Drainage basin rainfall runoff forecasting method and system based on neural network

Abstract

The invention relates to the technical field of rainfall runoff simulation, and discloses a drainage basin rainfall runoff forecasting method and system based on a neural network, and the method comprises the steps: obtaining terrain basic data, an actually measured rainfall runoff process diagram and historical rainfall runoff data; inputting the historical rainfall runoff data into a back propagation neural network to obtain initial model parameters of a rainfall runoff model; according to the initial model parameters, obtaining optimal model parameters through a precision optimal algorithm; taking the optimal model parameters as preset parameters of a rainfall runoff model to obtain an optimal rainfall runoff model; and calculating rainfall runoff data by using the optimal rainfall runoff model and a preset runoff transmission model according to the terrain basic data and the actually measured rainfall runoff process diagram. According to the method, parameter correction is performed on the hydrological model by combining the artificial neural network, and the runoff prediction result accuracy is improved.

Description

Technical Field

[0001] The invention relates to the technical field of rainfall runoff simulation, and in particular to a basin rainfall runoff forecasting method and system based on a neural network. Background Art

[0002] Hydrological models are systems used to generate various calculation results. They can simulate complex hydrological processes and then generate simulated runoff processes in watersheds. To meet application requirements, the model must be fully interpreted based on the correlation between the input of the model's internal system and its results. Hydrological models can be divided into empirical models, conceptual models, and distributed models according to their development history. Empirical models have a long history of development and are widely used because they have fewer input factors and are easier to master. However, they lack a physical basis and have limited fitting capabilities for hydrological processes. Conceptual models consider hydrological processes from multiple aspects such as physics, chemistry, biology, ecology, and fluid mechanics to establish input-output relationships. However, due to their relatively simple input-output relationships, they can only be applied to watersheds with simple hydrological processes and strong regularity. A series of distributed hydrological models have been developed for complex watersheds. The model divides the watershed into grids, assumes that the watershed response function of each grid is the same, and focuses the simulation on the spatial variation analysis and prediction of input factors. Distributed hydrological models have high simulation accuracy, but are limited by complex input factors. Rainfall is the main input factor of the model. The main method of obtaining rainfall information is that the Meteorological Bureau regularly releases regional rainfall maps obtained by satellite remote sensing and radar. However, due to the insufficient spatial resolution of this data, it is difficult to use it for accurate rainfall forecasting, resulting in a decrease in the accuracy of basin simulation. With the development of data mining technology, artificial intelligence technology has gradually merged with hydrological models, and the model has been trained through a large number of data sets to improve the forecast accuracy of the model.

[0003] There have been many studies on the application of data mining technology in the field of hydrology and water resources, and it has a certain guiding role in the development of water resources allocation, hydrological forecasting and other fields. At present, the combination of artificial intelligence and hydrological models is mostly to apply artificial intelligence methods to the parameter calibration and input factor selection of traditional models, such as artificial neural networks (ANN), support vector machines (SVM), genetic algorithms (GA), etc., to improve the simulation accuracy of the model. However, this combination is mainly based on the analysis of traditional conceptual hydrological models, which has the problem of limited parameter information and cannot fully reflect the dynamic changes in space, resulting in inaccurate prediction results. Summary of the invention

[0004] The present invention provides a basin rainfall runoff forecasting method, system, electronic equipment and storage medium based on a neural network, so as to improve the prediction accuracy of a hydrological model.

[0005] In the first aspect, in order to solve the above technical problems, the present invention provides a basin rainfall runoff forecasting method based on a neural network, comprising: Obtain basic terrain data, measured rainfall-runoff process maps and historical rainfall-runoff data; Inputting the historical rainfall-runoff data into a back propagation neural network to obtain initial model parameters of a rainfall-runoff model; According to the initial model parameters, the optimal model parameters are obtained by using the best accuracy algorithm; Using the optimal model parameters as preset parameters of the rainfall-runoff model to obtain the optimal rainfall-runoff model; According to the basic terrain data and the measured rainfall-runoff process diagram, rainfall-runoff data are calculated using the optimal rainfall-runoff model and a preset runoff transmission model.

[0006] In an optional implementation, the inputting of the historical rainfall-runoff data into a back propagation neural network to obtain initial model parameters of the rainfall-runoff model includes: The historical rainfall runoff data include historical rain gauge rainfall data and historical outlet measured flow; By means of parameter sensitivity analysis, the parameters of the rainfall-runoff model are selected, and then the initial values ​​of these parameters are randomly set as the initial weights and thresholds of the back propagation neural network; Inputting the historical rain gauge rainfall data and the historical outlet measured flow into a rainfall runoff model to perform rainfall runoff simulation; Calculate the error value of the back propagation neural network according to the simulation result; The back propagation neural network performs weight and threshold training through the gradient descent method, completes parameter optimization, and obtains initial model parameters of the rainfall runoff model.

[0007] In an optional implementation, obtaining optimal model parameters by using an optimal accuracy algorithm based on the initial model parameters includes: ; in, is the relative error of runoff, is the runoff volume at the i-th moment calculated by the model, is the measured runoff at the i-th moment, N is the total amount of runoff data; ; in, is the Nasher efficiency coefficient, is the average value of the measured runoff at the i-th moment, is the runoff volume at the i-th moment calculated by the model, is the measured runoff at the i-th moment; The runoff relative error and the Nasher efficiency coefficient are used as rainfall fitting accuracy indicators of the optimal accuracy algorithm; If the rainfall fitting accuracy index does not meet the preset threshold, the initial model parameters are input into the reverse prediction neural network for iteration, and the optimal model parameters are re-output.

[0008] In an optional implementation, the step of using the optimal model parameters as preset parameters of a rainfall-runoff model to obtain an optimal rainfall-runoff model includes: According to the optimal model parameters, a confluence relationship matrix between hypercubes is calculated, and an optimal rainfall-runoff model is determined according to the confluence relationship matrix; The optimal rainfall-runoff model is a confluence relationship matrix. If there are n hypercubes, the confluence relationship matrix C is a Matrix, where It represents the proportion of water flow contribution of the j-th hypercube to the i-th hypercube.

[0009] In an optional implementation, the calculating and obtaining rainfall runoff data according to the terrain basic data and the measured rainfall runoff process diagram using an optimal rainfall runoff model and a preset runoff transmission model includes: According to the basic terrain data, the watershed is divided into a number of hypercubes, and the terrain factor of each hypercube is calculated; Calculating the water storage capacity of the hypercube based on the terrain factors; According to the measured rainfall runoff process diagram, the rainfall is divided into a number of hypercubes, and the inflow of each cube is calculated; If the water storage capacity is less than the inflow, the outflow of each hypercube is calculated according to the terrain factor and the inflow; ; in, is the outflow of the ith unit at the jth moment, is the drainage area of ​​the ith unit, is the inflow of the ith unit at the jth moment; inputting the outflow into an optimal rainfall-runoff model to output a hypercube flow; inputting the hypercube flow into a runoff transport model to output rainfall runoff data; The runoff transport model is used to simulate the flow of water in a region, and the rainfall runoff data is obtained by solving the linear trough storage equation and the water balance equation.

[0010] In an optional implementation, before dividing the rainfall into a plurality of hypercubes according to the measured rainfall-runoff process diagram and calculating the inflow of each cube, the method further includes: A series of distributed rain gauges are set at the center point of the top surface of each hypercube; ; in, For flow rate, is the rainfall in the ith area into the flow, is the outlet interception flow of the ith area, and n is the number of rain gauges set.

[0011] In an optional embodiment, the terrain factors include slope, shape factor, water flow path length, and water flow path slope.

[0012] In a second aspect, the present invention provides a basin rainfall runoff forecasting system based on a neural network, comprising: Data acquisition module, used to obtain basic terrain data, measured rainfall runoff process diagrams and historical rainfall runoff data; An initial parameter module, used for inputting the historical rainfall-runoff data into a back propagation neural network to obtain initial model parameters of a rainfall-runoff model; A parameter optimization module, used to obtain optimal model parameters according to the initial model parameters through an optimal accuracy algorithm; A model optimization module, used for taking the optimal model parameters as preset parameters of a rainfall-runoff model to obtain an optimal rainfall-runoff model; The result output module is used to calculate the rainfall runoff data according to the terrain basic data and the measured rainfall runoff process diagram, using the optimal rainfall runoff model and the preset runoff transmission model.

[0013] In a third aspect, the present invention also provides an electronic device comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, it implements any one of the above-mentioned neural network-based basin rainfall runoff forecasting methods.

[0014] In a fourth aspect, the present invention further provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute any one of the above-mentioned neural network-based watershed rainfall runoff forecasting methods.

[0015] Compared with the prior art, the present invention has the following beneficial effects: the present invention discloses a method and system for basin rainfall runoff forecasting based on neural network, the method comprising obtaining terrain basic data, measured rainfall runoff process diagram and historical rainfall runoff data; inputting the historical rainfall runoff data into the back propagation neural network to obtain the initial model parameters of the rainfall runoff model; according to the initial model parameters, the optimal model parameters are obtained by the best accuracy algorithm; the optimal model parameters are used as the preset parameters of the rainfall runoff model to obtain the optimal rainfall runoff model; according to the terrain basic data and the measured rainfall runoff process diagram, the rainfall runoff data are calculated by using the optimal rainfall runoff model and the preset runoff transmission model. By constructing a rainfall-runoff model based on DRIVE and RAPID coupling, combined with hypercube terrain factors, parameters and surface rainfall data, the accurate simulation of the flow of each sub-region in the basin is realized. The method can effectively capture the process of converting rainfall events into surface runoff, and accurately calculate the water flow transfer between each hypercube through the confluence relationship matrix, thereby improving the accuracy of flood forecasting and water resources management. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a schematic flow chart of a method for forecasting rainfall and runoff in a watershed based on a neural network provided by the first embodiment of the present invention; Figure 2 It is a schematic diagram of the structure of a basin rainfall runoff forecasting system based on a neural network provided in the second embodiment of the present invention. DETAILED DESCRIPTION

[0017] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0018] Hydrological models are systems used to generate various calculation results. They can simulate complex hydrological processes and then generate simulated runoff processes in watersheds. To meet application requirements, the model must be fully interpreted based on the correlation between the input of the model's internal system and its results. Hydrological models can be divided into empirical models, conceptual models, and distributed models according to their development history. Empirical models have a long history of development and are widely used because they have fewer input factors and are easier to master. However, they lack a physical basis and have limited fitting capabilities for hydrological processes. Conceptual models consider hydrological processes from multiple aspects such as physics, chemistry, biology, ecology, and fluid mechanics to establish input-output relationships. However, due to their relatively simple input-output relationships, they can only be applied to watersheds with simple hydrological processes and strong regularity. A series of distributed hydrological models have been developed for complex watersheds. The model divides the watershed into grids, assumes that the watershed response function of each grid is the same, and focuses the simulation on the spatial variation analysis and prediction of input factors. Distributed hydrological models have high simulation accuracy, but are limited by complex input factors. Rainfall is the main input factor of the model. The main method of obtaining rainfall information is that the Meteorological Bureau regularly releases regional rainfall maps obtained by satellite remote sensing and radar. However, due to the insufficient spatial resolution of this data, it is difficult to use it for accurate rainfall forecasting, resulting in a decrease in the accuracy of basin simulation. With the development of data mining technology, artificial intelligence technology has gradually merged with hydrological models, and the model has been trained through a large number of data sets to improve the forecast accuracy of the model.

[0019] There have been many studies on the application of data mining technology in the field of hydrology and water resources, and it has a certain guiding role in the development of water resources allocation, hydrological forecasting and other fields. At present, the combination of artificial intelligence and hydrological models is mostly to apply artificial intelligence methods to the parameter calibration and input factor selection of traditional models, such as artificial neural networks (ANN), support vector machines (SVM), genetic algorithms (GA), etc., to improve the simulation accuracy of the model. However, this combination is mainly based on the analysis of traditional conceptual hydrological models, which has the problem of limited parameter information and cannot fully reflect the dynamic changes in space, resulting in inaccurate prediction results.

[0020] Reference Figure 1 The first embodiment of the present invention provides a method for forecasting rainfall runoff in a watershed based on a neural network, comprising the following steps: S11, obtaining basic terrain data, measured rainfall-runoff process diagram and historical rainfall-runoff data; S12, inputting the historical rainfall-runoff data into a back propagation neural network to obtain initial model parameters of a rainfall-runoff model; S13, obtaining optimal model parameters through an optimal precision algorithm according to the initial model parameters; S14, using the optimal model parameters as preset parameters of a rainfall-runoff model to obtain an optimal rainfall-runoff model; S15, according to the basic terrain data and the measured rainfall-runoff process diagram, using the optimal rainfall-runoff model and the preset runoff transmission model, calculate and obtain rainfall-runoff data.

[0021] In step S11, basic terrain data, measured rainfall-runoff process diagrams and historical rainfall-runoff data are obtained.

[0022] In one embodiment, the basic terrain data includes a digital elevation model, a topographic map, vector data, and remote sensing data. The digital elevation model can be obtained from the official website of ASTER GDEM (Advanced Spaceborne Thermal Emission and Reflection Radiometer Global Digital Elevation Model), which is a public source. GIS software can be used to download and process these data. The topographic map and vector data are downloaded and processed by GIS software in the National Administration of Surveying, Mapping and Geoinformation, local surveying and mapping departments or geographic information service centers. Remote sensing data comes from satellite images. The measured rainfall runoff process map is recorded by hydrological stations operated by national or local water conservancy departments and meteorological bureaus, including rainfall, flow and other data. Historical rainfall runoff data is recorded in the China Meteorological Bureau or local meteorological stations for a long time. The required historical rainfall runoff data can also be found through the global data set provided by the World Meteorological Organization.

[0023] It is worth mentioning that the Digital Elevation Model (DEM) is a key data set used in the Geographic Information System (GIS) to represent the topography of the earth's surface. DEM represents the elevation information of the ground through a set of ordered numerical arrays, each of which represents the altitude of a specific location. These data can be used to generate topographic maps, conduct terrain analysis, simulate water flow paths, and evaluate land use. It is usually stored in a regular rectangular grid, but the grid points are not feature points and cannot accurately represent the micro-topographic features. Alternatively, an irregular triangulated network can be used for storage, which can more accurately represent complex terrain and avoid loss of interpolation accuracy.

[0024] In step S12, the historical rainfall-runoff data is input into a back propagation neural network to obtain initial model parameters of the rainfall-runoff model.

[0025] In one implementation, the historical rainfall runoff data is input into a back propagation neural network to obtain initial model parameters of a rainfall runoff model, including: the historical rainfall runoff data includes historical rain gauge rainfall data and historical outlet measured flow; the parameters of the rainfall runoff model are selected through parameter sensitivity analysis, and then the initial values ​​of these parameters are randomly set as initial weights and thresholds of the back propagation neural network; the historical rain gauge rainfall data and the historical outlet measured flow are input into the rainfall runoff model to perform rainfall runoff simulation; the error value of the back propagation neural network is calculated according to the simulation results; the back propagation neural network performs weight and threshold training through the gradient descent method to complete parameter optimization and obtain the initial model parameters of the rainfall runoff model.

[0026] In one implementation, the network weights and bias terms are randomly initialized; the input data is passed through the network to calculate the output of each node; the difference between the predicted value and the actual value is calculated using the loss function; the gradient of the loss with respect to each weight and bias is calculated by the chain rule; the weights and biases are updated to minimize the loss function; the forward propagation and backpropagation processes are repeated until a predetermined number of training rounds is reached or convergence conditions are met. For the activation function, Rectified LinearUnit is selected, the mean square error is used as the loss function, and the stochastic gradient descent method is used as the optimization algorithm. The model performance is evaluated on the validation set. If the model performs poorly on the validation set, consider adjusting the network structure, hyperparameters, or regularizing.

[0027] In step S13, the optimal model parameters are obtained according to the initial model parameters through the precision optimization algorithm.

[0028] In one implementation, obtaining optimal model parameters by using an optimal precision algorithm based on the initial model parameters includes: ; in, is the relative error of runoff, is the runoff volume at the i-th moment calculated by the model, is the measured runoff at the i-th moment, N is the total amount of runoff data; ; in, is the Nasher efficiency coefficient, is the average value of the measured runoff at the i-th moment, is the runoff volume at the i-th moment calculated by the model, is the measured runoff at the i-th moment; The runoff relative error and the Nasher efficiency coefficient are used as rainfall fitting accuracy indicators of the optimal accuracy algorithm; If the rainfall fitting accuracy index does not meet the preset threshold, the initial model parameters are input into the reverse prediction neural network for iteration, and the optimal model parameters are re-output.

[0029] It is worth noting that the relative error of runoff is a simple way to measure the difference between the model simulation results and the observed data. A smaller value means that the absolute deviation between the model prediction and the measured data is smaller. The Nash-Sutcliffe efficiency coefficient is a more complex evaluation criterion. It not only considers the average deviation between the model prediction and the observed data, but also considers their variance relative to the mean of the observed data. The value range of the Nash-Sutcliffe efficiency coefficient is usually from negative infinity to 1. When the Nash-Sutcliffe efficiency coefficient is close to 1, it indicates that the model has a high degree of fit; if the Nash-Sutcliffe efficiency coefficient is 0, it means that the performance of the model is not better than using the mean of the observed data as a prediction; if the Nash-Sutcliffe efficiency coefficient is less than 0, it means that the performance of the model is even worse than a simple mean prediction. The relative error of runoff and the Nash-Sutcliffe efficiency coefficient are used as the loss function of the back-propagation neural network. The calculated accuracy index is compared with the preset threshold. If the accuracy index meets the requirements, the current parameters are the optimal solution.

[0030] In step S14, the optimal model parameters are used as preset parameters of the rainfall-runoff model to obtain the optimal rainfall-runoff model.

[0031] It is worth noting that the rainfall-runoff model adopts the DRIVE model, which is a distributed rainfall-runoff model mainly used to simulate the rainfall-runoff process in the basin. It combines multiple factors such as topography, soil type, and land use, and can provide detailed water flow path analysis and runoff generation prediction. The DRIVE model is designed to improve the accuracy of flood forecasting and support water resources management and environmental impact assessment. It first divides the basin into multiple sub-areas (hypercubes), each of which has its own specific topography, soil and land use characteristics. In this way, the model can more accurately capture the impact of spatial variability on runoff generation. The model considers a variety of hydrological processes, including rainfall infiltration, surface runoff, groundwater runoff, etc., and truly simulates the actual hydrological process. In principle, the model uses the Green-Ammert method to calculate runoff generation. The water flow path is determined based on the terrain data, the confluence relationship matrix is ​​constructed, the water flow transfer between the sub-areas is calculated, and the flow at the outlet of the basin is finally obtained. At the same time, the model is calibrated using historical observation data, and the parameters are adjusted to improve the simulation accuracy. The performance of the model is evaluated through the validation data set to ensure the reliability of the model under different conditions. In the present invention, it is used to calculate the total flow rate.

[0032] It is worth noting that Darcy's law is the basic equation that describes the flow of fluids in porous media. The law shows that the seepage rate of water in saturated soil is proportional to the hydraulic gradient. In groundwater dynamics, Darcy's law is used to calculate the flow of groundwater in aquifers. For example, when simulating the impact of different soil types on groundwater recharge in a basin, Darcy's law can be used to estimate the groundwater flow rate in different areas. In the rainfall-runoff process, Darcy's law can be used to simulate the soil infiltration process. When the rainfall exceeds the infiltration capacity of the soil, the excess water will form surface runoff. The permeability of the soil directly affects the rate at which surface runoff is generated. As the basic principle of the infiltration rate of water in the DRIVE model, it is used in the present invention to calculate the total flow rate.

[0033] In step S15, rainfall runoff data is calculated based on the terrain basic data and the measured rainfall runoff process diagram using the optimal rainfall runoff model and the preset runoff transmission model.

[0034] In one embodiment, the rainfall runoff data is calculated based on the terrain basic data and the measured rainfall runoff process map using an optimal rainfall runoff model and a preset runoff transmission model, including: According to the basic terrain data, the watershed is divided into a number of hypercubes, and the terrain factor of each hypercube is calculated; the water storage capacity of the hypercube is calculated according to the terrain factor; according to the measured rainfall runoff process diagram, the rainfall is divided into a number of hypercubes, and the inflow of each cube is calculated; if the water storage capacity is less than the inflow, the outflow of each hypercube is calculated according to the terrain factor and the inflow; ; in, is the outflow of the ith unit at the jth moment, is the drainage area of ​​the ith unit, is the inflow of the ith unit at the jth moment; Input the outflow into the optimal rainfall-runoff model to output the hypercube flow; input the hypercube flow into the runoff transport model to output rainfall-runoff data; The runoff transport model is used to simulate the flow of water in a region, and the rainfall runoff data is obtained by solving the linear trough storage equation and the water balance equation.

[0035] It is worth noting that the runoff transport model adopts the rapid model. This is a model for parallel calculation of runoff transport, which is designed to simulate the flow process of surface water in a watershed. This model is particularly suitable for runoff simulation of large-scale watersheds and long time series, and can efficiently process large amounts of data. Its core is the river confluence algorithm based on the Muskingum method, which is a method widely used in hydrology to simulate the movement of water in rivers. Furthermore, the core of the Muskingum method is to solve the finite difference formula of the flow by solving the linear channel storage equation and the water balance equation. The runoff transport model can obtain rainfall runoff data by inputting the hypercube flow output. The parameters required by the runoff transport model can be optimized using the same optimization method as the optimal model parameters of the rainfall runoff data, or the particle swarm optimization algorithm can be used for parameter optimization.

[0036] It is worth noting that, before dividing the rainfall into a plurality of hypercubes according to the measured rainfall runoff process diagram and calculating the inflow of each cube, the method further includes: A series of distributed rain gauges are set at the center point of the top surface of each hypercube; ; in, For flow rate, is the rainfall in the ith area into the flow, is the outlet interception flow of the ith area, and n is the number of rain gauges set.

[0037] It is worth noting that rainfall is usually highly variable in space, especially in areas with complex terrain or changeable climate. A single rain gauge cannot accurately reflect the actual rainfall distribution in the entire basin or region. Hydrological models are very sensitive to the quality of input data. Data collected using distributed rain gauges can provide more refined inputs for the model, thereby improving the accuracy of the output of the rainfall runoff model. In the present invention, the use of rain gauges as a data source for cube inflow calculations has higher reliability than the data source obtained from a single rain gauge. At the same time, rain gauge data can be used in combination with other hydrological observation data to more fully understand the water cycle process. This helps to establish more complex hydrological models and improve the understanding and prediction capabilities of hydrological phenomena. At the same time, for areas without data, rain gauges help establish basic rainfall data records. This helps to understand the climate characteristics, rainfall patterns and seasonal changes in the area. In areas without data, data obtained by installing rain gauges can also be used to preliminarily calibrate model parameters to improve the applicability of the model in the local area.

[0038] It is worth noting that the present invention improves the computational efficiency of the hydrological model by using hypercube units instead of traditional Thiessen polygons. The core advantage of this method is that it simplifies the process of spatial data processing and analysis while maintaining effective capture of terrain features. Each hypercube unit can be regarded as a small square of the same size, covering the entire study area. Since each unit is of regular shape, the relationship between adjacent units is clear and consistent, so it can be optimized at the algorithm level, such as parallel processing and vectorization operations. For large-scale watershed simulation, this structured grid system can better utilize modern computer hardware resources, such as multi-core processors and graphics processing units. Despite the use of simple hypercube units, the invention does not ignore the importance of terrain factors. Key terrain factors such as slope, shape factor, water flow path length and water flow path slope can still be accurately calculated through high-resolution digital elevation model data and assigned to each hypercube unit. In this way, the details of terrain information can be retained even under a regular grid structure. When processing rainfall data, the observation data provided by distributed rain gauges can be assigned to corresponding hypercube units. This not only improves the spatial resolution of rainfall data, but also makes the interpolation of rainfall fields more intuitive and efficient.

[0039] It is worth noting that the terrain factors include slope, shape factor, flow path length, and flow path slope. These factors have an impact on the formation, convergence, and transmission of water flow. Among them, slope refers to the degree of ground inclination, usually expressed as a percentage or angle. The greater the slope, the faster the speed of surface water flow, thereby reducing the chance of water infiltration and increasing runoff. Shape factors usually refer to the morphological characteristics of the basin, such as the aspect ratio and circularity of the basin. The shape of the basin can affect the concentration and speed of water flow. For example, a narrow and long basin causes water to converge to the outlet faster, while a circular or nearly circular basin causes the water flow to be more evenly distributed. The basin morphological characteristics are abstracted as shape factors for the convenience of subsequent calculations. Flow path length The flow path length is the shortest surface distance from a point to the outlet of the basin. A longer flow path means that the water needs to travel a longer distance to reach the outlet of the basin, which will result in more water infiltration and evaporation losses, thereby reducing the final runoff. The flow path slope is the average slope along the direction of the flow path. Similar to slope, the slope of a flow path affects the velocity and energy of water flow, which in turn affects the erosion process and runoff formation. Steeper path slopes result in faster water flow and greater erosion. Topographic factors are used to define hydrological response units in sub-basins to improve the accuracy of rainfall-runoff model predictions.

[0040] In summary, the invention proposes a method for improving the optimization of hydrological model parameters, aiming to improve the accuracy and reliability of runoff prediction. In this method, the core is to optimize the model parameters by using high-precision rainfall data and terrain factors, so as to better simulate the surface runoff process. First, the invention uses a distributed rain gauge network to collect detailed rainfall information. This distributed monitoring system can capture the spatial variability of rainfall and provide more accurate input data, which is particularly important for areas with complex terrain or variable climate. Secondly, the invention considers multiple terrain factors, including slope, shape factor, water flow path length, and water flow path slope. These terrain factors have a significant impact on the formation, convergence and transmission of water flow. By incorporating them into the model, the actual hydrological process can be more realistically reflected. In order to evaluate the performance of the model, the invention introduces two key indicators: runoff relative error (RE) and Nash-Sutcliffe efficiency coefficient (NSE). These two indicators measure the difference between the model prediction and the measured runoff from the perspective of absolute error and relative error, respectively. When the initial model parameters cannot meet the preset accuracy requirements, the invention adopts an iterative optimization method based on back-propagation neural network to adjust the parameters. This method uses the powerful nonlinear fitting ability of neural networks to gradually approach the optimal solution through continuous learning and adjustment. The entire optimization process includes initializing parameters, running hydrological models, calculating accuracy indicators, checking whether threshold conditions are met, and if not, updating parameters through neural networks until satisfactory accuracy or the maximum number of iterations is achieved. In addition, the invention also emphasizes the importance of installing rain gauges in areas without data. In these areas where historical data records are lacking, rain gauges not only provide basic data support, but also provide valuable real-time information for flood forecasting, water resources management, scientific research, etc. In this way, even in the case of scarce data, an effective hydrological model can be established to provide a scientific basis for local flood control and disaster reduction, agricultural irrigation, urban planning, etc. In general, this invention improves the prediction accuracy of rainfall runoff models by comprehensively using advanced monitoring technology, complex terrain factor analysis, and intelligent optimization algorithms.

[0041] Reference Figure 2 The second embodiment of the present invention provides a basin rainfall runoff forecasting system based on a neural network, comprising: Data acquisition module, used to obtain basic terrain data, measured rainfall runoff process diagrams and historical rainfall runoff data; An initial parameter module, used for inputting the historical rainfall-runoff data into a back propagation neural network to obtain initial model parameters of a rainfall-runoff model; A parameter optimization module, used to obtain optimal model parameters according to the initial model parameters through an optimal accuracy algorithm; A model optimization module, used for taking the optimal model parameters as preset parameters of a rainfall-runoff model to obtain an optimal rainfall-runoff model; The result output module is used to calculate the rainfall runoff data according to the terrain basic data and the measured rainfall runoff process diagram, using the optimal rainfall runoff model and the preset runoff transmission model.

[0042] Preferably, the data acquisition module is used to: Obtain basic terrain data, measured rainfall-runoff process maps, and historical rainfall-runoff data.

[0043] Preferably, the initial parameter module is used to: The historical rainfall-runoff data is input into the back propagation neural network to obtain the initial model parameters of the rainfall-runoff model, including: The historical rainfall runoff data include historical rain gauge rainfall data and historical outlet measured flow; By means of parameter sensitivity analysis, the parameters of the rainfall-runoff model are selected, and then the initial values ​​of these parameters are randomly set as the initial weights and thresholds of the back propagation neural network; Inputting the historical rain gauge rainfall data and the historical outlet measured flow into a rainfall runoff model to perform rainfall runoff simulation; Calculate the error value of the back propagation neural network according to the simulation result; The back propagation neural network performs weight and threshold training through the gradient descent method, completes parameter optimization, and obtains initial model parameters of the rainfall runoff model.

[0044] Preferably, the parameter optimization module is used to: According to the initial model parameters, the optimal model parameters are obtained by using the best accuracy algorithm, including: ; in, is the relative error of runoff, is the runoff volume at the i-th moment calculated by the model, is the measured runoff at the i-th moment, N is the total amount of runoff data; ; in, is the Nasher efficiency coefficient, is the average value of the measured runoff at the i-th moment, is the runoff volume at the i-th moment calculated by the model, is the measured runoff at the i-th moment; The runoff relative error and the Nasher efficiency coefficient are used as rainfall fitting accuracy indicators of the optimal accuracy algorithm; If the rainfall fitting accuracy index does not meet the preset threshold, the initial model parameters are input into the reverse prediction neural network for iteration, and the optimal model parameters are re-output.

[0045] Preferably, the model optimization module is used to: The optimal model parameters are used as preset parameters of the rainfall-runoff model to obtain the optimal rainfall-runoff model, including: According to the optimal model parameters, a confluence relationship matrix between hypercubes is calculated, and an optimal rainfall-runoff model is determined according to the confluence relationship matrix; The optimal rainfall-runoff model is a confluence relationship matrix. If there are n hypercubes, the confluence relationship matrix C is a Matrix, where It represents the proportion of water flow contribution of the j-th hypercube to the i-th hypercube.

[0046] Preferably, the result output module is used to: According to the terrain basic data and the measured rainfall runoff process diagram, the rainfall runoff data is calculated using the optimal rainfall runoff model and the runoff transmission model, including: According to the basic terrain data, the watershed is divided into a number of hypercubes, and the terrain factor of each hypercube is calculated; Calculating the water storage capacity of the hypercube based on the terrain factors; According to the measured rainfall runoff process diagram, the rainfall is divided into a number of hypercubes, and the inflow of each cube is calculated; If the water storage capacity is less than the inflow, the outflow of each hypercube is calculated according to the terrain factor and the inflow; ; in, is the outflow of the ith unit at the jth moment, is the drainage area of ​​the ith unit, is the inflow of the ith unit at the jth moment; inputting the outflow into an optimal rainfall-runoff model to output a hypercube flow; inputting the hypercube flow into a runoff transport model to output rainfall runoff data; The runoff transport model is used to simulate the flow of water in a region, and the rainfall runoff data is obtained by solving the linear trough storage equation and the water balance equation.

[0047] Preferably, before dividing the rainfall into a plurality of hypercubes according to the measured rainfall runoff process diagram and calculating the inflow of each cube, the method further comprises: A series of distributed rain gauges are set at the center point of the top surface of each hypercube; ; in, For flow rate, is the rainfall in the ith area into the flow, is the outlet interception flow of the ith area, and n is the number of rain gauges set.

[0048] Preferably, the terrain factors include slope, shape factor, water flow path length, and water flow path slope.

[0049] It should be noted that the neural network-based watershed rainfall runoff forecasting system provided in an embodiment of the present invention is used to execute all the process steps of the neural network-based watershed rainfall runoff forecasting method in the above-mentioned embodiment. The working principles and beneficial effects of the two correspond one to one, and thus will not be repeated here.

[0050] The embodiment of the present invention also provides an electronic device. The electronic device includes: a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a result output program. When the processor executes the computer program, the steps in the above-mentioned embodiments of the neural network-based watershed rainfall runoff forecasting method are implemented, such as Figure 1 Alternatively, when the processor executes the computer program, the functions of the modules / units in the above-mentioned device embodiments are realized, such as the data acquisition module.

[0051] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to implement the present invention. The one or more modules / units may be a series of computer program instruction segments capable of implementing specific functions, which are used to describe the execution process of the computer program in the electronic device.

[0052] The electronic device may be a computing device such as a desktop computer, a notebook, a PDA, and a smart tablet. The electronic device may include, but is not limited to, a processor and a memory. Those skilled in the art will appreciate that the above components are merely examples of electronic devices and do not constitute a limitation on the electronic device. The electronic device may include more or fewer components than the above components, or may combine certain components, or different components. For example, the electronic device may also include input and output devices, network access devices, buses, etc.

[0053] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the electronic device, and uses various interfaces and lines to connect various parts of the entire electronic device.

[0054] The memory can be used to store the computer program and / or module, and the processor realizes various functions of the electronic device by running or executing the computer program and / or module stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0055] Wherein, if the module / unit integrated in the electronic device is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0056] It should be noted that the device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the accompanying drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art may understand and implement it without paying any creative effort.

[0057] The specific embodiments described above further illustrate the purpose, technical solutions and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. It is particularly pointed out that for those skilled in the art, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of protection of the present invention.

Claims

1. A basin rainfall runoff forecasting method based on neural network, characterized in that: include: Obtain basic terrain data, measured rainfall-runoff process maps and historical rainfall-runoff data; Inputting the historical rainfall-runoff data into a back propagation neural network to obtain initial model parameters of a rainfall-runoff model; According to the initial model parameters, the optimal model parameters are obtained by using the best accuracy algorithm; Using the optimal model parameters as preset parameters of the rainfall-runoff model to obtain the optimal rainfall-runoff model; According to the basic terrain data and the measured rainfall-runoff process diagram, rainfall-runoff data are calculated using the optimal rainfall-runoff model and a preset runoff transmission model.

2. The method for basin rainfall runoff forecasting based on neural network according to claim 1 is characterized in that: The historical rainfall-runoff data is input into the back propagation neural network to obtain the initial model parameters of the rainfall-runoff model, including: The historical rainfall runoff data include historical rain gauge rainfall data and historical outlet measured flow; By means of parameter sensitivity analysis, the parameters of the rainfall-runoff model are selected, and then the initial values ​​of these parameters are randomly set as the initial weights and thresholds of the back propagation neural network; Inputting the historical rain gauge rainfall data and the historical outlet measured flow into a rainfall runoff model to perform rainfall runoff simulation; Calculate the error value of the back propagation neural network according to the simulation result; The back propagation neural network performs weight and threshold training through the gradient descent method, completes parameter optimization, and obtains initial model parameters of the rainfall runoff model.

3. The method for basin rainfall runoff forecasting based on neural network according to claim 1 is characterized in that: The step of obtaining the optimal model parameters by using the best accuracy algorithm according to the initial model parameters includes: ； in, is the relative error of runoff, is the runoff volume at the i-th moment calculated by the model, is the measured runoff at the i-th moment, N is the total amount of runoff data; ； in, is the Nasher efficiency coefficient, is the average value of the measured runoff at the i-th moment, is the runoff volume at the i-th moment calculated by the model, is the measured runoff at the i-th moment; The runoff relative error and the Nasher efficiency coefficient are used as rainfall fitting accuracy indicators of the optimal accuracy algorithm; If the rainfall fitting accuracy index does not meet the preset threshold, the initial model parameters are input into the reverse prediction neural network for iteration, and the optimal model parameters are re-output.

4. The method for basin rainfall runoff forecasting based on neural network according to claim 1 is characterized in that: The step of using the optimal model parameters as preset parameters of a rainfall-runoff model to obtain an optimal rainfall-runoff model comprises: According to the optimal model parameters, a confluence relationship matrix between hypercubes is calculated, and an optimal rainfall-runoff model is determined according to the confluence relationship matrix; The optimal rainfall-runoff model is a confluence relationship matrix. If there are n hypercubes, the confluence relationship matrix C is a Matrix, where It represents the proportion of water flow contribution of the j-th hypercube to the i-th hypercube.

5. The method for basin rainfall runoff forecasting based on neural network according to claim 1 is characterized in that: The method of calculating rainfall runoff data based on the terrain basic data and the measured rainfall runoff process diagram using an optimal rainfall runoff model and a preset runoff transmission model includes: According to the basic terrain data, the watershed is divided into a number of hypercubes, and the terrain factor of each hypercube is calculated; Calculating the water storage capacity of the hypercube based on the terrain factors; According to the measured rainfall runoff process diagram, the rainfall is divided into a number of hypercubes, and the inflow of each cube is calculated; If the water storage capacity is less than the inflow, the outflow of each hypercube is calculated according to the terrain factor and the inflow; ； in, is the outflow of the ith unit at the jth moment, is the drainage area of ​​the ith unit, is the inflow of the ith unit at the jth moment; inputting the outflow into an optimal rainfall-runoff model to output a hypercube flow; inputting the hypercube flow into a runoff transport model to output rainfall runoff data; The runoff transport model is used to simulate the flow of water in a region, and the rainfall runoff data is obtained by solving the linear trough storage equation and the water balance equation.

6. The method for basin rainfall runoff forecasting based on neural network according to claim 5 is characterized in that: Before dividing the rainfall into a plurality of hypercubes according to the measured rainfall runoff process diagram and calculating the inflow of each cube, the method further comprises: A series of distributed rain gauges are set at the center point of the top surface of each hypercube; ； in, For flow rate, is the rainfall in the ith area into the flow, is the outlet interception flow of the ith area, and n is the number of rain gauges set.

7. The method for basin rainfall runoff forecasting based on neural network according to claim 5 is characterized in that: The terrain factors include slope, shape factor, water flow path length, and water flow path slope.

8. A basin rainfall runoff forecasting system based on neural network, characterized in that: include: Data acquisition module, used to obtain basic terrain data, measured rainfall runoff process diagram and historical rainfall runoff data; An initial parameter module, used for inputting the historical rainfall-runoff data into a back propagation neural network to obtain initial model parameters of a rainfall-runoff model; A parameter optimization module, used to obtain optimal model parameters according to the initial model parameters through an optimal accuracy algorithm; A model optimization module, used for taking the optimal model parameters as preset parameters of a rainfall-runoff model to obtain an optimal rainfall-runoff model; The result output module is used to calculate the rainfall runoff data according to the terrain basic data and the measured rainfall runoff process diagram, using the optimal rainfall runoff model and the preset runoff transmission model.

9. An electronic device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the neural network-based basin rainfall runoff forecasting method as described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the basin rainfall runoff forecasting method based on neural network as described in any one of claims 1 to 7.

Images