A distributed hydrological forecasting method based on simultaneous flow analysis and confluence

Abstract

The present invention provides a distributed hydrological forecasting method using a simultaneous modeling and converging approach, comprising the following steps: analyzing and studying the converging flow relationships of each grid in a watershed, dividing all grids into multiple levels; establishing a hydrological forecasting model; the hydrological forecasting model including an underground runoff converging model and a surface runoff converging model of a combined runoff generation model; training the hydrological forecasting model to obtain a trained hydrological forecasting model; and using the trained hydrological forecasting model to forecast outflow flow for each grid in the study watershed. The distributed hydrological forecasting method using a simultaneous modeling and converging approach proposed by the present invention is a simultaneous modeling and converging approach that can forecast outflow flow for each grid in the watershed and has the advantages of high hydrological forecasting precision and accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of hydrology and water resources, and in particular to a distributed hydrological forecasting method in a simultaneous flow generation and merging mode. Background Art

[0002] The VIC model (Variable Infiltration Capacity Model) is a semi-distributed hydrological model used to simulate hydrological processes and energy balances in the land surface. By accounting for the spatial variation in land cover and soil heterogeneity, the VIC model can simulate hydrological processes over large areas without detailed topographic data. The model comprehensively simulates hydrological processes such as infiltration, soil moisture dynamics, evaporation, and plant transpiration, enabling accurate hydrological responses. This is crucial for improving the accuracy of flood simulations.

[0003] When using the VIC model for flood simulation, the first-run flow simulation and then-merge flow method is currently used. The first-run flow simulation and then-merge flow method involves establishing hydrological routing models for the main and tributary rivers of the calculated river section. Using these models, the flood outflow process for each river section is obtained. Then, at the final outflow section, the flood outflow processes of all river sections are linearly superimposed to obtain the outflow process. This first-run flow simulation and then-merge flow method is suitable for basins where the main and tributary rivers have little mutual influence, but it has certain limitations. Furthermore, the first-run flow simulation and then-merge flow method uses a unit line for confluence, which can only obtain the flow process at the outlet section of the basin, not the flow process at the outlet of each river section. Since the flow processes at different river sections within the basin have different characteristics, the first-run flow simulation and then-merge flow method cannot meet the requirements for calculating the flow process at any river section within the basin. How to accurately calculate the flow process at any river section within the basin is a key issue that needs to be addressed. Summary of the Invention

[0004] In view of the defects of the existing technology, the present invention provides a distributed hydrological forecasting method with a simultaneous runoff and confluence mode, which can effectively solve the above problems.

[0005] The technical solution adopted in the present invention is as follows:

[0006] The present invention provides a distributed hydrological forecasting method in a simultaneous flow analysis and confluence mode, comprising the following steps:

[0007] Step S1, gridding the study basin and dividing the study basin into N grids;

[0008] Step S2: Analyze the confluence flow relationship of N grids, and divide all grids into M levels, which are represented as: first level R1, second level R2, ..., Mth level R M ;

[0009] There is no confluence relationship between the grids in the same level; the Mth level R M Each grid in the is a grid that does not have an upstream grid in the study basin and is a non-confluence grid; the M-1 level R M-1 Each grid in the first level R1 is a sink grid with at least one directly connected upstream grid, and the M-1th level R M-1 Each grid in the second level R2 has only one directly connected downstream grid, while the first level R1 has only one grid, which is the final sink grid of the other N-1 grids;

[0010] Step S3, establishing a hydrological forecast model; the hydrological forecast model includes an underground runoff convergence model and a surface runoff convergence model of a joint runoff generation model;

[0011] Step S4, determining a training period T; determining initial values ​​of various model parameters of the hydrological forecast model;

[0012] Step S5, training the hydrological forecast model to obtain the trained hydrological forecast model;

[0013] Step S5.1, follow the grid from the Mth level R M The calculation order for the first level R1 is to adopt a simultaneous calculation and merging method, and to sequentially calculate the underground runoff time series and the surface runoff time series of each grid at each moment t during the training period T through the hydrological forecast model, and to add the underground runoff time series and the surface runoff time series of each grid at each moment t during the training period T at each moment t to obtain the outflow flow time series of each grid at each moment t during the training period T;

[0014] Step S5.2, selecting a landmark grid in the study basin; evaluating whether the currently obtained hydrological forecast model meets the accuracy requirements by analyzing the outflow flow time series of the landmark grid at each time t during the training period T; if not, adjusting the model parameters of the hydrological forecast model and returning to step S5.1; if so, obtaining the trained hydrological forecast model;

[0015] Step S6: using the trained hydrological forecast model to forecast the outflow flow for each grid in the study basin.

[0016] Preferably, step S2 is specifically as follows:

[0017] Step S2.1, among the N grids in the study basin, find the grid with the largest cumulative flow and set its level to level 1 R1;

[0018] Step S2.2, set the initial value of the rank variable RANK to 1;

[0019] Step S2.3: In the study basin, locate the grid of RANK level, called grid g; traverse the eight neighborhoods of grid g that have no assigned rank. If the outflow of the traversed grid flows into grid g, set the rank of the traversed grid to RANK+1;

[0020] Step S2.4: If all N grids have been assigned ranks, the grid rank calculation ends; otherwise, set RANK = RANK + 1 and return to step S2.3.

[0021] Preferably, the underground runoff confluence model of the combined runoff generation model includes the underground runoff confluence sub-model of the non-confluence grid shown in formula (1) and the underground runoff confluence sub-model of the confluence grid shown in formula (2):

[0022]

[0023] in:

[0024] is the underground runoff of grid j at time t;

[0025] KGG is the underground runoff decay coefficient, which is a model parameter that needs to be adjusted;

[0026] is the underground runoff of grid j at time t-1, and its initial value is Set to 0;

[0027] is the underground runoff generated by grid j at time t, which is predicted by the runoff model;

[0028] is the underground runoff inflow of grid j at time t, which is determined by formula (3):

[0029]

[0030] in: represents the underground runoff generated by the upstream grid i directly connected to grid j at time t; upi represents the number of upstream grids i directly connected to grid j.

[0031] Preferably, the surface runoff confluence model includes a surface runoff confluence sub-model of a non-confluence grid as shown in formula (4) and a surface runoff confluence sub-model of a confluence grid as shown in formula (5):

[0032]

[0033] in:

[0034] L is the grid length;

[0035] K is the conversion coefficient between flow rate and flow velocity; x is the flow rate specific gravity coefficient; is the conversion index between flow rate and flow velocity; K, x, are the model parameters that need to be adjusted;

[0036] is the surface runoff of grid j at time t;

[0037] is the surface runoff of grid j at time t-1, and its initial value is Set to 0;

[0038] Δt is the time interval between two adjacent moments;

[0039] represents the surface runoff of the upstream grid i directly connected to grid j at time t; upi represents the number of upstream grids i directly connected to grid j;

[0040] Represents the surface runoff of the upstream grid i directly connected to grid j at time t-1.

[0041] Preferably, step S5 is specifically as follows:

[0042] Step S5.1: Obtain vegetation data, soil data, and meteorological data for each grid j in the study basin at each time t during the training period T, input them into the runoff generation model, and obtain the time series of underground runoff generated by each grid j at each time t during the training period T. t=1,2,...,T;

[0043] Step S5.2, determine the initial values ​​of the model parameters of the hydrological forecast model; the model parameters include: underground runoff decay coefficient KGG, conversion coefficient K between flow and velocity, flow weight coefficient x, conversion index between flow and velocity

[0044] Step S5.3, for the Mth level R M Calculate for each grid j in:

[0045] Using the underground runoff confluence submodel of the non-confluence grid shown in formula (1), the M-th level R M Perform underground runoff routing on each grid j in the grid and obtain the M-th level R M The underground runoff time series of each grid j at time t during the training period T is t=1,2,...,T;

[0046] The surface runoff confluence submodel of the non-confluence grid shown in formula (4) is used to calculate the M-th level R M Surface runoff routing is performed on each grid j in the M-th level R M The surface runoff time series of each grid j at time t during the training period T is t=1,2,...,T;

[0047] Using formula (6), for the Mth level R M The underground runoff time series of each grid j at time t during the training period T is and surface runoff time series Add at each moment t to obtain the outflow flow time series of each grid j at each moment t during the training period T t=1,2,...,T;

[0048]

[0049] Step S5.4, for the M-1th level R M-1 Calculate for each grid j in:

[0050] Using the underground runoff confluence submodel of the confluence grid shown in formula (2), the M-1 level R M-1 Perform underground runoff routing on each grid j in the grid and obtain the M-1 level R M-1 The underground runoff time series of each grid j at time t during the training period T is t=1,2,...,T;

[0051] The surface runoff confluence submodel of the confluence grid shown in formula (5) is used to calculate the M-1 level R M-1 Surface runoff routing is performed on each grid j in the grid, and the M-1 level R is obtained. M-1 The surface runoff time series of each grid j at time t during the training period T is t=1,2,...,T;

[0052] Using formula (6), for the M-1 level R M-1 The underground runoff time series of each grid j at time t during the training period T is and surface runoff time series Add at each moment t to obtain the outflow flow time series of each grid j at each moment t during the training period T t=1,2,...,T;

[0053] Step S5.5: Using the method of step S5.4, the M-2 level R M-2 Calculate each grid j in the M-3 level RM-3 The calculation is performed for each grid j in the first level R1, and so on, until the calculation of the grid j in the first level R1 is completed;

[0054] Step S5.6, select a landmark grid in the study basin; evaluate whether the current hydrological forecast model meets the accuracy requirements by analyzing the outflow flow time series of the landmark grid at each time t during the training period T; if not, adjust the model parameters of the hydrological forecast model, including the underground runoff decay coefficient KGG, the conversion coefficient K between flow and velocity, the flow weight coefficient x, and the conversion index between flow and velocity Then return to step S5.3; if satisfied, the trained hydrological forecast model is obtained.

[0055] Preferably, step S5.6 is specifically as follows:

[0056] Among the N grids in the study basin, the grid where the hydrological station is located is selected as the landmark grid, denoted as landmark grid k;

[0057] Obtain the time series of outflow flow observations at each time t of the landmark grid k during the training period T Obtain the time series of outflow flow observations of each grid in the study basin at each time t during the training period T, and take the average of the outflow flow observations of all grids at the same time t to obtain the average time series of all outflow flow observations

[0058] Through this round of calculation, the outflow flow time series of the landmark grid k at each time t during the training period T is obtained.

[0059]

[0060] Using formula (7), we can get the Nash efficiency coefficient NSE:

[0061]

[0062] The Nash efficiency coefficient NSE is used to evaluate whether the currently obtained hydrological forecast model meets the accuracy requirements.

[0063] The distributed hydrological forecasting method provided by the present invention in a simultaneous flow analysis and confluence mode has the following advantages:

[0064] The present invention proposes a distributed hydrological forecasting method with a simultaneous simulation and merging method, which has the following characteristics: the present invention adopts a simultaneous simulation and merging method to realize the forecast of the outflow flow of each grid in the basin, and has the advantages of high hydrological forecasting precision and high accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1A flow chart of a distributed hydrological forecasting method with simultaneous flow analysis and merging provided by the present invention;

[0066] Figure 2 A schematic diagram of the grid calculation sequence for the Daning River Basin obtained in an embodiment of the present invention;

[0067] Figure 3 This is the runoff process diagram of the Daning River Basin outlet during the training period obtained by the traditional first-run and then-merge method;

[0068] Figure 4 This is a flow chart of the outlet of the Daning River Basin during the training period obtained by the simultaneous simulation and merging method of the present invention;

[0069] Figure 5 The outflow flow process diagram of each grid during the training period T is obtained by using the simultaneous simulation and merging method of the present invention;

[0070] Figure 6 In order to obtain the outflow flow process diagram of the basin outlet during the verification period by using the traditional first-run and then-synthesis method;

[0071] Figure 7 The outflow flow process diagram of the basin outlet during the verification period is obtained by using the simultaneous simulation and synthesis method of the present invention. DETAILED DESCRIPTION

[0072] In order to make the technical problems, technical solutions and beneficial effects solved by the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0073] This invention proposes a distributed hydrological forecasting method using a simultaneous flow prediction and confluence approach, with the following features: This method employs a simultaneous flow prediction and confluence approach to forecast outflow for each grid in a watershed, offering the advantages of high hydrological forecast precision and accuracy. This method also effectively increases hydrological data within the watershed, providing valuable insights for water resources management. The invention boasts clear concepts, easy operation, and strong practicality.

[0074] See Figure 1 The present invention provides a distributed hydrological forecasting method in a simultaneous flow analysis and confluence mode, comprising the following steps:

[0075] Step S1, gridding the study basin and dividing the study basin into N grids;

[0076] Specifically, based on the distribution of watersheds and spatial scale requirements, the watersheds can be gridded to obtain a watershed grid network.

[0077] Step S2: Analyze the confluence flow relationship of N grids, and divide all grids into M levels, which are represented as: first level R1, second level R2, ..., Mth level R M ;

[0078] There is no confluence relationship between the grids in the same level; the Mth level R M Each grid in the is a grid that does not have an upstream grid in the study basin and is a non-confluence grid; the M-1 level R M-1 Each grid in the first level R1 is a sink grid with at least one directly connected upstream grid, and the M-1th level R M-1 Each grid in the second level R2 has only one directly connected downstream grid, while the first level R1 has only one grid, which is the final sink grid of the other N-1 grids;

[0079] As a specific implementation method, step S2 is specifically as follows:

[0080] Step S2.1, among the N grids in the study basin, find the grid with the largest cumulative flow and set its level to level 1 R1;

[0081] One way to obtain the grid with the maximum cumulative flow is to estimate the grid with the maximum cumulative flow based on the terrain characteristics of each grid in the study basin. Usually, the grid at the lowest terrain position is the grid with the maximum cumulative flow, and it is usually the outlet grid of the entire basin.

[0082] Step S2.2, set the initial value of the rank variable RANK to 1;

[0083] Step S2.3: In the study basin, locate the grid of RANK level, called grid g; traverse the eight neighborhoods of grid g that have no assigned rank. If the outflow of the traversed grid flows into grid g, set the rank of the traversed grid to RANK+1;

[0084] Step S2.4: If all N grids have been assigned ranks, the grid rank calculation ends; otherwise, set RANK = RANK + 1 and return to step S2.3.

[0085] In the present invention, the function of determining each grid level is: in the subsequent process, the calculation order of each grid is determined according to the grid level, that is, from the Mth level R M Calculation is performed on each grid in the order of the first level R1.

[0086] Step S3, establishing a hydrological forecast model; the hydrological forecast model includes an underground runoff convergence model and a surface runoff convergence model of a joint runoff generation model;

[0087] The underground runoff confluence model of the combined runoff generation model uses a linear reservoir for confluence, specifically including the underground runoff confluence sub-model of the non-confluence grid shown in formula (1) and the underground runoff confluence sub-model of the confluence grid shown in formula (2):

[0088]

[0089] in:

[0090] is the underground runoff of grid j at time t;

[0091] KGG is the underground runoff decay coefficient, which is a model parameter that needs to be adjusted;

[0092] is the underground runoff of grid j at time t-1, and its initial value is Set to 0;

[0093] is the underground runoff generated by grid j at time t, which is predicted by the runoff model;

[0094] is the underground runoff inflow of grid j at time t, which is determined by formula (3):

[0095]

[0096] in: represents the underground runoff generated by the upstream grid i directly connected to grid j at time t; upi represents the number of upstream grids i directly connected to grid j.

[0097] The surface runoff confluence model includes a surface runoff confluence sub-model of a non-confluence grid as shown in formula (4) and a surface runoff confluence sub-model of a confluence grid as shown in formula (5):

[0098]

[0099] in:

[0100] L is the grid length;

[0101] K is the conversion coefficient between flow rate and flow velocity; x is the flow rate specific gravity coefficient; is the conversion index between flow rate and flow velocity; K, x, are the model parameters that need to be adjusted;

[0102] is the surface runoff of grid j at time t;

[0103] is the surface runoff of grid j at time t-1, and its initial value is Set to 0;

[0104] Δt is the time interval between two adjacent moments;

[0105] represents the surface runoff of the upstream grid i directly connected to grid j at time t; upi represents the number of upstream grids i directly connected to grid j;

[0106] Represents the surface runoff of the upstream grid i directly connected to grid j at time t-1.

[0107] Step S4, determining a training period T; determining initial values ​​of various model parameters of the hydrological forecast model;

[0108] Step S5, training the hydrological forecast model to obtain the trained hydrological forecast model;

[0109] Step ①, follow the grid from the Mth level R M The calculation order for the first level R1 is to adopt a simultaneous calculation and merging method, and to sequentially calculate the underground runoff time series and the surface runoff time series of each grid at each moment t during the training period T through the hydrological forecast model, and to add the underground runoff time series and the surface runoff time series of each grid at each moment t during the training period T at each moment t to obtain the outflow flow time series of each grid at each moment t during the training period T;

[0110] Step 2: Select a landmark grid in the study basin; evaluate whether the currently obtained hydrological forecast model meets the accuracy requirements by analyzing the outflow flow time series of the landmark grid at each time t during the training period T; if not, adjust the model parameters of the hydrological forecast model and return to step 1; if it does, obtain the trained hydrological forecast model;

[0111] Step S5 is specifically as follows:

[0112] Step S5.1: Obtain vegetation data, soil data, and meteorological data for each grid j in the study basin at each time t during the training period T, input them into the runoff generation model, and obtain the time series of underground runoff generated by each grid j at each time t during the training period T. t=1,2,...,T;

[0113] In this step, the runoff model may adopt the VIC runoff model.

[0114] Step S5.2, determine the initial values ​​of the model parameters of the hydrological forecast model; the model parameters include: underground runoff decay coefficient KGG, conversion coefficient K between flow and velocity, flow weight coefficient x, conversion index between flow and velocity

[0115] Step S5.3, for the Mth level R M Calculate for each grid j in:

[0116] Using the underground runoff confluence submodel of the non-confluence grid shown in formula (1), the M-th level R M Perform underground runoff routing on each grid j in the grid and obtain the M-th level R M The underground runoff time series of each grid j at time t during the training period T is t=1,2,...,T;

[0117] The surface runoff confluence submodel of the non-confluence grid shown in formula (4) is used to calculate the M-th level R M Surface runoff routing is performed on each grid j in the M-th level R M The surface runoff time series of each grid j at time t during the training period T is t=1,2,...,T;

[0118] Using formula (6), for the Mth level R M The underground runoff time series of each grid j at time t during the training period T is and surface runoff time series Add at each moment t to obtain the outflow flow time series of each grid j at each moment t during the training period T t=1,2,...,T;

[0119]

[0120] Step S5.4, for the M-1th level R M-1 Calculate for each grid j in:

[0121] Using the underground runoff confluence submodel of the confluence grid shown in formula (2), the M-1 level R M-1 Perform underground runoff routing on each grid j in the grid and obtain the M-1 level R M-1 The underground runoff time series of each grid j at time t during the training period T is t=1,2,...,T;

[0122] The surface runoff confluence submodel of the confluence grid shown in formula (5) is used to calculate the M-1 level R M-1Surface runoff routing is performed on each grid j in the grid, and the M-1 level R is obtained. M-1 The surface runoff time series of each grid j at time t during the training period T is t=1,2,...,T;

[0123] Using formula (6), for the M-1 level R M-1 The underground runoff time series of each grid j at time t during the training period T is and surface runoff time series Add at each moment t to obtain the outflow flow time series of each grid j at each moment t during the training period T t=1,2,...,T;

[0124] Step S5.5: Using the method of step S5.4, the M-2 level R M-2 Calculate each grid j in the M-3 level R M-3 The calculation is performed for each grid j in the first level R1, and so on, until the calculation of the grid j in the first level R1 is completed;

[0125] Step S5.6, select a landmark grid in the study basin; evaluate whether the current hydrological forecast model meets the accuracy requirements by analyzing the outflow flow time series of the landmark grid at each time t during the training period T; if not, adjust the model parameters of the hydrological forecast model, including the underground runoff decay coefficient KGG, the conversion coefficient K between flow and velocity, the flow weight coefficient x, and the conversion index between flow and velocity Then return to step S5.3; if satisfied, the trained hydrological forecast model is obtained.

[0126] Step S5.6 is specifically as follows:

[0127] Among the N grids in the study basin, the grid where the hydrological station is located is selected as the landmark grid, denoted as landmark grid k;

[0128] Obtain the time series of outflow flow observations at each time t of the landmark grid k during the training period T Obtain the time series of outflow flow observations of each grid in the study basin at each time t during the training period T, and take the average of the outflow flow observations of all grids at the same time t to obtain the average time series of all outflow flow observations

[0129] Through this round of calculation, the outflow flow time series of the landmark grid k at each time t during the training period T is obtained.

[0130]

[0131] Using formula (7), we can get the Nash efficiency coefficient NSE:

[0132]

[0133] The Nash efficiency coefficient NSE is used to evaluate whether the hydrological forecast model currently obtained meets the accuracy requirements. The larger the value of the Nash efficiency coefficient NSE, the higher the accuracy of the hydrological forecast model.

[0134] Step S6: using the trained hydrological forecast model to forecast the outflow flow for each grid in the study basin.

[0135] Specifically, by training the hydrological forecast model, the trained hydrological forecast model is obtained, and its model parameters include underground runoff decay coefficient KGG, conversion coefficient K between flow and flow velocity, flow weight coefficient x, conversion index between flow and flow velocity are all determined and no longer change; then, the hydrological forecast model that has been trained is used to forecast the outflow flow for each grid in the study basin.

[0136] Specifically, when conducting hydrological forecasts for the study basin, vegetation, soil, and meteorological data are first obtained for each grid during the forecast period. These data are then input into the VIC runoff model to obtain the time series of groundwater runoff generated by each grid j at each time t during the forecast period. Next, the groundwater runoff confluence model and surface runoff confluence model of the trained joint runoff model are used to perform simultaneous runoff analysis for each grid, following the algorithmic sequence defined above. This yields the outflow discharge for each grid at each time t during the forecast period.

[0137] The following takes the Daning River Basin as an example to introduce an embodiment:

[0138] Step 1: Take the Daning River Basin as the study basin and divide the study basin into N grids;

[0139] Step 2: Based on ArcGIS, the DEM data of the watershed is processed and all grids are divided into several levels according to the level division method mentioned above. Then, the order of flow calculation is determined according to the level of each grid. Figure 2 The figure shows a schematic diagram of the grid calculation sequence in the Daning River Basin.

[0140] Step 3: Determine the training period T as 2014-5-1 to 2014-10-31, and use the hydrological forecast model and the training method of the hydrological forecast model described above to obtain a trained hydrological forecast model that meets the accuracy requirements.

[0141] The hydrological forecast model trained by the present invention can be obtained as follows: Figure 4As shown in Figure 1, the runoff process of the basin outlet during the training period T is shown in Figure 2. As a comparison, the traditional first-run and then-merge method is used to obtain the following Figure 3 The runoff process at the Daning River basin outlet is shown from May 1, 2014 to October 31, 2014.

[0142] Furthermore, the hydrological forecast model trained by the present invention can also obtain the outflow flow process of each grid during the training period T. For example, taking September 2, 2014 as an example, the outflow flow process of each grid during the training period T can be obtained. Figure 5 The outflow flow process of each grid in the basin is shown.

[0143] Step 4: Select a suitable verification period to verify the trained hydrological forecast model of the present invention.

[0144] Since the training period is the flood season of the year, the validation period is also selected during the flood season (May-October).

[0145] This example selects May 2016 to October 2016 as the verification period. Figure 6 As shown in the figure, the outflow process diagram of the basin outlet during the verification period is obtained by using the traditional first-run and then-combination method. Figure 7 As shown, the outflow flow process diagram of the basin outlet during the verification period is obtained by using the simultaneous simulation and combination method of the present invention.

[0146] Furthermore, the Nash efficiency coefficients of the hydrological forecast model trained by the present invention and the traditional first-run and then-merge method are calculated respectively.

[0147] The Nash efficiency coefficient of the present invention's simultaneous performance and merging method is 0.926 during the training period and 0.805 during the verification period.

[0148] The traditional first-play-then-merge confluence method is adopted, with a Nash efficiency coefficient of 0.904 during the training period and a Nash efficiency coefficient of 0.739 during the validation period.

[0149] It can be seen from the Nash efficiency coefficient results that the simultaneous simulation and merging method of the present invention has a higher Nash efficiency coefficient, better simulation results and can obtain the flow process of each grid. Therefore, the simultaneous simulation and merging method of the present invention is suitable for the Daning River Basin.

[0150] The present invention proposes a distributed hydrological forecasting method with a simultaneous flow analysis and confluence method, which has the following characteristics:

[0151] (1) In the present invention, only one hydrological calculation model is established for all grids in the watershed, i.e., the hydrological forecast model in the hydrological forecast stage after training is completed. By using this hydrological calculation model, hydrological calculation in the training stage and hydrological forecast in the forecast stage can be quickly implemented. Since the present invention does not require the establishment of a hydrological calculation model for each river section and tributary, the present invention can simplify the system complexity and improve the efficiency of hydrological calculation in the training stage and hydrological forecast in the hydrological forecast stage.

[0152] (2) In the present invention, the calculation order is determined according to the level of each grid in the basin; according to the calculation order, underground runoff calculation and surface runoff calculation are performed on each grid in turn, and then the underground runoff calculation results and surface runoff calculation results of each grid are superimposed, so that the outflow flow of each grid can be obtained; and when each grid is calculated according to the calculation order, the underground runoff calculation results and surface runoff calculation results of the upstream grid are used as inputs of the hydrological calculation model of the downstream grid directly connected to it, and then the underground runoff calculation and surface runoff calculation are performed on the downstream grid. Therefore, the present invention realizes the simultaneous calculation and merging of the flow of each grid in the basin, which can improve the convergence calculation accuracy of the outflow flow of each grid.

[0153] (3) In the present invention, during the hydrological forecasting stage, the outflow flow of each grid in the basin can be forecasted, thereby improving the accuracy of the hydrological forecast.

[0154] (4) The hydrological forecast model of the present invention comprehensively considers underground runoff and surface runoff affected by the runoff generation model, thereby improving the accuracy of the outflow flow of each grid in the basin.

[0155] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A distributed hydrological forecasting method with simultaneous prediction and convergence, characterized in that: The following steps are involved: Step S1, gridding the study basin and dividing the study basin into N grids; Step S2: Analyze the confluence flow relationship of N grids, and divide all grids into M levels, which are represented as: first level R1, second level R2, ..., Mth level R M ; There is no confluence relationship between the grids in the same level; the Mth level R M Each grid in the is a grid that does not have an upstream grid in the study basin and is a non-confluence grid; the M-1 level R M-1 Each grid in the first level R1 is a sink grid with at least one directly connected upstream grid, and the M-1th level R M-1 Each grid in the second level R2 has only one directly connected downstream grid, while the first level R1 has only one grid, which is the final sink grid of the other N-1 grids; Step S3, establishing a hydrological forecast model; the hydrological forecast model includes an underground runoff convergence model and a surface runoff convergence model of a joint runoff generation model; The underground runoff confluence model of the combined runoff generation model includes the underground runoff confluence sub-model of the non-confluence grid shown in formula (1) and the underground runoff confluence sub-model of the confluence grid shown in formula (2): in: is the underground runoff of grid j at time t; KGG is the underground runoff decay coefficient, which is a model parameter that needs to be adjusted; is the underground runoff of grid j at time t-1, and its initial value is Set to 0; is the underground runoff generated by grid j at time t, which is predicted by the runoff model; is the underground runoff inflow of grid j at time t, which is determined by formula (3): in: represents the underground runoff generated by the upstream grid i directly connected to grid j at time t; upi represents the number of upstream grids i directly connected to grid j; The surface runoff confluence model includes a surface runoff confluence sub-model of a non-confluence grid as shown in formula (4) and a surface runoff confluence sub-model of a confluence grid as shown in formula (5): in: L is the grid length; K is the conversion coefficient between flow rate and flow velocity; x is the flow rate specific gravity coefficient; is the conversion index between flow rate and flow velocity; K, x, are the model parameters that need to be adjusted; is the surface runoff of grid j at time t; is the surface runoff of grid j at time t-1, and its initial value is Set to 0; Δt is the time interval between two adjacent moments; represents the surface runoff of the upstream grid i directly connected to grid j at time t; upi represents the number of upstream grids i directly connected to grid j; represents the surface runoff of the upstream grid i directly connected to grid j at time t-1; Step S4, determining a training period T; determining initial values ​​of various model parameters of the hydrological forecast model; Step S5, training the hydrological forecast model to obtain the trained hydrological forecast model; Step S5.1, follow the grid from the Mth level R M The calculation order for the first level R1 is to adopt a simultaneous calculation and merging method, and to sequentially calculate the underground runoff time series and the surface runoff time series of each grid at each moment t during the training period T through the hydrological forecast model, and to add the underground runoff time series and the surface runoff time series of each grid at each moment t during the training period T at each moment t to obtain the outflow flow time series of each grid at each moment t during the training period T; Step S5.2, selecting a landmark grid in the study basin; evaluating whether the currently obtained hydrological forecast model meets the accuracy requirements by analyzing the outflow flow time series of the landmark grid at each time t during the training period T; if not, adjusting the model parameters of the hydrological forecast model and returning to step S5.1; if so, obtaining the trained hydrological forecast model; Step S6: using the trained hydrological forecast model to forecast the outflow flow for each grid in the study basin.

2. A distributed hydrological forecasting method with simultaneous simulation and merging according to claim 1, characterized in that: Step S2 is specifically as follows: Step S2.1, among the N grids in the study basin, find the grid with the largest cumulative flow and set its level to level 1 R1; Step S2.2, set the initial value of the rank variable RANK to 1; Step S2.3: In the study basin, locate the grid of RANK level, called grid g; traverse the eight neighborhoods of grid g that have no assigned rank. If the outflow of the traversed grid flows into grid g, set the rank of the traversed grid to RANK+1; Step S2.4: If all N grids have been assigned ranks, the grid rank calculation ends; otherwise, set RANK = RANK + 1 and return to step S2.

3.

3. The distributed hydrological forecasting method of claim 1, wherein: Step S5 is specifically as follows: Step S5.1: Obtain vegetation data, soil data, and meteorological data for each grid j in the study basin at each time t during the training period T, input them into the runoff generation model, and obtain the time series of underground runoff generated by each grid j at each time t during the training period T. Step S5.2, determine the initial values ​​of the model parameters of the hydrological forecast model; the model parameters include: underground runoff decay coefficient KGG, conversion coefficient K between flow and velocity, flow weight coefficient x, conversion index between flow and velocity Step S5.3, for the Mth level R M Calculate for each grid j in: Using the underground runoff confluence submodel of the non-confluence grid shown in formula (1), the M-th level R M Perform underground runoff routing on each grid j in the grid and obtain the M-th level R M The underground runoff time series of each grid j at time t during the training period T is The surface runoff confluence submodel of the non-confluence grid shown in formula (4) is used to calculate the M-th level R M Surface runoff routing is performed on each grid j in the M-th level R M The surface runoff time series of each grid j at time t during the training period T is Using formula (6), for the Mth level R M The underground runoff time series of each grid j at time t during the training period T is and surface runoff time series Add at each moment t to obtain the outflow flow time series of each grid j at each moment t during the training period T Step S5.4, for the M-1th level R M-1 Calculate for each grid j in: Using the underground runoff confluence submodel of the confluence grid shown in formula (2), the M-1 level R M-1 Perform underground runoff routing on each grid j in the grid and obtain the M-1 level R M-1 The underground runoff time series of each grid j at time t during the training period T is The surface runoff confluence submodel of the confluence grid shown in formula (5) is used to calculate the M-1 level R M-1 Surface runoff routing is performed on each grid j in the grid, and the M-1 level R is obtained. M-1 The surface runoff time series of each grid j at time t during the training period T is Using formula (6), for the M-1 level R M-1 The underground runoff time series of each grid j at time t during the training period T is and surface runoff time series Add at each moment t to obtain the outflow flow time series of each grid j at each moment t during the training period T Step S5.5: Using the method of step S5.4, the M-2 level R M-2 Calculate each grid j in the M-3 level R M-3 The calculation is performed for each grid j in the first level R1, and so on, until the calculation of the grid j in the first level R1 is completed; Step S5.6, select a landmark grid in the study basin; evaluate whether the current hydrological forecast model meets the accuracy requirements by analyzing the outflow flow time series of the landmark grid at each time t during the training period T; if not, adjust the model parameters of the hydrological forecast model, including the underground runoff decay coefficient KGG, the conversion coefficient K between flow and velocity, the flow weight coefficient x, and the conversion index between flow and velocity Then return to step S5.3; if satisfied, the trained hydrological forecast model is obtained.

4. The distributed hydrological forecasting method of claim 3, wherein: Step S5.6 is specifically as follows: Among the N grids in the study basin, the grid where the hydrological station is located is selected as the landmark grid, denoted as landmark grid k; Obtain the time series of outflow flow observations at each time t of the landmark grid k during the training period T Obtain the time series of outflow flow observations of each grid in the study basin at each time t during the training period T, and take the average of the outflow flow observations of all grids at the same time t to obtain the average time series of all outflow flow observations Through this round of calculation, the outflow flow time series of the landmark grid k at each time t during the training period T is obtained. Using formula (7), we can get the Nash efficiency coefficient NSE: The Nash efficiency coefficient NSE is used to evaluate whether the currently obtained hydrological forecast model meets the accuracy requirements.

Images