Runoff prediction method and device, storage medium and computer device

Abstract

The application provides a runoff prediction method and device, a storage medium and a computer device, and comprises the following steps: obtaining historical hydro-meteorological data, historical climate index data, historical runoff return data of a plurality of preset historical periods and historical runoff data of a target basin; the historical runoff return data is runoff data of the preset historical period simulated by a plurality of different runoff simulation models according to the historical hydro-meteorological data and the historical climate index data; the historical runoff data is measured runoff data of the historical period corresponding to the historical runoff return data; training a plurality of initial runoff prediction models according to the historical hydro-meteorological data, the historical climate index data, the historical runoff return data and the historical runoff data to obtain a plurality of runoff prediction models; inputting basin hydro-meteorological data, basin climate index data and a plurality of runoff return data of the target basin into the plurality of runoff prediction models to obtain runoff data prediction data.

Description

Technical Field

[0001] This application relates to the technical field of runoff prediction, specifically to a runoff prediction method, apparatus, storage medium, and computer equipment. Background Technology

[0002] Existing runoff prediction techniques drive Earth system numerical models by describing and solving mathematical and physical equations to systematically simulate and predict watershed hydrological processes, including runoff. This method is based on strong physical mechanisms, but the uncertainty of the results is greatly affected by model structure, parameters, and initial conditions. However, the accuracy of this model in capturing regional local-scale climate and hydrological process characteristics is low. Therefore, Earth system numerical models cannot accurately predict long-term runoff data; that is, the longer the prediction time corresponding to the prediction results output by the Earth system numerical model, the lower the accuracy of the prediction results. Summary of the Invention

[0003] The purpose of this application is to overcome the shortcomings and deficiencies of the prior art and provide a runoff prediction method, device, storage medium and computer equipment that can obtain highly accurate runoff prediction data.

[0004] A first aspect of this application provides a runoff prediction method, including:

[0005] The process involves acquiring historical hydrological and meteorological data, historical climate index data, historical runoff return data for multiple preset historical periods, and historical runoff data for multiple preset historical periods for the target watershed. The historical runoff return data consists of runoff data for preset historical periods simulated by several different runoff simulation models based on the historical hydrological and meteorological data and historical climate index data. The historical runoff data is the measured runoff data for the historical periods corresponding to the historical runoff return data.

[0006] Based on the historical hydrological and meteorological data, the historical climate index data, the historical runoff return data, and the historical runoff data, several initial runoff prediction models are trained to obtain several runoff prediction models.

[0007] The target watershed's hydrological and meteorological data, watershed climate index data, and multiple runoff return data are input into several runoff prediction models to obtain runoff prediction data.

[0008] A second aspect of this application provides a runoff prediction device, comprising:

[0009] The historical data acquisition module is used to acquire historical hydrological and meteorological data, historical climate index data, historical runoff return data for multiple preset historical periods, and historical runoff data for multiple preset historical periods of the target watershed; wherein, the historical runoff return data are runoff data for preset historical periods simulated by several different runoff simulation models based on the historical hydrological and meteorological data and historical climate index data; the historical runoff data are measured runoff data for the historical periods corresponding to the historical runoff return data;

[0010] The prediction model acquisition module is used to train several initial runoff prediction models based on the historical hydrological and meteorological data, the historical climate index data, the historical runoff return data, and the historical runoff data, so as to obtain several runoff prediction models.

[0011] The target watershed's hydrological and meteorological data, watershed climate index data, and multiple runoff return data are input into several runoff prediction models to obtain runoff prediction data.

[0012] A third aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the runoff prediction method as described above.

[0013] A fourth aspect of this application provides a computer device including a storage device, a processor, and a computer program stored in the storage device and executable by the processor, wherein the processor executes the computer program to implement the steps of the runoff prediction method as described above.

[0014] Compared to existing technologies, this application first acquires historical hydrological and meteorological data, historical climate index data, and historical runoff data for multiple preset historical periods for the target watershed, as well as runoff data for preset historical periods simulated based on the historical hydrological and meteorological data and historical climate index data. Then, it trains models based on the historical hydrological and meteorological data, the historical climate index data, the historical runoff return data, and the historical runoff data to obtain several runoff prediction models. Finally, based on the watershed hydrological and meteorological data, watershed climate index data, and runoff prediction models of the target watershed, it corrects and optimizes multiple runoff return data to obtain multiple runoff prediction data with higher accuracy.

[0015] To provide a clearer understanding of this application, the specific embodiments of this application will be described below in conjunction with the accompanying drawings. Attached Figure Description

[0016] Figure 1 This is a flowchart of a runoff prediction method according to an embodiment of this application.

[0017] Figure 2 This is a schematic diagram of a runoff prediction model for a runoff prediction method according to an embodiment of this application.

[0018] Figure 3 This is a schematic diagram of the module connections of a runoff prediction data acquisition device according to an embodiment of this application.

[0019] 100. Medium- and long-term runoff forecast data acquisition device; 101. Historical data acquisition module; 102. Prediction model acquisition module; 103. Runoff forecast data acquisition module. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0021] It should be understood that the described embodiments are merely some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of the embodiments of this application.

[0022] In the following description, when referring to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. In the description of this application, it should be understood that the terms "first," "second," "third," etc., are used only to distinguish similar objects and are not necessarily used to describe a specific order or sequence, nor should they be construed as indicating or implying relative importance. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances. The singular forms "a," "the," and "the" used in this application and the appended claims are also intended to include the plural forms, unless the context clearly indicates otherwise. The word "if" as used herein can be interpreted as "when," "when," or "in response to determination."

[0023] Furthermore, in the description of this application, unless otherwise stated, "multiple" means two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. The character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0024] Please see Figure 1 This is a flowchart of a runoff prediction method according to an embodiment of this application, including:

[0025] S1: Obtain historical hydrological and meteorological data, historical climate index data, historical runoff return data for multiple preset historical periods, and historical runoff data for multiple preset historical periods for the target watershed; wherein, the historical runoff return data are runoff data for preset historical periods simulated by several different runoff simulation models based on the historical hydrological and meteorological data and historical climate index data; the historical runoff data are measured runoff data for the historical periods corresponding to the historical runoff return data.

[0026] Historical hydro-meteorological data refers to hydro-meteorological data obtained from historical data of the target watershed, which may include precipitation data, runoff data, etc.; historical climate index data refers to climate index data obtained from historical data of the target watershed, which may include atmospheric circulation index, sea surface temperature index, etc.; historical runoff data refers to runoff data obtained from historical data of the target watershed.

[0027] It should be noted that the preset historical period is later than the historical hydrological and meteorological data and historical climate index data. That is, the historical runoff report data is data simulated based on the historical hydrological and meteorological data and historical climate index data, and belongs to the future prediction data corresponding to the historical hydrological and meteorological data and historical climate index data. Historical runoff data, on the other hand, is measured runoff data from the same historical period as the historical runoff report data. In other words, the historical runoff data is later than the historical hydrological and meteorological data and historical climate index data, and the historical runoff data is at least 3 days later than the historical hydrological and meteorological data and historical climate index data. For example, the historical hydrological and meteorological data and historical climate index data are from January 1st to March 5th, while the historical runoff report data and historical runoff data are from March 9th of the same year onwards. In this embodiment, the historical runoff report data refers to simulated medium- and long-term runoff data, and specifically, multiple medium- and long-term runoff data are simulated from the historical hydrological and meteorological data and historical climate index data using different runoff simulation models.

[0028] S2: Based on the historical hydrological and meteorological data, the historical climate index data, the historical runoff return data, and the historical runoff data, train several initial runoff prediction models to obtain several runoff prediction models.

[0029] In one feasible embodiment, step S2 includes:

[0030] S21: Use the historical hydrological and meteorological data, the historical climate index data, and the historical runoff report data as input samples, and use the historical runoff data as output samples.

[0031] S22: Divide the input samples and the output samples to obtain input training samples, input validation samples, output training samples and output validation samples.

[0032] S23: Train the several initial runoff prediction models based on the input training samples and the output training samples to obtain several candidate runoff prediction models.

[0033] S24: Optimize the plurality of candidate runoff prediction models based on the input verification samples and the output verification samples to obtain the plurality of runoff prediction models.

[0034] In this embodiment, in order to improve the prediction accuracy of each runoff prediction model, the historical hydrological and meteorological data, the historical climate index data, the multiple historical runoff return data, and the historical runoff data are divided into training samples and validation samples. The training samples are used to train the initial runoff prediction model, and the validation samples are used to validate and optimize the trained initial runoff prediction model to obtain a runoff prediction model with high prediction accuracy.

[0035] S3: Input the target watershed hydrological and meteorological data, watershed climate index data, and multiple runoff return data into several runoff prediction models to obtain runoff prediction data.

[0036] Watershed hydrological and meteorological data refers to hydrological and meteorological data for a target watershed within a predetermined period of time. Watershed climate index data refers to watershed climate index data for a predetermined period of time. The predetermined "predetermined period of time" can refer to the last few days, weeks, or months. Multiple runoff reports increase the reference data for runoff prediction models in forecasting future runoff changes, reducing the impact of the forecast timeframe on the prediction results, and particularly enabling the acquisition of highly accurate medium- to long-term runoff data predictions. Here, "medium- to long-term" refers to a forecast period of several weeks or months.

[0037] Compared to existing technologies, this application first acquires historical hydrological and meteorological data, historical climate index data, and historical runoff data for multiple preset historical periods for the target watershed, as well as runoff data for preset historical periods simulated based on the historical hydrological and meteorological data and historical climate index data. Then, it trains models based on the historical hydrological and meteorological data, the historical climate index data, the historical runoff return data, and the historical runoff data to obtain several runoff prediction models. Finally, based on the watershed hydrological and meteorological data, watershed climate index data, and runoff prediction models of the target watershed, it corrects and optimizes multiple runoff return data to obtain multiple runoff prediction data with higher accuracy.

[0038] One approach is to increase the range of prediction results by using multiple runoff prediction data, or to obtain the most accurate final runoff prediction data from multiple runoff prediction data by taking the average, maximum, median, or other values ​​of multiple runoff prediction data.

[0039] Please see Figure 2 In one feasible embodiment, each of the initial runoff prediction models includes a Long Short-Term Memory (LSTM) network encoder, a first fully connected layer, a repeating input layer, an LSTM network decoder, a time distribution layer, and a second fully connected layer. Specifically, the LSTM network encoder is an LSTM Encoder, the first fully connected layer is Dense_1, the repeating input layer is RepeatVector, the LSTM network decoder is an LSTM Decoder, the time distribution layer is TimeDistributed, and the second fully connected layer is Dense_2.

[0040] like Figure 2 As shown, during training, multiple historical runoff return data of the target watershed are the original members of the runoff forecast, and the historical runoff data are the new members of the runoff forecast; during application, multiple runoff return data of the target watershed are the original members of the runoff forecast, and multiple runoff prediction data are the new members of the runoff forecast.

[0041] S2: The step of training several initial runoff prediction models to obtain several runoff prediction models based on the historical hydrological and meteorological data, the historical climate index data, the multiple historical runoff return data, and the historical runoff data includes:

[0042] S201: Input historical hydrological and meteorological data and historical climate index data into the Long Short-Term Memory Network Encoder for encoding to obtain the historical hydrological and meteorological time series vector and historical climate index time series vector output by the Long Short-Term Memory Network Encoder.

[0043] The role of the Long Short-Term Memory (LSTM) network encoder is to encode the input sequence into a fixed-shape state, which contains all the information of the input sequence. Historical hydrological and meteorological data and historical climate index data are the input sequences for the LSM network encoder, respectively.

[0044] S202: Input the monthly time series of the historical runoff return data into the first fully connected layer to obtain the time series information of the return data output by the first fully connected layer.

[0045] The first fully connected layer is used to integrate the features of the input data. For example, when the monthly time series of historical runoff return data is input to the first fully connected layer, the first fully connected layer integrates the features of the monthly time series of historical runoff return data to obtain the time series information of the return data.

[0046] S203: Input the historical hydrological and meteorological time series vector, the historical climate index time series vector, and the time series information of the reported data into the repeated input layer for repeated time dimension expansion processing to obtain the time series expansion information output by the repeated input layer.

[0047] The role of the repeated input layer is to repeat the input sequence to expand the receptive field of the time step. Specifically, the repeated input layer repeats the input sequence a specified number of times along the time dimension, thereby expanding the length of the time step to a preset foresight period. For example, if the foresight period is set to n months, the repeated input layer will repeat the input sequence multiple times along the time dimension until the time step length reaches n months. After expanding the time step length, the repeated input layer outputs the temporal expansion information in the form of a three-dimensional tensor.

[0048] S204: Input the timing extension information into the Long Short-Term Memory (LSTM) network decoder for decoding to obtain the timing features output by the LSM network decoder.

[0049] The role of the Long Short-Term Memory (LSTM) network decoder is to decode the intermediate representation information generated by the encoder, thereby generating the target sequence features.

[0050] S205: The temporal features are input to the second fully connected layer through the time distribution layer, so that the second fully connected layer is applied to each time step of the temporal features to obtain training prediction data.

[0051] The role of the time distribution layer is to apply a regular layer to each time step of the time series data. In this embodiment, the time distribution layer applies the second fully connected layer to each time step of the time series feature. The time distribution layer causes the second fully connected layer to process the time series feature at each time step, rather than processing it after the entire time series feature has ended. This allows the initial runoff prediction model to be better trained and learned based on historical hydrological and meteorological data, historical climate index data, historical runoff return data, and the dynamic changes of historical runoff data over time, so as to obtain highly accurate training prediction data.

[0052] S206: Based on the differences between the training prediction data and the historical runoff data, adjust the model parameters of the corresponding initial runoff prediction model to obtain several runoff prediction models.

[0053] In this embodiment, by using an initial runoff prediction model that includes a long short-term memory network encoder, a first fully connected layer, a repetitive input layer, a long short-term memory network decoder, a temporal distribution layer, and a second fully connected layer, the temporal features of the training data can be extracted and dynamically learned, resulting in a runoff prediction model with high prediction accuracy.

[0054] In one feasible embodiment, step S204 includes:

[0055] S2041: Construct a loss function based on the training prediction data and the historical runoff data to obtain the data difference.

[0056] The loss function is a formula used to calculate the root mean square error between training prediction data and historical runoff data; that is, the root mean square error calculated by it is the data difference.

[0057] S2042: If the data difference is greater than a preset difference threshold, adjust the model parameters of the corresponding initial runoff prediction model to update the training prediction data until the data difference is less than or equal to the difference threshold, and obtain the corresponding runoff prediction model.

[0058] Each time the model parameters are adjusted, the initial runoff prediction model can output new training prediction data, thus obtaining new data differences. When the new data differences are less than or equal to the preset difference threshold, it means that the prediction results of the initial runoff prediction model after adjusting the model parameters have small errors compared with the actual results, and therefore it can be used as a runoff prediction model.

[0059] In this embodiment, a loss function is constructed to obtain the data difference, and then the model training is determined based on the data difference and the difference threshold to obtain a runoff prediction model with high prediction accuracy.

[0060] Please see Figure 3 The second embodiment of this application provides a runoff prediction data acquisition device 100, including:

[0061] The historical data acquisition module 101 is used to acquire historical hydrological and meteorological data, historical climate index data, historical runoff return data for multiple preset historical periods, and historical runoff data for multiple preset historical periods of the target watershed; wherein, the historical runoff return data are runoff data for preset historical periods simulated by several different runoff simulation models based on the historical hydrological and meteorological data and historical climate index data; and the historical runoff data are measured runoff data for the historical periods corresponding to the historical runoff return data.

[0062] The prediction model acquisition module 102 is used to train several initial runoff prediction models based on the historical hydrological and meteorological data, the historical climate index data, the historical runoff return data and the historical runoff data, to obtain several runoff prediction models.

[0063] The runoff prediction data acquisition module 103 is used to input the target watershed hydrological and meteorological data, watershed climate index data and multiple runoff return data into several runoff prediction models to obtain runoff prediction data.

[0064] In one feasible embodiment, each of the initial runoff prediction models includes a long short-term memory network encoder, a first fully connected layer, a repetitive input layer, a long short-term memory network decoder, a temporal distribution layer, and a second fully connected layer.

[0065] The prediction model acquisition module 102 includes:

[0066] The time-series vector extraction submodule is used to input historical hydrological and meteorological data and historical climate index data into the long short-term memory network encoder for encoding, and to obtain the historical hydrological and meteorological time-series vector and historical climate index time-series vector output by the long short-term memory network encoder.

[0067] The time series information acquisition submodule for return data is used to input the monthly time series of the historical runoff return data into the first fully connected layer to obtain the time series information of return data output by the first fully connected layer;

[0068] The time-series extension submodule is used to input the historical hydrological and meteorological time-series vector, the historical climate index time-series vector, and the time-series information of the reported data into the repeated input layer for repeated time-dimensional extension processing, so as to obtain the time-series extension information output by the repeated input layer.

[0069] The timing feature acquisition module is used to input the timing extension information into the long short-term memory network decoder for decoding, and obtain the timing features output by the long short-term memory network decoder.

[0070] The training prediction data acquisition submodule is used to input the temporal features through the time distribution layer to the second fully connected layer, so as to apply the second fully connected layer to each time step of the temporal features to obtain training prediction data;

[0071] The prediction model acquisition submodule is used to adjust the model parameters of the corresponding initial runoff prediction model based on the data differences between each of the training prediction data and the historical runoff data, so as to obtain several runoff prediction models.

[0072] In one feasible embodiment, the prediction model acquisition submodule includes:

[0073] The data difference acquisition unit is used to acquire the data difference between the training prediction data and the historical runoff data according to the loss function.

[0074] The model parameter update unit is used to adjust the model parameters of the corresponding initial runoff prediction model if the data difference is greater than a preset difference threshold, so as to update the training prediction data until the data difference is less than or equal to the difference threshold, and obtain the corresponding runoff prediction model.

[0075] In a feasible embodiment of sterol consumption, the prediction model acquisition module 102 includes:

[0076] The sample acquisition submodule is used to take the historical hydrological and meteorological data, the historical climate index data and the historical runoff report data as input samples, and the historical runoff data as output samples.

[0077] The sample partitioning submodule is used to partition the input samples and the output samples to obtain input training samples, input validation samples, output training samples and output validation samples;

[0078] The model training submodule is used to train the plurality of initial runoff prediction models based on the input training samples and the output training samples, thereby obtaining a plurality of candidate runoff prediction models;

[0079] The model validation and optimization submodule is used to validate and optimize the plurality of candidate runoff prediction models based on the input validation samples and the output validation samples, so as to obtain the plurality of runoff prediction models.

[0080] It should be noted that the runoff prediction device provided in the second embodiment of this application is only illustrated by the above-described division of functional modules when executing the runoff prediction method. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the runoff prediction device provided in the second embodiment of this application and the runoff prediction method in the first embodiment of this application belong to the same concept, and its implementation process is detailed in the method embodiment, which will not be repeated here.

[0081] A third aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the runoff prediction method as described above.

[0082] A fourth aspect of this application provides a computer device including a storage device, a processor, and a computer program stored in the storage device and executable by the processor, wherein the processor executes the computer program to implement the steps of the runoff prediction method as described above.

[0083] The device embodiments described above are merely illustrative. The components described as separate parts may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this application according to actual needs. Those skilled in the art can understand and implement this without any inventive effort.

[0084] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0085] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 The computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function selected in one or more boxes.

[0086] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function selected in one or more boxes.

[0087] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0088] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, like read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0089] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0090] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0091] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method of predicting runoff, characterized by, The method comprises: obtaining historical hydro-meteorological data, historical climate index data, historical runoff return data of a plurality of preset historical periods, and historical runoff data of the plurality of preset historical periods of a target basin; wherein the historical runoff return data is runoff data of the preset historical periods simulated by a plurality of different runoff simulation models according to the historical hydro-meteorological data and the historical climate index data; and the historical runoff data is measured runoff data of the historical periods corresponding to the historical runoff return data; training a plurality of initial runoff prediction models according to the historical hydro-meteorological data, the historical climate index data, the historical runoff return data, and the historical runoff data to obtain a plurality of runoff prediction models; wherein each initial runoff prediction model comprises a long short-term memory network encoder, a first fully connected layer, a repeated input layer, a long short-term memory network decoder, a time distribution layer, and a second fully connected layer; the step of training a plurality of initial runoff prediction models according to the historical hydro-meteorological data, the historical climate index data, the plurality of historical runoff return data, and the historical runoff data to obtain a plurality of runoff prediction models comprises: inputting the historical hydro-meteorological data and the historical climate index data into the long short-term memory network encoder for encoding to obtain historical hydro-meteorological time series vectors and historical climate index time series vectors output by the long short-term memory network encoder; inputting monthly time series of the historical runoff return data into the first fully connected layer to obtain return data time series information output by the first fully connected layer; inputting the historical hydro-meteorological time series vectors, the historical climate index time series vectors, and the return data time series information into the repeated input layer for time dimension repeated expansion processing to obtain time series expansion information output by the repeated input layer; inputting the time series expansion information into the long short-term memory network decoder for decoding to obtain time series features output by the long short-term memory network decoder; inputting the time series features into the second fully connected layer through the time distribution layer to apply the second fully connected layer to each time step of the time series features to obtain training prediction data; adjusting model parameters of the corresponding initial runoff prediction model according to data differences between each training prediction data and historical runoff data to obtain a plurality of runoff prediction models; inputting basin hydro-meteorological data, basin climate index data, and a plurality of runoff return data of a target basin into a plurality of runoff prediction models to obtain runoff data prediction data.

2. The runoff prediction method according to claim 1, characterized by, The step of adjusting model parameters of the corresponding initial runoff prediction model according to data differences between the training prediction data and the historical runoff data to obtain a plurality of runoff prediction models comprises: obtaining the data differences between the training prediction data and the historical runoff data according to a loss function; if the data differences are greater than a preset difference threshold, adjusting the model parameters of the corresponding initial runoff prediction model to update the training prediction data until the data differences are less than or equal to the difference threshold to obtain the corresponding runoff prediction model.

3. The runoff prediction method according to claim 1, characterized by, The step of training a plurality of initial runoff prediction models according to the historical hydro-meteorological data, the historical climate index data, the historical runoff return period data and the historical runoff data comprises: inputting the historical hydro-meteorological data, the historical climate index data and the historical runoff return period data as input samples and inputting historical runoff data as output samples; dividing the input samples and the output samples to obtain input training samples, input validation samples, output training samples and output validation samples; training the plurality of initial runoff prediction models according to the input training samples and the output training samples to obtain a plurality of candidate runoff prediction models; verifying and optimizing the plurality of candidate runoff prediction models according to the input validation samples and the output validation samples to obtain the plurality of runoff prediction models.

4. A runoff prediction device, characterized by, comprise: a historical data acquisition module, configured to acquire historical hydro-meteorological data, historical climate index data, historical runoff return period data of a plurality of preset historical periods and historical runoff data of the plurality of preset historical periods of a target river basin; wherein the historical runoff return period data is runoff data of a preset historical period simulated by a plurality of different runoff simulation models according to the historical hydro-meteorological data and the historical climate index data; and the historical runoff data is measured runoff data of a historical period corresponding to the historical runoff return period data; a prediction model acquisition module, configured to train a plurality of initial runoff prediction models according to the historical hydro-meteorological data, the historical climate index data, the historical runoff return period data and the historical runoff data to obtain a plurality of runoff prediction models; wherein each initial runoff prediction model comprises a long short-term memory network encoder, a first full connection layer, a repeated input layer, a long short-term memory network decoder, a time distribution layer and a second full connection layer; the step of training a plurality of initial runoff prediction models according to the historical hydro-meteorological data, the historical climate index data, the plurality of historical runoff return period data and the historical runoff data to obtain a plurality of runoff prediction models comprises: inputting the historical hydro-meteorological data and the historical climate index data into the long short-term memory network encoder for encoding to obtain historical hydro-meteorological time series vectors and historical climate index time series vectors output by the long short-term memory network encoder; inputting monthly time series of the historical runoff return period data into the first full connection layer to obtain return period data time series information output by the first full connection layer; inputting the historical hydro-meteorological time series vectors, the historical climate index time series vectors and the return period data time series information into the repeated input layer for time dimension repeated expansion processing to obtain time series expansion information output by the repeated input layer; inputting the time series expansion information into the long short-term memory network decoder for decoding to obtain time series features output by the long short-term memory network decoder; The time sequence feature is input to the second full connection layer through the time distribution layer to apply the second full connection layer to each time step of the time sequence feature to obtain training prediction data; According to the data difference between each training prediction data and historical runoff data, the model parameters of the corresponding initial runoff prediction model are adjusted to obtain a plurality of runoff prediction models; The runoff prediction data acquisition module is configured to input the basin hydro-meteorological data, basin climate index data, and a plurality of runoff return data of a target basin to a plurality of runoff prediction models to obtain runoff data prediction data.

5. The runoff prediction device according to claim 4, characterized by Each initial runoff prediction model includes a long short-term memory network encoder, a first full connection layer, a repeated input layer, a long short-term memory network decoder, a time distribution layer, and a second full connection layer; The prediction model acquisition module includes: The time sequence vector extraction submodule is configured to input historical hydro-meteorological data and historical climate index data to the long short-term memory network encoder for encoding to obtain historical hydro-meteorological time sequence vectors and historical climate index time sequence vectors output by the long short-term memory network encoder; The return data time sequence information acquisition submodule is configured to input the monthly time sequence of the historical runoff return data to the first full connection layer to obtain return data time sequence information output by the first full connection layer; The time sequence expansion submodule is configured to input the historical hydro-meteorological time sequence vectors, the historical climate index time sequence vectors, and the return data time sequence information to the repeated input layer for time dimension repeated expansion processing to obtain time sequence expansion information output by the repeated input layer; The time sequence feature acquisition module is configured to input the time sequence expansion information to the long short-term memory network decoder for decoding to obtain time sequence features output by the long short-term memory network decoder; The training prediction data acquisition submodule is configured to input the time sequence features to the second full connection layer through the time distribution layer to apply the second full connection layer to each time step of the time sequence features to obtain training prediction data; The prediction model acquisition submodule is configured to adjust model parameters of a corresponding initial runoff prediction model according to a data difference between each training prediction data and historical runoff data to obtain a plurality of runoff prediction models.

6. The runoff prediction apparatus according to claim 5, characterized by The prediction model acquisition submodule includes: The data difference acquisition unit is configured to acquire the data difference between the training prediction data and the historical runoff data according to a loss function; The model parameter updating unit is configured to adjust the model parameters of the corresponding initial runoff prediction model if the data difference is greater than a preset difference threshold to update the training prediction data until the data difference is less than or equal to the difference threshold to obtain the corresponding runoff prediction model.

7. The runoff prediction device according to claim 4, characterized by The prediction model acquisition module includes: The sample acquisition submodule is configured to take the historical hydro-meteorological data, the historical climate index data, and the historical runoff return data as input samples and take historical runoff data as output samples. a sample division sub-module, configured to divide the input sample and the output sample to obtain input training samples, input validation samples, output training samples and output validation samples; a model training sub-module, configured to train the plurality of initial runoff prediction models according to the input training samples and the output training samples to obtain a plurality of candidate runoff prediction models; a model verification and optimization sub-module, configured to verify and optimize the plurality of candidate runoff prediction models according to the input validation samples and the output validation samples to obtain the plurality of runoff prediction models.

8. A computer readable storage medium storing a computer program, characterized in that: The computer program, when executed by a processor, implements the steps of the runoff prediction method according to any one of claims 1 to 3.

9. A computer device, comprising: A computer program product, comprising a storage and a processor, and a computer program stored in the storage and executable by the processor, wherein the processor implements the steps of the runoff prediction method according to any one of claims 1 to 3 when executing the computer program.

Images