Parallel computing method for flood forecast of ultra-large river network

Abstract

The invention discloses a parallel computing method for flood forecasting of an ultra-large river network, which comprises the following steps of: dividing sub-region boundaries in a research region according to gradient abrupt change points of a reference terrain and confluence points of tributaries, and ensuring that complete water flow confluence ports exist in the sub-regions; constructing a hydrodynamic continuous equation and a hydrodynamic momentum equation of each sub-region, discretizing the hydrodynamic continuous equation and the hydrodynamic momentum equation, and calculating the water depth h and the flow velocity u of each space node in the sub-region; building a parallel computing framework, and inputting the discretized hydrodynamic continuous equation and hydrodynamic momentum equation into each computing node of the parallel computing framework; after each calculation node calculates the water depth and the flow velocity of the space node, the water depth and the flow velocity are accumulated, and flood forecast data of the river channel main stream in the research area are calculated. According to the method, the ultra-large river network is divided into a plurality of sub-regions and parallel calculation is carried out, so that the calculation duration of flood forecasting is greatly shortened, the urgent demand of real-time flood forecasting is effectively met, and more precious time is won for flood control and disaster reduction decisions.

Description

Technical Field

[0001] The invention relates to the field of flood forecasting, and in particular to a parallel computing method for flood forecasting of a super-large river network. Background Art

[0002] Flood forecasting for super-large river networks is crucial for flood prevention and disaster reduction. However, due to the large scale of river networks, complex water flow movement, and huge data volume, traditional serial calculation methods face problems such as low computational efficiency and long time consumption when processing flood forecasting for super-large river networks, and it is difficult to meet the needs of real-time flood forecasting and decision support. Therefore, it is necessary to propose an efficient parallel calculation method for flood forecasting for super-large river networks. Summary of the invention

[0003] In view of the above-mentioned deficiencies in the prior art, the present invention provides a parallel computing method for flood forecasting in a super-large river network, which significantly improves the calculation rate of flood forecasting and ensures the accuracy of forecast results.

[0004] In order to achieve the above-mentioned object of the invention, the technical solution adopted by the present invention is:

[0005] A parallel computing method for flood forecasting of a super-large river network is provided, which comprises the following steps:

[0006] S1: Determine the study area for flood forecasting based on the distribution of the river network, and divide the sub-area boundaries within the study area based on the sudden change points of terrain slope and the confluence points of tributaries to ensure that there are complete water confluence outlets in the sub-areas;

[0007] S2: construct the hydrodynamic continuity equation and hydrodynamic momentum equation of each sub-region, and discretize them. Use the discretized hydrodynamic continuity equation and hydrodynamic momentum equation to calculate the water depth h and flow velocity u of each spatial node in the sub-region;

[0008] S3: Build a parallel computing framework, input the discretized hydrodynamic continuity equation and hydrodynamic momentum equation into each computing node of the parallel computing framework; calculate the computational complexity index in each sub-area, allocate computing nodes according to the computational complexity index, balance the data computational complexity of each computing node, and input the water depth and flow velocity data collected by the spatial node into the computing node;

[0009] S4: After each computing node calculates the water depth and flow velocity of the spatial node, the maximum value of the water depth and flow velocity of each computing node is screened out Based on the number of sub-areas N in the study area, the water depth and flow velocity are accumulated to calculate the flood forecast data of the mainstream river in the study area;

[0010]

[0011] Among them, H is the water depth of the main river channel for flood forecasting, and U is the flow velocity of the main river channel for flood forecasting.

[0012] Furthermore, step S1 includes:

[0013] S11: Determine the research area for flood forecasting according to the distribution of the river network, obtain the confluence points of all tributaries within the research area, and calculate the water flow convergence coefficient C of each confluence point p ;

[0014]

[0015] Among them, p is the number of the confluence point, j is the river channel number related to the confluence point, J is the number of river channels related to the confluence point, Q j is the flow rate of river channel j, L j is the length of river channel j, and D is the distance from the confluence point to the river channel outlet;

[0016] S12: Set the threshold C of the water flow convergence coefficient 阈值 , and calculate the significance coefficient X of the water flow convergence coefficient p ;

[0017]

[0018] S13: Compare the significance coefficient X p with the significance coefficient threshold X0; if X p ≥X0, then determine that the confluence point p is used as the boundary division point; if X p <X0, then determine that the confluence point p cannot be used as the boundary division point, and the confluence point p;

[0019] S14: Mark all the terrain slope mutation points within the research area, and take the midpoint between two adjacent boundary division points along the main river channel direction as the boundary starting point of the sub-region; draw a perpendicular line to the main river channel between two adjacent boundary division points, and take the terrain slope mutation points between two adjacent perpendicular lines as the boundary end point of the sub-region;

[0020] S15: Connect the corresponding boundary starting points and boundary end points along both sides of the main river channel to form the boundary line of the sub-region, and the area between two adjacent boundary lines is used as the sub-region, and the boundary division points within the sub-region are used as the water flow outlets of the sub-region.

[0021] Furthermore, step S2 includes:

[0022] S21: Construct the hydrodynamic continuity equation for each sub-region;

[0023]

[0024] Among them, h is the water depth of the tributary in the sub-area, t is the time of monitoring the water depth of the tributary, u is the flow velocity at the water depth h, x is the distance between the water depth monitoring point of the tributary and the water flow confluence outlet, and q is the inflow of the tributary;

[0025] S22: construct the hydrodynamic momentum equation for each sub-region;

[0026]

[0027] Among them, g is the acceleration of gravity, C is the Xie Cai coefficient, and R is the hydraulic radius;

[0028] S23: discretize the hydrodynamic continuity equation and set the spatial step Δx and the time step Δt;

[0029]

[0030] Among them, i is the spatial node number of the tributary in the sub-area, w is the time step number of the monitored water depth, is the water depth monitored at the i-th spatial node and the w-th time step, is the flow velocity monitored at the ith spatial node and the ith spatial node time step;

[0031] S24: Discretize the hydrodynamic momentum equation;

[0032]

[0033] S25: Calculate Xiecai coefficient C; n is the roughness coefficient of water flow;

[0034] S26: Substitute the Xie Cai coefficient C into the discretized hydrodynamic continuity equation and hydrodynamic momentum equation to calculate the water depth h and flow velocity u of each spatial node in the sub-area;

[0035] S27: correcting the boundary data between adjacent sub-regions;

[0036]

[0037] Among them, i′ and i are spatial nodes on the boundary line of two adjacent sub-regions, is the corrected water depth and flow velocity of spatial node i, α and β are the boundary correction coefficients of water depth and flow velocity respectively.

[0038] Further, step S3 includes:

[0039] S31: building a parallel computing framework, the parallel computing framework includes n computing nodes, the number of computing nodes is equal to the number of sub-regions N, and the discretized hydrodynamic continuity equation and hydrodynamic momentum equation are input into each computing node of the parallel computing framework;

[0040] S32: Calculate the computational complexity index F in the sub-region according to the number of spatial nodes I set in the sub-region and the number of time steps W set in a data monitoring cycle. n ;

[0041]

[0042] Among them, γ1 and γ2 are the influence weights of water depth and flow velocity calculation on the computational complexity, γ1+γ2=1;

[0043] S33: Calculate the total computational complexity index in the study area And calculate the ideal computational complexity index for each computing node

[0044] S34: Calculate the computational complexity index of each sub-region in the study area, and sort the computational complexity indexes from the largest to the smallest, to obtain a computational complexity index sequence (F1, F2, ..., F N );

[0045] S35: compare the size of F1 and F0;

[0046] If F1≤F0, the water depth and flow velocity data in each sub-area are directly input into the corresponding calculation node, and step S4 is executed;

[0047] If F1>F0, execute step S36;

[0048] S36: According to the computational complexity index sequence (F1, F2, ..., F N ) traverses the first computational complexity index F1 and the Nth computational complexity index F N , assign some of the spatial nodes in the sub-region corresponding to the first computational complexity index F1 to the computational complexity index F N In the corresponding computing node, return to step S34; assign the computational complexity index F N The spatial node in the corresponding calculation node is the spatial node where the current water depth or flow velocity is the minimum value in the sub-area;

[0049] S37: until all computational complexity indicators in the computational complexity indicator sequence are less than or equal to F0, then the water depth and flow velocity data in each sub-area are input into the corresponding computing node, and step S4 is executed.

[0050] The beneficial effects of the present invention are as follows: by dividing the super-large river network into multiple sub-areas and implementing parallel calculations, the present invention greatly reduces the calculation time of flood forecasting, effectively meets the urgent needs of real-time flood forecasting, and wins more valuable time for flood prevention and disaster reduction decision-making. It also uses boundary data correction processing and synchronization mechanisms to ensure the continuity and coordination of water flow movement between sub-areas, thereby effectively ensuring the accuracy of flood forecast results for the entire super-large river network. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 Flowchart of the parallel computing method for flood forecasting in very large river networks. DETAILED DESCRIPTION

[0052] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0053] like Figure 1 As shown, a parallel computing method for flood forecasting of a super-large river network includes the following steps:

[0054] S1: Determine the study area for flood forecasting based on the distribution of the river network, and divide the sub-area boundaries within the study area based on the sudden change points of terrain slope and the confluence points of tributaries to ensure that there are complete water confluence outlets in the sub-areas;

[0055] Step S1 includes:

[0056] S11: Determine the study area for flood forecasting based on the distribution of the river network, obtain the confluence points of all tributaries in the study area, and calculate the water flow convergence coefficient C at each confluence point p ;

[0057]

[0058] Where p is the number of the confluence point, j is the number of the river channel associated with the confluence point, J is the number of the river channels associated with the confluence point, and Q j is the flow rate of river j, L j is the length of river channel j, D is the distance from the confluence point to the river outlet;

[0059] S12: Set the threshold C of the water flow convergence coefficient 阈值 , calculate the significance coefficient X of the water flow convergence coefficient p ;

[0060]

[0061] S13: Compare the significance coefficient X p with the significance coefficient threshold X0; if X p ≥X0, then determine that the confluence point p is used as the boundary division point; if X p <X0, then determine that the confluence point p cannot be used as the boundary division point, and the confluence point p;

[0062] S14: Mark all the terrain slope mutation points in the research area. Take the midpoint between two adjacent boundary division points along the main stream direction of the river as the boundary starting point of the sub-region; draw a perpendicular line to the main stream of the river between two adjacent boundary division points, and take the terrain slope mutation points between two adjacent perpendicular lines as the boundary end point of the sub-region;

[0063] S15: Connect the corresponding boundary starting points and boundary end points along both sides of the main stream of the river to form the boundary line of the sub-region. The area between two adjacent boundary lines is used as the sub-region, and the boundary division point in the sub-region is used as the water flow outlet of the sub-region.

[0064] S2: Construct the hydrodynamic continuity equation and hydrodynamic momentum equation for each sub-region, and perform discretization. Use the discretized hydrodynamic continuity equation and hydrodynamic momentum equation to calculate the water depth h and flow velocity u of each spatial node in the sub-region.

[0065] Step S2 specifically includes:

[0066] S21: Construct the hydrodynamic continuity equation for each sub-region;

[0067]

[0068] where h is the water depth of the tributary in the sub-region, t is the time for monitoring the water depth of the tributary, u is the flow velocity at the water depth h, x is the position of the water depth monitoring point of the tributary from the water flow outlet, and q is the inflow of the tributary;

[0069] S22: Construct the hydrodynamic momentum equation for each sub-region;

[0070]

[0071] where g is the acceleration due to gravity, C is the Chezy coefficient, and R is the hydraulic radius;

[0072] S23: Discretize the hydrodynamic continuity equation, and set the spatial step size Δx and time step size Δt;

[0073]

[0074] where i is the spatial node number on the tributary in the sub-region, and w is the time step number for monitoring the water depth, is the water depth monitored at the i-th spatial node and the w-th time step, is the flow velocity monitored at the ith spatial node and the ith spatial node time step;

[0075] S24: Discretize the hydrodynamic momentum equation;

[0076]

[0077] S25: Calculate Xiecai coefficient C; n is the roughness coefficient of water flow;

[0078] S26: Substitute the Xie Cai coefficient C into the discretized hydrodynamic continuity equation and hydrodynamic momentum equation to calculate the water depth h and flow velocity u of each spatial node in the sub-area;

[0079] S27: During the calculation process, there is water flow interaction between sub-regions, that is, the transmission of boundary data is crucial. Correct the boundary data between adjacent sub-regions;

[0080]

[0081] Among them, i′ and i are spatial nodes on the boundary line of two adjacent sub-regions, is the corrected water depth and flow velocity of spatial node i, α and β are the boundary correction coefficients of water depth and flow velocity respectively.

[0082] S3: Build a parallel computing framework, input the discretized hydrodynamic continuity equation and hydrodynamic momentum equation into each computing node of the parallel computing framework; calculate the computational complexity index in each sub-region, and allocate computing nodes according to the computational complexity index to balance the data computational complexity of each computing node. Reasonable allocation based on computational complexity avoids simple average allocation strategy.

[0083] Step S3 specifically includes:

[0084] S31: building a parallel computing framework, the parallel computing framework includes n computing nodes, the number of computing nodes is equal to the number of sub-regions N, and the discretized hydrodynamic continuity equation and hydrodynamic momentum equation are input into each computing node of the parallel computing framework;

[0085] S32: Calculate the computational complexity index F in the sub-region according to the number of spatial nodes I set in the sub-region and the number of time steps W set in a data monitoring cycle. n ;

[0086]

[0087] Among them, γ1 and γ2 are the influence weights of water depth and flow velocity calculation on the computational complexity, γ1+γ2=1, and γ1=0.5 and γ2=0.5 are generally taken;

[0088] S33: Calculate the total computational complexity index in the study area And calculate the ideal computational complexity index for each computing node

[0089] S34: Calculate the computational complexity index of each sub-region in the study area, and sort the computational complexity indexes from the largest to the smallest, to obtain a computational complexity index sequence (F1, F2, ..., F N );

[0090] S35: compare the size of F1 and F0;

[0091] If F1≤F0, the water depth and flow velocity data in each sub-area are directly input into the corresponding calculation node, and step S4 is executed;

[0092] If F1>F0, execute step S36;

[0093] S36: According to the computational complexity index sequence (F1, F2, ..., F N ) traverses the first computational complexity index F1 and the Nth computational complexity index F N , assign some of the spatial nodes in the sub-region corresponding to the first computational complexity index F1 to the computational complexity index F N In the corresponding computing node, return to step S34 and assign the computational complexity index F N The spatial node in the corresponding calculation node is the spatial node where the current water depth or flow velocity is the minimum value in the sub-area;

[0094] S37: until all computational complexity indicators in the computational complexity indicator sequence are less than or equal to F0, then the water depth and flow velocity data in each sub-area are input into the corresponding computing node, and step S4 is executed.

[0095] S4: After each computing node calculates the water depth and flow velocity of the spatial node, the maximum value of the water depth and flow velocity of each computing node is screened out Based on the number of sub-areas N in the study area, the water depth and flow velocity are accumulated to calculate the flood forecast data of the mainstream river in the study area;

[0096]

[0097] Among them, H is the water depth of the mainstream of the river in the flood forecast, and U is the flow velocity of the mainstream of the river in the flood forecast.

[0098] The present invention significantly reduces the calculation time of flood forecasting by dividing the super-large river network into multiple sub-regions and implementing parallel computing, effectively meeting the urgent needs of real-time flood forecasting and gaining more valuable time for flood prevention and disaster reduction decision-making. It also uses boundary data correction processing and synchronization mechanisms to ensure the continuity and coordination of water flow movement between sub-regions, thereby effectively ensuring the accuracy of flood forecasting results for the entire super-large river network.

Claims

1. A parallel computing method for flood forecasting in a super-large river network, characterized in that: The following steps are involved: S1: Determine the study area for flood forecasting based on the distribution of the river network, and divide the sub-area boundaries within the study area based on the sudden change points of terrain slope and the confluence points of tributaries to ensure that there are complete water confluence outlets in the sub-areas; S2: construct the hydrodynamic continuity equation and hydrodynamic momentum equation of each sub-region, and discretize them. Use the discretized hydrodynamic continuity equation and hydrodynamic momentum equation to calculate the water depth h and flow velocity u of each spatial node in the sub-region; S3: Build a parallel computing framework, input the discretized hydrodynamic continuity equation and hydrodynamic momentum equation into each computing node of the parallel computing framework; calculate the computational complexity index in each sub-area, allocate computing nodes according to the computational complexity index, balance the data computational complexity of each computing node, and input the water depth and flow velocity data collected by the spatial node into the computing node; S4: After each computing node calculates the water depth and flow velocity of the spatial node, the maximum value of the water depth and flow velocity of each computing node is screened out Based on the number of sub-areas N in the study area, the water depth and flow velocity are accumulated to calculate the flood forecast data of the mainstream river in the study area; Among them, H is the water depth of the mainstream of the river in the flood forecast, and U is the flow velocity of the mainstream of the river in the flood forecast.

2. The parallel computing method for flood forecasting of a super-large river network according to claim 1 is characterized in that: The step S1 comprises: S11: Determine the study area for flood forecasting based on the distribution of the river network, obtain the confluence points of all tributaries in the study area, and calculate the water flow convergence coefficient C at each confluence point p ; Where p is the number of the confluence point, j is the number of the river channel associated with the confluence point, J is the number of the river channels associated with the confluence point, and Q j is the flow rate of river j, L j is the length of river channel j, D is the distance from the confluence point to the river outlet; S12: Set the threshold C of the water flow convergence coefficient 阈值 , calculate the significance coefficient X of the water flow convergence coefficient p ; S13: Compare the significance coefficient X p with the significance coefficient threshold X0; if X p ≥ X0, then determine that the confluence point p is used as the boundary division point; if X p < X0, then determine that the confluence point p cannot be used as the boundary division point, and the confluence point p; S14: Mark all the sudden changes in terrain slope in the study area, and take the midpoint between two adjacent boundary dividing points along the river mainstream as the boundary starting point of the sub-area; draw a vertical line along the river mainstream along two adjacent boundary dividing points, and take the sudden changes in terrain slope between the two adjacent vertical lines as the boundary end point of the sub-area; S15: Connect the corresponding boundary starting points and boundary end points along the two sides of the river mainstream to form the boundary line of the sub-region, the area between two adjacent boundary lines is used as the sub-region, and the boundary dividing point within the sub-region is used as the water flow confluence outlet of the sub-region.

3. The parallel computing method for flood forecasting of a super-large river network according to claim 2 is characterized in that: The step S2 comprises: S21: construct the hydrodynamic continuity equation for each sub-region; Among them, h is the water depth of the tributary in the sub-area, t is the time of monitoring the water depth of the tributary, u is the flow velocity at the water depth h, x is the distance between the water depth monitoring point of the tributary and the water flow confluence outlet, and q is the inflow of the tributary; S22: construct the hydrodynamic momentum equation for each sub-region; Among them, g is the acceleration of gravity, C is the Xie Cai coefficient, and R is the hydraulic radius; S23: discretize the hydrodynamic continuity equation and set the spatial step Δx and the time step Δt; Among them, i is the spatial node number of the tributary in the sub-area, w is the time step number of the monitored water depth, is the water depth monitored at the i-th spatial node and the w-th time step, is the flow velocity monitored at the ith spatial node and the ith spatial node time step; S24: Discretize the hydrodynamic momentum equation; S25: Calculate Xiecai coefficient C; n is the roughness coefficient of water flow; S26: Substitute the Xie Cai coefficient C into the discretized hydrodynamic continuity equation and hydrodynamic momentum equation to calculate the water depth h and flow velocity u of each spatial node in the sub-area; S27: correcting the boundary data between adjacent sub-regions; Among them, i′ and i are spatial nodes on the boundary line of two adjacent sub-regions, h i ' w 、u i ' w is the corrected water depth and flow velocity of spatial node i, α and β are the boundary correction coefficients of water depth and flow velocity respectively.

4. The parallel computing method for flood forecasting of a super-large river network according to claim 3 is characterized in that: The step S3 comprises: S31: building a parallel computing framework, the parallel computing framework includes n computing nodes, the number of computing nodes is equal to the number of sub-regions N, and the discretized hydrodynamic continuity equation and hydrodynamic momentum equation are input into each computing node of the parallel computing framework; S32: Calculate the computational complexity index F in the sub-region according to the number of spatial nodes I set in the sub-region and the number of time steps W set in a data monitoring cycle. n ; Among them, γ1 and γ2 are the influence weights of water depth and flow velocity calculation on the computational complexity, γ1+γ2=1; S33: Calculate the total computational complexity index in the study area And calculate the ideal computational complexity index for each computing node S34: Calculate the computational complexity index of each sub-region in the study area, and sort the computational complexity indexes from the largest to the smallest, to obtain a computational complexity index sequence (F1, F2, ..., F N ); S35: compare the size of F1 and F0; If F1≤F0, the water depth and flow velocity data in each sub-area are directly input into the corresponding calculation node, and step S4 is executed; If F1>F0, execute step S36; S36: According to the computational complexity index sequence (F1, F2, ..., F N ) traverses the first computational complexity index F1 and the Nth computational complexity index F N , assign some of the spatial nodes in the sub-region corresponding to the first computational complexity index F1 to the computational complexity index F N In the corresponding computing node, return to step S34; assign the computational complexity index F N The spatial node in the corresponding calculation node is the spatial node where the current water depth or flow velocity is the minimum value in the sub-area; S37: until all computational complexity indicators in the computational complexity indicator sequence are less than or equal to F0, then the water depth and flow velocity data in each sub-area are input into the corresponding computing node, and step S4 is executed.

Images