Flood forecasting method and system, storage medium and computing equipment

Abstract

The invention relates to the field of flood forecasting, and discloses a flood forecasting method and system, a storage medium and computing equipment, and the method comprises the steps: topological modeling: taking a river catchment area surrounded by a water diversion line and a catchment area outlet hydrometric station as nodes of a graph, constructing a connection relation of edges according to the actual flow direction of a river and the connection relation of branches, and carrying out the topological modeling; therefore, topological graph structure data adaptive to a hydrological mechanism is constructed; converting month number characteristics, generating a dynamic periodic timestamp vector, splicing the dynamic periodic timestamp vector with topological graph structure data, and inputting an attention neural network for training; the attention neural network architecture comprises two independent feature channels which are used for processing historical data and forecast data respectively, the network supports a variable modular architecture, and a flood forecast result can be output according to needs after a model is trained. According to the method, the flood prediction precision and interpretability are improved, and a high-precision and interpretable space-time joint flood prediction scheme is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of flood forecasting, and in particular to a flood forecasting method, system, storage medium and computing device of a topological modeling-spatiotemporal coupling attention neural network. Background Art

[0002] Flood disasters are one of the most destructive natural disasters globally. Floods can cause economic losses, casualties and ecological damage. Timely and accurate flood prediction can provide flood control decisions and reduce losses.

[0003] The mainstream flood prediction methods mainly include hydrological mechanism models and data-driven models. The mechanism models infer the evolution of rivers in the basin by simulating hydrological processes. The data-driven models directly mine spatiotemporal patterns from data without explicit physical equations. With the development of sensor technology, satellite remote sensing and the construction of hydrological stations, relevant meteorological and hydrological data are becoming increasingly rich, and data-driven models have developed rapidly.

[0004] Most of the existing data-driven methods simply start from the time series perspective, rely on historical data for prediction, and ignore the spatial structure information in the basin. CN111915058A provides a flood detection method implemented by a long short-term memory network and transfer learning, and provides a BiGRU multi-step prediction method, system and storage medium applied to flood prediction. These mainly capture correlations at the time series level and lack the capture of spatial correlations. The graph structures of the existing methods using graph neural network models to capture spatial correlations generally use rectangular segmentation or directly use hydrological stations as nodes, without considering natural watershed boundaries and main-branch relationships, and the model interpretability is poor. For example, the prediction method proposed in the public literature "FloodGNN-GRU: a spatio-temporal graph neural network for flood prediction" constructs a graph structure by rectangularly segmenting the basin.

[0005] In summary, there are obvious deficiencies in the existing flood prediction technology in capturing the spatial correlations of data-driven models. Summary of the Invention

[0006] Aiming at the above problems, the purpose of the present invention is to provide a flood forecasting method, system, storage medium and computing device of a topological modeling-spatiotemporal coupling attention neural network, which combines the construction method of a topological graph structure constructed by hydrological mechanism and a spatiotemporal attention convolutional network model, improves the flood prediction accuracy and interpretability, and realizes a high-precision and interpretable spatiotemporal joint flood prediction scheme.

[0007] To achieve the above object, in the first aspect, the technical solution adopted by the present invention is: a flood forecasting method, which includes: using the river catchment area surrounded by the watershed line and the hydrological measurement station at the outlet of the catchment area as the nodes of the topological graph, and constructing the connection relationship of the topological graph based on the actual flow direction of the river and the connection relationship of each tributary, so that the organizational form of the topological graph structure is adapted to the hydrological mechanism; converting the monthly features, generating a dynamic periodic timestamp vector according to the bidirectional embedding of the sine function and the cosine function, and splicing it with the topological graph structure data to generate spatio-temporal coupled features; inputting the spatio-temporal coupled feature data into the spatio-temporal graph attention convolutional network training with two independent feature channels; the two independent feature channels in the spatio-temporal graph attention neural network architecture respectively process historical data and forecast data, and this network supports a variable modular architecture; forming data of the same type as the training model by topological construction and spatio-temporal coupling of the meteorological data and historical flow of the period to be forecast in a certain basin, and inputting it into the trained spatio-temporal attention convolutional network to obtain the flood forecasting result of this basin.

[0008] Further, the spatio-temporal graph attention convolutional network with two independent feature channels includes multiple cascaded STACNN modules; Each STACNN module includes: a temporal attention module, a spatial attention module, a spatial convolutional module, a temporal convolutional module, a residual connection module, and an output connection module; The input data is respectively transmitted to the temporal attention module and the residual connection module. After the different graph structure data are assigned different weights in the time dimension by the temporal attention module, it is multiplied by the input data and then transmitted to the spatial attention module; among them, the input data of the first-level STACNN module is the encoded graph structure data, and the input data of each subsequent level of STACNN module is the output data of the previous level of STACNN module; After the spatial attention module assigns different weights to the received data in the spatial dimension, it is multiplied by the input data and then processed by the spatial convolutional module and the temporal convolutional module in sequence, added to the data output by the residual connection module, and output after passing through the output connection module as the input data of the next-level STACNN module.

[0009] Further, the temporal attention module is:

[0010] Among them, is the input data of the STACNN module; is the time attention score matrix after being normalized by the Sigmoid function; is the output data processed by the temporal attention module.

[0011] Further, the spatial attention module is:

[0012] Among them, is the output data of the spatial attention module; is the spatial attention score matrix after being normalized by the Softmax function; S is the output data of this module; are learnable parameters; is used to extract features in the time dimension, is used to transform the matrix dimension, mapping it from the channel dimension to the time dimension, is used to extract features in the channel dimension.

[0013] Furthermore, the spatial convolution module adopts a graph convolution operation simplified based on Chebyshev polynomials, combines the topological structure of the graph with node features, obtains neighbor node information through message passing, aggregates and updates the aggregation result for the obtained information, and finally obtains complex structured information in the graph.

[0014] Furthermore, the spatio-temporal attention convolutional network adopts the mean squared error MSE as the loss function.

[0015] Furthermore, it also includes an evaluation method for the prediction ability of the spatio-temporal attention convolutional network: The Nash efficiency coefficient NSE is used to evaluate the prediction results of the hydrological process; NSE = 1 indicates that the model simulation value is exactly the same as the observed value, and the model prediction effect is perfect; NSE = 0 indicates that the model prediction effect is the same as directly using the average value of the observed values as the prediction value; NSE < 0 indicates that the model prediction effect is worse than directly using the average value of the observed values as the prediction value.

[0016] In a second aspect, the technical solution adopted by the present invention is as follows: A flood forecasting system, which includes: a topological modeling module, using the river catchment area surrounded by the watershed divide line and the hydrological gauging station at the outlet of the catchment area as the nodes of the topological graph, and constructing the connection relationship of the topological graph based on the actual flow direction of the river and the connection relationship of each tributary, so that the organizational form of the topological graph structure is adapted to the hydrological mechanism; a spatio-temporal coupling module, converting the month characteristics, generating a dynamic periodic timestamp vector according to the bidirectional embedding of the sine function and the cosine function, and splicing it with the topological graph structure data to generate spatio-temporal coupled characteristics; a spatio-temporal attention convolutional network module, inputting the spatio-temporal coupled feature data into the spatio-temporal graph attention convolutional network training with two independent feature channels; the two independent feature channels in the spatio-temporal graph attention neural network architecture respectively process historical data and forecast data, and this network supports a variable modular architecture; a prediction module, forming data of the same type as the training model by performing topological construction and spatio-temporal coupling on the meteorological data and historical flow of the meteorological data required for the forecast period of a certain basin, and inputting it into the trained spatio-temporal attention convolutional network to obtain the flood forecast result of the basin.

[0017] In a third aspect, the technical solution adopted by the present invention is as follows: A computer-readable storage medium storing one or more programs, the one or more programs including instructions that, when executed by a computing device, cause the computing device to execute any of the above methods.

[0018] In a fourth aspect, the technical solution adopted by the present invention is as follows: A computing device, which includes: one or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs include instructions for executing any of the above methods.

[0019] Due to the above technical solutions adopted by the present invention, it has the following advantages: The present invention combines the construction method of the topological graph structure constructed based on the hydrological mechanism and the spatio-temporal attention convolutional network model, improves the flood prediction accuracy and interpretability, and realizes a high-precision and interpretable spatio-temporal joint flood prediction scheme. Description of the Drawings

[0020] Figure 1 is the overall flowchart of the flood forecasting method in the embodiment of the present invention; Figure 2 is the structural schematic diagram of the spatio-temporal attention convolutional network in the embodiment of the present invention. Detailed Embodiments

[0021] In order to overcome the shortcomings of existing data-driven time-series flood prediction models and improve flood prediction accuracy and interpretability, the present invention provides a flood forecasting method, system, medium and device based on a graph structure and a spatio-temporal attention convolutional network, and sets a spatio-temporal attention mechanism to jointly model complex spatio-temporal dependencies.

[0022] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some but not all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the described embodiments of the present invention fall within the protection scope of the present invention.

[0023] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or combinations thereof.

[0024] In one embodiment of the present invention, as Figure 1 shown, a flood forecasting method is provided, which is a method for processing various features in a basin into a graph structure and using a data-driven spatio-temporal attention convolutional network to forecast river water levels and flows. In this embodiment, the method includes the following steps: The first step is to construct a graph structure through topological modeling. The nodes of the graph are the river catchments surrounded by the watershed divide and the hydrological stations at the outlets of the catchments. Each node is associated with the flow and water level characteristics of the hydrological station, and at the same time, the weighted average and cumulative values of meteorological characteristics are obtained by associating the corresponding area of the catchment with the overlapping area of the reanalysis grid generated based on satellite remote sensing data. The edges of the graph: Based on the actual flow direction of the river and the connection relationship of each tributary, a connection relationship is constructed to make the organizational form of the graph structure adapt to the hydrological mechanism. For catchments without hydrological stations at the outlets, they are merged downstream along the river distribution to ensure that the characteristics of all nodes are kept complete.

[0025] The second step is spatio-temporal coupling. Feature extraction is performed on the month in which the data is located, and a dynamic periodic timestamp vector is generated according to the bidirectional embedding of the sine function and the cosine function, and is spliced with the topological graph structure data to generate spatio-temporally coupled features.

[0026] In the third step, the spatio-temporal coupled feature data is input into the spatio-temporal graph attention convolutional network with double independent feature channels (DFC-STAGCN) for training; the two independent feature channels in the spatio-temporal graph attention neural network architecture process historical data and forecast data respectively, and the network supports a variable modular architecture.

[0027] In this embodiment, as Figure 2 shown, a spatio-temporal graph attention convolutional network with double independent feature channels (DFC-STAGCN) is constructed, which contains multiple cascaded spatio-temporal attention convolutions, namely STACNN modules. Each STACNN module is composed of a temporal attention module, a spatial attention module, a spatial convolution module, a temporal convolution module, a residual connection module and an output connection module: (1) Temporal attention module: In the temporal dimension, there is a correlation between different features in the front and back time periods. For example, rainfall in the previous period will affect the flow of the gauging station and the soil water content of the corresponding catchment area, and the correlations in different situations are also different. The attention mechanism can adjust the different weights between different data through training. The corresponding formula is as follows:

[0028] Among them, is the input data of the STACNN module where it is located; is the learnable parameter; is used to extract features in the node dimension, is used to transform the matrix dimension, mapping it from the channel dimension to the node dimension, is used to extract features in the channel dimension; Sigmoid The activation function is used to limit the result range; is the temporal attention score matrix normalized by the Sigmoid function.

[0029] (2) Spatial attention module: In the graph constructed based on the catchment area and the hydrological gauging station, all nodes are part of the same interconnected basin, which causes different features between different nodes to affect each other. Considering that the geographical locations, shapes, meteorological conditions, etc. of different nodes are different, we have a high degree of confidence that this influence is not uniform. The attention mechanism can automatically capture the influence with different weights between different nodes through training. The corresponding formula is as follows:

[0030] Among them, is the input data of the STACNN module where it is located; is the input data after passing through the temporal attention module; is the learnable parameter; For extracting features in the time dimension, For transforming the matrix dimension and mapping it from the channel dimension to the time dimension, For extracting features in the channel dimension; Sigmoid The activation function is used to limit the result range; It is through Softmax The spatial attention score matrix after being normalized by the function.

[0031] Multiply the spatial attention matrix output by the module with the original data to obtain the data after passing through the spatial attention module .

[0032] (3) Spatial convolution module: The graph convolutional network transfers the convolution operation from the image domain to the graph domain. It can directly run on the irregular graph structure without converting it into regular grid data, thereby obtaining the complex structured information in the graph. It combines the topological structure of the graph with the node features, obtains neighbor node information through message passing, aggregates and updates the aggregation result, and stacks multiple layers to expand the node's view for learning. To reduce the computational complexity and simplify the calculation, this module adopts a graph convolution operation simplified based on the Chebyshev polynomial:

[0033] Among them, is the graph convolution kernel, is the polynomial coefficient, represents performing the graph convolution operation; is the Chebyshev polynomial, recursively defined as , where . K is the order of the Chebyshev polynomial; is the normalized Laplacian matrix.

[0034] Among them: .

[0035] (4) Temporal convolution module: The temporal convolution module is located at the end of the entire module to ensure that all nodes can obtain the corresponding spatial dependencies:

[0036] Among them, are the parameters of the temporal convolution kernel; represents the standard convolution operation; ReLU is the activation function.

[0037] (5) Residual connection module: The residual connection adds the input directly to the output, avoiding the problems of gradient vanishing or explosion and accelerating the training of the model. If the dimensions of the input and output are different, an additional convolutional layer is required for adjustment:

[0038] where, are the convolutional kernel parameters; are the bias parameters.

[0039] (6) Output connection module: The output connection module includes a ReLU activation function and a layer normalization module, which can provide non-linear transformation for the previous modules and maintain stability, slowing down the problems of gradient vanishing or gradient explosion.

[0040] In the fourth step, the meteorological data and historical flow of the forecast period in a certain basin, along with the meteorological data, are topologically constructed and spatio-temporally coupled through the first and second steps to construct spatio-temporal graph structure data of the same type as the training model, which is used as training data to input into the spatio-temporal attention convolutional network constructed in the third step for neural network training to obtain the flood forecast model of this basin.

[0041] In the above embodiments, the spatio-temporal attention convolutional network uses the mean squared error MSE as the loss function.

[0042] In the above embodiments, after the fourth step is completed, it also includes an evaluation step of the prediction ability of the spatio-temporal attention convolutional network. Specifically, the evaluation method is: using the Nash-Sutcliffe efficiency coefficient NSE to evaluate the prediction results of the hydrological process. Specifically: NSE = 1 indicates that the model simulation value is completely consistent with the observed value, and the model prediction effect is perfect; NSE = 0 indicates that the model prediction effect is the same as using the average value of the observed values directly as the prediction value; NSE < 0 indicates that the model prediction effect is worse than using the average value of the observed values directly as the prediction value.

[0043] In order to make the objectives, technical solutions and advantages of the present invention more clear and understandable, the following further describes the present invention in detail with reference to embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0044] (1) Establish a training dataset. First, unify the reanalyzed grid meteorological data and the catchment areas divided by the water divide line into the same geographic projection coordinate system WGS84. Perform operations such as weighted averaging or weighted summation on the grid meteorological data according to the corresponding area of the catchment area to obtain the meteorological data characteristics corresponding to each catchment area and its own location and area. Each catchment area is associated with the flow and water level characteristics of its outlet hydrological station. For catchment areas without an outlet hydrological station, merge them downstream along the river distribution to ensure that the characteristics of all nodes are complete. In this way, the nodes of the graph structure are constructed. Construct the edge connection relationship of the graph through the actual flow direction of the rivers between different catchment areas and the connection relationship of each tributary. For the constructed dataset, perform missing value filling and data standardization processing. Set the prediction time window size of the training data according to the length of time to be predicted, the storage capacity of the device, and the computing power of the device. Process the training data into the data format required by the spatio-temporal attention convolutional model.

[0045] (2) Establish a spatio-temporal graph attention convolutional network with two independent feature channels based on the STACNN module. Each STACNN module consists of a time attention module, a spatial attention module, a spatial convolutional module, a time convolutional module, a residual connection module, and an output connection module. The overall spatio-temporal attention convolutional neural network consists of a periodic encoding module, several STACNN modules, and a fully connected output module. The specific network structure can refer to the Figure 1 structure schematic diagram in the attached drawing description.

[0046] (3) Select the mean squared error MSE as the loss function of the model:

[0047] where n is the number of samples, is the i-th true value, is the i-th predicted value.

[0048] (4) Establish a training method for the training data in (1) and the spatio-temporal attention convolutional neural network in (2). Iteratively learn the model parameters through the graphics card. Each iteration adjusts the model parameters to reduce the mean squared error of the loss function between the predicted value and the true value generated during the training process. Generate the most suitable model parameters for the corresponding basin through multiple trainings and iterations.

[0049] (5) Evaluate the prediction ability of the model. Evaluate the prediction results of the hydrological process through the Nash-Sutcliffe efficiency coefficient NSE. NSE is a statistical index used to evaluate the performance of hydrological models or other prediction models. It measures the degree of fit between the model simulation value and the observed value, and its value range is .

[0050] NSE = 1 indicates that the model simulation value is exactly the same as the observed value, and the model prediction effect is perfect.

[0051] NSE = 0 indicates that the model prediction effect is equivalent to directly using the average value of the observed values as the prediction value.

[0052] NSE < 0 indicates that the model prediction effect is worse than directly using the average value of the observed values as the prediction value.

[0053] The calculation formula of the Nash equilibrium efficiency coefficient (NSE) is as follows:

[0054] where represents the observed value at the "i" - th time step; represents the simulated value at the i - th time step; represents the average value of the observed values; n represents the total number of time steps.

[0055] (6) After training and evaluation, the parameters of the spatio - temporal attention convolutional neural network need to be fixed and saved to generate an inference model for predicting future floods. At the same time, the feature data required for predicting future floods is constructed according to the construction method of the training data. Combining the feature data and the inference model can predict the floods in this basin in the future.

[0056] In an embodiment of the present invention, a flood forecasting system is provided, which includes: A topological modeling module, using the river catchment surrounded by the watershed divide and the hydrological gauging station at the catchment outlet as the nodes of the topological graph, and constructing the connection relationship of the topological graph based on the actual flow direction of the river and the connection relationship of each tributary, so that the organizational form of the topological graph structure is adapted to the hydrological mechanism; A spatio - temporal coupling module, converting the monthly features, generating a dynamic periodic timestamp vector according to the bidirectional embedding of the sine function and the cosine function, and splicing it with the topological graph structure data to generate spatio - temporal coupled features; A spatio - temporal attention convolutional network module, inputting the spatio - temporal coupled feature data into a spatio - temporal graph attention convolutional network with two independent feature channels for training; the two independent feature channels in the spatio - temporal graph attention neural network architecture respectively process historical data and forecast data, and this network supports a variable modular architecture; A prediction module, forming data of the same type as the training model by performing topological construction and spatio - temporal coupling on the meteorological data and historical flow of the period to be forecasted in a certain basin, and inputting it into the trained spatio - temporal attention convolutional network to obtain the flood forecasting result of this basin.

[0057] In the above - mentioned embodiment, the spatio - temporal graph attention convolutional network with two independent feature channels includes multiple cascaded STACNN modules; Each STACNN module includes: a temporal attention module, a spatial attention module, a spatial convolution module, a temporal convolution module, a residual connection module, and an output connection module; The input data is transmitted to the temporal attention module and the residual connection module respectively. After the temporal attention module assigns different weights to different graph structure data in the temporal dimension, it multiplies with the input data and transmits it to the spatial attention module; wherein, the input data of the first-level STACNN module is the encoded graph structure data, and the input data of each subsequent level of STACNN module is the output data of the previous level of STACNN module; After the spatial attention module assigns different weights to the received data in the spatial dimension, multiplies with the input data, and then processes it through the spatial convolution module and the temporal convolution module in sequence, it adds with the data output by the residual connection module, and outputs after passing through the output connection module, serving as the input data of the next-level STACNN module.

[0058] In this embodiment, the temporal attention module is:

[0059] Wherein, is the input data of the STACNN module; is the temporal attention score matrix normalized by the Sigmoid function; is the output data processed by the temporal attention module.

[0060] In this embodiment, the spatial attention module is:

[0061] Wherein, is the output data of the spatial attention module; is the spatial attention score matrix normalized by the Softmax function; S is the output data of this module; is the learnable parameter; is used to extract features in the temporal dimension, is used to transform the matrix dimension, mapping it from the channel dimension to the temporal dimension, is used to extract features in the channel dimension.

[0062] In this embodiment, the spatial convolution module adopts a graph convolution operation simplified based on Chebyshev polynomials, combines the topological structure of the graph with node features, obtains neighbor node information through message passing, aggregates and updates the aggregation result for the obtained information, and finally obtains the complex structured information in the graph.

[0063] In the above embodiments, the spatio-temporal attention convolutional network uses the mean squared error (MSE) as the loss function.

[0064] In the above embodiments, it further includes an evaluation method for the prediction ability of the spatio-temporal attention convolutional network: The Nash-Sutcliffe Efficiency (NSE) coefficient is used to evaluate the prediction results of the hydrological process; NSE = 1 indicates that the model simulation value completely coincides with the observed value, and the model prediction effect is perfect; NSE = 0 indicates that the model prediction effect is the same as directly using the average value of the observed values as the predicted value; NSE < 0 indicates that the model prediction effect is worse than directly using the average value of the observed values as the predicted value.

[0065] The system provided in this embodiment is used to execute the above method embodiments. For the specific process and detailed content, please refer to the above embodiments and will not be elaborated here.

[0066] In an embodiment of the present invention, a computing device is provided. The computing device may be a terminal, and it may include: a processor, a communication interface, a memory, a display screen, and an input device. Among them, the processor, the communication interface, and the memory complete communication with each other through a communication bus. The processor is used to provide computing and control capabilities. The memory includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. When the computer program is executed by the processor, it realizes the methods in the above embodiments; the internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface is used to communicate with an external terminal in a wired or wireless manner. The wireless manner can be achieved through WIFI, a carrier network, NFC (Near Field Communication), or other technologies. The display screen may be a liquid crystal display screen or an electronic ink display screen. The input device may be a touch layer covering the display screen, or a button, a trackball, or a touchpad provided on the housing of the computing device, or an external keyboard, touchpad, or mouse, etc. The processor can call the logical instructions in the memory.

[0067] In addition, when the logical instructions in the above-mentioned memory can be implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs that can store program codes.

[0068] In an embodiment of the present invention, there is provided a computer program product, where the computer program product includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer is capable of executing the methods provided in the above-mentioned method embodiments.

[0069] In an embodiment of the present invention, there is provided a non-transitory computer-readable storage medium that stores server instructions, and the computer instructions cause the computer to execute the methods provided in the above-mentioned embodiments.

[0070] For the computer-readable storage medium provided in the above-mentioned embodiment, its implementation principle and technical effects are similar to those of the above-mentioned method embodiment, and will not be elaborated here.

[0071] This application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of this application. It should be understood that each process and / or block in the flowchart and / or block diagram, and the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in one process Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0072] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to work in a particular manner, such that the instructions stored in the computer-readable memory produce a manufacture including an instruction device that implements in the process Figure 1 one process or a plurality of processes and / or blocks Figure 1 the functions specified in one block or a plurality of blocks.

[0073] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus, so that a series of operation steps are executed on the computer or other programmable apparatus to produce a computer-implemented process, and thus the instructions executed on the computer or other programmable apparatus provide for implementing in the process Figure 1 one process or a plurality of processes and / or blocks Figure 1 the steps of the functions specified in one block or a plurality of blocks.

[0074] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A flood forecasting method, characterized in that, Including: Taking the river catchment area surrounded by the watershed divide and the hydrological station at the outlet of the catchment area as the nodes of the topological graph, and constructing the connection relationship of the topological graph according to the actual flow direction of the river and the connection relationship of each tributary, so that the organizational form of the topological graph structure is adapted to the hydrological mechanism; Converting the monthly characteristics, generating a dynamic periodic timestamp vector according to the bidirectional embedding of the sine function and the cosine function, and splicing it with the topological graph structure data to generate spatio-temporal coupled characteristics; Inputting the spatio-temporal coupled characteristic data into the spatio-temporal graph attention convolutional network training with two independent feature channels; the two independent feature channels in the spatio-temporal graph attention neural network architecture respectively process historical data and forecast data, and this network supports a variable modular architecture; Forming data of the same type as the training model through topological construction and spatio-temporal coupling of the meteorological data and historical flow of the period to be forecast in a certain river basin, and inputting it into the trained spatio-temporal attention convolutional network to obtain the flood forecast result of this river basin.

2. The flood forecasting method according to claim 1, wherein The spatio-temporal graph attention convolutional network with two independent feature channels includes multiple cascaded STACNN modules; Each STACNN module includes: a temporal attention module, a spatial attention module, a spatial convolution module, a temporal convolution module, a residual connection module, and an output connection module; Transmitting the input data to the temporal attention module and the residual connection module respectively. After the temporal attention module assigns different weights to different graph structure data in the time dimension, it multiplies with the input data and transmits it to the spatial attention module; among them, the input data of the first-level STACNN module is the encoded graph structure data, and the input data of each subsequent level of STACNN module is the output data of the previous level of STACNN module; The spatial attention module assigns different weights to the received data in the spatial dimension, then multiplies with the input data and is processed by the spatial convolution module and the temporal convolution module in sequence, adds with the data output by the residual connection module, and outputs after passing through the output connection module as the input data of the next-level STACNN module.

3. The flood forecasting method according to claim 2, characterized in that The temporal attention module is: ； Among them, is the input data of the STACNN module; is the time attention score matrix normalized by the Sigmoid function; is the output data processed by the time attention module.

4. The flood forecasting method according to claim 3, wherein The spatial attention module is: ； Among them, is the output data of the spatial attention module; is the spatial attention score matrix normalized by the Softmax function; is the output data of this module; are learnable parameters; is used to extract features in the time dimension, is used to transform the matrix dimension and map it from the channel dimension to the time dimension, is used to extract features in the channel dimension.

5. The flood forecasting method according to claim 2, wherein The spatial convolution module adopts a graph convolution operation simplified based on Chebyshev polynomials, combines the topological structure of the graph with the node features, obtains neighbor node information through message passing, aggregates and updates the aggregation result, and finally obtains the complex structured information in the graph.

6. The flood forecasting method according to claim 1, wherein The spatio-temporal attention convolutional network uses the mean squared error MSE as the loss function.

7. The flood forecasting method according to claim 1, wherein It also includes an evaluation method for the prediction ability of the spatio-temporal attention convolutional network: Using the Nash efficiency coefficient NSE to evaluate the prediction result of the hydrological process; NSE = 1 indicates that the model simulation value is exactly the same as the observed value, and the model prediction effect is perfect; NSE = 0 indicates that the model prediction effect is the same as the effect of directly using the average value of the observed values as the prediction value; NSE < 0 indicates that the model prediction effect is worse than the effect of directly using the average value of the observed values as the prediction value.

8. A flood forecasting system, characterized in that, Including: The topological modeling module uses the river catchment area surrounded by the watershed divide and the hydrological gauging station at the outlet of the catchment area as the nodes of the topological graph, and constructs the connection relationship of the topological graph based on the actual flow direction of the river and the connection relationship of each tributary, so that the organizational form of the topological graph structure is adapted to the hydrological mechanism; The spatio-temporal coupling module converts the monthly characteristics, generates a dynamic periodic timestamp vector according to the bidirectional embedding of the sine function and the cosine function, and splices it with the topological graph structure data to generate spatio-temporal coupled characteristics; The spatio-temporal attention convolutional network module inputs the spatio-temporal coupled feature data into the spatio-temporal graph attention convolutional network training with two independent feature channels; the two independent feature channels in the spatio-temporal graph attention neural network architecture process historical data and forecast data respectively, and the network supports a variable modular architecture; The prediction module forms data of the same type as the training model by topological construction and spatio-temporal coupling of the meteorological data and historical flow of the time period to be forecasted in a certain basin, and inputs it into the trained spatio-temporal attention convolutional network to obtain the flood forecast result of the basin.

9. A computer-readable storage medium storing one or more programs, characterized in that, The one or more programs include instructions that, when executed by a computing device, cause the computing device to perform any of the methods described in claims 1 to 7.

10. A computing device, characterized in that, Comprising: One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs include instructions for performing any of the methods described in claims 1 to 7.

Images