Drainage basin runoff prediction method and system
By gridding the watershed and combining a polynomial regression model with a variety of meteorological data, the problems of spatial heterogeneity and single meteorological factors in watershed runoff prediction are solved, and runoff prediction with higher accuracy and efficiency is achieved, supporting water resource scheduling and risk warning within the watershed.
Patent Information
- Application Number
- CN202510879508.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-27
- Publication Date
- 2025-10-14
AI Technical Summary
Existing basin runoff prediction methods do not fully consider the spatial heterogeneity within the basin, resulting in large deviations in the prediction results. They also rely on a single meteorological factor, resulting in low reliability of the prediction results. The complex calculation model makes it difficult to achieve timely and rapid data feedback.
The watershed is gridded, and multiple meteorological state models are established. Meteorological forecasts are performed for each watershed grid center, and runoff data are predicted using a polynomial regression model. Combined with multiple meteorological variables such as temperature, air pressure, rainfall, and humidity, the first and second cycles are coordinated to match the meteorological forecast frequency with the runoff forecast frequency.
It achieves runoff prediction with higher accuracy and efficiency, provides more fine-grained prediction results, and supports water resource scheduling and risk warning in different areas within the basin.
Smart Images
Figure CN120782045A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of watershed runoff prediction, and in particular to a watershed runoff prediction method and system. Background Art
[0002] With the frequent occurrence of extreme weather events and the increasing demand for water resources regulation, the importance of river basin hydrological forecasting technology has become increasingly prominent.
[0003] Existing methods for predicting runoff in a river basin typically rely on traditional empirical models or hydrological models based on a single meteorological data source. These methods have limitations, such as failing to fully account for spatial heterogeneity within the river basin, leading to large deviations in prediction results; considering only a single meteorological factor, such as temperature or precipitation, resulting in low reliability; or having complex computational models that hinder timely and rapid feedback of forecast data.
[0004] Therefore, there is an urgent need for a basin runoff prediction method that can integrate multi-dimensional meteorological information, have higher spatial resolution and timeliness, and improve the accuracy and efficiency of runoff prediction. Summary of the Invention
[0005] The purpose of the present invention is to provide a method and system for predicting watershed runoff, which can improve the accuracy and efficiency of runoff prediction.
[0006] The present invention is achieved through the following technical solutions:
[0007] The present invention first provides a basin runoff prediction method, which performs gridding processing on the basin to be predicted to obtain multiple basin grids, and obtains the center of each basin grid. The following operations are performed on each basin grid:
[0008] Establishing a plurality of meteorological state models for the center of the watershed grid, each meteorological state model is used to describe a type of meteorological data of the center of the watershed grid;
[0009] Periodically predicting multiple meteorological forecast data of the center of the watershed grid based on multiple meteorological state models at intervals of the first period;
[0010] The runoff data at the center of the watershed grid is periodically predicted based on multiple meteorological forecast data and initial runoff data at the center of the watershed grid at intervals of a second period, wherein the second period is greater than the first period and is an integer multiple of the first period.
[0011] Preferably, the method for performing gridding processing on the watershed to be predicted is:
[0012] Covering the entire watershed to be predicted by a first grid, where the first grid is a minimum square covering the entire watershed to be predicted;
[0013] Dividing the first grid into a plurality of second grids, wherein the second grids are squares having a side length smaller than that of the first grid;
[0014] The portion of the watershed to be predicted covered by each second grid constitutes a watershed grid.
[0015] Preferably, the method for obtaining the center of the watershed grid is: obtaining the geometric center of the watershed grid.
[0016] Preferably, the method for obtaining the weather forecast data is to perform the following operations for each type of weather data:
[0017] Obtaining initial meteorological data of the center of the watershed grid, and obtaining a time difference between a time point of the initial meteorological data and a predicted time point;
[0018] The meteorological state model is solved based on the initial meteorological data and the time difference to obtain meteorological forecast data at the forecast time point.
[0019] Preferably, the method for establishing the meteorological state model is to establish a fitting polynomial:
[0020]
[0021] Among them, Y b is the meteorological forecast data of the center of the watershed grid b, Y 0,b is the initial meteorological data of the center of the watershed grid b, t is the time difference between the time point of the initial meteorological data and the predicted time point, Δt is the fitting time step, is the coefficient of the kth term in the fitting polynomial of the center of the watershed grid b, The operator for rounding up.
[0022] Preferably, the coefficient of the kth term in the fitting polynomial of the center of the watershed grid b is The method to obtain is:
[0023] Divide a natural year into multiple sub-periods;
[0024] Perform the following processing for each sub-time period:
[0025] Acquire historical meteorological data of multiple time points in the sub-time period, and establish multiple historical meteorological data pairs, each historical meteorological data pair including historical meteorological data of any two time points and a time difference between the two time points;
[0026] Substitute all the historical meteorological data into the fitting polynomial for fitting, and obtain the coefficient of the kth term in the fitting polynomial for the corresponding sub-time period.
[0027] When obtaining the weather forecast data, obtain the sub-time period to which the current time belongs, and set the coefficient of the kth term in the corresponding fitting polynomial to Substitute into the meteorological state model.
[0028] Preferably, the types of meteorological forecast data include temperature forecast data, air pressure forecast data, rainfall forecast data and humidity forecast data.
[0029] Preferably, the method for predicting the runoff data is:
[0030] Get the time difference t' from the current time point to the predicted target time point;
[0031] Get the initial runoff data Q0 at the current time point;
[0032] Obtain all the weather forecast data between the current time point and the forecast target time point;
[0033] Predict the runoff data Q based on the polynomial regression model t′ :
[0034]
[0035] Among them, T1 represents the first cycle, R n represents the nth rainfall forecast data between the current time point and the forecast target time point, α m is the mth order coefficient of the polynomial of rainfall prediction data, M is the total number of terms of the polynomial of rainfall prediction data, T n represents the nth temperature prediction data between the current time point and the predicted target time point, β i is the i-th order coefficient of the polynomial of temperature prediction data, I is the total number of terms of the polynomial of temperature prediction data, P n represents the nth air pressure prediction data between the current time point and the predicted target time point, γ j is the j-th order coefficient of the polynomial of the air pressure prediction data, J is the total number of terms of the polynomial of the air pressure prediction data, H n represents the nth humidity prediction data between the current time point and the predicted target time point, δ l is the lth order coefficient of the polynomial of humidity prediction data, and L is the total number of terms of the polynomial of humidity prediction data;
[0036] The values of each coefficient are determined through fitting training.
[0037] Preferably, for each type of weather forecast data, the total number of terms of the polynomial corresponding to each type of weather forecast data in the polynomial regression model is determined by the following method:
[0038] Create the test polynomial:
[0039]
[0040] ω=α,β,γ,δ;
[0041] X=R,T,P,H;
[0042] Wherein, X represents the type of weather forecast data, X n The nth weather forecast data between the current time point and the forecast target time point, ω y is the y-th order coefficient of the polynomial;
[0043] Determine the values of each coefficient when Y=y through fitting training, y=1,2,3;
[0044] The test polynomial is tested by F groups of test data, and the prediction error v of the fth group of test data when Y is y is obtained respectively. f,y ,f=1,2,…,F;
[0045] Get the error judgment parameter ε when Y is y y ′:
[0046] ε y ′=var(ε f ,f=1,2,…,F);
[0047] Among them, var(.) represents the function of finding the variance;
[0048] Get ε y The minimum value of ′, the corresponding value of y is the total number of terms of the polynomial when the type of weather forecast data is X.
[0049] The present invention further provides a watershed runoff prediction system, which is applied to the above-mentioned watershed runoff prediction method, comprising:
[0050] The gridding module is used to grid the watershed to be predicted, obtain multiple watershed grids, and obtain the centers of the watershed grids respectively:
[0051] A meteorological state modeling module is used to establish multiple meteorological state models for the center of the watershed grid, each meteorological state model is used to describe a type of meteorological data of the center of the watershed grid;
[0052] A meteorological forecast module, configured to periodically forecast a plurality of meteorological forecast data of the center of the watershed grid according to a plurality of meteorological state models at intervals of a first period;
[0053] The runoff prediction module is used to periodically predict the runoff data of the center of the watershed grid based on multiple meteorological forecast data and initial runoff data of the center of the watershed grid at intervals of a second period, wherein the second period is greater than the first period and is an integer multiple of the first period.
[0054] The technical solution of the present invention has at least the following advantages and beneficial effects:
[0055] The present invention grids the watershed to be predicted and constructs a meteorological model with the center of each grid as the base point, thereby achieving fine-grained meteorological data modeling for each local area within the watershed, which helps to obtain finer-grained runoff predictions based on finer-grained meteorological forecasts.
[0056] The present invention establishes multiple meteorological models for each grid center to describe different types of meteorological data, thereby improving the model's ability to express complex meteorological changes and its prediction accuracy.
[0057] The present invention achieves a reasonable match between the meteorological forecast frequency and the runoff forecast frequency by coordinating the first cycle and the second cycle, thereby improving the adaptability of the model on a time scale and the stability of the forecast.
[0058] The present invention can output independent runoff prediction results for each grid area, providing more refined decision-making support for water resource scheduling and risk warning in different areas within the basin. BRIEF DESCRIPTION OF THE DRAWINGS
[0059] Figure 1 A schematic flow chart of a watershed runoff prediction method provided in Example 1 of the present invention;
[0060] Figure 2 A schematic diagram of the principle of a watershed runoff prediction system provided in Example 2 of the present invention. DETAILED DESCRIPTION
[0061] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of 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. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.
[0062] Example 1
[0063] This embodiment provides a method for predicting runoff in a watershed. Figure 1 .
[0064] In this embodiment, the watershed to be predicted is first gridded to obtain multiple watershed grids, and the centers of the watershed grids are obtained respectively. As an example of this embodiment, the method for gridding the watershed to be predicted can be:
[0065] Covering the entire watershed to be predicted by a first grid, where the first grid is a minimum square covering the entire watershed to be predicted;
[0066] Dividing the first grid into a plurality of second grids, wherein the second grids are squares having a side length smaller than that of the first grid;
[0067] The portion of the watershed to be predicted covered by each second grid constitutes a watershed grid.
[0068] At the same time, the method for obtaining the center of the watershed grid is: obtaining the geometric center of the watershed grid.
[0069] In the above scheme, by constructing the minimum square covering the entire basin to be predicted as the first grid, it can effectively avoid the omission of areas caused by irregular boundaries and ensure that the entire basin range is included in the subsequent modeling and prediction process. The use of a unified square grid as the division basis circumvents the geometric operation problems of complex polygonal grids, is easy to implement, has high computational efficiency, and is suitable for engineering deployment and large-scale automated modeling. In subsequent data analysis, each basin area covered by the second grid is treated as an independent grid, and then local meteorological models or hydrological models are established and updated separately, thereby achieving more refined spatial modeling and prediction, which helps to identify local anomalies or extreme changes.
[0070] Next, perform the following operations for each watershed grid:
[0071] First, multiple meteorological state models are established for the center of the watershed grid, and each meteorological state model is used to describe a type of meteorological data of the center of the watershed grid;
[0072] Then, at intervals of the first period, multiple meteorological forecast data of the center of the watershed grid are periodically predicted according to multiple meteorological state models;
[0073] As a preferred solution of this embodiment, the weather forecast data method is to perform the following operations for each type of weather data:
[0074] Obtaining initial meteorological data of the center of the watershed grid, and obtaining a time difference between a time point of the initial meteorological data and a predicted time point;
[0075] The meteorological state model is solved based on the initial meteorological data and the time difference to obtain meteorological forecast data at the forecast time point.
[0076] Specifically, the method for establishing the meteorological state model is to establish a fitting polynomial:
[0077]
[0078] Among them, Y b is the meteorological forecast data of the center of the watershed grid b, Y 0,b is the initial meteorological data of the center of the watershed grid b, t is the time difference between the time point of the initial meteorological data and the predicted time point, Δt is the fitting time step, is the coefficient of the kth term in the fitting polynomial of the center of the watershed grid b, The operator for rounding up.
[0079] This embodiment avoids cross-interference between multiple meteorological elements by modeling and predicting each type of meteorological data separately, making the model structure clear and the prediction mechanism more targeted. The model strategy can be flexibly adjusted according to the characteristics of different meteorological elements. Modeling is also based on the time difference t between the initial meteorological data and the prediction target time point, and fitting is performed through a fitting model based on an appropriate fitting time step Δt. It can flexibly respond to prediction needs of different time scales, including short-term and medium- and long-term meteorological forecasts. The value of Δt can be selected according to needs, so that the model complexity can be flexibly adjusted, thereby achieving a trade-off between model accuracy and computational overhead, meeting the modeling requirements of different river basins and different weather evolution rates. For example, the smaller Δt, the higher the accuracy but the greater the computational cost. Appropriately increasing Δt can reduce computational cost.
[0080] On this basis, the coefficient of the kth term in the fitting polynomial of the center of the watershed grid b is The method to obtain is:
[0081] Divide a natural year into multiple sub-periods;
[0082] Perform the following processing for each sub-time period:
[0083] Acquire historical meteorological data of multiple time points in the sub-time period, and establish multiple historical meteorological data pairs, each historical meteorological data pair including historical meteorological data of any two time points and a time difference between the two time points;
[0084] Substitute all the historical meteorological data into the fitting polynomial for fitting, and obtain the coefficient of the kth term in the fitting polynomial for the corresponding sub-time period.
[0085] When obtaining the weather forecast data, obtain the sub-time period to which the current time belongs, and set the coefficient of the kth term in the corresponding fitting polynomial to Substitute into the meteorological state model.
[0086] Since the meteorological change trends in different time periods are significantly different, uniformly using the fitting coefficients for the whole year may lead to prediction deviations. Here, a natural year is divided into multiple sub-time periods (such as months, weeks or other divisions), and polynomial coefficients are fitted for each sub-time period respectively, so that the model can fully consider the meteorological change patterns of different seasons or time periods, and improve the time series dynamic adaptability of the prediction model. By using the coefficients in the model parameters By linking the model to time periods, when faced with climate anomalies or localized sudden changes, the model can automatically select parameters that are more appropriate for the current climate context, reducing forecast volatility and error. Furthermore, while the model uses different fitting coefficients for each sub-time period, the overall model structure remains consistent, helping to improve forecast accuracy without introducing complex structures and facilitating model maintenance, debugging, and widespread application.
[0087] Finally, the runoff data at the center of the watershed grid is periodically predicted based on the multiple meteorological forecast data and initial runoff data at the center of the watershed grid at intervals of a second period, wherein the second period is greater than the first period and is an integer multiple of the first period.
[0088] As a preferred solution of this embodiment, the types of meteorological forecast data include temperature forecast data, air pressure forecast data, rainfall forecast data and humidity forecast data.
[0089] Based on the above scheme, the method for predicting the runoff data is preferably:
[0090] Get the time difference t' from the current time point to the predicted target time point;
[0091] Get the initial runoff data Q0 at the current time point;
[0092] Obtain all the weather forecast data between the current time point and the forecast target time point;
[0093] Predict the runoff data Q based on the polynomial regression model t′ :
[0094]
[0095] Among them, T1 represents the first cycle, R n represents the nth rainfall forecast data between the current time point and the forecast target time point, α m is the mth order coefficient of the polynomial of rainfall prediction data, M is the total number of terms of the polynomial of rainfall prediction data, T n represents the nth temperature prediction data between the current time point and the predicted target time point, β iis the i-th order coefficient of the polynomial of the temperature prediction data, I is the total number of items of the polynomial of the temperature prediction data, P is the total number of items of the polynomial of the pressure prediction data, H is the total number of items of the polynomial of the humidity prediction data, and Y is the total number of items of the polynomial of the weather prediction data. n represents the n-th weather prediction data between the current time point and the prediction target time point, and ω represents the type of the weather prediction data. j is the j-th order coefficient of the polynomial of the pressure prediction data, J is the total number of items of the polynomial of the pressure prediction data, H is the total number of items of the polynomial of the humidity prediction data, and Y is the total number of items of the polynomial of the weather prediction data. n represents the n-th weather prediction data between the current time point and the prediction target time point, and ω represents the type of the weather prediction data. l is the l-th order coefficient of the polynomial of the humidity prediction data, L is the total number of items of the polynomial of the humidity prediction data, P is the total number of items of the polynomial of the pressure prediction data, H is the total number of items of the polynomial of the humidity prediction data, and Y is the total number of items of the polynomial of the weather prediction data.
[0096] The values of the respective coefficients are determined by fitting training.
[0097] At this time, for each type of the weather prediction data, the total number of items of the polynomial corresponding to each type of the weather prediction data in the polynomial regression model is determined by the following method:
[0098] A test polynomial is established:
[0099]
[0100] ω = α, β, γ, δ;
[0101] X = R, T, P, H;
[0102] wherein X represents the type of the weather prediction data, X n is the n-th weather prediction data between the current time point and the prediction target time point, and ω represents the type of the weather prediction data. y is the y-th order coefficient of the polynomial;
[0103] The values of the respective coefficients when Y = y are determined by fitting training, and y = 1, 2, 3.
[0104] The prediction error ε f,y of the f-th group of test data when Y = y is obtained by testing the test polynomial by the f-th group of test data, and f = 1, 2, …, F.
[0105] The error evaluation parameter ε y ′ when Y = y is obtained.
[0106] ε y ′ = var(ε f , f = 1, 2, …, F);
[0107] wherein var(.) represents a function of calculating variance;
[0108] The minimum value of ε y ′ is obtained, and the value of y corresponding thereto is the total number of items of the polynomial when the type of the weather prediction data is X.
[0109] The method simultaneously introduces multiple key meteorological variables such as temperature, air pressure, rainfall and humidity, and builds a polynomial regression model based on these data to predict runoff, which can more comprehensively reflect the combined effect of meteorological factors on the evolution process of runoff, and significantly improve the accuracy and reliability of the prediction results. In the prediction process, meteorological prediction data is obtained based on the first period T1 segmentation, and the meteorological influence of each time point is gradually accumulated within the entire runoff prediction period, so that the model can capture the gradual influence of various types of meteorological prediction data on runoff over time, and improve the sensitivity to dynamic changes.
[0110] On the other hand, for each type of meteorological prediction data, the embodiment automatically selects the optimal order of the polynomial based on error evaluation, which is limited to 1-3 orders, effectively avoiding the problem of overfitting or underfitting of the model, and enhancing the generalization ability and adaptability of the model. When obtaining the optimal order, each type of meteorological prediction data is calculated and processed respectively, and the method mainly uses polynomial fitting of historical data for calculation. By obtaining the error distribution of test data, specifically the variance of the error of multiple test data, the error fluctuation is judged, and the type of polynomial with smaller error fluctuation is considered to be more consistent with the data relationship between the meteorological prediction data and the runoff prediction, so that the polynomial regression model can adaptively select the function form with the most suitable complexity when fitting various types of meteorological prediction data. Based on the variance of the error, the stability and consistency of the prediction results of different polynomial models can be more objectively reflected.
[0111] Embodiment 2
[0112] The embodiment provides a watershed runoff prediction system, which is applied to a watershed runoff prediction method of the above-mentioned embodiment, and refers to Figure 2 , comprising:
[0113] The gridding module is configured to perform gridding processing on the to-be-predicted watershed to obtain a plurality of watershed grids and obtain the center of each watershed grid.
[0114] The meteorological state modeling module is configured to establish a plurality of meteorological state models for the center of each watershed grid, and each meteorological state model is configured to describe one type of meteorological data of the center of each watershed grid.
[0115] The meteorological prediction module is configured to periodically predict a plurality of types of meteorological prediction data of the center of each watershed grid according to the plurality of meteorological state models at intervals of a first period.
[0116] a runoff prediction module configured to periodically predict runoff data of the center of the watershed grid based on a plurality of weather forecast data and initial runoff data of the center of the watershed grid at intervals of a second period, the second period being greater than the first period and being an integer multiple of the first period.
[0117] The preferred embodiments of the present application have been described above with the preferred embodiments, and are not intended to limit the present application. The present application can be variously changed and modified by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall fall within the scope of the present application.
Claims
1. A basin runoff prediction method, characterized in that: Grid the watershed to be predicted to obtain multiple watershed grids, and obtain the center of each watershed grid. Perform the following operations on each watershed grid: Establishing a plurality of meteorological state models for the center of the watershed grid, each meteorological state model is used to describe a type of meteorological data of the center of the watershed grid; Periodically predicting multiple meteorological forecast data of the center of the watershed grid based on multiple meteorological state models at intervals of the first period; The runoff data at the center of the watershed grid is periodically predicted based on multiple meteorological forecast data and initial runoff data at the center of the watershed grid at intervals of a second period, wherein the second period is greater than the first period and is an integer multiple of the first period.
2. A basin runoff prediction method according to claim 1, characterized in that: The method for performing grid processing on the watershed to be predicted is: Covering the entire watershed to be predicted by a first grid, where the first grid is a minimum square covering the entire watershed to be predicted; Dividing the first grid into a plurality of second grids, wherein the second grids are squares having a side length smaller than that of the first grid; The portion of the watershed to be predicted covered by each second grid constitutes a watershed grid.
3. A basin runoff prediction method according to claim 1, characterized in that: The method for obtaining the center of the watershed grid is: obtaining the geometric center of the watershed grid.
4. A basin runoff prediction method according to claim 1, characterized in that: The method for obtaining the weather forecast data is to perform the following operations for each type of weather data: Obtaining initial meteorological data of the center of the watershed grid, and obtaining a time difference between a time point of the initial meteorological data and a predicted time point; The meteorological state model is solved based on the initial meteorological data and the time difference to obtain meteorological forecast data at the forecast time point.
5. A basin runoff prediction method according to claim 4, characterized in that: The method for establishing the meteorological state model is to establish a fitting polynomial: Among them, Y b is the meteorological forecast data of the center of the watershed grid b, Y 0,b is the initial meteorological data of the center of the watershed grid b, t is the time difference between the time point of the initial meteorological data and the predicted time point, Δt is the fitting time step, is the coefficient of the kth term in the fitting polynomial of the center of the watershed grid b, The operator for rounding up.
6. A basin runoff prediction method according to claim 5, characterized in that: The coefficient of the kth term in the fitting polynomial of the center of the watershed grid b The method to obtain is: Divide a natural year into multiple sub-periods; Perform the following processing for each sub-time period: Acquire historical meteorological data of multiple time points in the sub-time period, and establish multiple historical meteorological data pairs, each historical meteorological data pair including historical meteorological data of any two time points and a time difference between the two time points; Substitute all the historical meteorological data into the fitting polynomial for fitting, and obtain the coefficient of the kth term in the fitting polynomial for the corresponding sub-time period. When obtaining the weather forecast data, obtain the sub-time period to which the current time belongs, and set the coefficient of the kth term in the corresponding fitting polynomial to Substitute into the meteorological state model.
7. A basin runoff prediction method according to claim 1, characterized in that: The types of meteorological forecast data include temperature forecast data, air pressure forecast data, rainfall forecast data and humidity forecast data.
8. A basin runoff prediction method according to claim 7, characterized in that: The method for predicting the runoff data is: Get the time difference t' from the current time point to the predicted target time point; Get the initial runoff data Q0 at the current time point; Obtain all the weather forecast data between the current time point and the forecast target time point; Predict the runoff data Q based on the polynomial regression model t′ : Among them, T1 represents the first cycle, R n represents the nth rainfall forecast data between the current time point and the forecast target time point, α m is the mth order coefficient of the polynomial of rainfall prediction data, M is the total number of terms of the polynomial of rainfall prediction data, T n represents the nth temperature prediction data between the current time point and the predicted target time point, β i is the i-th order coefficient of the polynomial of temperature prediction data, I is the total number of terms of the polynomial of temperature prediction data, P n represents the nth air pressure prediction data between the current time point and the predicted target time point, γ j is the j-th order coefficient of the polynomial of the air pressure prediction data, J is the total number of terms of the polynomial of the air pressure prediction data, H n represents the nth humidity forecast data between the current time point and the forecast target time point, δ l is the lth order coefficient of the polynomial of humidity prediction data, and L is the total number of terms of the polynomial of humidity prediction data; The values of each coefficient are determined through fitting training.
9. A basin runoff prediction method according to claim 7, characterized in that: For each type of weather forecast data, the total number of terms of the polynomial corresponding to each type of weather forecast data in the polynomial regression model is determined by the following method: Create the test polynomial: ω=α,β,γ,δ; X=R,T,P,H; Wherein, X represents the type of weather forecast data, X n The nth weather forecast data between the current time point and the forecast target time point, ω y is the y-th order coefficient of the polynomial; Determine the values of each coefficient when Y=y through fitting training, y=1,2,3; The test polynomial is tested by F groups of test data, and the prediction error ε of the fth group of test data when Y is y is obtained respectively. f,y ,f=1,2,…,F; Get the error judgment parameter ε when Y is y y ′: e y ′=var(ε f ,f=1,2,…,F); Among them, var(.) represents the function of finding variance; Get ε y The minimum value of ′, the corresponding value of y is the total number of terms of the polynomial when the type of weather forecast data is X.
10. A watershed runoff prediction system, applied to a watershed runoff prediction method according to any one of claims 1 to 9, characterized in that: include: The gridding module is used to perform gridding processing on the watershed to be predicted, obtain multiple watershed grids, and obtain the centers of the watershed grids respectively; A meteorological state modeling module is used to establish multiple meteorological state models for the center of the watershed grid, each meteorological state model is used to describe a type of meteorological data of the center of the watershed grid; A meteorological forecast module, configured to periodically forecast a plurality of meteorological forecast data of the center of the watershed grid according to a plurality of meteorological state models at intervals of a first period; The runoff prediction module is used to periodically predict the runoff data of the center of the watershed grid based on multiple meteorological forecast data and initial runoff data of the center of the watershed grid at intervals of a second period, wherein the second period is greater than the first period and is an integer multiple of the first period.