River network aquatic habitat response rapid prediction method, device and equipment and storage medium

Abstract

The application discloses a river network water habitat response rapid prediction method and device, equipment and a storage medium, relates to the technical field of data processing, and the method comprises the steps that river network hydrological data and river channel basic geographic data are collected and pretreated; a river network water power numerical model is established according to the pretreated river network hydrological data and river channel basic geographic data; the river network water power result is determined according to the river network water power numerical model; the river network water power result is converted into a first river network water habitat response index; the first runoff scenario is taken as an input variable, and the first river network water habitat response index is taken as an output variable, and a structured time sequence sample set is constructed; a long short-term memory network model is trained by using the structured time sequence sample set; and the trained long short-term memory network model is used to predict a second river network water habitat response index corresponding to a second runoff scenario. The application can quickly and efficiently predict the river network water habitat response index, and the index can be applied to multi-scenario ecological regulation decision-making.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a method, apparatus, equipment and storage medium for rapid prediction of river network aquatic habitat response. Background Technology

[0002] River aquatic habitats are vital spatial carriers for the survival, reproduction, foraging, migration, and overwintering of aquatic organisms. Their quality and suitability are closely related to the hydrodynamic conditions within the river channel. Generally, water depth, flow velocity, water level, and their spatiotemporal distribution characteristics are key factors determining the state of aquatic habitats. When the inflow process of a river changes, the hydrodynamic structure within the river network will adjust accordingly. The flow conditions at different river sections, different distributary areas, and different nodes will exhibit varying degrees of change, thus affecting the range and quality of usable habitat space for aquatic organisms. Summary of the Invention

[0003] The main objective of this application is to propose a method, apparatus, device, and storage medium for rapid prediction of river network aquatic habitat response, so as to efficiently predict the river network aquatic habitat response.

[0004] To achieve the above objectives, one aspect of this application proposes a rapid prediction method for river network aquatic habitat response, the method comprising the following steps: Collect and preprocess river network hydrological data and basic river channel geographic data; A numerical model of river network hydrodynamics is established based on the preprocessed river network hydrological data and the basic geographic data of the river channel. The hydrodynamic results of the river network are determined based on the numerical model of river network hydrodynamics. The hydrodynamic results of the river network are converted into the first river network aquatic habitat response index; Using the first runoff scenario as the input variable and the first river network aquatic habitat response index as the output variable, a structured time series sample set is constructed. The structured temporal sample set was used to train a long short-term memory network model; The trained Long Short-Term Memory (LSTM) network model was used to predict the response indicators of the second river network aquatic habitat corresponding to the second runoff scenario.

[0005] In some embodiments, the collection and preprocessing of river network hydrological data and basic river channel geographic data includes the following steps: Suppose the river network has p control boundary sections, and historical data covers... If there are 12 months in a year, then the 12th month... The boundary section at the first Year The average monthly runoff is denoted as ;in, =1,2,...,p; =1,2,..., =1,2,...12; Definition of the first The first cross-section The multi-year average monthly runoff is: ; The multi-year average monthly runoff is standardized and missing values ​​are filled to obtain the pre-processed average monthly runoff. For any runoff scenario s to be simulated, define the corresponding monthly-scale runoff input vector. for: ; in, Indicates the month; Indicating the first case under scenario s The pre-treated monthly average runoff at each control section; Collect the basic geographic data of the river channel : ; Where R represents the total number of river segments; Let r be the length of the r-th river segment; The characteristic width of the river section; This refers to riverbed or cross-sectional elevation information; is the Manning roughness coefficient.

[0006] In some embodiments, establishing a river network hydrodynamic numerical model based on the preprocessed river network hydrological data and the basic geographic data of the river channel includes the following steps: Based on the processed river network topology, cross-sectional data and topographic elevation in the basic geographic data of the river channel, the study area is divided into river segments and grid units. Two-dimensional shallow water equations are adopted as hydrodynamic governing equations, and momentum equations are defined. The Manning formula is used as the subsoil resistance term; Apply flow boundary conditions to the upstream boundary and water level boundary conditions to the downstream boundary.

[0007] In some embodiments, converting the river network hydrodynamic results into a first river network aquatic habitat response index includes the following steps: Based on the flow velocity in the river network hydrodynamic results, each of the grid cells is... Determining the suitability of the flow rate specifically includes: Let the suitable flow velocity range for the target organism be... The optimal flow rate is Define the flow rate suitability Represented as: ; in, For grid cells In runoff scenario The suitability of the flow rate, Let the velocity of grid cell j be the velocity under runoff scenario s. Based on the water depth in the river network hydrodynamic results, each of the grid cells is... Determining the suitability of water depth specifically includes: Let the suitable water depth range be... The optimal water depth is Define the water depth suitability for: ; in, For grid cells In runoff scenario Suitable water depth For grid cells In runoff scenario The water depth is below; Based on flow velocity suitability and water depth suitability Determine the grid cells In runoff scenario The comprehensive aquatic habitat suitability index ; According to the comprehensive aquatic habitat suitability index The river network-scale index and the average aquatic habitat index of the entire river network were respectively determined as the first river network aquatic habitat response index; River network scale index for: ; The average aquatic habitat index of the entire river network for: ; in, For the first The area of ​​each grid cell, The number of the grid cells.

[0008] In some embodiments, constructing a structured time-series sample set using a first runoff scenario as the input variable and the first river network aquatic habitat response index as the output variable includes the following steps: For the Given samples, define the input feature vector. for: ; in, For the first The sample corresponds to the first Monthly runoff at each control section; Monthly encoding information; In the input feature vector Add extended feature vectors The enhanced input vector is obtained. ; Wherein, the enhanced input vector for: ; ; If the comprehensive index of the entire river network is used as the output variable, then the first... The labels for each sample are represented as follows: or ; If indicators from multiple river segments or multiple spatial partitions are used as the output variables, then the output vector is defined as follows: ;in, To determine the dimension of the output index; Construct the structured time series sample set as follows: ,or, ; in, The total number of samples; For the input variables; The output variable is denoted as .

[0009] In some embodiments, training a long short-term memory network model using the structured temporal sample set includes the following steps: According to the The input feature vector or the enhanced input vector of each sample constructs an input time series sequence with a time window length of T. Definition of the first The output label corresponding to each sample is the first river network aquatic habitat response index at the target time. The long short-term memory network model is trained based on the input time sequence and the output label.

[0010] In some embodiments, the method further includes the following steps: The long short-term memory network model was trained using the mean squared error loss function. The parameter set of the Long Short-Term Memory network model is updated using the backpropagation time-travel algorithm; wherein, in the e-th iteration, the parameter update expression is: ; The learning rate; The mean squared error loss function is set of parameters. The gradient.

[0011] To achieve the above objectives, another aspect of this application provides a rapid prediction device for river network aquatic habitat response, the device comprising: The data processing unit is used to collect and preprocess river network hydrological data and basic river channel geographic data; The model building unit is used to establish a river network hydrodynamic numerical model based on the preprocessed river network hydrological data and the basic geographic data of the river channel. The hydrodynamic determination unit is used to determine the hydrodynamic results of the river network based on the numerical model of river network hydrodynamics. The index determination unit is used to convert the river network hydrodynamic results into a first river network aquatic habitat response index. The sample set construction unit is used to construct a structured time series sample set with the first runoff scenario as the input variable and the first river network aquatic habitat response index as the output variable. The model training unit is used to train a long short-term memory network model using the structured temporal sample set. The indicator prediction unit is used to predict the response indicators of the second river network aquatic habitat corresponding to the second runoff scenario using the trained long short-term memory network model.

[0012] To achieve the above objectives, another aspect of this application provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the above-described method.

[0013] To achieve the above objectives, another aspect of the embodiments of this application proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method.

[0014] To achieve the above objectives, another aspect of this application provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0015] The embodiments of this application include at least the following beneficial effects: This application provides a method, apparatus, equipment, and storage medium for rapid prediction of river network aquatic habitat responses. The proposed method involves: collecting and preprocessing river network hydrological data and basic river channel geographic data; establishing a river network hydrodynamic numerical model based on the preprocessed data; determining the river network hydrodynamic results based on the model; converting these results into a first river network aquatic habitat response index; constructing a structured time-series sample set using a first runoff scenario as input and the first river network aquatic habitat response index as output; training a long short-term memory (LSTM) network model using this set; and using the trained LSM model to predict the second river network aquatic habitat response index corresponding to a second runoff scenario. This application can rapidly and efficiently predict river network aquatic habitat response indices, which can be applied to scenarios such as river network ecological flow regulation, aquatic habitat suitability assessment, watershed water resource allocation optimization, ecological restoration effect analysis, and multi-scenario ecological scheduling decision-making. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 A flowchart illustrating the rapid prediction method for river network aquatic habitat response provided in this application embodiment; Figure 2 Example flowchart of the rapid prediction method for river network aquatic habitat response provided in the embodiments of this application; Figure 3 A schematic diagram of the structure of the rapid prediction device for river network aquatic habitat response provided in an embodiment of this application; Figure 4 This is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit it. In the following description, when referring to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with those of this application; they are merely examples of apparatuses and methods consistent with some aspects of the embodiments of this application as detailed in the appended claims.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0020] Before providing a detailed description of the embodiments of this application, some related technologies involved in the embodiments of this application will be described first, as follows: How to analyze the hydrodynamic evolution of river networks and further assess the response of aquatic habitats based on changes in inflow has become an important technical issue in river ecological protection and river network water environment management.

[0021] Under natural conditions, the inflow process of river networks exhibits significant temporal variability, showing not only interannual differences in abundance and scarcity but also substantial fluctuations between different months within the year. In areas affected by reservoir scheduling, dam control, water diversion projects, river regulation, and water intake activities, the inflow process is further influenced by human activities, making the hydrodynamic state within the river network even more complex. Especially for complex river network systems with multiple inflow boundaries, multiple river branches, multi-level confluence relationships, and mutual backwater effects between upstream and downstream areas, the response of the same inflow change at different spatial locations is inconsistent, often exhibiting strong nonlinearity and spatiotemporal heterogeneity. Therefore, conducting aquatic habitat response analysis for complex river networks requires not only considering the hydrodynamic process itself but also comprehensively taking into account multiple inflow boundaries, the river network topology, and the continuous temporal variation characteristics.

[0022] In engineering practice, river network aquatic habitat analysis has strong application requirements. For example, in ecological flow protection projects, it is necessary to analyze changes in water depth, flow velocity, and habitat suitability in target river sections or key areas under different flow conditions to determine the discharge process that meets ecological protection requirements. In water resource optimization and scheduling management, it is necessary to compare the differences in the ecological state of the river network under different scheduling schemes, different inflow combinations, and different monthly operating conditions to support scheduling decisions. In river ecological restoration, river-lake connectivity, water system improvement, and ecological impact assessment of water-related projects, it is also usually necessary to assess the degree of improvement and changing trend of river network aquatic habitats before and after project implementation or under different schemes. In ecological risk early warning and emergency analysis, it is also necessary to complete rapid assessment of multiple possible inflow scenarios within a short period of time. All of the above application scenarios indicate that river network aquatic habitat analysis is not merely confined to the academic research level, but is closely related to ecological scheduling, engineering demonstration, and management decision-making, and has obvious practical engineering requirements.

[0023] Currently, river network aquatic habitat analysis is typically based on hydrodynamic numerical simulation. In practical applications, technicians generally first collect basic information such as river cross-section data, topographic data, boundary inflow data, roughness parameters, and control structure operation data for the study area. Then, they establish a one-dimensional or two-dimensional hydrodynamic model to calculate the water level, depth, and flow velocity of each river segment or grid unit under different inflow conditions. Based on this, they further analyze the aquatic habitat status under different conditions using single-factor fitness curves, comprehensive fitness functions, weighted available area, or other ecological evaluation indicators. This technical approach has a strong physical basis and can realistically reflect the impact of inflow changes on hydrodynamic conditions and aquatic habitat status; therefore, it has been widely used in ecological flow assessment, ecological restoration evaluation, and river network ecological impact analysis.

[0024] However, in practical engineering, river network aquatic habitat analysis is often not conducted for a single working condition, but rather requires comparison across multiple scenarios. For example, it's necessary to analyze the response differences under different hydrological conditions such as low water, normal water, and high water; to analyze the ecological effects under different months, different boundary water inflow combinations, and different engineering scheduling methods; and to conduct batch comparisons of different target schemes to identify sensitive areas, critical periods, and optimal control methods. In this context, traditional methods typically require setting boundaries, calculating models, and extracting indicators separately for each scenario. As the number of scenarios increases, the river network structure becomes more complex, or the analysis period lengthens, the number of model runs increases significantly, and the workload for data processing, result extraction, and scheme comparison also rises substantially, resulting in a long overall analysis cycle that fails to meet the engineering requirements for high efficiency, multi-scenario analysis, and rapid evaluation.

[0025] On the other hand, the response processes of aquatic habitats in complex river networks not only exhibit spatial complexity but also significant temporal series characteristics. The state of aquatic habitats at a given moment within a river network is often related not only to current inflow conditions but also to previous inflow processes and the resulting hydrodynamic conditions. Especially when conducting continuous analysis on a monthly scale, the changes in inflow between different months exhibit continuity and correlation, and the corresponding aquatic habitat responses also possess certain lag and dynamic transmission characteristics. While traditional scenario-by-scenario and condition-by-condition analysis methods can obtain results for each time period, they struggle to efficiently extract the mapping patterns between continuous runoff processes and aquatic habitat responses, thus limiting their application in batch prediction and rapid deployment.

[0026] With the development of data-driven technologies and intelligent prediction methods, fitting complex nonlinear relationships and time series characteristics using machine learning models has gradually become a feasible technical approach to improve analytical efficiency. By sampling existing runoff scenarios, hydrodynamic simulation results, and aquatic habitat evaluation results, a mapping model between input variables and target responses can be constructed. This allows for the rapid output of corresponding aquatic habitat response results without repeatedly running the complete hydrodynamic numerical model given a new inflow scenario. This approach provides a new direction for the rapid analysis of complex river networks under multiple scenarios.

[0027] However, in the scenario of rapid prediction of river network aquatic habitats, how to effectively integrate river network hydrodynamic simulation, aquatic habitat index construction, time-series sample organization, and machine learning proxy modeling, and how to ensure improved prediction efficiency while maintaining the adaptability of the analysis process to actual engineering needs, remain technical problems that require further research and solutions in this field. Therefore, it is necessary to propose a rapid prediction method for river network aquatic habitat response analysis to adapt to the actual needs under multiple scenarios, time periods, and engineering application conditions.

[0028] The relevant technologies are described below: (1) Existing methods for analyzing river network aquatic habitats typically begin by establishing a one-dimensional or two-dimensional hydrodynamic numerical model based on the river channel morphology, cross-sectional data, bed elevation, roughness coefficient, and upstream and downstream boundary conditions in the study area. The model is then used to calculate hydrodynamic results such as water level, water depth, and flow velocity under different flow conditions, which serve as the basis for subsequent analysis of river network aquatic habitat changes.

[0029] (2) In the prior art, after obtaining the hydrodynamic simulation results, factors related to the habitat requirements of the target aquatic organisms, such as water depth, flow velocity, and water level, are usually selected to establish a single-factor suitability function or a comprehensive suitability evaluation method, and then the aquatic habitat suitability index at the river section level, cross section level, or regional level is calculated to reflect the changes in aquatic habitat quality under different water inflow conditions.

[0030] (3) In existing engineering applications, in order to compare the impact of different hydrological year types, different boundary flow combinations or different scheduling schemes on aquatic habitats, it is usually necessary to conduct hydrodynamic simulation and aquatic habitat evaluation for each scenario, and then make a horizontal comparison of the results under each scenario, so as to identify the changing trends, sensitive areas and suitability differences of river network aquatic habitats under different working conditions.

[0031] (4) With the development of data-driven technology, existing technical solutions have begun to apply machine learning methods to the prediction of water-related factors such as runoff, water level, and water quality. By establishing a mapping relationship between input variables and target variables, the prediction efficiency is improved. Some neural network models and time series prediction models have shown strong expressive ability for complex nonlinear relationships and time series characteristics, providing a certain technical foundation for the rapid prediction of river network aquatic habitats.

[0032] The drawbacks of related technologies: (1) When conducting river network aquatic habitat analysis, the existing technology usually requires running the hydrodynamic model separately for each group of inflow scenarios and further completing the aquatic habitat evaluation. When the analysis object is a complex river network consisting of multiple boundaries, multiple river sections and multiple time periods, the number of calculations required is large, the overall calculation cost is high, and the analysis cycle is long, which is not conducive to quickly completing the comparison and dynamic evaluation of multiple schemes in engineering practice.

[0033] (2) Most existing technologies rely on scenario-by-scenario simulation to obtain analysis results. When long-term continuous scenario analysis needs to be carried out on a monthly or finer time scale, the workload of model building, boundary setting, calculation and solution and result processing will increase significantly as the number of input scenarios increases, which limits its application efficiency in engineering scheduling, scheme selection and rapid response scenarios.

[0034] (3) Although existing technologies can reflect the relationship between hydrodynamic changes in river networks and aquatic habitat status to a certain extent, they often still rely on a large number of numerical simulation results to support the significant spatiotemporal differences between different river sections and different zones in complex river networks. There is a lack of an efficient method that can directly and quickly deduce the aquatic habitat response results from runoff scenarios. Therefore, there are still shortcomings in multi-scenario batch analysis.

[0035] (4) In the existing technology, data-driven methods are mainly applied to the prediction of single hydrological or environmental elements. There are few complete technical routes oriented towards “runoff scenario input - river network aquatic habitat response output”. The process of combining hydrodynamic simulation, aquatic habitat index extraction and time series machine learning proxy modeling has not been effectively combined. Therefore, it is difficult to simultaneously take into account the analysis efficiency, the ability to express time series response features and the need for rapid prediction in engineering applications.

[0036] The technical problem this application aims to solve is that existing methods for analyzing river network aquatic habitats typically rely on hydrodynamic numerical models and aquatic habitat evaluation models to perform calculations on a scenario-by-scenario and operational-by-operational-condition basis. When analysis is required for different months, different inflow combinations, different hydrological conditions, or different scheduling schemes, it is often necessary to repeatedly set boundary conditions, solve models, and extract results, resulting in a large overall computational load, long analysis cycle, and low efficiency. This makes it difficult to meet the application needs of complex river networks in multi-scenario comparison, rapid assessment, and engineering auxiliary decision-making. Furthermore, complex river network systems typically have characteristics such as multiple inflow boundaries, multiple distributary channels, and complex hydraulic connections. The responses of different river sections and regions to changes in inflow exhibit significant spatiotemporal differences and nonlinear characteristics. While existing technologies can reflect the impact of changes in inflow on hydrodynamic conditions such as water depth and flow velocity, as well as aquatic habitat status, to some extent, there is a lack of a technical solution that can effectively combine runoff scenarios, hydrodynamic simulation results, and aquatic habitat response indicators, and further establish rapid predictive relationships. Therefore, it is difficult to achieve efficient characterization and batch prediction of river network aquatic habitat response processes. Furthermore, existing data-driven methods primarily focus on predicting single hydrological elements such as runoff and water level. They lack targeted surrogate modeling methods for the entire technical chain of "runoff scenario input – river network aquatic habitat response output," particularly lacking an effective representation of monthly continuous inflow processes and their corresponding dynamic aquatic habitat responses. Therefore, this application aims to provide a rapid prediction method for river network aquatic habitat responses based on machine learning surrogate models, addressing the technical problems of low efficiency in multi-scenario analysis, difficulty in rapidly characterizing complex river network response patterns, and difficulty in efficiently obtaining aquatic habitat prediction results in existing technologies.

[0037] This application provides a method, apparatus, device, and storage medium for rapid prediction of river network aquatic habitat response, relating to the field of data processing technology. The method, apparatus, device, and storage medium provided in this application can be applied to a terminal, a server, or software running on a terminal or server. In some embodiments, the terminal can be a smartphone, tablet, laptop, desktop computer, smart speaker, smartwatch, or vehicle terminal, but is not limited thereto; the server can be configured as an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the server can also be a node server in a blockchain network; the software can be an application implementing the rapid prediction method for river network aquatic habitat response, but is not limited to the above forms.

[0038] This application can be applied to numerous general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics devices, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0039] Reference Figure 1 This application provides a method for rapid prediction of river network aquatic habitat response, which may include, but is not limited to, steps S100 to S160, as follows: S100: Collect and preprocess river network hydrological data and basic river channel geographic data; S110: Establish a numerical model of river network hydrodynamics based on the preprocessed river network hydrological data and the basic geographic data of the river channel; S120: Determine the hydrodynamic results of the river network based on the aforementioned river network hydrodynamic numerical model; S130: Convert the aforementioned river network hydrodynamic results into a first river network aquatic habitat response index; S140: Using the first runoff scenario as the input variable and the first river network aquatic habitat response index as the output variable, construct a structured time series sample set; S150: Train a long short-term memory network model using the structured temporal sample set; S160: Use the trained Long Short-Term Memory Network model to predict the response index of the second river network aquatic habitat corresponding to the second runoff scenario.

[0040] Optionally, the collection and preprocessing of river network hydrological data and basic river channel geographic data includes the following steps: Suppose the river network has p control boundary sections, and historical data covers... If there are 12 months in a year, then the 12th month... The boundary section at the first Year The average monthly runoff is denoted as ;in, =1,2,...,p; =1,2,..., =1,2,...12; Definition of the first The first cross-section The multi-year average monthly runoff is: ; The multi-year average monthly runoff is standardized and missing values ​​are filled to obtain the pre-processed average monthly runoff. For any runoff scenario s to be simulated, define the corresponding monthly-scale runoff input vector. for: ; in, Indicates the month; Indicating the first case under scenario s The pre-treated monthly average runoff at each control section; Collect the basic geographic data of the river channel : ; Where R represents the total number of river segments; Let r be the length of the r-th river segment; The characteristic width of the river section; This refers to riverbed or cross-sectional elevation information; is the Manning roughness coefficient.

[0041] Optionally, the step of establishing a river network hydrodynamic numerical model based on the preprocessed river network hydrological data and the basic geographical data of the river channel includes the following steps: Based on the processed river network topology, cross-sectional data and topographic elevation in the basic geographic data of the river channel, the study area is divided into river segments and grid units. Two-dimensional shallow water equations are adopted as hydrodynamic governing equations, and momentum equations are defined. The Manning formula is used as the subsoil resistance term; Apply flow boundary conditions to the upstream boundary and water level boundary conditions to the downstream boundary.

[0042] Optionally, the step of converting the river network hydrodynamic results into a first river network aquatic habitat response index includes the following steps: Based on the flow velocity in the river network hydrodynamic results, each of the grid cells is... Determining the suitability of the flow rate specifically includes: Let the suitable flow velocity range for the target organism be... The optimal flow rate is Define the flow rate suitability Represented as: ; in, For grid cells In runoff scenario The suitability of the flow rate, Let the velocity of grid cell j be the velocity under runoff scenario s. Based on the water depth in the river network hydrodynamic results, each of the grid cells is... Determining the suitability of water depth specifically includes: Let the suitable water depth range be... The optimal water depth is Define the water depth suitability for: ; in, For grid cells In runoff scenario Suitable water depth For grid cells In runoff scenario The water depth is below; Based on flow velocity suitability and water depth suitability Determine the grid cells In runoff scenario The comprehensive aquatic habitat suitability index ; According to the comprehensive aquatic habitat suitability index The river network-scale index and the average aquatic habitat index of the entire river network were respectively determined as the first river network aquatic habitat response index; River network scale index for: ; The average aquatic habitat index of the entire river network for: ; in, For the first The area of ​​each grid cell, The number of the grid cells.

[0043] Optionally, the step of constructing a structured time-series sample set using the first runoff scenario as the input variable and the first river network aquatic habitat response index as the output variable includes the following steps: For the Given samples, define the input feature vector. for: ; in, For the first The sample corresponds to the first Monthly runoff at each control section; Monthly encoding information; In the input feature vector Add extended feature vectors The enhanced input vector is obtained. ; Wherein, the enhanced input vector for: ; ; If the comprehensive index of the entire river network is used as the output variable, then the first... The labels for each sample are represented as follows: or ; If indicators from multiple river segments or multiple spatial partitions are used as the output variables, then the output vector is defined as follows: ;in, To determine the dimension of the output index; Construct the structured time series sample set as follows: ,or, ; in, The total number of samples; For the input variables; The output variable is denoted as .

[0044] Optionally, training the long short-term memory network model using the structured time-series sample set includes the following steps: According to the The input feature vector or the enhanced input vector of each sample constructs an input time series sequence with a time window length of T. Definition of the first The output label corresponding to each sample is the first river network aquatic habitat response index at the target time. The long short-term memory network model is trained based on the input time sequence and the output label.

[0045] Optionally, the method further includes the following steps: The long short-term memory network model was trained using the mean squared error loss function. The parameter set of the Long Short-Term Memory network model is updated using the backpropagation time-travel algorithm; wherein, in the e-th iteration, the parameter update expression is: ; The learning rate; The mean squared error loss function is set of parameters. The gradient.

[0046] The following sections will provide a detailed description and explanation of some optional embodiments of this application, using specific application examples.

[0047] This embodiment discloses a rapid prediction method for river network aquatic habitat response based on a machine learning surrogate model. The method uses basic data such as river network boundary runoff, hydrogeometry, and cross-sectional parameters as the main input information. It uses a coupled numerical model of river network hydrodynamics and aquatic habitat as the basic model and a machine learning surrogate model as the rapid prediction module. It constructs a collaborative technical route of "runoff scenario generation—hydrodynamic process simulation—aquatic habitat response extraction—surrogate model training—rapid prediction output" (e.g.,...). Figure 2 As shown in the figure, this method enables efficient simulation and rapid prediction of aquatic habitat response processes under different inflow conditions in complex river networks. This method can be applied to scenarios such as ecological flow regulation in river networks, aquatic habitat suitability assessment, watershed water resource allocation optimization, ecological restoration effect analysis, and multi-scenario ecological scheduling decision-making. The method includes the following steps: (1) collecting and preprocessing river network hydrological and basic geographic data; (2) establishing a river network hydrodynamic numerical model; (3) constructing river network aquatic habitat response indicators; (4) constructing a training sample set for the surrogate model; (5) training the LSTM machine learning surrogate model; and (6) performing rapid prediction of river network aquatic habitat response.

[0048] Specifically, this embodiment may include the following steps: Step S1: Collect and preprocess river network hydrological and basic geographic data.

[0049] Step S1 is used to construct the basic dataset required for training the river network hydrodynamic model and the surrogate model. The basic data includes multi-year runoff data, river cross-section data, river network topology data, bed elevation data, roughness parameters, and calculation boundary condition data.

[0050] Among them, multi-year runoff data are used to reflect the interannual variation trend and intra-annual distribution characteristics of the main control sections of the river network; river channel cross-section data, river network topology data, and roughness parameters are used to establish the river network hydrodynamic model. This paper uses multi-year runoff data of the Dongjiang River Basin and its main control sections as a basis to analyze annual runoff and monthly average runoff, and selects typical dry years, normal water years, and wet water years to construct monthly runoff scenarios.

[0051] S1.1 Data Acquisition.

[0052] Suppose the river network has p control boundary sections, and historical data covers... If there are 12 months in a year, then the 12th month... The boundary section at the first Year The average monthly runoff is recorded as follows: ; in =1,2,...,p; =1,2,..., =1,2,...12.

[0053] Simultaneously, basic geographic data of the river channel was collected:

[0054] In the formula, R represents the total number of river segments; Let r be the length of the r-th river segment; The characteristic width of the river section; This refers to riverbed or cross-sectional elevation information; is the Manning roughness coefficient.

[0055] S1.2 Runoff Statistical Analysis.

[0056] The annual runoff at each control section is statistically analyzed, and the first section is defined as follows: The cross-section at the first The total annual runoff is:

[0057] Definition of the first The first cross-section The multi-year average monthly runoff is:

[0058] Definition of the first The multi-year average annual runoff of each cross section is:

[0059] The annual runoff variation coefficient can be further calculated to characterize the degree of interannual runoff fluctuation.

[0060] By using annual runoff, monthly average runoff, and variation constants, the interannual variation characteristics and monthly scale distribution patterns of water inflow at the river network boundary are identified, and typical dry years, normal water years, and high water years are selected accordingly.

[0061] S1.3 Data quality control and standardization processing.

[0062] The raw runoff, water level, cross-sectional, and topographic data are subjected to missing value checks, outlier identification, and scale standardization. Preferably, the input variables are standardized to facilitate subsequent machine learning model training; the standardization formula is as follows:

[0063] In the formula These are the original variables; The mean of the variable; The standard deviation is the variable. These are the standardized variables.

[0064] For missing monthly runoff data, interpolation between adjacent months or the multi-year average of the same month can be used to fill the gaps, for example:

[0065] or:

[0066] in The value represents the monthly runoff after the repair.

[0067] S1.4 Constructing a monthly-scale runoff scenario.

[0068] For any scenario s to be simulated, its monthly runoff input vector is expressed as:

[0069] In the formula Indicates the month; Indicating the first case under scenario s The average monthly runoff at each control section.

[0070] Preferably, typical dry years, normal water years, and high water years can be used as representative scenarios to form multiple sets of monthly runoff samples required for surrogate model training and prediction.

[0071] Step S2: Establish a numerical model of river network hydrodynamics.

[0072] Step S2 is used to simulate the changes in hydrodynamic elements such as water level, velocity, and depth within the river network under a given monthly runoff scenario. Specifically, based on the analysis of runoff variation characteristics, a hydrodynamic model of the Dongjiang River network is constructed. River cross-sectional data and roughness parameters are input, and different runoff scenarios are set as boundary conditions to simulate the changes in water level and velocity within the river network under different runoff conditions.

[0073] S2.1 River network generalization and network division.

[0074] Based on the river network topology, cross-sectional data, and topographic elevation, the study area is divided into river segments and discretized into grid cells. For a two-dimensional implementation, the study area can be discretized into M grid cells, each denoted as . The area is denoted as .

[0075] S2.2 Hydrodynamic control equations.

[0076] Preferably, a two-dimensional shallow water equation is used as the hydrodynamic governing equation. The continuity equation is:

[0077] The momentum equations are as follows:

[0078]

[0079] In the formula For water depth; They are respectively , directional velocity component; For source and sink items; It is the acceleration due to gravity; This refers to the water surface elevation. Density of water; This represents the shear stress component of the subbed.

[0080] The subgrade resistance term is preferably expressed using the Manning formula as follows:

[0081]

[0082] in is the Manning roughness coefficient.

[0083] S2.3 Boundary condition settings.

[0084] Apply flow boundary conditions to the upstream boundary:

[0085] That is, in the context Month Next, the The inflow at each upstream control section is taken as the monthly runoff scenario value.

[0086] Water level boundary conditions can be applied to the downstream boundary:

[0087] in This is a process for controlling the boundary water level downstream.

[0088] S2.4 Hydrodynamic simulation results output.

[0089] For each grid cell To obtain stable or time-averaged hydrodynamic results:

[0090] And further define the flow velocity modulus as:

[0091] and This will serve as the basis for constructing aquatic habitat response indicators in step S3.

[0092] Step S3: Construct river network aquatic habitat response indicators.

[0093] Step S3 is used to transform the hydrodynamic results into quantifiable aquatic habitat response indicators. Specifically, based on the hydrodynamic simulation results, hydrodynamic indicators closely related to the aquatic habitat, such as flow velocity and water depth, are selected to analyze the aquatic habitat conditions of the river network.

[0094] S3.1 Construction of the single-factor fitness function.

[0095] For each grid cell We construct flow velocity suitability function and water depth suitability function respectively.

[0096] 1) Flow rate suitability.

[0097] Let the suitable flow velocity range for the target organism be... The optimal flow rate is Then the suitability of the flow rate It can be represented as a piecewise function:

[0098] 2) Suitable water depth.

[0099] Let the suitable water depth range be... The optimal water depth is Then the water depth is suitable for:

[0100] In the formula For grid cells In runoff scenario The suitability of the flow rate; For grid cells In runoff scenario The appropriate depth of water.

[0101] S3.2 Construction of the comprehensive aquatic habitat index.

[0102] Preferably, the geometric mean method is used to construct the unit comprehensive aquatic habitat suitability index. :

[0103] Alternatively, a weighted average can be used:

[0104] In the formula ; .

[0105] Calculation of indicators at the scale of the S3.3 river section or the entire river network.

[0106] To obtain river network-scale metrics that can serve as labels for machine learning outputs, a weighted available area metric is preferred:

[0107] In the formula: For the context Weighted usable area of ​​the entire river network; For the first The area of ​​each grid cell.

[0108] It can also calculate the average aquatic habitat index for the entire river network:

[0109] In practice, the output label can be a single comprehensive indicator, such as... or Alternatively, it can be an index vector corresponding to multiple key river segments. This not only conforms to the expression of "spatial distribution characteristics" and "comparative analysis of multiple river segments", but also facilitates machine learning modeling.

[0110] Step S4: Construct a training sample set for the proxy model.

[0111] Step S4 is used to establish the correspondence between input and output samples required for the machine learning agent model. This paper uses monthly runoff scenarios as input variables, river network aquatic habitat indicators as output variables, and hydrodynamic and aquatic habitat numerical simulation results as reference ground truth.

[0112] S4.1 Input feature vector construction.

[0113] For the Given samples, define the input feature vector as:

[0114] In the formula For the first The sample corresponds to the first Monthly runoff at each control section; This is the month-coded information.

[0115] Furthermore, extended features such as typical year-type coding, upstream and downstream flow ratio, and total inflow can be added, for example:

[0116] This results in an enhanced input vector:

[0117] S4.2 Output label vector construction.

[0118] If the comprehensive index of the entire river network is used as the output, then the first... The label for each sample can be written as:

[0119] or:

[0120] If the indicators of multiple river segments or multiple spatial partitions are used as the joint output, then the output vector is defined as follows:

[0121] in To determine the dimension of the output index.

[0122] S4.3 Sample set definition.

[0123] The structured time series sample set is constructed as follows:

[0124] In the formula The total number of samples; Input features; This is the standard output obtained from numerical simulation.

[0125] Preferably, the sample set is divided into a training set and a validation set:

[0126] For example:

[0127] in .

[0128] Step S5: Train the machine learning agent model.

[0129] Step S5 is used to learn the training sample set constructed in step S4 using a long short-term memory network model, and establish a nonlinear dynamic mapping relationship between the monthly runoff time series input and the river network aquatic habitat response output, thereby obtaining a surrogate model that can be used for rapid prediction.

[0130] S5.1 Construct the input sequence for the LSTM model.

[0131] For the k-th training sample, assuming a time window length of T, its input time series sequence is represented as:

[0132] in, The input feature vector of the k-th sample at time step t can be represented as:

[0133] In the formula, p represents the number of control boundary sections; Let be the monthly runoff of the k-th sample at the t-th time step and the i-th boundary section.

[0134] Furthermore, auxiliary features such as month number, typical year type identifier, total inflow, and upstream-downstream flow ratio can be added to the input feature vector to enhance the model's ability to identify seasonality and scenario differences. At this point, the input vector can be expanded to:

[0135] in, For month coding, To assist in the characteristics of the context.

[0136] S5.2 Construct an LSTM model to output labels.

[0137] For the k-th sample, the corresponding output label can be the river network aquatic habitat response index at the target time, denoted as:

[0138] In the formula, r is the dimension of the output index; Let j be the response index for the aquatic habitat.

[0139] Preferably, the output label may be at least one of the following: 1) Comprehensive aquatic habitat index of the entire river network; 2) Weighted usable area WUA ; 3) Aquatic habitat response indicators for several key river sections; 4) Average suitability index of each zone.

[0140] If only a single indicator is predicted, the output can be written as:

[0141] or:

[0142] S5.3 Establish the LSTM network structure.

[0143] An LSTM network consists of an input layer, LSTM hidden layers, and an output layer. For any time step t, the LSTM unit selectively retains historical and current input information through forget gates, input gates, and output gates. Let the input at time step t be... The hidden state at the previous moment was The previous cell state was The LSTM cell calculation process is as follows.

[0144] 1) Gate of Oblivion:

[0145] In the formula Output for the forget gate; Here is the forget gate weight matrix; Forget gate bias term; ( ) is the Sigmoid activation function.

[0146] The forget gate is used to control the state of the cell in the previous time step. How much information is retained?

[0147] 2) Input Gate:

[0148]

[0149] In the formula For input gate output; Candidate cell state; , These are the corresponding weight matrices; , These are the bias terms; tanh( ) is the hyperbolic tangent activation function.

[0150] The input gate is used to determine the degree to which the current input information is written to the cell state.

[0151] 3) Cell status update:

[0152] in, This represents the Hadamard element-wise product.

[0153] This formula indicates that the current cell state is composed of "retained historical memories" and "candidate memories formed by the current new input".

[0154] 4) Output gate:

[0155]

[0156] In the formula Output gate output; Hide the state at the current time step; , These represent the output gate weight matrix and bias term, and the hidden state, respectively. Used to characterize temporal features up to the current time step.

[0157] S5.4 output layer construction.

[0158] Hidden state of the last time step of the LSTM Inputting a fully connected layer yields predicted values ​​for the aquatic response of the river network:

[0159] In the formula This is the predicted output for the k-th sample; This is the output layer weight matrix; This is the output layer bias term.

[0160] If the output is a single aquatic habitat index, it can be written as:

[0161] This formula shows that LSTM can predict the river network aquatic habitat response under a target month or target scenario by extracting the runoff time series features of the first T time steps.

[0162] S5.5 loss function construction.

[0163] To train the LSTM surrogate model, the error between the true label and the predicted value is used as the objective function. The mean squared error loss function is preferred.

[0164] In the formula, N is the total number of training samples; This is the actual output; The model predicts the output; It is the Euclidean norm.

[0165] When the output is a single metric, the loss function can be simplified to:

[0166] S5.6 model training method.

[0167] The backpropagation time-travel algorithm is used to update the LSTM network parameters. Let the model parameter set be:

[0168] Then, in the e-th iteration, the parameter update formula can be expressed as:

[0169] In the formula The learning rate; For the loss function with respect to the parameter set The gradient.

[0170] Preferably, the Adam optimization algorithm is used to iteratively update the parameters to improve training convergence speed and stability. Furthermore, batch size, number of iterations, and early stopping mechanisms can be set to prevent model overfitting and improve generalization performance.

[0171] S5.7 Model Validation and Optimization.

[0172] The training sample set is divided into training, validation, and test sets, and the accuracy of the trained LSTM surrogate model is evaluated. Mean absolute error, root mean square error, and coefficient of determination are preferred as evaluation metrics.

[0173]

[0174]

[0175] In the formula, This represents the mean of the true values ​​in the test set.

[0176] When the LSTM model meets the preset accuracy requirements on the test set, it is determined as the optimal surrogate model for rapid prediction of river network aquatic habitat response.

[0177] Step S6: Conduct rapid prediction of river network aquatic habitat response.

[0178] Step S6 is used to quickly predict new runoff scenarios using a trained surrogate model without repeatedly running the complex hydrodynamic numerical model, thereby supporting multi-scenario analysis. This is the main technical effect of "improving the efficiency of multi-runoff scenario analysis and achieving rapid prediction".

[0179] S6.1 New Scenario Input.

[0180] For the lunar-scale runoff scenario to be predicted Construct input variables:

[0181] If standardization was used in step S1, the same transformation is performed synchronously on the new input:

[0182] S6.2 Fast Prediction.

[0183] Input the new scenario into the optimal agent model:

[0184] If the output is a single comprehensive index, then:

[0185] or:

[0186] If the output is a multi-segment index vector, then:

[0187] S6.3 Scenario Comparison Analysis.

[0188] To quantify the degree of change of the scenario to be predicted relative to the baseline scenario 0, the relative rate of change can be calculated:

[0189] or:

[0190] In the formula or These are reference values ​​under the baseline scenario; or This represents the relative rate of change.

[0191] The aforementioned rate of change can be used to identify aquatic habitat improvement zones, degradation zones, and sensitive river sections under different water inflow scenarios.

[0192] S6.4 Format of prediction results output.

[0193] Preferably, the rapid prediction results include at least one of the following: 1) Overall comprehensive aquatic habitat index of the river network; 2) River section-level aquatic habitat response indicators; 3) Suitable area of ​​aquatic habitat at the zoning level; 4) Ranking results of change rates between scenarios; 5) Rapid evaluation results for ecological scheduling or scheme comparison.

[0194] In summary, this embodiment includes the following key technical solutions: (1) A rapid prediction overall technical route for river network aquatic habitat response analysis is provided, namely, taking river network hydrological and basic geographic data as the basic input, river network hydrodynamic simulation results and aquatic habitat indicators as the intermediate bridge, and machine learning proxy model as the core prediction unit to realize the rapid output from runoff scenario to river network aquatic habitat response results.

[0195] (2) A method for preprocessing and constructing scenarios for river network hydrology and basic geographic data is provided, namely, collecting, organizing and standardizing river network boundary runoff data, topographic data, river channel cross-section data, roughness parameters and boundary conditions, and constructing a monthly runoff scenario input system suitable for subsequent simulation and prediction.

[0196] (3) A numerical simulation method for river network hydrodynamics based on runoff scenario is provided, which takes runoff scenarios under different months, different inflow combinations or different hydrological conditions as boundary inputs, and calculates hydrodynamic elements such as water level, water depth and flow velocity through river network hydrodynamic model, so as to provide basic data support for the response analysis of river network aquatic habitat.

[0197] (4) A method for constructing a river network aquatic habitat response index is provided, namely, extracting key factors related to aquatic habitat changes based on the results of river network hydrodynamic simulation, establishing single-factor suitability index and / or comprehensive aquatic habitat response index, realizing the quantitative representation of hydrodynamic results to aquatic habitat state, and using it as the output target of the surrogate model.

[0198] (5) A training sample construction method for “runoff scenario input - aquatic habitat response output” is provided, that is, to establish a one-to-one correspondence between the input parameters under different runoff scenarios and the corresponding hydrodynamic simulation results and aquatic habitat evaluation results, forming a structured sample set for use in training machine learning proxy models.

[0199] (6) A method for rapid prediction of river network aquatic habitat response using LSTM long short-term memory network is provided. That is, the continuous monthly runoff sequence is input into the LSTM surrogate model, and its ability to learn the time series dependence and nonlinear response law is used to establish the mapping relationship between runoff process and river network aquatic habitat response. Thus, the prediction results can be obtained quickly without repeating the complete hydrodynamic model under new runoff scenarios.

[0200] This embodiment includes at least the following beneficial effects: (1) Originality: This embodiment organically combines river network hydrodynamic simulation, aquatic habitat response analysis, and machine learning surrogate modeling to construct a technical route of "runoff scenario input - hydrodynamic process characterization - aquatic habitat index extraction - rapid prediction by surrogate model". Compared with the traditional method that relies solely on numerical simulation and scenario-by-scenario analysis, this embodiment is the first to combine the need for rapid prediction of river network aquatic habitat response with time-series machine learning methods, which can establish a mapping relationship between monthly runoff processes and river network aquatic habitat response, thus forming a new technical solution suitable for complex river network ecological analysis.

[0201] (2) Value: This embodiment addresses practical needs such as river network ecological protection, ecological flow assurance, water resource allocation, ecological restoration evaluation, and ecological impact analysis of water-related projects. It can provide efficient technical support for the analysis of river network aquatic habitat changes under multiple water inflow scenarios. This technical solution can not only serve river network ecological management and engineering scheme comparison, but also provide a basis for ecological scheduling optimization, sensitive area identification, and water ecological risk assessment. Therefore, it has high engineering application value and promotion value.

[0202] (3) Practicality: This embodiment uses existing hydrodynamic simulation and aquatic habitat evaluation results as the training basis. After the model training is completed, it can directly and quickly predict new runoff scenarios without having to run the complete hydrodynamic numerical model repeatedly for each set of conditions. This significantly reduces the amount of computation, shortens the analysis cycle, and improves work efficiency. Especially in scenarios involving multiple months, multiple boundary combinations, and rapid comparison of multiple schemes, this embodiment can better meet the actual requirements for rapid assessment and timely response in engineering practice, and has significant practical effects.

[0203] (4) Systematic Approach: This embodiment is not a single prediction model or a single-step improvement, but rather a complete technical system encompassing basic data acquisition and preprocessing, construction of river network hydrodynamic models, establishment of aquatic habitat response indicators, construction of training sample sets, training of machine learning surrogate models, and rapid prediction output. The steps are logically clear and closely connected, reflecting both the physical basis of hydrodynamic processes and aquatic habitat response analysis, and the technical advantages of rapid prediction by machine learning. It possesses strong integrity, completeness, and feasibility.

[0204] Reference Figure 3 This application also provides a rapid prediction device for river network aquatic habitat response, which can realize the above-mentioned rapid prediction method for river network aquatic habitat response. The device includes: The data processing unit is used to collect and preprocess river network hydrological data and basic river channel geographic data; The model building unit is used to establish a river network hydrodynamic numerical model based on the preprocessed river network hydrological data and the basic geographic data of the river channel. The hydrodynamic determination unit is used to determine the hydrodynamic results of the river network based on the numerical model of river network hydrodynamics. The index determination unit is used to convert the river network hydrodynamic results into a first river network aquatic habitat response index. The sample set construction unit is used to construct a structured time series sample set with the first runoff scenario as the input variable and the first river network aquatic habitat response index as the output variable. The model training unit is used to train a long short-term memory network model using the structured temporal sample set. The indicator prediction unit is used to predict the response indicators of the second river network aquatic habitat corresponding to the second runoff scenario using the trained long short-term memory network model.

[0205] It is understood that the content of the above method embodiments is applicable to the present device embodiments. The specific functions implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0206] This application also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the method of this application. This electronic device can be any smart terminal, including tablet computers, in-vehicle computers, etc.

[0207] It is understood that the content of the above method embodiments is applicable to the device embodiments. The specific functions implemented by the device embodiments are the same as those of the methods of this application, and the beneficial effects achieved are the same as those achieved by the methods of this application.

[0208] Figure 4 The hardware structure of an electronic device according to another embodiment is illustrated. The electronic device includes: The processor 101 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application. The memory 102 can be implemented as a read-only memory (ROM), static storage device, dynamic storage device, or random access memory (RAM). The memory 102 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 102 and is called and executed by the processor 101. Input / output interface 103 is used to implement information input and output; The communication interface 104 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.). Bus 105 transmits information between various components of the device (e.g., processor 101, memory 102, input / output interface 103, and communication interface 104); The processor 101, memory 102, input / output interface 103 and communication interface 104 are connected to each other within the device via bus 105.

[0209] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method of this application.

[0210] It is understood that the content of the above method embodiments is applicable to this storage medium embodiment. The specific functions implemented in this storage medium embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0211] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0212] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0213] The embodiments described in this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

[0214] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of this application, and may include more or fewer steps than shown, or combine certain steps, or different steps.

[0215] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0216] Those skilled in the art will understand that all or some of the steps in the methods disclosed above, as well as the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, or suitable combinations thereof.

[0217] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0218] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0219] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of the units described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0220] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0221] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0222] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes multiple instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0223] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the claims of the present application.

Claims

1. A rapid prediction method for the response of river network aquatic habitats, characterized in that, The method includes the following steps: Collect and preprocess river network hydrological data and basic river channel geographic data; A numerical model of river network hydrodynamics is established based on the preprocessed river network hydrological data and the basic geographic data of the river channel. The hydrodynamic results of the river network are determined based on the numerical model of river network hydrodynamics. The hydrodynamic results of the river network are converted into the first river network aquatic habitat response index; Using the first runoff scenario as the input variable and the first river network aquatic habitat response index as the output variable, a structured time series sample set is constructed. The structured temporal sample set was used to train a long short-term memory network model; The trained Long Short-Term Memory (LSTM) network model was used to predict the response indicators of the second river network aquatic habitat corresponding to the second runoff scenario.

2. The rapid prediction method for river network aquatic habitat response according to claim 1, characterized in that, The collection and preprocessing of river network hydrological data and basic river channel geographic data includes the following steps: Suppose the river network has p control boundary sections, and historical data covers... If there are 12 months in a year, then the 12th month... The boundary section at the first Year The average monthly runoff is denoted as ;in, =1,2,...,p; =1,2,..., =1,2,...12; Definition of the first The first cross-section The multi-year average monthly runoff is: ; The multi-year average monthly runoff is standardized and missing values ​​are filled to obtain the pre-processed average monthly runoff. For any runoff scenario s to be simulated, define the corresponding monthly-scale runoff input vector. for: ； in, Indicates the month; Indicating the first case under scenario s The pre-treated monthly average runoff at each control section; Collect the basic geographic data of the river channel : ； Where R represents the total number of river segments; Let r be the length of the r-th river segment; The characteristic width of the river section; This refers to riverbed or cross-sectional elevation information; is the Manning roughness coefficient.

3. The rapid prediction method for river network aquatic habitat response according to claim 1, characterized in that, The process of establishing a river network hydrodynamic numerical model based on the preprocessed river network hydrological data and the basic geographical data of the river channel includes the following steps: Based on the processed river network topology, cross-sectional data and topographic elevation in the basic geographic data of the river channel, the study area is divided into river segments and grid units. Two-dimensional shallow water equations are adopted as hydrodynamic governing equations, and momentum equations are defined. The Manning formula is used as the subsoil resistance term; Apply flow boundary conditions to the upstream boundary and water level boundary conditions to the downstream boundary.

4. The rapid prediction method for river network aquatic habitat response according to claim 3, characterized in that, The process of converting the river network hydrodynamic results into a first river network aquatic habitat response index includes the following steps: Based on the flow velocity in the river network hydrodynamic results, each of the grid cells is... Determining the suitability of the flow rate specifically includes: Let the suitable flow velocity range for the target organism be... The optimal flow rate is Define the flow rate suitability Represented as: ； in, For grid cells In runoff scenario The suitability of the flow rate, Let the velocity of grid cell j be the velocity under runoff scenario s. Based on the water depth in the river network hydrodynamic results, each of the grid cells is... Determining the suitability of water depth specifically includes: Let the suitable water depth range be... The optimal water depth is Define the water depth suitability for: ； in, For grid cells In runoff scenario Suitable water depth For grid cells In runoff scenario The water depth is below; Based on flow velocity suitability and water depth suitability Determine the grid cells In runoff scenario The comprehensive aquatic habitat suitability index ; According to the comprehensive aquatic habitat suitability index The river network-scale index and the average aquatic habitat index of the entire river network were respectively determined as the first river network aquatic habitat response index; River network scale index for: ； The average aquatic habitat index of the entire river network for: ； in, For the first The area of ​​each grid cell, The number of the grid cells.

5. The rapid prediction method for river network aquatic habitat response according to claim 4, characterized in that, The construction of a structured time-series sample set, using the first runoff scenario as the input variable and the first river network aquatic habitat response index as the output variable, includes the following steps: For the Given samples, define the input feature vector. for: ； in, For the first The sample corresponds to the first Monthly runoff at each control section; Monthly encoding information; In the input feature vector Add extended feature vectors The enhanced input vector is obtained. ; Wherein, the enhanced input vector for: ； ； If the comprehensive index of the entire river network is used as the output variable, then the first... The labels for each sample are represented as follows: or ; If indicators from multiple river segments or multiple spatial partitions are used as the output variables, then the output vector is defined as follows: ;in, To determine the dimension of the output index; Construct the structured time series sample set as follows: ,or, ; in, The total number of samples; For the input variables; The output variable is denoted as .

6. The rapid prediction method for river network aquatic habitat response according to claim 5, characterized in that, The training of the long short-term memory network model using the structured time-series sample set includes the following steps: According to the The input feature vector or the enhanced input vector of each sample constructs an input time series sequence with a time window length of T. Definition of the first The output label corresponding to each sample is the first river network aquatic habitat response index at the target time. The long short-term memory network model is trained based on the input time sequence and the output label.

7. The rapid prediction method for river network aquatic habitat response according to claim 6, characterized in that, The method further includes the following steps: The long short-term memory network model was trained using the mean squared error loss function. The parameter set of the Long Short-Term Memory network model is updated using the backpropagation time-travel algorithm; wherein, in the e-th iteration, the parameter update expression is: ; The learning rate; The mean squared error loss function is set of parameters. The gradient.

8. A rapid prediction device for river network aquatic habitat response, characterized in that, The device includes: The data processing unit is used to collect and preprocess river network hydrological data and basic river channel geographic data; The model building unit is used to establish a river network hydrodynamic numerical model based on the preprocessed river network hydrological data and the basic geographic data of the river channel. The hydrodynamic determination unit is used to determine the hydrodynamic results of the river network based on the numerical model of river network hydrodynamics. The index determination unit is used to convert the river network hydrodynamic results into a first river network aquatic habitat response index. The sample set construction unit is used to construct a structured time series sample set with the first runoff scenario as the input variable and the first river network aquatic habitat response index as the output variable. The model training unit is used to train a long short-term memory network model using the structured temporal sample set. The indicator prediction unit is used to predict the response indicators of the second river network aquatic habitat corresponding to the second runoff scenario using the trained long short-term memory network model.

9. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 7.

Images