Non-data watershed runoff prediction method based on double-layer LSTM transfer learning

Abstract

The invention relates to a data-free watershed runoff prediction method based on double-layer LSTM transfer learning. The method comprises the following steps: firstly, obtaining site data information of a research area, and preprocessing to obtain training data; and then, inputting the training data into an initial double-layer LSTM model to obtain initial predicted runoff data. And then, determining a loss function based on the initial predicted runoff data and the actual runoff data, and adjusting hyper-parameters of the initial double-layer LSTM model based on the loss function to obtain a target double-layer LSTM model. And finally, inputting to-be-predicted data-free watershed site meteorological data into the target double-layer LSTM model to obtain target data-free watershed runoff data.

Description

Technical Field

[0001] The present application relates to the technical field of environmental engineering and water resources management, and in particular to a data-free basin runoff prediction method based on double-layer LSTM transfer learning. Background Art

[0002] Runoff simulation is a very important technical support for water resources management in river basins, and it is also a very important link in the field of hydrological forecasting. Improving the quality of runoff simulation forecasting is still a very important research field. In recent years, affected by factors such as global warming and human activities, extreme weather events have occurred frequently, and flood and drought disasters have intensified, causing extremely serious losses to the economies and societies of various countries. However, in many parts of the world, there is simply not enough spatial density to conduct sufficient long-term river flow observations, so hydrological measurement data cannot be obtained in real time. The lack of long-term runoff data poses a major challenge to the early warning and effective monitoring of hydrological extreme events such as droughts and floods. Therefore, it is of great scientific significance and practical application value to propose a river basin runoff simulation method that can adapt to the changing environment, thereby improving the accuracy of river basin runoff simulation.

[0003] In the existing technology, with the development of artificial intelligence, machine learning has been applied to the monitoring and management of water resources in various ways, and deep learning based on artificial intelligence has proved its effectiveness in revealing the relationship between hydrological variables. Common flow prediction models such as artificial neural networks (ANN) have made significant progress in predicting and simulating hydrological processes and capturing non-stationarity, dynamics and nonlinearity in data sets. Long short-term memory networks (LSTM) regularly produce reliable results in local and regional rainfall-runoff applications around the world. The Qinghai-Tibet Plateau in my country is a major water tower in the world, and its runoff is crucial to regional hydrological processes and water supply for a large population downstream. However, the flow simulation and prediction of process-based hydrological models in the Qinghai-Tibet Plateau have been limited by low site density, short observation time, complex terrain with multiple low-temperature hydrogeological attributes (such as permafrost and glaciers), and flow observations. Hydrological models with physical basis are difficult to apply. Existing applications such as VIC model and HBV model are mostly applied in basins with little data, and models cannot be constructed in basins without data. And general deep learning models are also difficult to accurately simulate runoff in high-altitude cold basins with little data because they require sufficient training data.

[0004] Therefore, in the relevant technology, there is an urgent need for a method that can solve the problem that the hydrological model cannot simulate the basin runoff due to the lack of measured data. Summary of the invention

[0005] Based on this, it is necessary to provide a data-free basin runoff prediction method based on double-layer LSTM transfer learning that can solve the problem that the hydrological model cannot simulate the basin runoff without measured data in response to the above technical problems.

[0006] In the first aspect, the present application provides a data-free basin runoff prediction method based on double-layer LSTM transfer learning. The method comprises:

[0007] Obtain data from the study area sites, perform preprocessing, and obtain training data;

[0008] Inputting the training data into an initial two-layer LSTM model to obtain initial predicted runoff data;

[0009] Determine a loss function based on the initial predicted runoff data and the actual runoff data, and adjust the hyperparameters of the initial two-layer LSTM model based on the loss function to obtain a target two-layer LSTM model;

[0010] The meteorological data of the watershed site to be predicted with no data is input into the target two-layer LSTM model to obtain the runoff data of the target watershed with no data.

[0011] Optionally, in one embodiment of the present application, the obtaining of the site data in the research area and preprocessing includes:

[0012] The data of the stations in the study area are checked for accuracy and averaged.

[0013] Optionally, in one embodiment of the present application, the obtaining of the site data in the research area and preprocessing further includes:

[0014] Correlation analysis is used to determine the characteristic data most relevant to runoff prediction in the site data of the study area as training data.

[0015] Optionally, in one embodiment of the present application, the hyperparameters of the initial two-layer LSTM model include model batch size, sliding window function, gradient descent algorithm, number of hidden layer neurons and learning rate.

[0016] Optionally, in one embodiment of the present application, determining the loss function based on the initial predicted runoff data and the actual runoff data includes:

[0017] A root mean square error is determined based on the initial predicted runoff data and the actual runoff data.

[0018] Optionally, in one embodiment of the present application, the method further includes:

[0019] Data from multiple study area sites were selected for combined training, and the length of the training period was adjusted to evaluate the migration effect of the target two-layer LSTM model.

[0020] Optionally, in one embodiment of the present application, the method further includes:

[0021] The BTOP hydrological model is used to predict the runoff flow data of the data-free basin and compared with the runoff data of the target data-free basin to verify the migration effect of the target two-layer LSTM model.

[0022] In the second aspect, the present application also provides a data-free basin runoff prediction device based on double-layer LSTM transfer learning. The device comprises:

[0023] The data acquisition module is used to obtain the data of the sites in the study area, perform preprocessing, and obtain training data;

[0024] An initial predicted runoff data determination module is used to input the training data into an initial double-layer LSTM model to obtain initial predicted runoff data;

[0025] A model training module, used for determining a loss function based on the initial predicted runoff data and the actual runoff data, and adjusting the hyperparameters of the initial two-layer LSTM model based on the loss function to obtain a target two-layer LSTM model;

[0026] The data-free basin runoff prediction module is used to input the meteorological data of the basin site to be predicted into the target double-layer LSTM model to obtain the target data-free basin runoff data.

[0027] In a third aspect, the present application further provides a computer device, wherein the computer device comprises a memory and a processor, wherein the memory stores a computer program, and the processor executes the steps of the methods described in the above embodiments.

[0028] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the method described in each of the above embodiments are implemented.

[0029] The above-mentioned data-free basin runoff prediction method based on double-layer LSTM transfer learning first obtains the data of the site in the study area, performs preprocessing, and obtains training data. Then, the training data is input into the initial double-layer LSTM model to obtain the initial predicted runoff data. Then, the loss function is determined based on the initial predicted runoff data and the actual runoff data, and the hyperparameters of the initial double-layer LSTM model are adjusted based on the loss function to obtain the target double-layer LSTM model. Finally, the meteorological data of the site of the basin to be predicted without data is input into the target double-layer LSTM model to obtain the target data-free basin runoff data. That is to say, the two-layer LSTM model is trained in a basin with measured data, so that the two-layer LSTM model can use the precipitation-runoff relationship of other stations to infer the runoff time series of stations without measured data. When the trained two-layer LSTM model is migrated to any basin without measured data that has a certain similarity to the study area, the CMFD dataset is combined to supplement the precipitation data in the meteorological data used for prediction. The basin runoff data can be predicted without the need for station data and measured flow data, which solves the problem that the hydrological model cannot be simulated without measured data. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 This is an application environment diagram of a data-free watershed runoff prediction method based on double-layer LSTM transfer learning in one embodiment;

[0031] Figure 2 A schematic diagram of a flow chart of a data-free watershed runoff prediction method based on double-layer LSTM transfer learning in one embodiment;

[0032] Figure 3 It is a structural block diagram of a data-free watershed runoff prediction device based on double-layer LSTM transfer learning in one embodiment;

[0033] Figure 4 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0034] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0035] The data-free basin runoff prediction method based on double-layer LSTM transfer learning provided in the embodiment of the present application can be applied to Figure 1In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or it can be placed on the cloud or other network servers. Among them, the terminal 102 can be but is not limited to various personal computers, laptops, smart phones, tablets, Internet of Things devices and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart car-mounted devices, etc. Portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. The server 104 can be implemented with an independent server or a server cluster consisting of multiple servers.

[0036] In one embodiment, Figure 2 As shown in the figure, a data-free basin runoff prediction method based on double-layer LSTM transfer learning is provided. Figure 1 The server in the example is used to illustrate the following steps:

[0037] S201: Obtain site data in the study area, perform preprocessing, and obtain training data.

[0038] In the embodiment of the present application, firstly, the site data of the study area are collected, that is, the available data of the study area, including precipitation data, China regional high spatiotemporal resolution surface meteorological driven data set (CMFD) data, temperature data, soil moisture, snow depth, ice cover, and other data closely related to the unique hydrological processes in high-cold areas, and the collected data are preprocessed to ensure that the data used for model training is accurate.

[0039] S203: Input the training data into an initial two-layer LSTM model to obtain initial predicted runoff data.

[0040] In the embodiment of the present application, an initial double-layer LSTM model is constructed, wherein the core structure of each LSTM network framework is divided into four parts, namely, a forget gate, an input gate, a cell state, and an output gate. The training data is input into the initial double-layer LSTM model, and according to the characteristics of LSTM processing sequence data, a sliding window of appropriate length is constructed as the input feature of the model, and the time range of the model training period, validation period, and test period is set to capture the temporal dependency. At the same time, L2 regularization is applied in each layer of LSTM structure to prevent overfitting, ensuring that the model can still maintain stability and accuracy in an environment lacking a large amount of training data. In a specific application, the training data, i.e., the site data used for training, is spliced ​​into a complete long sequence data, including meteorological data and actual runoff data, and the actual runoff data refers to the historical runoff data actually measured, and the initial double-layer LSTM model is input, the data length of the training period and the validation period is selected, the hyperparameters of the model are set, the model is run, and the initial predicted runoff data is output.

[0041] S205: Determine a loss function based on the initial predicted runoff data and the actual runoff data, and adjust the hyperparameters of the initial two-layer LSTM model based on the loss function to obtain a target two-layer LSTM model.

[0042] In the embodiment of the present application, the initial predicted runoff data and the actual runoff data are used to determine the loss function, and the hyperparameters of the initial two-layer LSTM model are continuously adjusted based on the change of the loss function until the value of the loss function becomes smaller and smaller and tends to be stable, the model training is completed, and the target two-layer LSTM model is obtained. In a specific application, different hydrological stations are selected as training sites in turn to train the initial two-layer LSTM model.

[0043] S207: Input the meteorological data of the watershed site to be predicted without data into the target two-layer LSTM model to obtain the runoff data of the target watershed without data.

[0044] In an embodiment of the present application, the meteorological data of the site in the watershed to be predicted without data is input into the trained target two-layer LSTM model, and the simulated predicted runoff data of the target watershed without data is output.

[0045] In the above-mentioned data-free basin runoff prediction method based on double-layer LSTM transfer learning, first, the data of the site in the study area is obtained and preprocessed to obtain training data. Then, the training data is input into the initial double-layer LSTM model to obtain initial predicted runoff data. Then, the loss function is determined based on the initial predicted runoff data and the actual runoff data, and the hyperparameters of the initial double-layer LSTM model are adjusted based on the loss function to obtain the target double-layer LSTM model. Finally, the meteorological data of the site of the data-free basin to be predicted is input into the target double-layer LSTM model to obtain the target data-free basin runoff data. That is to say, the two-layer LSTM model is trained in a basin with measured data, so that the two-layer LSTM model can use the precipitation-runoff relationship of other stations to infer the runoff time series of stations without measured data. When the trained two-layer LSTM model is migrated to any basin without measured data that has a certain similarity to the study area, the CMFD dataset is combined to supplement the precipitation data in the meteorological data used for prediction. The basin runoff data can be predicted without the need for station data and measured flow data, which solves the problem that the hydrological model cannot be simulated without measured data.

[0046] In one embodiment of the present application, the obtaining of the site data in the research area and the preprocessing includes:

[0047] The data of the stations in the study area are checked for accuracy and averaged.

[0048] In one embodiment of the present application, the collected data of the research area sites are strictly quality controlled and accuracy tested, including outlier detection, missing value filling and data consistency check. At the same time, according to the location of the research area site, the site control area is determined by ARCGIS software, the data of the site control area is extracted, and the data is averaged.

[0049] In this embodiment, by performing accuracy detection and surface averaging on the site data in the study area, it is possible to ensure that the input data is accurate and consistent in time and space resolution, so as to facilitate subsequent processing and analysis.

[0050] In one embodiment of the present application, the obtaining of the site data in the research area and the preprocessing further includes:

[0051] Correlation analysis is used to determine the characteristic data most relevant to runoff prediction in the site data of the study area as training data.

[0052] In one embodiment of the present application, correlation analysis is used to screen out the feature data most relevant to runoff prediction in the data of the study area sites, that is, the features that have the greatest impact on runoff prediction, as training data. In a specific application, the initially obtained data include meteorological data, longitude and latitude, DEM data, soil data, evapotranspiration, potential evapotranspiration data, leaf area data, etc. After correlation analysis, it is determined that longitude and latitude, DEM data, and soil data have little effect on the results, and evapotranspiration, potential evapotranspiration, and leaf area index have a certain correlation with meteorological data. Therefore, the most useful meteorological data for runoff prediction is selected as input, and the meteorological data includes daily precipitation, CMFD daily precipitation, and daily average temperature.

[0053] In this embodiment, by using correlation analysis to determine the characteristic data most relevant to runoff prediction in the site data of the study area as training data, the complexity of the model can be reduced and the prediction accuracy can be improved.

[0054] In one embodiment of the present application, the hyperparameters of the initial two-layer LSTM model include model batch size, sliding window function, gradient descent algorithm, number of hidden layer neurons and learning rate.

[0055] In one embodiment of the present application, the hyperparameters of the initial two-layer LSTM model include model batch size, history window, gradient descent method, the number of neurons contained in a single hidden layer, learning rate, etc. During the model training process, the order of determining the hyperparameters is: model batch size → optimal history window size → gradient descent algorithm → number of neurons in the hidden layer → learning rate. In addition, when performing transfer learning, the model can be adjusted to be adaptive according to the specific conditions of the target site, mainly adjusting the learning rate and batch size of the model.

[0056] In one embodiment of the present application, determining the loss function based on the initial predicted runoff data and the actual runoff data includes:

[0057] A root mean square error is determined based on the initial predicted runoff data and the actual runoff data.

[0058] In one embodiment of the present application, actual runoff data is input into the model as a simulation target, the model is trained and verified, the root mean square error (RMSE) between the initial predicted runoff data and the actual runoff data is selected as the loss function, and continuous iterative training is performed to adjust the parameters of the model. Specifically, the calculation formula of the root mean square error (RMSE) is as follows.

[0059]

[0060] Among them, Q obs,i is the actual runoff data, Q sim,i is the initial predicted runoff data.

[0061] In one embodiment of the present application, the method further includes:

[0062] Data from multiple study area sites were selected for combined training, and the length of the training period was adjusted to evaluate the migration effect of the target two-layer LSTM model.

[0063] In one embodiment of the present application, data from multiple study area sites are selected to perform combined training on the model, and the length of the training period is adjusted to evaluate the migration effect of the target two-layer LSTM model. Specifically, different combination schemes are shown in the following table.

[0064]

[0065] At the same time, we use a variety of evaluation indicators such as RMSE (Root Mean Square Error), NSE (Nash Sutcliffeefficiency), KGE (Kling–Gupta Efficiency) and RBias (Relative Bias) to comprehensively evaluate the simulation results, analyze the performance differences of the model in different scenarios and their causes, and iteratively optimize the model based on the feedback of the experimental results, continuously adjust the model's hyperparameters, and obtain the optimal hyperparameter combination of the model.

[0066] In one embodiment of the present application, the method further includes:

[0067] The BTOP hydrological model is used to predict the runoff flow data of the data-free basin and compared with the runoff data of the target data-free basin to verify the migration effect of the target two-layer LSTM model.

[0068] In one embodiment of the present application, the BTOP (Block-wise use of the TOPMODEL) hydrological model is a distributed basin hydrological model based on physical mechanisms, which is used for distributed hydrological simulation of large basins. It divides the entire basin into several natural sub-basins according to the terrain heterogeneity, and believes that areas with the same terrain index in the basin also have the same hydrological response. The relatively few tuning parameters of the BTOP hydrological model can effectively reduce the computational burden and uncertainty of the model, and has been widely used in more than 2,000 basins, including basin water resources management, flood and drought forecasting, and water environment simulation. The BTOP hydrological model is a useful tool for runoff simulation and integrated water resources management applications. The BTOP hydrological model is used to predict the runoff flow data of the data-free basin, and compared with the prediction results of the target two-layer LSTM model to verify the migration effect of the target two-layer LSTM model. Similarly, the validation indicators used are NSE (Nash Sutcliffe efficiency), RMSE (Root Mean Square Error), KGE (Kling–Gupta Efficiency) and RBias (Relative Bias), which are widely used in the field of hydrological model evaluation. The specific calculation formula is as follows.

[0069]

[0070] Among them, Q obs,i is the actual runoff data, Q sim,i is the initial predicted runoff data, r is the correlation coefficient, β is the deviation ratio, and γ is the variability ratio.

[0071] In this embodiment, the BTOP hydrological model is used to predict the runoff flow data of the data-free basin, and compared with the runoff data of the target data-free basin to verify the migration effect of the target two-layer LSTM model. It can be verified that the two-layer LSTM model driven only by meteorological data is generally superior to the physics-based BTOP hydrological model, and the credibility of the target two-layer LSTM model is verified.

[0072] The following is a specific example to illustrate the specific implementation steps of the data-free basin runoff prediction method based on double-layer LSTM transfer learning of the present application. First, S301, obtain the site data of the study area, perform preprocessing, and obtain training data. Specifically, S303-S305, perform accuracy detection and surface averaging on the site data of the study area. Use correlation analysis to determine the feature data in the site data of the study area that is most relevant to runoff prediction as training data.

[0073] Afterwards, S307, the training data is input into the initial two-layer LSTM model to obtain initial predicted runoff data. S309, a loss function is determined based on the initial predicted runoff data and the actual runoff data, and the hyperparameters of the initial two-layer LSTM model are adjusted based on the loss function to obtain a target two-layer LSTM model. Among them, S311, the hyperparameters of the initial two-layer LSTM model include model batch size, sliding window function, gradient descent algorithm, number of hidden layer neurons and learning rate. S313, a root mean square error is determined based on the initial predicted runoff data and the actual runoff data.

[0074] After that, in S315, data from multiple study area sites are selected for combined training, and the length of the training period is adjusted to evaluate the migration effect of the target two-layer LSTM model. In S317, the BTOP hydrological model is used to predict the runoff flow data of the watershed without data, and compared with the runoff data of the target watershed without data to verify the migration effect of the target two-layer LSTM model. Finally, in S319, the meteorological data of the watershed site to be predicted without data is input into the target two-layer LSTM model to obtain the runoff data of the target watershed without data.

[0075] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0076] Based on the same inventive concept, the embodiment of the present application also provides a device for predicting runoff in a watershed without data based on double-layer LSTM transfer learning, which is used to implement the method for predicting runoff in a watershed without data based on double-layer LSTM transfer learning. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more embodiments of the device for predicting runoff in a watershed without data based on double-layer LSTM transfer learning provided below can be referred to the limitations of the method for predicting runoff in a watershed without data based on double-layer LSTM transfer learning above, and will not be repeated here.

[0077] In one embodiment, Figure 3 As shown, a data-free basin runoff prediction device 300 based on double-layer LSTM transfer learning is provided, comprising: a data acquisition module 301, an initial predicted runoff data determination module 303, a model training module 305 and a data-free basin runoff prediction module 307, wherein:

[0078] The data acquisition module 301 is used to acquire the site data in the study area, perform preprocessing, and obtain training data.

[0079] The initial predicted runoff data determination module 303 is used to input the training data into the initial two-layer LSTM model to obtain initial predicted runoff data.

[0080] The model training module 305 is used to determine a loss function based on the initial predicted runoff data and the actual runoff data, and adjust the hyperparameters of the initial two-layer LSTM model based on the loss function to obtain a target two-layer LSTM model.

[0081] The data-free basin runoff prediction module 307 is used to input the meteorological data of the basin site to be predicted into the target double-layer LSTM model to obtain the target data-free basin runoff data.

[0082] In one embodiment of the present application, the data acquisition module is further used to:

[0083] The data of the stations in the study area are checked for accuracy and averaged.

[0084] In one embodiment of the present application, the data acquisition module is further used to:

[0085] Correlation analysis is used to determine the characteristic data most relevant to runoff prediction in the site data of the study area as training data.

[0086] In one embodiment of the present application, the hyperparameters of the initial two-layer LSTM model include model batch size, sliding window function, gradient descent algorithm, number of hidden layer neurons and learning rate.

[0087] In one embodiment of the present application, the model training module is also used to:

[0088] A root mean square error is determined based on the initial predicted runoff data and the actual runoff data.

[0089] In one embodiment of the present application, the model training module is also used to:

[0090] Data from multiple study area sites were selected for combined training, and the length of the training period was adjusted to evaluate the migration effect of the target two-layer LSTM model.

[0091] In one embodiment of the present application, the model training module is also used to:

[0092] The BTOP hydrological model is used to predict the runoff flow data of the data-free basin and compared with the runoff data of the target data-free basin to verify the migration effect of the target two-layer LSTM model.

[0093] Each module in the above-mentioned data-free watershed runoff prediction device based on double-layer LSTM transfer learning can be implemented in whole or in part by software, hardware and their combination. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the corresponding operations of each of the above modules.

[0094] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Figure 4As shown. The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. 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 of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a method for predicting runoff in a watershed without data based on double-layer LSTM transfer learning is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a key, trackball or touchpad set on the computer device shell, or an external keyboard, touchpad or mouse.

[0095] Those skilled in the art will understand that Figure 4 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0096] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0097] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0098] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0099] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0100] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0101] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0102] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A data-free basin runoff prediction method based on double-layer LSTM transfer learning, characterized in that: The method comprises: Obtain data from the study area sites, perform preprocessing, and obtain training data; Inputting the training data into an initial two-layer LSTM model to obtain initial predicted runoff data; Determine a loss function based on the initial predicted runoff data and the actual runoff data, and adjust the hyperparameters of the initial two-layer LSTM model based on the loss function to obtain a target two-layer LSTM model; The meteorological data of the watershed site to be predicted with no data is input into the target two-layer LSTM model to obtain the runoff data of the target watershed with no data.

2. According to claim 1, the method for predicting runoff in a watershed without data based on double-layer LSTM transfer learning is characterized in that: The acquisition of site data in the study area and preprocessing include: The data of the stations in the study area are checked for accuracy and averaged.

3. The method for predicting runoff in a watershed without data based on double-layer LSTM transfer learning according to claim 2 is characterized in that: The acquisition of the site data in the study area and preprocessing also include: Correlation analysis is used to determine the characteristic data most relevant to runoff prediction in the site data of the study area as training data.

4. The method for predicting runoff in a watershed without data based on double-layer LSTM transfer learning according to claim 1 is characterized in that: The hyperparameters of the initial two-layer LSTM model include model batch size, sliding window function, gradient descent algorithm, number of hidden layer neurons and learning rate.

5. The method for predicting runoff in a watershed without data based on double-layer LSTM transfer learning according to claim 1 is characterized in that: Determining the loss function based on the initial predicted runoff data and the actual runoff data includes: A root mean square error is determined based on the initial predicted runoff data and the actual runoff data.

6. The data-free basin runoff prediction method based on double-layer LSTM transfer learning according to claim 1 is characterized in that: The method further comprises: Data from multiple study area sites were selected for combined training, and the length of the training period was adjusted to evaluate the migration effect of the target two-layer LSTM model.

7. The method for predicting runoff in a watershed without data based on double-layer LSTM transfer learning according to claim 1 is characterized in that: The method further comprises: The BTOP hydrological model is used to predict the runoff flow data of the data-free basin and compared with the runoff data of the target data-free basin to verify the migration effect of the target two-layer LSTM model.

8. A data-free basin runoff prediction device based on double-layer LSTM transfer learning, characterized in that: The device comprises: The data acquisition module is used to obtain the data of the sites in the study area, perform preprocessing, and obtain training data; An initial predicted runoff data determination module is used to input the training data into an initial double-layer LSTM model to obtain initial predicted runoff data; A model training module, used for determining a loss function based on the initial predicted runoff data and the actual runoff data, and adjusting the hyperparameters of the initial two-layer LSTM model based on the loss function to obtain a target two-layer LSTM model; The data-free basin runoff prediction module is used to input the meteorological data of the basin site to be predicted into the target double-layer LSTM model to obtain the target data-free basin runoff data.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images