A hydrological forecast data-driven modeling sample synthesis method and system

Abstract

The present invention discloses a method and system for synthesizing hydrological forecast data-driven modeling samples, the method comprising: randomly sampling meteorological factors to obtain meteorological factor sampling sequences; dividing basin historical observation data into verification period data and test period data; using verification period data to verify basin hydrological process simulation models, the basin hydrological process simulation models include physical hydrological models and structural deviation simulation models, using test period data to test basin hydrological process simulation models; inputting meteorological factor sampling sequences into the verified and tested basin hydrological process simulation models to obtain corresponding hydrological process simulation results; the meteorological factor sampling sequences, underlying surface environmental data and hydrological process simulation results are finally hydrological forecast data-driven modeling samples. The present invention introduces a structural deviation simulation model to make the synthesized hydrological forecast data-driven modeling samples closer to the actual basin hydrological process.

Description

Technical Field

[0001] The present invention relates to the technical field of hydrological forecasting, and in particular to a hydrological forecasting data-driven modeling sample synthesis method and system. Background Art

[0002] The hydrological forecast model constructed using data-driven modeling technology has outstanding forecasting capabilities for hydrological processes. Relying on big data level samples to carry out data-driven modeling can significantly improve the accuracy of hydrological forecasts. However, the sample size in the field of hydrological forecasting is still difficult to reach the big data level in the theoretical sense. Objective factors such as equipment failure, transmission anomalies, and extreme weather also limit the number of effective samples used in data-driven modeling. Therefore, sample synthesis is an effective tool to meet the large sample requirements of data-driven modeling.

[0003] At present, physical hydrological models are mostly used to simulate the hydrological process of the basin, so as to achieve data-driven modeling sample synthesis. However, physical hydrological models have structural deviations, and there is a certain gap between the sample synthesis results and the actual hydrological process, which will have an adverse effect on the effect of data-driven modeling. Summary of the invention

[0004] The embodiments of the present invention provide a hydrological forecast data-driven modeling sample synthesis method and system to solve the problem that the physical hydrological model in the prior art has structural deviations, resulting in a certain gap between the sample synthesis results and the actual hydrological process.

[0005] On the one hand, an embodiment of the present invention provides a hydrological forecast data-driven modeling sample synthesis method, comprising:

[0006] Randomly sample meteorological factors to obtain a meteorological factor sampling sequence;

[0007] Obtain historical observation data of the basin including meteorological factor data, underlying surface environment data and hydrological sequence data;

[0008] The historical observation data of the basin are divided into verification period data and test period data;

[0009] The verification period data is used to verify the basin hydrological process simulation model, which includes a physical hydrological model and a structural deviation simulation model. Both the verification period data and the test period data contain driving factors and simulation objects. In the process of verifying the basin hydrological process simulation model, the driving factors in the verification period data are used as input and the simulation objects in the verification period data are used as output to verify the physical hydrological model, and the driving factors in the verification period data are used as input and the structural deviation of the physical hydrological model in the verification period is used as output to verify the structural deviation simulation model.

[0010] Input the driving factors in the test period data into a verified physical hydrological model to obtain the simulation results of the physical hydrological model in the test period; input the driving factors in the test period data into a verified structural deviation simulation model to obtain the structural simulation deviation of the physical hydrological model in the test period; determine the simulation results of the basin hydrological process simulation model in the test period according to the simulation results of the physical hydrological model in the test period and the structural simulation deviation of the physical hydrological model in the test period;

[0011] The meteorological factor sampling sequence is input into the verified and tested watershed hydrological process simulation model to obtain the corresponding hydrological process simulation results. The meteorological factor sampling sequence, underlying surface environmental data and hydrological process simulation results are sorted out to obtain the hydrological forecast data-driven modeling sample.

[0012] On the other hand, an embodiment of the present invention provides a hydrological forecast data-driven modeling sample synthesis system, comprising:

[0013] A random sampling module is used to randomly sample meteorological factors to obtain a meteorological factor sampling sequence;

[0014] The historical data acquisition module is used to obtain the basin historical observation data including meteorological factor data, underlying surface environment data and hydrological sequence data;

[0015] A data division module is used to divide the historical observation data of the watershed into verification period data and test period data;

[0016] A model verification module is used to verify the basin hydrological process simulation model using the verification period data. The basin hydrological process simulation model includes a physical hydrological model and a structural deviation simulation model. Both the verification period data and the test period data contain driving factors and simulation objects. In the process of verifying the basin hydrological process simulation model, the driving factors in the verification period data are used as input and the simulation objects in the verification period data are used as output to verify the physical hydrological model. The driving factors in the verification period data are used as input and the structural deviation of the physical hydrological model in the verification period is used as output to verify the structural deviation simulation model.

[0017] A model testing module is used to input the driving factors in the test period data into a verified physical hydrological model to obtain the simulation results of the physical hydrological model in the test period, input the driving factors in the test period data into a verified structural deviation simulation model to obtain the structural simulation deviation of the physical hydrological model in the test period, and determine the simulation results of the basin hydrological process simulation model in the test period according to the simulation results of the physical hydrological model in the test period and the structural simulation deviation of the physical hydrological model in the test period;

[0018] The sample generation module is used to input the meteorological factor sampling sequence into the verified and tested watershed hydrological process simulation model to obtain the corresponding hydrological process simulation results, organize the meteorological factor sampling sequence, underlying surface environmental data and hydrological process simulation results, and obtain the hydrological forecast data-driven modeling samples.

[0019] The hydrological forecast data-driven modeling sample synthesis method and system of the present invention have the following advantages:

[0020] 1. The synthetic hydrological forecast data-driven modeling samples are closer to the actual basin hydrological process.

[0021] 2. Data-driven modeling samples for synthesizing hydrological forecasts can be developed for any hydrological process. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0023] Figure 1 A flowchart of a hydrological forecast data-driven modeling sample synthesis method provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0024] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0025] Figure 1 A flow chart of a hydrological forecast data driven modeling sample synthesis method provided in an embodiment of the present invention. An embodiment of the present invention provides a hydrological forecast data driven modeling sample synthesis method, comprising:

[0026] Randomly sample meteorological factors to obtain a meteorological factor sampling sequence;

[0027] Obtain historical observation data of the basin including meteorological factor data, underlying surface environment data and hydrological sequence data;

[0028] The historical observation data of the basin are divided into verification period data and test period data;

[0029] The verification period data is used to verify the basin hydrological process simulation model, which includes a physical hydrological model and a structural deviation simulation model. Both the verification period data and the test period data contain driving factors and simulation objects. In the process of verifying the basin hydrological process simulation model, the driving factors in the verification period data are used as input and the simulation objects in the verification period data are used as output to verify the physical hydrological model, and the driving factors in the verification period data are used as input and the structural deviation of the physical hydrological model in the verification period is used as output to verify the structural deviation simulation model.

[0030] Input the driving factors in the test period data into a verified physical hydrological model to obtain the simulation results of the physical hydrological model in the test period; input the driving factors in the test period data into a verified structural deviation simulation model to obtain the structural simulation deviation of the physical hydrological model in the test period; determine the simulation results of the basin hydrological process simulation model in the test period according to the simulation results of the physical hydrological model in the test period and the structural simulation deviation of the physical hydrological model in the test period;

[0031] The meteorological factor sampling sequence is input into the verified and tested watershed hydrological process simulation model to obtain the corresponding hydrological process simulation results. The meteorological factor sampling sequence, underlying surface environmental data and hydrological process simulation results are sorted out to obtain the hydrological forecast data-driven modeling sample.

[0032] Exemplarily, the historical observation data of the watershed can be divided into historical observation data corresponding to multiple similar underlying surface environments according to the periodic and seasonal characteristics of the watershed hydrological process. Before verifying the watershed hydrological process simulation model, the historical observation data of the watershed of a similar underlying surface can be extracted as the verification period data, and then the physical hydrological model and the structural deviation simulation model can be verified based on the verification period data. During the verification process, the structural deviation of the physical hydrological model of the verification period can be determined according to the measured data of the simulation object in the verification period data and the physical hydrological simulation results of the verification period. Specifically, the structural deviation of the physical hydrological model of the verification period can be obtained by subtracting the simulation results of the physical hydrological model of the verification period from the measured data of the simulation object of the verification period, and then the structural deviation simulation model can be verified based on the structural deviation of the physical hydrological model of the verification period and the driving factors of the verification period.

[0033] In an embodiment of the present invention, the physical hydrological model may adopt the MIKE SHE model or other physical hydrological models with a physical basis, and the structural deviation simulation model may adopt a machine learning model or other data-driven model.

[0034] In an embodiment of the present invention, historical observation data of a watershed with a similar underlying surface can also be extracted as test period data. The test period data also includes driving factors and simulation objects. Before the meteorological factor sampling sequence is input into the verified watershed hydrological process simulation model, it also includes: using the test period data to test the verified watershed hydrological process simulation model. When the accuracy of the watershed hydrological process simulation model meets the requirements, the meteorological factor sampling sequence is input into the verified watershed hydrological process simulation model.

[0035] Specifically, after determining the hydrological process simulation results of the test period, it is also necessary to compare the hydrological process simulation results with the actual measured data of the simulation objects in the test period data to determine whether the accuracy of the basin hydrological process simulation model meets the requirements.

[0036] During the test, the hydrological process during the test period and the structural deviation during the test period can be added together to obtain the simulation results of the hydrological process during the test period.

[0037] In the above method, the hydrological process simulation results are compared with the actual measured data of the simulation object in the test period data to determine whether the accuracy meets the requirements, including: using the accuracy evaluation index to determine the accuracy of the basin hydrological process simulation model; comparing the accuracy of the basin hydrological process simulation model with the set accuracy threshold to determine whether the accuracy of the basin hydrological process simulation model meets the requirements. If the accuracy meets the requirements, the next step is carried out, otherwise the verification and test of the basin hydrological process simulation model are repeated until the accuracy of the basin hydrological process simulation model meets the requirements.

[0038] In actual processing, the accuracy evaluation index can be any one or more of RMSE, MAE and MAPE. For example, the calculation formulas of RMSE, MAE, MAPE and other indicators are as follows:

[0039]

[0040]

[0041]

[0042] Where N is the length of the hydrological process during the test period, x(t) is the simulation object in the test period data, It is the simulation result of the hydrological process during the test period.

[0043] In a possible embodiment, the method for randomly sampling meteorological factors includes: obtaining historical observation data of meteorological factors; normalizing the historical observation data of meteorological factors; establishing a time series generation model; using the normalized historical observation data of meteorological factors to train the time series generation model; using the trained time series generation model to randomly sample meteorological factors to obtain random sampling results of meteorological factors; and denormalizing the random sampling results of meteorological factors to obtain a meteorological factor sampling sequence.

[0044] Exemplarily, when normalizing the historical observation data of meteorological factors, linear normalization, mean normalization or other normalization methods can be used to normalize the historical observation data of meteorological factors.

[0045] For example, linear normalization is expressed as:

[0046]

[0047] Among them, x' represents the normalized result of the historical observation data of meteorological factors, x represents the historical observation data of meteorological factors, and x max and x min They respectively represent the maximum and minimum values ​​of the historical observation data of meteorological factors.

[0048] When establishing a time series generation model, the TimeGAN model or other models can be used. If the TimeGAN model is used, the operator needs to provide constructor parameters. The constructor parameters that need to be provided in the present invention include the meteorological factor dimension and the time window size, and other constructor parameters can use the default parameters of the TimeGAN model.

[0049] When using the trained time series generation model to randomly sample meteorological factors, the operator is also required to specify the time step. After the time step is determined, the time series generation model randomly samples the meteorological factors according to the time step.

[0050] When the random sampling results of meteorological factors are denormalized, the following formula can be used:

[0051]

[0052] in, Represents the meteorological factor sampling sequence.

[0053] The embodiment of the present invention also provides a hydrological forecast data driven modeling sample synthesis system, comprising:

[0054] A random sampling module is used to randomly sample meteorological factors to obtain a meteorological factor sampling sequence;

[0055] The historical data acquisition module is used to obtain the basin historical observation data including meteorological factor data, underlying surface environment data and hydrological sequence data;

[0056] A data division module is used to divide the historical observation data of the watershed into verification period data and test period data;

[0057] A model verification module is used to verify the basin hydrological process simulation model using the verification period data. The basin hydrological process simulation model includes a physical hydrological model and a structural deviation simulation model. Both the verification period data and the test period data contain driving factors and simulation objects. In the process of verifying the basin hydrological process simulation model, the driving factors in the verification period data are used as input and the simulation objects in the verification period data are used as output to verify the physical hydrological model. The driving factors in the verification period data are used as input and the structural deviation of the physical hydrological model in the verification period is used as output to verify the structural deviation simulation model.

[0058] A model testing module is used to input the driving factors in the test period data into a verified physical hydrological model to obtain the simulation results of the physical hydrological model in the test period, input the driving factors in the test period data into a verified structural deviation simulation model to obtain the structural simulation deviation of the physical hydrological model in the test period, and determine the simulation results of the basin hydrological process simulation model in the test period according to the simulation results of the physical hydrological model in the test period and the structural simulation deviation of the physical hydrological model in the test period;

[0059] The sample generation module is used to input the meteorological factor sampling sequence into the verified and tested watershed hydrological process simulation model to obtain the corresponding hydrological process simulation results, organize the meteorological factor sampling sequence, underlying surface environmental data and hydrological process simulation results, and obtain the hydrological forecast data-driven modeling samples.

[0060] Experimental Description

[0061] 1. Collect historical observation data of the basin. The historical observation data of meteorological factors that need to be collected include hydrological and meteorological data such as runoff, precipitation, evaporation pressure, temperature, and underlying surface data such as topography, vegetation cover, and land use. For example: the runoff data can use the daily runoff data of the Xianyang Hydrological Station in the Weihe River Basin in Shaanxi Province from January 1960 to December 2021, and the meteorological factors can use the data of the time period corresponding to the runoff data. The main data types and data magnitudes are shown in Table 1 below. The underlying surface data uses the underlying surface data corresponding to the control basin of the Xianyang Hydrological Station. Table 2 is a partial sequence of meteorological factors selected from the Xianyang Station.

[0062] Table 1 Basic information of meteorological observation variables

[0063]

[0064] Table 2 Meteorological factor sequence

[0065]

[0066] 2. Random sampling of meteorological factors: Randomly sample the meteorological factor data in the historical observation data of the basin.

[0067] (1) The 7 types of meteorological data in Table 2 are normalized, and the normalized meteorological series are shown in Table 3.

[0068] Table 3 Normalized sequence of meteorological factors

[0069]

[0070] (2) Determine the constructor parameters of the TimeGAN model. The parameters that need to be input include the meteorological factor dimension, time window size and time step. The meteorological factor dimension is the number of meteorological observation variables input. It can be seen from Table 1 that the meteorological factor dimension is 7. The time window size is set according to the periodic characteristics of the meteorological factor sequence. The present invention uses daily-scale meteorological data, and assumes that the meteorological factor sequence has a regularity characteristic with a period of 5 days, so the time window size is set to 5. The time step refers to the output sequence length of the meteorological sequence after sampling by the TimeGAN model. Assuming that the sequence length of the meteorological sequence with an original sequence length of 5 is 10 after sampling, the time step is set to 10. As shown in Table 4, this is the normalized sequence of meteorological sampling after sampling in Table 3.

[0071] Table 4 Normalized sequence of meteorological sampling

[0072]

[0073] (3) Denormalizing the normalized meteorological sampling sequence in Table 4 can obtain the meteorological sampling sequence, as shown in Table 5.

[0074] Table 5 Meteorological sampling sequence

[0075]

[0076] 3. Simulation of watershed hydrological processes.

[0077] (1) The historical observation data of the basin are collected for simulating the hydrological process of the basin. The historical observation data of the basin are shown in Tables 6 and 7 below.

[0078] Table 6 Basin historical observation data information

[0079]

[0080]

[0081] Table 7 Different land use types and their Manning coefficients in the study area

[0082]

[0083] (2) The historical observation data used in the present invention is shown in Table 8, and the first three rows are divided into verification period data, and the last two rows are divided into test period data.

[0084] Table 8 Historical observation data

[0085]

[0086] Note: Land use and digital elevation data are high-dimensional data and cannot be displayed in this table.

[0087] The verification period data and the test period data are divided into driving factors and simulation objects respectively, where the driving factors are AVG_PS, AVG_T, RHU, P208, P820, P2020, L_EVP, ZB, TL, and DEM, and the simulation object is Q.

[0088] (3) The driving factors of the verification period and the required underlying surface data such as topography, vegetation cover, and land use are input into the MIKE SHE model to calibrate the parameters of the MIKE SHE model. The parameters and calibration results of the MIKE SHE model are shown in Table 9. After parameter calibration, the runoff simulation results of the MIKE SHE model for the verification period are obtained. Table 10 shows the comparison results between the measured and simulated values ​​of the MIKE SHE model for the verification period, and the deviation values ​​between the measured and predicted values ​​are obtained.

[0089] Table 9 Results of parameter calibration of MIKE SHE model

[0090]

[0091] Table 10 Comparison results of measured and simulated values ​​of MIKE SHE model during the verification period

[0092]

[0093] (4) The driving factors in Table 8 and the prediction deviations in Table 10 constitute the runoff prediction samples and are input into the machine learning model SVR, and the parameters of the SVR model are calibrated. The parameters of the SVR model and the calibration results are shown in Table 11. The prediction results of the SVR model for the deviation Qe during the calibration period are shown in Table 12.

[0094] Table 11 SVR model parameters

[0095]

[0096] Table 12 Comparison between prediction and actual measurement of deviation Qe by SVR during the calibration period

[0097]

[0098] (5) The driving factors in the test period data are input into the MIKE SHE and SVR models with calibrated parameters to obtain the runoff prediction results and prediction deviations for the test period. The prediction results for the test period are added to the prediction deviations for the test period to obtain the hydrological process simulation results for the test period, as shown in Table 13 below.

[0099] Table 13 Comparison between simulation results and actual measurements during the test period

[0100]

[0101] (6) Use RMSE, MAE, MAPE and other indicators to compare the test period simulation with the test period actual measurement to determine whether the accuracy meets the requirements. The prediction evaluation indicators are shown in Table 14. Set the threshold of each indicator. If the calculated indicator value meets the threshold setting, the test period simulation result meets the accuracy requirements. If not, recalibrate the parameters of the MIKE SHE model and the SVR model until the accuracy meets the requirements. After the accuracy meets the requirements, the combined MIKE SHE model and the SVR model is the basin hydrological process simulation model of this patent.

[0102] Table 14 Evaluation indexes of runoff simulation results during the test period

[0103]

[0104] (7) The meteorological factor sampling sequence in Table 5 is input into the basin hydrological process simulation model to obtain the hydrological process simulation results of the meteorological sampling sequence, and the meteorological sampling sequence is combined with the simulated runoff results to obtain the runoff prediction samples shown in Table 15. Combined with the underlying surface data such as topography, vegetation coverage, and land use as the final hydrological forecast data-driven modeling samples, the practical application purpose of the present invention is achieved.

[0105] Table 15 Runoff prediction samples

[0106]

[0107] Note: Land use and digital elevation data are high-dimensional data and cannot be displayed in this table.

[0108] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0109] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A hydrological forecast data-driven modeling sample synthesis method, characterized in that: include: Randomly sample meteorological factors to obtain a meteorological factor sampling sequence; Obtain historical observation data of the basin including meteorological factor data, underlying surface environment data and hydrological sequence data; Dividing the historical observation data of the watershed into verification period data and test period data; The verification period data is used to verify the basin hydrological process simulation model, the basin hydrological process simulation model includes a physical hydrological model and a structural deviation simulation model, the verification period data and the test period data both include driving factors and simulation objects, in the process of verifying the basin hydrological process simulation model, the driving factors in the verification period data are used as input, and the simulation objects in the verification period data are used as output, the physical hydrological model is verified, and the driving factors in the verification period data are used as input, and the structural deviation of the physical hydrological model in the verification period is used as output, the structural deviation simulation model is verified; Input the driving factors in the test period data into a verified physical hydrological model to obtain the simulation results of the test period physical hydrological model; input the driving factors in the test period data into a verified structural deviation simulation model to obtain the structural simulation deviation of the test period physical hydrological model; determine the test period simulation results of the basin hydrological process simulation model according to the simulation results of the test period physical hydrological model and the structural simulation deviation of the test period physical hydrological model; The meteorological factor sampling sequence is input into the verified and tested watershed hydrological process simulation model to obtain the corresponding hydrological process simulation results. The meteorological factor sampling sequence, underlying surface environmental data and hydrological process simulation results are sorted out to obtain a hydrological forecast data-driven modeling sample.

2. A hydrological forecast data-driven modeling sample synthesis method according to claim 1, characterized in that: After determining the simulation results of the basin hydrological process simulation model test period, it also includes: The hydrological process simulation results are compared with the actual measured data of the simulation objects in the test period data to determine whether the accuracy of the basin hydrological process simulation model meets the requirements.

3. A hydrological forecast data-driven modeling sample synthesis method according to claim 2, characterized in that: Comparing the hydrological process simulation results with the actual measured data of the simulation object in the test period data to determine whether the accuracy meets the requirements, including: Using accuracy evaluation index to determine the accuracy of the watershed hydrological process simulation model; The accuracy of the watershed hydrological process simulation model is compared with a set accuracy threshold to determine whether the accuracy of the watershed hydrological process simulation model meets the requirements.

4. A hydrological forecast data-driven modeling sample synthesis method according to claim 3, characterized in that: The accuracy evaluation index is any one or more of RMSE, MAE and MAPE.

5. The hydrological forecast data-driven modeling sample synthesis method according to claim 1, characterized in that: The physical hydrological model is a MIKESHE model, and the structural deviation simulation model is a machine learning model.

6. A hydrological forecast data-driven modeling sample synthesis method according to claim 1, characterized in that: The method of randomly sampling the meteorological factors includes: Obtain historical observation data of meteorological factors; Normalizing the historical observation data of the meteorological factors; Build a time series generation model; Using the normalized historical observation data of the meteorological factors to train the time series generation model; The trained time series generation model is used to randomly sample meteorological factors to obtain meteorological factor random sampling results; The meteorological factor random sampling results are denormalized to obtain the meteorological factor sampling sequence.

7. A hydrological forecast data-driven modeling sample synthesis method according to claim 6, characterized in that: The normalizing of the historical observation data of the meteorological factors includes: The historical observation data of the meteorological factors are normalized by linear normalization or mean normalization.

8. The hydrological forecast data-driven modeling sample synthesis method according to claim 6, characterized in that: The time series generation model is a TimeGAN model.

9. A hydrological forecast data-driven modeling sample synthesis system, characterized in that: include: A random sampling module is used to randomly sample meteorological factors to obtain a meteorological factor sampling sequence; The historical data acquisition module is used to obtain the basin historical observation data including meteorological factor data, underlying surface environment data and hydrological sequence data; A data division module, used for dividing the historical observation data of the watershed into verification period data and test period data; A model verification module, used for verifying a basin hydrological process simulation model using the verification period data, wherein the basin hydrological process simulation model includes a physical hydrological model and a structural deviation simulation model, and the verification period data and the test period data both include driving factors and simulation objects. In the process of verifying the basin hydrological process simulation model, the driving factors in the verification period data are used as input, and the simulation objects in the verification period data are used as output to verify the physical hydrological model, and the driving factors in the verification period data are used as input, and the structural deviation of the physical hydrological model in the verification period is used as output to verify the structural deviation simulation model; A model testing module, used for inputting the driving factors in the test period data into a verified physical hydrological model to obtain the simulation results of the physical hydrological model in the test period, inputting the driving factors in the test period data into a verified structural deviation simulation model to obtain the structural simulation deviation of the physical hydrological model in the test period, and determining the simulation results of the test period of the basin hydrological process simulation model according to the simulation results of the physical hydrological model in the test period and the structural simulation deviation of the physical hydrological model in the test period; The sample generation module is used to input the meteorological factor sampling sequence into the verified and tested watershed hydrological process simulation model to obtain the corresponding hydrological process simulation results, organize the meteorological factor sampling sequence, underlying surface environmental data and hydrological process simulation results, and obtain hydrological forecast data-driven modeling samples.

Images