Meteorological-hydrological-hydrodynamic force coupled small watershed mountain torrent disaster forecasting method

Abstract

The invention discloses a meteorological-hydrological-hydrodynamic coupling small watershed mountain torrent disaster forecasting method, which comprises the following steps of: constructing a disaster-inducing factor and disaster-pregnancy environment multi-dimensional quantitative index system, combining a three-dimensional water system model and a smoothed particle hydrodynamics (SPH) hydrodynamic model, constructing a disaster forecasting ontology and extracting feature data; according to the method, features and target variables are optimized through machine learning and deep learning, the extracted features are trained, and a disaster scene is constructed based on historical mountain torrent disaster data, so that a chain relation of hydro meteorological data-hydrodynamic characteristics-water level flow characterization-disaster conditions is established. According to the method, through fine hydrodynamic simulation, sediment deposition and impurity influences are accurately considered, detailed flood flowing path and speed prediction is provided in combination with a meteorological-hydrological model, high-precision and real-time mountain torrent risk prediction and construction of various disaster scenes are achieved, and the accuracy and response speed of mountain torrent prediction are effectively improved.

Description

Technical Field

[0001] The invention relates to the field of hydrological prediction and hydrodynamic simulation, and in particular to a meteorological-hydrological-hydrodynamic coupling small watershed flash flood disaster forecasting method. Background Art

[0002] Flash floods are common and extremely destructive natural disasters in mountainous areas. They occur frequently, develop rapidly, and are difficult to warn, posing a serious threat to human life and property safety and social and economic development. Traditional flash flood warning methods mainly rely on meteorological data (such as rainfall) and hydrological data (such as river water level, flow, etc.) for prediction, while taking into account the topography and underlying surface factors of the basin.

[0003] Statistical methods analyze historical meteorological, hydrological data and flash flood records to establish statistical models for prediction. This type of method is simple to operate and has fast calculation speed, but the prediction accuracy is limited by the representativeness of historical data. It is difficult to adapt to climate change and complex geographical environment, and it cannot capture the complex nonlinear relationship of flash flood disasters. In recent years, machine learning methods have been widely used in flash flood warning. By learning a large amount of meteorological and hydrological data, complex nonlinear relationships are established to improve prediction accuracy. This type of method has overcome the shortcomings of traditional methods to a certain extent, but its performance is highly dependent on the quality and quantity of training data, and the interpretability of the model is poor, making it difficult to provide an intuitive understanding of the physical process. The prediction ability of pure data-driven models may be significantly reduced when facing extreme meteorological conditions and unknown environmental changes, limiting their application in dynamically changing environments. The hydrodynamic model simulates the hydraulic characteristics of the river under different flow rates and water levels based on the principles of fluid mechanics, and can accurately predict the flow path and inundation range of floods. However, traditional physical models cannot directly incorporate hydrological and meteorological elements such as rainfall, runoff, and soil permeability. They have limitations in the initial data import and parameter calibration of complex river topography and riverbed characteristics, which increases the difficulty and cost of practical applications. Summary of the invention

[0004] In order to solve the above problems, the present invention proposes a meteorological-hydrological-hydrodynamic coupling small watershed flash flood disaster forecasting method, combining the SPH hydrodynamic model and the machine learning algorithm, to construct a chain relationship of hydrological and meteorological data-hydrodynamic characteristics-water level and flow characterization-disaster situation (flooding range, flooding depth and flash flood propagation speed). This method organically combines meteorological data, hydrological data and hydrodynamic characteristics, comprehensively considers multiple factors such as rainfall, hydrological monitoring data, terrain elevation, river section, riverbed characteristics, etc., combines hydrodynamic coupling mechanism modeling and machine learning algorithms, and simulates the occurrence and development process of flash flood disasters more comprehensively and accurately. Compared with ordinary machine learning algorithms, the meteorological-hydrological-hydrodynamic coupling method not only improves the prediction accuracy, but also enhances the interpretability and robustness of the model, ensuring excellent prediction performance in complex river environments and changeable meteorological conditions, and improving the reliability and practicality of the early warning system.

[0005] The present invention is realized by the following technical scheme: a meteorological-hydrological-hydrodynamic coupling small watershed flash flood disaster forecasting method, comprising the following steps: A meteorological-hydrological-hydrodynamic coupling small watershed flash flood disaster forecasting method comprises the following steps: Step S1, comprehensively collect multi-source data and obtain three-dimensional spatial data through satellite images; Step S2, cleaning the data to remove noise and missing values, and performing normalization and standardization to unify the data scale; Step S3, selecting and quantifying the main disaster-causing factors and disaster-pregnant environmental factors, comprehensively analyzing and constructing a multi-dimensional quantitative indicator system; Step S4, constructing a three-dimensional water system model and a three-dimensional river hydrodynamic model, integrating terrain data and dividing the water system into segments; Step S5, using the smooth particle method SPH to construct a hydrodynamic model of segmented nodes; combining hydrological data and meteorological data to simulate the movement trajectory of floods under different scenarios; Step S6, after running the SPH model, analyze the key indicators of the flood and extract and analyze the water flow characteristic indicators of each section of the water system, select the corresponding characteristic data as the secondary collection data, and integrate the data into the prediction model input variables after data preprocessing; Step S7, using the water flow characteristic index obtained from the SPH model as the target variable, optimizing the features and the target variable and training the extracted features, and constructing a chain relationship of hydrological and meteorological data-hydrodynamic characteristics-water level and flow representation-disaster situation; Step S8, verifying the model and optimizing the algorithm to obtain a trained comprehensive model; Step S9, deploy the trained comprehensive model to the real-time data processing system to forecast and issue warning information in real time.

[0006] Furthermore, the step S1 includes the following sub-steps: Step S1-1, collecting meteorological data, including but not limited to rainfall, rainfall intensity, wind speed, and wind direction, to obtain meteorological conditions that affect the occurrence of flash floods; Step S1-2, collecting hydrological data, including but not limited to river water level, water flow velocity, and soil moisture, for analyzing hydrological processes and their impact on floods; Step S1-3, collecting terrain data, including but not limited to terrain elevation, slope, and vegetation coverage, for analyzing the terrain characteristics and vegetation conditions of the watershed; Step S1-4, collecting historical flash flood disaster data, including but not limited to the time, location, intensity, and scope of disaster impact, to establish a historical basis for the occurrence of disasters; Step S1-5, collecting watershed water system maps and riverbed maps through satellite images, obtaining river channel morphology, sediment distribution and impurity information, and providing spatial data support for the construction of a three-dimensional water system model.

[0007] Furthermore, the step S2 includes the following sub-steps: Step S2-1, cleaning the collected data to remove noise and missing values; Step S2-2, performing data normalization processing to unify the data scale; Step S2-3, smoothing the time series data to eliminate short-term fluctuations; Step S2-4, standardize the riverbed morphology and sediment distribution data to ensure the input data quality of the hydrodynamic model.

[0008] Furthermore, the step S3 includes the following sub-steps: Step S3-1, selecting major disaster-causing factors, including but not limited to rainfall, rainfall intensity, rainfall duration, temperature, soil moisture and vegetation cover, and identifying their relationship with flash flood disasters through quantitative analysis; combining historical meteorological data from different regions to fit the intensity, frequency and spatial distribution model of rainfall; Step S3-2, selecting disaster-prone environmental factors in the watershed, including but not limited to terrain elevation, slope, soil type, vegetation coverage, and underlying surface influencing factors of the watershed; Step S3-3, construct a multidimensional quantitative indicator system based on the multivariate analysis of meteorological factors and disaster-prone environmental factors.

[0009] Furthermore, the step S4 includes the following sub-steps: Step S4-1, using high-resolution satellite image data, extracting spatial information of the watershed system, obtaining the precise shape and distribution of the river channel, and generating a watershed system map; Step S4-2, based on the watershed system map, a data base in tiff format is generated, and data information is imported: digital terrain elevation, river section distribution, riverbed topography, riverbank embankment engineering data, sediment distribution and impurity accumulation, to generate a three-dimensional model of the watershed system, which fully reflects the three-dimensional structure of the river; Step S4-3, integrating terrain data to ensure the accuracy and completeness of the 3D model. The model needs to be further refined for complex riverbed areas; Step S4-4, water system division and node definition, specifically: divide the water system in the basin into n sections, ensure that each section can reflect the actual terrain, basin structure and hydrodynamic characteristics; specifically, it should be segmented according to the length of the river channel, the slope of the basin, and the terrain undulation to adapt to the water flow conditions in different areas; preferably, when the water system is segmented, adaptive segmentation is performed according to the hydrodynamic characteristics of different areas in the basin; n nodes are set in each water system segment to describe key variables: changes in water flow, flow rate and water level; nodes are set at key positions of the river channel, including but not limited to river channel bends and confluences.

[0010] Furthermore, the step S5 includes the following sub-steps: Step S5-1, setting boundary conditions and initial conditions for each section of the water system, specifically: Step S5-1a, according to the meteorological data and hydrological data, set appropriate boundary conditions for the start and end of each water system segment; Step S5-1b, setting the initial water level state and flow rate state according to historical data; Step S5-1c, perform particle initialization on the segmented river model, divide the three-dimensional water system model into several particles, each particle represents a fluid unit in the water body, and assign physical properties to each particle: initial position, velocity, mass and density; particles are used to represent fluids, and each particle carries the physical quantities of the fluid: mass and velocity. The particles simulate the fluid movement through interaction forces, thereby simulating the flow and propagation process of floods; ensure that the distribution of particles can be refined to local areas; Step S5-2, selecting a kernel function, selecting a kernel function suitable for flash flood simulation, used to define the interaction range between particles, and determining the support radius of the kernel function to balance calculation accuracy and efficiency; Step S5-3, perform density calculation, for each particle, calculate its local density, the formula is as follows: ; in, For particles The density of For particles Quality, is the kernel function, is the position of the ith particle, is the position of the jth particle, is the nuclear radius; Step S5-4, perform pressure calculation, and use the state equation to calculate the pressure of each particle. The formula is as follows: ; in, For particles pressure, is the speed of sound, is the reference density; Step S5-5, perform force calculation to calculate the pressure between particles and viscosity , the formula is as follows: ; ; in, is the viscosity coefficient, and For particles and speed, is the pressure of the ith particle, is the pressure of the jth particle; Step S5-6, performing external force and gravity calculation to calculate the external force on the particle; Step S5-7, calculate the total force and acceleration, combine all forces, and calculate the total acceleration of the particle. The formula is as follows: ; in, is the gravitational force acting on the i-th particle; is the mass of the ith particle; Step S5-8, perform time integration, and use the time stepping method to update the velocity and position of the particle. The formula is as follows: ; ; in, is the velocity of the ith particle at time t, is the acceleration of the ith particle at time t, is the time position, is the time step; Step S5-9, performing boundary condition processing to prevent particles from crossing solid boundaries, and using mirror particles or reflection force methods to process boundary interactions.

[0011] Furthermore, the step S6 includes the following sub-steps: Step S6-1, running the smooth particle method SPH model to calculate the water flow trajectory under different rainfall intensities, durations and terrain conditions; simulating the propagation process of flood waves, and analyzing the impact of the interaction between the propagation path of flood waves and the terrain and river channels on flood flow through the model output; Step S6-2, based on the data output by the SPH model, calculate the propagation mode of mountain torrents under different rainfall intensities, durations and terrain conditions, and analyze the key indicators of floods: expansion speed, overflow locations, and river flooding areas; extract and analyze the flow characteristics of each section of the water system, including key indicators: maximum water level, maximum flow, flood propagation speed, and expected flooding area; Step S6-3, using the key indicator parameters derived in step S6-2 as secondary collection data, performing noise reduction, normalization and standardization on the data; Step S6-4, using a feature selection algorithm to screen the pre-processed secondary collection data and select the feature value with the greatest predictive ability.

[0012] Furthermore, the step S7 includes the following sub-steps: Step S7-1, based on statistical methods or machine learning algorithms, establish a nonlinear relationship model between each disaster-causing factor and disaster-pregnant environmental factor in the index system in step S3-3 and the characteristic value in step S6-4, screen the factors that have the greatest impact on the occurrence of flash floods through a feature selection algorithm, quantify each factor, and calculate the sensitivity of each disaster-causing factor and disaster-pregnant environmental factor; Step S7-2: using the flooding range, flooding depth and propagation speed obtained from the SPH model as target variables, optimizing features and target variables through machine learning and deep learning, and training the extracted features to establish a multi-factor multi-variable prediction model; Step S7-3, constructing a chain relationship of hydrological and meteorological data - hydrodynamic characteristics - water level and flow representation - disaster situation; Step S7-4, by analyzing the rainfall, flow changes and disaster impacts in historical data, establish the flash flood disaster occurrence mechanism under different scenarios; Step S7-5, using the generated disaster scenario, simulate the flash flood response under different rainfall intensities and durations, and provide multiple disaster scenario predictions.

[0013] Furthermore, the step S8 includes the following sub-steps: Step S8-1, using the cross-validation method to optimize the model parameters, prevent overfitting, and ensure the generalization ability of the model in different watershed environments; Step S8-2, improving the stability and accuracy of the model through an integrated learning method; Step S8-3, achieving synchronous prediction of multiple target variables to ensure consistency and relevance of prediction results; Step S8-4, using an independent test data set to verify the prediction performance of the model, and evaluating the accuracy and applicability of the model by comparing the prediction results with the actual observation data; Step S8-5, adjust the model structure and parameters according to the verification results to optimize the combination of hydrodynamic simulation and machine learning algorithm.

[0014] Furthermore, the step S9 includes the following sub-steps: Step S9-1, deploying the trained comprehensive model to a real-time data processing system, wherein the model can receive and process the latest data input in real time; Step S9-2, receiving real-time data, combining the output of the hydrodynamic model, conducting flash flood risk assessment and forecasting, and timely updating the current flood dynamic forecast results; Step S9-3, combined with the SPH model, realizes the three-dimensional visualization of flood evolution, and conveys the warning information through the simulation results through visualization technology.

[0015] The beneficial effects of the present invention are as follows: Based on the combination of meteorological and hydrological information and hydrodynamic models, the present invention uses historical disaster situations and real-time monitoring data for continuous iterative calibration. On the one hand, it gives full play to the advantages of high-precision simulation and interpretability of the physical model for the flood propagation process. On the other hand, it uses machine learning algorithms to deal with complex nonlinear relationships and extreme conditions, thus forming a more robust and timely flash flood prediction system. At the same time, by establishing a situational flash flood disaster scenario library, it provides emergency departments with accurate simulation predictions under different rainfall types, river conditions and land use changes, significantly improving the accuracy, rapid response capability and practicality of flash flood forecasts in small watersheds in mountainous areas. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 The present invention is an overall flow chart of a method for predicting flash flood disasters in a small watershed based on meteorological-hydrological-hydrodynamic coupling. DETAILED DESCRIPTION

[0017] Exemplary embodiments will now be described in detail, examples of which are shown in the accompanying drawings. In the following description and drawings, unless otherwise indicated, the same numerals in different drawings represent the same or similar elements.

[0018] The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Instead, they are merely examples of devices consistent with some aspects of the present invention as detailed in the appended claims. Each embodiment of this specification is described in a progressive manner.

[0019] like Figure 1 As shown, the following technical solution is proposed. According to a meteorological-hydrological-hydrodynamic coupling small watershed flash flood disaster forecasting method provided in this application, the following steps are included: Step S1, comprehensively collect multi-source data, including obtaining meteorological data such as rainfall, rainfall intensity, temperature, humidity, and hydrological data such as river water level, water flow velocity, and flow, and obtaining three-dimensional spatial data such as river basin water system, terrain elevation, slope, etc. through satellite images; Step S2, cleaning the data to remove noise and missing values, normalizing and standardizing the data to unify the data scale, so as to ensure the quality of the input data of the hydrodynamic model; Step S3, selecting and quantifying the main disaster-causing factors and disaster-pregnant environmental factors that affect the occurrence of flash floods, and constructing a multidimensional quantitative indicator system based on the comprehensive analysis of these multiple factors and environmental elements; Step S4, constructing a three-dimensional water system model, establishing a three-dimensional water dynamic model of the river based on digital terrain elevation, watershed water system map, river section data, riverbed topographic map, and riverbank embankment engineering data, integrating terrain data, and dividing the water system into sections; Step S5, using the smooth particle method (SPH) to construct a hydrodynamic model of segmented nodes to simulate the movement, propagation and overflow of water during flash floods. Combining hydrological data (such as water level and flow in the basin) and meteorological data (such as rainfall and intensity), the movement trajectory of floods under different scenarios is simulated to reflect the generation and evolution of flash floods; Step S6, after running the SPH model, analyze the key indicators such as the flood expansion speed, possible overflow location, and river inundation area, extract and analyze the water flow characteristics of each section of the water system, such as the maximum water level, maximum flow, flood propagation speed, and expected inundation area, so as to quantify the response of disaster-causing factors and disaster-prone environments in different scenarios, select the corresponding characteristic data as the secondary collection data, and integrate them into the prediction model input variables after data preprocessing; Step S7, the flooding range, flooding depth and propagation speed obtained from the SPH model are used as target variables, the features and target variables are optimized through machine learning and deep learning, and the extracted features are trained to construct a chain relationship of hydrological and meteorological data-hydrodynamic characteristics-water level and flow representation-disaster situation (flooding range, flooding depth and flash flood propagation speed); Step S8, model verification and algorithm optimization to ensure the reliability and accuracy of the prediction model; Step S9, real-time forecasting and release, realize the practical application of the model and provide early warning information.

[0020] Furthermore, the step S1 is specifically as follows: Step S1-1, collecting meteorological data, including rainfall, rainfall intensity, wind speed, wind direction, etc., to obtain meteorological conditions that affect the occurrence of flash floods; Step S1-2, collecting hydrological data, including river water level, water flow velocity, soil moisture, etc., for analyzing hydrological processes and their impact on floods; Step S1-3, collecting terrain data, including terrain elevation, slope, vegetation coverage, etc., to understand the terrain characteristics and vegetation conditions of the watershed; Step S1-4, collecting historical flash flood disaster data, including the time, location, intensity, and scope of disaster impact, to establish a historical basis for the occurrence of disasters; Step S1-5, collecting watershed water system maps and riverbed maps through satellite images, obtaining river channel morphology, sediment distribution and impurity information, and providing spatial data support for the construction of a three-dimensional water system model.

[0021] Furthermore, the step S2 is specifically as follows: Step S2-1, clean the collected data to remove noise and missing values ​​to ensure the integrity and accuracy of the data; Step S2-2, perform data normalization processing to unify the data scale to facilitate comparison and fusion between different data sources; Step S2-3, smoothing the time series data to eliminate short-term fluctuations and improve data stability; Step S2-4, standardize the riverbed morphology and sediment distribution data to ensure the input data quality of the hydrodynamic model.

[0022] Furthermore, the step S3 is specifically as follows: Step S3-1, select the main disaster-causing factors, such as rainfall, rainfall intensity, rainfall duration, temperature, soil moisture, vegetation cover, etc., and identify the relationship with the occurrence of flash floods through quantitative analysis; combine the historical meteorological data of different regions to fit the intensity, frequency and spatial distribution model of rainfall; Step S3-2, selecting disaster-prone environmental factors in the watershed, including terrain elevation, slope, soil type, vegetation coverage, etc., and considering the underlying surface influencing factors of the watershed; Step S3-3, constructing a multidimensional quantitative indicator system based on multivariate analysis of meteorological factors and disaster-prone environmental factors; Furthermore, the step S4 is specifically as follows: Step S4-1, using high-resolution satellite image data, extracting spatial information of the watershed system, obtaining the precise shape and distribution of the river channel, and generating a watershed system map; Step S4-2, based on the watershed water system map, a data base in tiff format is generated, and data information such as digital terrain elevation, river section distribution, riverbed topography, riverbank embankment engineering data, sediment distribution and impurity accumulation are imported to generate a three-dimensional model of the watershed water system, which fully reflects the three-dimensional structure of the river channel; Step S4-3, integrating terrain data to ensure the accuracy and completeness of the three-dimensional model, with special attention paid to the details of complex riverbed areas to improve the refinement of the model; Step S4-4, water system division and node definition, specifically: Divide the watershed system into n sections to ensure that each section can reflect the actual topography, watershed structure and hydrodynamic characteristics; specifically, the sections should be divided according to factors such as river length, watershed slope, and topographic undulation to adapt to the water flow conditions in different areas; preferably, when dividing the watershed system, adaptive segmentation can be performed according to the hydrodynamic characteristics of different areas in the watershed to improve the calculation accuracy and stability of the model; n nodes are set in each river system segment to describe key variables such as water flow changes, flow velocity and water level. Nodes can be set at key locations of the river, such as river bends, water confluences, etc.

[0023] Furthermore, the step S5 is specifically as follows: Step S5-1, setting boundary conditions and initial conditions for each section of the water system, specifically: Step S5-1a, according to meteorological data (rainfall, intensity) and hydrological data (water level, flow, etc.), set appropriate boundary conditions for the start and end of each water system segment; Step S5-1b, setting the initial water level, flow rate and other conditions according to historical data; Step S5-1c, perform particle initialization on the segmented river model, divide the three-dimensional water system model into several particles, each particle represents a fluid unit in the water body, and assign physical properties such as initial position, velocity, mass and density to each particle; particles are used to represent fluids, each particle carries physical quantities such as mass and velocity of the fluid, and particles interact with each other to simulate fluid motion, thereby simulating the flow and propagation of floods; ensure that the distribution of particles can be refined to local areas (such as river bends and complex terrain areas) to improve the adaptability of the model under different basin conditions; Step S5-2, selecting a kernel function, selecting a kernel function suitable for flash flood simulation, such as a cubic B-spline kernel function, to define the interaction range between particles and determine the support radius of the kernel function to balance calculation accuracy and efficiency; Step S5-3, perform density calculation, for each particle, calculate its local density, the formula is as follows:

[0024] in, For particles The density of For particles Quality, is the kernel function, is the position of the ith particle, is the position of the jth particle, is the nuclear radius; Step S5-4, perform pressure calculation, and use the state equation to calculate the pressure of each particle. The formula is as follows:

[0025] in, For particles pressure, is the speed of sound, is the reference density; Step S5-5, perform force calculation to calculate the pressure between particles and viscosity , the formula is as follows:

[0026]

[0027] in, is the viscosity coefficient, and For particles and speed, is the pressure of the ith particle, is the pressure of the jth particle; Step S5-6, performing external force and gravity calculation to calculate the external force on the particle; Step S5-7, calculate the total force and acceleration, combine all forces, and calculate the total acceleration of the particle. The formula is as follows: ; is the gravitational force acting on the i-th particle; is the mass of the ith particle; Step S5-8, perform time integration, and use a time stepping method (such as Leapfrog or Verlet integration) to update the velocity and position of the particle. The formula is as follows: ; ; in, is the velocity of the ith particle at time t, is the acceleration of the ith particle at time t, is the time position, is the time step; Step S5-9, performing boundary condition processing to realize boundary conditions such as riverbanks and levees, preventing particles from crossing solid boundaries, and using mirror particles or reflection force methods to process boundary interactions; Furthermore, SPH modeling was performed for each section of the water system, the outlet boundary of the previous section of the river was used as the inlet boundary of the next section, and the nodes at the junction of tributaries were modeled separately to ensure the continuity and accuracy of the hydrodynamic model.

[0028] Furthermore, step S6 quantifies the response characteristics of disaster-causing factors and disaster-prone environments in different scenarios by simulating the propagation and dynamic changes of floods, specifically: Step S6-1, run the smooth particle method (SPH) model to calculate the water flow trajectory under different rainfall intensities, durations and terrain conditions; simulate the propagation process of flood waves, considering the coupling effects of multiple factors such as precipitation, water level in the basin, terrain and soil moisture; analyze the impact of the propagation path of flood waves and the interaction between terrain and river channels on flood flow through model output; Step S6-2, based on the data output by the SPH model, calculate the propagation mode of mountain torrents under different rainfall intensities, durations and terrain conditions, analyze key indicators such as flood expansion speed, possible overflow locations, and river flooding areas; extract and analyze the flow characteristics of each section of the water system, such as the maximum water level, maximum flow, flood propagation speed, and expected flooding area; in addition to the flood propagation speed and flooding depth, modeling of factors such as sediment movement, reflection and refraction of flood waves, etc. can also be considered; Step S6-3, using the key indicator parameters derived in step S6-2 as secondary collection data, performing noise reduction, normalization and standardization on the data to ensure that these parameters can be effectively used in subsequent feature extraction and selection steps; Step S6-4, using feature selection algorithms (such as principal component analysis, genetic algorithm) to screen the pre-processed secondary collection data, select the most predictive features, such as the spatiotemporal distribution of extreme rainfall events, river water level and flow change curves, riverbed evolution curves, etc.

[0029] Furthermore, the step S7 is specifically as follows: Step S7-1, based on statistical methods or machine learning algorithms (such as regression analysis), establish a nonlinear relationship model between each disaster-causing factor and disaster-pregnant environmental factor in the index system in step S3-3 and the characteristic value in step S6-4, screen the factors that have the greatest impact on the occurrence of flash floods through a feature selection algorithm, quantify each factor, and calculate the sensitivity of each disaster-causing factor and disaster-pregnant environmental factor; Step S7-2, the flooding range, flooding depth and propagation speed obtained from the SPH model are used as target variables, the features and target variables are optimized through machine learning and deep learning (such as heuristic search, evolutionary strategy and Bayesian optimization), and the extracted features are trained to establish a multi-factor multi-variable prediction model; Preferably, multiple machine learning algorithms can be used for training and comparison respectively to select the most suitable model for the modeled watershed; Step S7-3, constructing a chain relationship of hydrological and meteorological data - hydrodynamic characteristics - water level and flow representation - disaster conditions (flooding range, flooding depth and flash flood propagation speed); Step S7-4, by analyzing the rainfall, flow changes and disaster impacts in historical data, establish the flash flood disaster occurrence mechanism under different scenarios; Step S7-5, using the generated disaster scenario, simulates flash flood responses under different rainfall intensities and durations, and provides multiple possible disaster scenario predictions.

[0030] Furthermore, the step S8 is specifically as follows: Step S8-1, using cross-validation and other methods to optimize model parameters, prevent overfitting, ensure the generalization ability of the model in different watershed environments, and improve the stability and reliability of the model; Step S8-2, further improving the stability and accuracy of the model through integrated learning methods such as Bagging and Boosting; Preferably, bagging is used to train multiple base learners through multiple sampling with replacement, and their prediction results are averaged or voted to reduce the variance of the model; Preferably, Boosting is used to reduce the bias of the model by gradually training the base learners, focusing on the wrong predictions of the previous learners; Furthermore, combining these two methods simultaneously can fully utilize the advantages of multiple base learners, reducing both the variance and bias of the model, thereby improving the generalization and robustness of the model, ensuring that it maintains excellent flash flood disaster prediction performance under different river basin environments and changeable meteorological conditions.

[0031] Step S8-3 realizes the synchronous prediction of multiple objective variables, ensures the consistency and relevance of the prediction results, and improves the model's adaptability to complex flood situations.

[0032] In step S8-4, an independent test data set is used to verify the prediction performance of the model, especially its performance in complex river environments. The accuracy and applicability of the model are evaluated by comparing the prediction results with the actual observation data.

[0033] In step S8-5, the model structure and parameters are adjusted according to the verification results, the combination of hydrodynamic simulation and machine learning algorithm is optimized, the overall prediction ability of the comprehensive model is improved, and its effectiveness in practical applications is ensured.

[0034] Furthermore, the step S9 is specifically as follows: Step S9-1, deploying the trained comprehensive model to the real-time data processing system to ensure that the model can receive and process the latest data input in real time; Step S9-2, receiving real-time data, combining it with the output of the hydrodynamic model, conducting flash flood risk assessment and forecasting, and updating the forecast results in a timely manner to reflect the current flood dynamics; Step S9-3, combining the SPH model to realize the three-dimensional visualization of flood evolution, and intuitively present the simulation results through visualization technology to help relevant departments and the public better understand and respond to flash flood disasters; Step S9-4, release forecast information through various channels, inform relevant departments and the public in a timely manner, and ensure accurate communication of warning information.

[0035] The preferred embodiments of the present invention are described in detail above; however, the present invention is not limited to the specific details in the above embodiments. Within the technical concept of the present invention, various equivalent transformations can be made to the technical solutions of the present invention, and these equivalent transformations all belong to the protection scope of the present invention.

Claims

1. A meteorological-hydrological-hydrodynamic coupling small watershed flash flood disaster forecasting method, characterized in that: The steps include: Step S1, comprehensively collect multi-source data and obtain three-dimensional spatial data through satellite images; Step S2, cleaning the data to remove noise and missing values, and performing normalization and standardization to unify the data scale; Step S3, selecting and quantifying the main disaster-causing factors and disaster-pregnant environmental factors, comprehensively analyzing and constructing a multi-dimensional quantitative indicator system; Step S4, constructing a three-dimensional water system model and a three-dimensional river hydrodynamic model, integrating terrain data and dividing the water system into segments; Step S5, using the smooth particle method SPH to construct a hydrodynamic model of segmented nodes; combining hydrological data and meteorological data to simulate the movement trajectory of floods under different scenarios; Step S6, after running the SPH model, analyze the key indicators of the flood and extract and analyze the water flow characteristic indicators of each section of the water system, select the corresponding characteristic data as the secondary collection data, and integrate the data into the prediction model input variables after data preprocessing; Step S7, using the water flow characteristic index obtained from the SPH model as the target variable, optimizing the features and the target variable and training the extracted features, and constructing a chain relationship of hydrological and meteorological data-hydrodynamic characteristics-water level and flow representation-disaster situation; Step S8, verifying the model and optimizing the algorithm to obtain a trained comprehensive model; Step S9, deploy the trained comprehensive model to the real-time data processing system to forecast and issue warning information in real time.

2. The meteorological-hydrological-hydrodynamic coupling small watershed flash flood disaster forecasting method according to claim 1 is characterized in that: The step S1 includes the following sub-steps: Step S1-1, collecting meteorological data, including but not limited to rainfall, rainfall intensity, wind speed, and wind direction, to obtain meteorological conditions that affect the occurrence of flash floods; Step S1-2, collecting hydrological data, including but not limited to river water level, water flow velocity, and soil moisture, for analyzing hydrological processes and their impact on floods; Step S1-3, collecting terrain data, including but not limited to terrain elevation, slope, and vegetation coverage, for analyzing the terrain characteristics and vegetation conditions of the watershed; Step S1-4, collecting historical flash flood disaster data, including but not limited to the time, location, intensity, and scope of disaster impact, to establish a historical basis for the occurrence of disasters; Step S1-5, collecting watershed water system maps and riverbed maps through satellite images, obtaining river channel morphology, sediment distribution and impurity information, and providing spatial data support for the construction of a three-dimensional water system model.

3. The meteorological-hydrological-hydrodynamic coupling small watershed flash flood disaster forecasting method according to claim 1 is characterized in that: The step S2 includes the following sub-steps: Step S2-1, cleaning the collected data to remove noise and missing values; Step S2-2, performing data normalization processing to unify the data scale; Step S2-3, smoothing the time series data to eliminate short-term fluctuations; Step S2-4, standardize the riverbed morphology and sediment distribution data to ensure the input data quality of the hydrodynamic model.

4. The meteorological-hydrological-hydrodynamic coupling small watershed flash flood disaster forecasting method according to claim 1 is characterized in that: The step S3 includes the following sub-steps: Step S3-1, selecting major disaster-causing factors, including but not limited to rainfall, rainfall intensity, rainfall duration, temperature, soil moisture and vegetation cover, and identifying their relationship with flash flood disasters through quantitative analysis; combining historical meteorological data from different regions to fit the intensity, frequency and spatial distribution model of rainfall; Step S3-2, selecting disaster-prone environmental factors in the watershed, including but not limited to terrain elevation, slope, soil type, vegetation coverage, and underlying surface influencing factors of the watershed; Step S3-3, construct a multidimensional quantitative indicator system based on the multivariate analysis of the disaster-causing factors and disaster-pregnant environmental factors in the indicator system.

5. The meteorological-hydrological-hydrodynamic coupling small watershed flash flood disaster forecasting method according to claim 1 is characterized in that: The step S4 includes the following sub-steps: Step S4-1, using high-resolution satellite image data to extract spatial information of the watershed system, obtain the precise shape and distribution of the river channel, and generate a watershed system map; Step S4-2, based on the watershed system map, a data base in tiff format is generated, and data information is imported: digital terrain elevation, river section distribution, riverbed topography, riverbank embankment engineering data, sediment distribution and impurity accumulation, to generate a three-dimensional model of the watershed system, which fully reflects the three-dimensional structure of the river; Step S4-3, integrating terrain data to ensure the accuracy and completeness of the 3D model. The model needs to be further refined for complex riverbed areas; Step S4-4, water system division and node definition, specifically: divide the water system in the basin into n sections, ensure that each section can reflect the actual terrain, basin structure and hydrodynamic characteristics; specifically, it should be segmented according to the length of the river channel, the slope of the basin, and the terrain undulation to adapt to the water flow conditions in different areas; preferably, when the water system is segmented, adaptive segmentation is performed according to the hydrodynamic characteristics of different areas in the basin; n nodes are set in each water system segment to describe key variables: changes in water flow, flow rate and water level; nodes are set at key positions of the river channel, including but not limited to river channel bends and confluences.

6. The meteorological-hydrological-hydrodynamic coupling small watershed flash flood disaster forecasting method according to claim 1 is characterized in that: The step S5 comprises the following sub-steps: Step S5-1, setting boundary conditions and initial conditions for each section of the water system, specifically: Step S5-1a, setting appropriate boundary conditions for the start and end of each water system segment according to meteorological data and hydrological data; Step S5-1b, setting the initial water level state and flow rate state according to historical data; Step S5-1c, perform particle initialization on the segmented river model, divide the three-dimensional water system model into several particles, each particle represents a fluid unit in the water body, and assign physical properties to each particle: initial position, velocity, mass and density; particles are used to represent fluids, and each particle carries the physical quantities of the fluid: mass and velocity. The particles simulate the fluid movement through interaction forces, thereby simulating the flow and propagation process of floods; ensure that the distribution of particles can be refined to local areas; Step S5-2, selecting a kernel function, selecting a kernel function suitable for flash flood simulation, for defining the range of interaction between particles, and determining a support radius of the kernel function to balance calculation accuracy and efficiency; Step S5-3, perform density calculation, for each particle, calculate its local density, the formula is as follows: ; in, For particles The density of For particles Quality, is the kernel function, is the position of the ith particle, is the position of the jth particle, is the nuclear radius; Step S5-4, perform pressure calculation, and use the state equation to calculate the pressure of each particle. The formula is as follows: ； in, For particles pressure, is the speed of sound, is the reference density; Step S5-5, perform force calculation to calculate the pressure between particles and viscosity , the formula is as follows: ； ； in, is the viscosity coefficient, and For particles and speed, is the pressure of the ith particle, is the pressure of the jth particle; Step S5-6, performing external force and gravity calculation to calculate the external force on the particle; Step S5-7, calculate the total force and acceleration, combine all forces, and calculate the total acceleration of the particle. The formula is as follows: ； in, is the gravitational force acting on the i-th particle; is the mass of the ith particle; Step S5-8, perform time integration, and use the time stepping method to update the velocity and position of the particle. The formula is as follows: ； ； in, is the velocity of the ith particle at time t, is the acceleration of the ith particle at time t, is the time position, is the time step; Step S5-9, performing boundary condition processing to prevent particles from crossing solid boundaries, and using mirror particles or reflection force methods to process boundary interactions.

7. The meteorological-hydrological-hydrodynamic coupling small watershed flash flood disaster forecasting method according to claim 1 is characterized in that: The step S6 comprises the following sub-steps: Step S6-1, running the smooth particle method SPH model to calculate the water flow trajectory under different rainfall intensities, durations and terrain conditions; simulating the propagation process of flood waves, and analyzing the impact of the interaction between the propagation path of flood waves and the terrain and river channels on flood flow through the model output; Step S6-2, based on the data output by the SPH model, calculate the propagation mode of flash floods under different rainfall intensities, durations and terrain conditions, and analyze the key indicators of floods: expansion speed, overflow locations, and river flooding areas; Extract and analyze the flow characteristics of each section of the river system, including key indicators: maximum water level, maximum flow, flood propagation speed, and expected flooded area; Step S6-3, using the key indicator parameters derived in step S6-2 as secondary collection data, performing noise reduction, normalization and standardization on the data; Step S6-4, using a feature selection algorithm to screen the pre-processed secondary collection data and select the feature value with the greatest predictive ability.

8. The meteorological-hydrological-hydrodynamic coupling small watershed flash flood disaster forecasting method according to claim 1 is characterized in that: The step S7 includes the following sub-steps: Step S7-1, based on statistical methods or machine learning algorithms, establish a nonlinear relationship model between each disaster-causing factor and disaster-pregnant environmental factor in the index system in step S3-3 and the characteristic value in step S6-4, screen the factors that have the greatest impact on the occurrence of flash floods through a feature selection algorithm, quantify each factor, and calculate the sensitivity of each disaster-causing factor and disaster-pregnant environmental factor; Step S7-2: using the flooding range, flooding depth and propagation speed obtained from the SPH model as target variables, optimizing features and target variables through machine learning and deep learning, and training the extracted features to establish a multi-factor multi-variable prediction model; Step S7-3, constructing a chain relationship of hydrological and meteorological data - hydrodynamic characteristics - water level and flow representation - disaster situation; Step S7-4, by analyzing the rainfall, flow changes and disaster impacts in historical data, establish the flash flood disaster occurrence mechanism under different scenarios; Step S7-5, using the generated disaster scenario, simulate the flash flood response under different rainfall intensities and durations, and provide multiple disaster scenario predictions.

9. The meteorological-hydrological-hydrodynamic coupling small watershed flash flood disaster forecasting method according to claim 1 is characterized in that: The step S8 includes the following sub-steps: Step S8-1, using the cross-validation method to optimize the model parameters, prevent overfitting, and ensure the generalization ability of the model in different watershed environments; Step S8-2, improving the stability and accuracy of the model through an integrated learning method; Step S8-3, achieving synchronous prediction of multiple target variables to ensure consistency and relevance of prediction results; Step S8-4, using an independent test data set to verify the prediction performance of the model, and evaluating the accuracy and applicability of the model by comparing the prediction results with the actual observation data; Step S8-5, adjust the model structure and parameters according to the verification results, and optimize the combination of hydrodynamic simulation and machine learning algorithm to improve the overall prediction ability of the model.

10. The meteorological-hydrological-hydrodynamic coupling small watershed flash flood disaster forecasting method according to claim 1, characterized in that: The step S9 includes the following sub-steps: Step S9-1, deploying the trained comprehensive model to a real-time data processing system, wherein the model can receive and process the latest data input in real time; Step S9-2, receiving real-time data, combining the output of the hydrodynamic model, conducting flash flood risk assessment and forecasting, and timely updating the current flood dynamic results; Step S9-3, combined with the SPH model, realizes the three-dimensional visualization of flood evolution, and conveys the warning information through the simulation results through visualization technology.

Images