peak water level of 208.68 m, an inundation area of ​​0.428 m², an average inundation time of 3.06 h, and an average inundation severity of 0.56. The predicted inundation area and the satellite floodplain display showed a 96.7% agreement. The unrevised group predicted a runoff of 20.7 m³ / s, a peak time of 1.34 h, a peak water level of 208.43 m, an inundation area of ​​0.386 km², an average inundation time of 4.25 h, and an average inundation severity of 0.72.

[0097] The comparison shows that the uncorrected group overestimated the peak runoff by 12.7% and the inundation area by 44.8% compared to the corrected group. By dynamically correcting the underlying surface parameters, the model significantly improved the accuracy of flood prediction for valley-type tailings ponds.

[0098] like Figure 8 As shown, during continuous monitoring from June 20 to July 5, 2024, the technical solution of this application provided real-time predictions of the rainfall-runoff process in the tailings dam basin. Measured data showed that the maximum rainfall occurred on June 27, with a cumulative 24-hour rainfall of 92.5 mm, corresponding to a peak measured runoff flow of 20.6 m³ / s and a peak flood level of 208.43 m. However, the prediction results, corrected based on the method of this application, showed a predicted peak flow of 21.21 m³ / s and a flood level of 208.48 m for that day, with deviations from the measured values ​​of 2.96% and 0.024%, respectively. Looking at the overall trend over 15 days, the curves of the measured and predicted flow rates showed a high degree of agreement during the rainfall-runoff generation phase, especially during the runoff generation delay after the rainfall peak on June 28. The method of this application predicted the flood peak to occur 1.8 hours after the start of rainfall, a difference of only 6 minutes from the measured 1.9 hours.

[0099] Further comparison of the prediction performance of the proposed method with that of the uncorrected parameter model revealed that during the rainfall event on June 29th (85.2 mm of rainfall), the uncorrected model, using the initial underlying surface parameters, predicted a peak flow of 20.2 m³ / s, an overestimation of 9.19% compared to the measured value of 18.5 m³ / s, and the peak flow occurred 42 minutes earlier. In contrast, the proposed method, by updating vegetation cover and soil moisture parameters in real-time using remote sensing data, predicted a peak flow of 18.7 m³ / s, with a deviation of only 1.08% from the measured value. The 15-day water level-time series data showed that the correlation coefficient between the water level fluctuation curve predicted by the proposed method and the measured data reached 0.98, with a root mean square error of 0.11 m, significantly better than the 0.41 m error of the uncorrected model. The data demonstrate that the proposed technical solution, through dynamic correction of underlying surface parameters, effectively improves the accuracy of runoff and water level prediction for valley-type tailings ponds under heavy rainfall conditions, providing precise flood warning support for the safe operation of tailings ponds during the flood season.

[0100] like Figure 9 As shown, taking June 27, 2024 as an example, the rainfall in the tailings dam basin reached 92.5 mm on that day. The flood inundation range predicted by the technical solution of this application was 0.375 km², while the measured floodplain range obtained by satellite monitoring was 0.372 km². After superimposing the two, it can be seen that the overlap area between the predicted range and the actual floodplain is 0.362 km², and the overlap area accounts for 94.47% of the satellite floodplain. In the steep valley area on the northeast side of the tailings dam, the predicted inundation boundary deviates from the actual inundation boundary in the satellite image by less than 5 meters, capturing the flood accumulation path under the constraints of terrain. Although the predicted range in the gentle slope area on the southwest side is slightly larger than the actual floodplain by 0.014 km², the main difference is distributed in the edge area where the inundation depth is less than 0.2 meters, which is a reasonable error caused by the terrain data resolution (5 meters). From the spatial distribution, the core inundation area of ​​the predicted range completely overlaps with the actual floodplain, with only slight differences in the terrain transition zone, which fully proves that the prediction accuracy of the flood inundation range of the technical solution of this application has reached the engineering application standard. In the gentle slope area on the southwest side, although the predicted range is slightly smaller than the actual range, the overall trend is consistent. The deviation mainly stems from the subtle changes in the water flow obstruction effect caused by local soil texture differences. The overall error is controlled within an acceptable range, fully demonstrating the high accuracy and reliability of the technical solution in predicting the flood inundation range.

[0101] This embodiment details how risk prediction based on dynamic correction of underlying surface parameters is significantly better than risk prediction without correction of underlying surface parameters, thus significantly improving the accuracy of flood risk forecasting and early warning for valley-type tailings dams and providing more reliable support for the safe operation and disaster prevention of tailings dams.

[0102] The above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. For those skilled in the art, the present invention can have various modifications and variations. Any changes, modifications, substitutions, integrations, and parameter changes made to these embodiments within the spirit and principles of the present invention, without departing from the principles and spirit of the present invention, through conventional substitutions or to achieve the same function, fall within the scope of protection of the present invention.

Claims

1. A method for predicting flood runoff in valley-type tailings dams based on dynamic correction of underlying surface parameters, characterized in that, include: S1. Acquire topographic and geomorphological data, vegetation and soil types, land use data, meteorological and hydrological data, tailings dam parameters, and historical flood data of valley-type tailings dams. Preprocess the collected multi-source data to construct a basic database. S2. Obtain the underlying surface parameters of the valley-type tailings dam, including vegetation interception parameters, vegetation cover index, soil type, soil saturated hydraulic conductivity and impermeable area ratio, and determine the initial value range of each underlying surface parameter. S3. By considering vegetation growth factors, soil moisture, and the operation of tailings ponds, differentiated runoff and runoff parameters are set for different regions. A dynamic correction model for underlying surface parameters is established, and the preliminary underlying surface parameters are dynamically corrected based on a dynamic update mechanism of real-time remote sensing data. S4. Construct a coupled hydrological-hydrodynamic model, simulate the watershed runoff formation process and time-varying characteristics through multi-risk flood scenarios, and output predicted values ​​of runoff flow, peak flood time and peak flood level. Verify and calibrate the hydrological-hydrodynamic model by comparing it with measured values ​​of floodplains in historical flood events from satellite remote sensing images. S5. Input real-time meteorological and hydrological data, topography, land use and vegetation coverage data into the hydrological-hydrodynamic model. Through geospatial analysis methods, based on the predicted runoff, flood peak time and flood peak water level, predict the inundation range, inundation time and inundation degree of flood runoff, and realize the prediction of flood runoff of valley-type tailings dams. The dynamic correction model for the underlying surface parameters is established in S3 as follows: Obtain comprehensive indicators of tailings dam operation status This includes tailings dam water level, tailings dam dry beach length, and flood discharge capacity of flood control structures; comprehensive vegetation growth indicators were obtained. This includes the rate of change in vegetation cover and the amount of vegetation height growth; a dynamic correction model for underlying surface parameters is constructed using a multilayer sensing mechanism, with the model input being... The hidden layer uses the LeakyReLU activation function, with the following formula: ,in For the first The weight matrix of the layer, For bias vectors, For the first The output of the hidden layer after processing by the activation function is expressed by the following formula: ,in , The input values ​​for the activation function. The initial input vector is used; the output layer uses a linear activation function to output the correction amount of the underlying surface parameters. The formula is: , The total number of network layers. This is the output layer weight matrix. This is the output layer bias vector. For the first The output of the layer is determined by the correction amount of the underlying surface parameters. And original underlay parameters Construct a dynamic correction model for underlying surface parameters, the formula is as follows: ,in To obtain the corrected underlying surface parameters, the correction amount is calculated by continuously inputting real-time tailings dam operation status and vegetation growth data, and the underlying surface parameters are dynamically corrected.

2. The method for predicting flood runoff of valley-type tailings dams based on dynamic correction of underlying surface parameters according to claim 1, characterized in that, In step S1, the collected multi-source data is preprocessed. Through a multi-source data association feature mining filtering algorithm, the correlation between topography, vegetation and soil type, land use, meteorology and hydrology, tailings dam parameters and historical flood data is analyzed to filter out feature data that has an impact on flood risk prediction. Using the multi-source data fusion mechanism of graph neural network, different types of data are mapped to a unified feature space, and the relationship between feature data is captured and fused using graph structure. The fused data is standardized to give it a unified dimension and scale. The filtered, fused and standardized data are used to build a basic database according to the database architecture and indexing method.

3. The method for predicting flood runoff in valley-type tailings dams based on dynamic correction of underlying surface parameters according to claim 1, characterized in that, When dynamically correcting the initially determined underlying surface parameters using the established dynamic correction model, the specific steps are as follows: Using a Geographic Information System (GIS), the valley-type tailings dam basin is divided into multiple sub-regions with different topographic, vegetation, and soil characteristics. For each sub-region, a local parameter adjustment model based on fuzzy inference is constructed. The sub-region's topographic slope, aspect, and vegetation cover are used as input variables. Membership functions are used to determine the membership degrees of each input variable in different fuzzy sets. Fuzzy inference is then performed on the input variables to derive the adjustment coefficients for the runoff generation and confluence parameters of each sub-region. For the initially determined underlying surface parameters, the global correction amount obtained from the established underlying surface parameter dynamic correction model is combined with the local adjustment coefficients obtained from the adjustment coefficients of each sub-region to correct the initially determined underlying surface parameters. The formula is as follows: ,in For the first The underlying surface parameters at the first The corrected values ​​for each sub-region For the underlying surface parameters in the first... The initial values ​​determined for each sub-region The first obtained by the dynamic correction model of the underlying surface parameters Global correction amount for each underlying surface parameter For the first The local adjustment coefficients of each sub-region form the settings for differentiated runoff generation parameters in different regions, and dynamically correct the initially determined underlying surface parameters.

4. The method for predicting flood runoff in valley-type tailings dams based on dynamic correction of underlying surface parameters according to claim 3, characterized in that, For each sub-region, a local parameter adjustment model based on fuzzy inference is constructed, specifically as follows: Determine the slope of the sub-region Slope aspect and vegetation coverage The feature vectors are defined, multiple fuzzy sets are set, and Gaussian membership functions are used to define them. Calculate the membership degree of each input variable in different fuzzy sets, where For input variable values, The membership function center value, To determine the standard deviation, a fuzzy rule base is constructed. A fuzzy synthesis inference algorithm is then used to infer and calculate the fuzzy rules, yielding the adjustment coefficients for the flow generation and merging parameters of each sub-region. The calculation formula is: ,in To determine the number of fuzzy rules, For the first The weight of a fuzzy rule, For the first The comprehensive membership degree of the input variables in the fuzzy rules is used to construct a local parameter adjustment model.

5. The method for predicting flood runoff of valley-type tailings dams based on dynamic correction of underlying surface parameters according to claim 1, characterized in that, The hydrological-hydrodynamic model in S4 includes a distributed hydrological model and a flood propagation hydrodynamic model, which are coupled to construct the hydrological-hydrodynamic model. The distributed hydrological model simulates the runoff generation and confluence process of the watershed and outputs the flood process line at the tailings dam inlet. The flood propagation hydrodynamic model simulates the evolution of flood in the tailings dam area based on the modified topography and inflow boundary, and calculates the inundation depth and flow velocity. The hydro-hydrodynamic model simulates multiple risk scenarios, including local dam failure and vegetation degradation, to obtain simulated values ​​of runoff generation and confluence processes in the watershed. By inverting historical flood events, the simulated values ​​are compared with the measured values ​​to verify and calibrate the hydro-hydrodynamic model.

6. The method for predicting flood runoff in valley-type tailings dams based on dynamic correction of underlying surface parameters according to claim 5, characterized in that, The distributed hydrological model simulates the runoff generation and confluence process of the watershed and outputs the flood hydrograph at the tailings dam inlet, specifically: The watershed is divided into grids, with each grid cell... Corresponding underlying surface parameters A runoff generation model was constructed, utilizing a deep residual shrinkage network to analyze meteorological and hydrological data, topographic data, and underlying surface parameters of the grid cells. Perform feature extraction to obtain feature vectors The formula is: ,in For rainfall, For evaporation, Given the slope, runoff generation was calculated using an infiltration model, taking into account the impact of vegetation interception on infiltration, and the corrected infiltration rate was calculated. The formula is: ,in For the duration of rainfall, For grid cells The formula for the water accumulation delay time is: , For the vegetation interception correction factor, obtain the first... Flow rate of grid cells The formula is: ,in For grid cells The area is calculated by summing the flow rates of all grid cells flowing into the tailings dam inlet and subtracting the discharge flow rate from the combined drainage wells of the flood discharge system that flows into the main tunnel. The discharge flow rate of the main tunnel is calculated as pressurized flow. The formula is: , ,in It is a resistance-corrected flow coefficient. For water flow section, The calculated head at the tunnel inlet. The tunnel has a longitudinal slope. The length of the tunnel. The height of the tunnel. For the Xie Cai coefficient, The drag coefficient, Given the local gravitational acceleration, the formula for outputting the flood hydrograph at the tailings dam inlet is: ,in The number of grid cells flowing into the tailings dam inlet. For the first The flow rate of each grid cell flowing into the tailings dam inlet. For the first The main tunnel discharge flow rate is associated with each grid cell flowing into the tailings dam inlet.

7. The method for predicting flood runoff in valley-type tailings dams based on dynamic correction of underlying surface parameters according to claim 5, characterized in that, The flood propagation hydrodynamic model, based on modified topography and inflow boundary, simulates the evolution of floodwaters in the tailings dam area, and calculates the inundation depth and flow velocity, specifically: In simulating flood evolution, a multi-scale spatiotemporal grid division technique is used to dynamically refine and coarsen the grid based on topographic changes and water flow velocity, thereby establishing the water flow motion equations: ,in Because of the water depth, , They are respectively , Flow velocity component in the direction, for directional unit width flow, for directional unit width flow, , The inflow and outflow rates of the grid cells are respectively used. Through iterative solutions, combined with corrected topography and inflow boundary conditions, the evolution of the flood in the tailings dam area is simulated; the current water depth is used as a reference. Subtract terrain elevation Calculate the flood depth The formula is: ; obtained by solving the equation , Flow velocity component in direction , To obtain the submerged flow velocity The formula is: .

8. The method for predicting flood runoff of valley-type tailings dams based on dynamic correction of underlying surface parameters according to claim 5, characterized in that, The hydro-hydrodynamic model simulates multiple risk scenarios to obtain runoff generation and confluence processes in the watershed, outputs simulated values, and verifies and calibrates the hydro-hydrodynamic model by comparing simulated and measured values ​​through historical flood event inversion. Specifically: The runoff generation and confluence processes of the watershed are obtained through a hydro-hydrodynamic model. The tailings dam inlet flood hydrograph output by the WRF-Hydro distributed hydrological model in the hydro-hydrodynamic model is used to simulate the inflow boundary of the flood propagation hydrodynamic model. The flood propagation hydrodynamic model simulates flood evolution and outputs simulated runoff. Peak flood time and flood peak water level Through inversion verification of historical flood events, measured runoff is obtained from historical flood events. Peak flood time Flood peak water level Construct an error evaluation function ,in , , The weights of each indicator are determined based on their importance to flood risk assessment. The parameters in the hydrological-hydrodynamic model are adjusted to optimize the comprehensive error assessment function. Minimum, enabling the verification and calibration of hydrological-hydrodynamic models.

9. The method for predicting flood runoff of valley-type tailings dams based on dynamic correction of underlying surface parameters according to claim 1, characterized in that, In S5, based on the predicted runoff flow... Peak flood time and flood peak water level The extent and duration of flooding can be determined by reversing these parameters, thus establishing the degree of inundation. Constructing terrain surfaces using digital elevation model data ,in Using geographical coordinates, the flood inundation area is determined by the following formula: in, For point The cross-sectional area of ​​the water passage, To preset the water level difference, ,point Within the flood zone, At time, point Those not submerged were identified by coordinating data points to determine the extent of flooding; the timing of the flood peak was also considered. The formula for determining the flooding time is: ,in, For point The time of submersion, This represents the shortest water flow path distance from this point to the starting point of the flood peak. This is a correction factor for water flow velocity. The average water flow velocity is used; the degree of inundation is determined by the inundation range and inundation time, using the following formula: in, To determine the degree of flooding, To submerge the water level, and These are the highest and lowest water levels in history. and These represent the longest and shortest flooding times in history, respectively. This is the highest runoff volume in history. , and These are the corresponding weighting coefficients.

Images