A branch point processing method and system driven by digital twin water network data

Through the data-driven point processing method, combined with CFD simulation and machine learning, the hydrodynamic characteristics of the point are predicted, and the problems of insufficient accuracy and low efficiency at the point in the traditional method are solved, and efficient calculation of complex river networks is achieved.

CN117933143BActive Publication Date: 2024-11-29HOHAI UNIV
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
CN202410168611.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-02-06
Publication Date
2024-11-29
Estimated Expiration
2044-02-06

AI Technical Summary

Technical Problem

Traditional river network hydrodynamic calculation ignores momentum and energy conservation at the sill point, resulting in insufficient calculation accuracy and low accuracy of three-dimensional hydrodynamic calculation efficiency, which is not suitable for the digital twin river basin requirements of complex river networks.

Method used

Using data-driven point processing method, by collecting and processing point morphological data, a point sample working condition database and hydrodynamic information database are established, combined with CFD simulation and machine learning models, predicting the hydrodynamic characteristics of the point, and using explicit finite Volume method improves calculation efficiency.

Benefits of technology

It realizes rapid and accurate prediction of the hydrodynamic structure of complex river network points, improves calculation accuracy and efficiency, and adapts to the scale and real-time requirements of digital twin water networks.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117933143B_ABST
    Figure CN117933143B_ABST
Patent Text Reader

Abstract

The present invention discloses a method and system for processing branch points driven by digital twin water network data, the method comprising: step 1, collecting and processing branch point morphological data; step 2, establishing a branch point sample condition library; step 3, collecting branch point boundary conditions; step 4, obtaining branch point fine hydrodynamic data; step 5, establishing a branch point hydrodynamic information library; step 6, extracting data, and using machine learning methods such as correlation analysis and random forest classification to establish a river network branch point hydrodynamic fast and accurate prediction model; step 7, coupling the river network branch point hydrodynamic fast and accurate prediction model with one-dimensional river channel calculation. The present invention realizes data-driven branch point calculation, has the advantages of wide application range, fast calculation efficiency, and high simulation accuracy, and meets the needs of smart water conservancy construction.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the field of hydraulics and river dynamics, and specifically relates to a branch point processing method in digital twin calculation of complex river networks. Background Art

[0002] A river network is a water system composed of crisscrossing rivers, and the intersection of rivers is called a fork. River network calculation consists of two parts: one-dimensional river channel calculation and fork calculation. In traditional river network hydrodynamic calculation, the Saint-Venant equation is usually used to solve the river implicitly, which has many iterations, large amount of calculation, and low efficiency. It is only suitable for long river sections and simple river network calculations. For forks, their shape and structure are often ignored, and they are generalized as a simple water storage unit. The calculation is performed only based on the mass conservation of water in and out of the fork without considering the conservation of momentum and energy. Therefore, the precise calculation of the hydrodynamics of forks has always been one of the key difficulties in river network calculations.

[0003] my country has a vast territory, extensive river basins, and complex river networks. Taking river basins as units, building a national water network smart water conservancy system, realizing digital twins, and improving water regime monitoring and reporting capabilities are current research hotspots. However, smart water conservancy management in river basins requires a larger river network calculation scale; digital twin river basins require real-time river network models and higher computing efficiency; in digital twin river basins, there are a large number of short river channels that are interconnected and have densely distributed branches, and the complexity of calculations has greatly increased compared to the past; at the same time, the number of branches in complex river networks has increased, and their proportion in river network calculations has increased. The local loss of momentum and energy at the branches will be amplified in the entire river network calculation with the cumulative effect, so higher calculation accuracy is required.

[0004] Since the traditional method only considers the conservation of mass at the fork, but not the conservation of momentum and energy, it does not meet the accuracy requirements of the digital twin of the complex river network; and the method of using three-dimensional hydrodynamic precision calculation can accurately describe the hydrodynamic structure of mass, momentum and energy conservation at the fork, but the calculation scale is large, the calculation efficiency is low, and it is not real-time. Therefore, there is an urgent need for an algorithm that can adapt to the calculation scale, calculation efficiency and calculation accuracy requirements of the digital twin water network, quickly and accurately predict the hydrodynamic structure of the river network fork, and calculate the hydrodynamic conditions such as the fork diversion ratio and head loss. Summary of the invention

[0005] Purpose of the invention: In order to overcome the above-mentioned technical deficiencies, the present invention provides a data-driven branch point processing method and system, which aims to calculate the hydrodynamic conditions of complex river networks and provide support for the construction of digital twin river basins.

[0006] Technical solution: A branch point processing method driven by digital twin water network data, including the following steps:

[0007] Step 1, collect and process the morphological data of the forks: collect the terrain data of the river width and the river angle through various methods, and divide the forks into three-way and multi-way forks according to the shape of the forks, split the multi-way forks into multiple three-way forks, and divide the three-way forks into diverging forks and converging forks according to the flow direction of the river. The diverging forks have one upstream river channel and two downstream river channels, and the converging forks have two upstream river channels and one downstream river channel. The cross section of the compound river channel is generalized into an isosceles trapezoid, and the morphological data of each fork is dimensionless;

[0008] Step 2: Establish a sample working condition library of the branch point: Combine the river channel width and river channel angle morphological data of the branch point x obtained in step 1 with the dimensionless parameters and store them in vector A x ; Summarize all the sample working conditions of the river network branches and establish the branch sample working condition library matrix A;

[0009] Step 3, collect the boundary conditions of the branch point: the hydrodynamic data include the upstream section water level-time function relationship H(t) and the flow-time function relationship Q(t), determine the flow characteristic quantity and set the dimensionless parameters of the upstream water level and flow, calculate the Froude number Fr, and determine the river flow coefficient;

[0010] Step 4, obtaining detailed hydrodynamic data of the branch point: according to the process parameters of the upstream section, the computational fluid dynamics (CFD) simulation is used to obtain the downstream water level and flow rate and non-dimensionalize them, and calculate the diversion ratio n and the head loss coefficient ξ;

[0011] Step 5: Establish a branch point hydrodynamic information database: Combine the hydrodynamic characteristics of branch point x obtained in steps 3 and 4 with the dimensionless parameters and store them in vector B x In , the information vectors of different branch points are combined into a matrix B to form a branch point hydrodynamic information database;

[0012] Step 6: extract data from the branch point sample operating condition database and the branch point hydrodynamic information database, divide the data set into a training set and a test set, use the training set data to train the machine learning model, use the test set data to perform parameter calibration verification, and establish a river network branch point hydrodynamic prediction model;

[0013] Step 7: Input the upstream one-dimensional river channel hydrodynamic data and river network branch point topographic data into the river network branch point hydrodynamic prediction model, use the model to output the downstream river channel hydrodynamic prediction value, input the obtained result as the boundary condition into the Saint-Venant equation of the downstream river channel, and use the explicit finite volume method to solve the one-dimensional river channel hydrodynamic conditions.

[0014] Among them, in step 1, the dimensionless transformation of branch point morphological data includes:

[0015] For the width of the river, the average width of the river is taken For the characteristic quantity, set the dimensionless parameter: is the width of channel i at branch point x;

[0016] For the channel angle, the channel angle itself is taken as a dimensionless number.

[0017] In step 3, for the upstream water level, set the dimensionless parameters: is the average width of the river channel, h u is the water level of the upstream river u;

[0018] For flow, take the average flow of upstream and downstream rivers For the characteristic quantity, set the dimensionless parameter: Q u is the flow rate of the upstream river u;

[0019] Calculate the Froude number Flow rate u The upstream river flow Q u and cross-sectional area A u Divide them to calculate that g is the acceleration due to gravity and h is u is the upstream river water level;

[0020] Determining the river flow coefficient includes: Based on the relationship between the Froude number Fr and the river flow pattern Determine the flow pattern coefficient

[0021] In step 4, the Reynolds time-averaged stress RANS equation is used to calculate the downstream water depth and flow rate. The RANS equation is closed by the turbulence model, where:

[0022] Continuity equation:

[0023]

[0024] Momentum equation:

[0025]

[0026] Among them, u i ,u j represents the Reynolds mean speed, ν is the dynamic viscosity coefficient, ν t is the eddy viscosity coefficient, ρ is the water density, p is the pressure, g j is the volume force, and g1=g2=0, g3=-g, i, j=1, 2, 3 represent the components on the x, y, z axes respectively.

[0027] Furthermore, in step 4, the dimensionless number of the downstream water level is calculated: is the average width of the river channel, h d is the water level of the downstream river channel d;

[0028] Calculate the dimensionless number for downstream flow: is the average flow rate of the river, Q d represents the flow rate of the downstream channel d;

[0029] For the diversion type branch point, the upstream river channel is ①, and the downstream river channels are ② and ③. The diversion ratio n and the head loss coefficient ξ are calculated according to the following two formulas: Among them, the subscripts ①, ②, and ③ are the river channel numbers, indicating the calculation for the river channel; Q is the flow rate; h is the river water level; P is the pressure; ρ is the water density; g is the gravitational acceleration; U is the flow velocity of the river section; K Q2 and K Q3 are dimensionless numbers of downstream flow of channels ② and ③ respectively;

[0030] For a confluence type branch point, the diversion ratio n = NaN; let the upstream river channel be ①' and ②', and the downstream river channel be ③', then the head loss coefficient

[0031] In step 6, training the machine learning model using the training set data includes:

[0032] Input training set branch point samples: river width B, river angle θ;

[0033] Input training set flow conditions: section water level-time function relationship H(t) and flow-time function relationship Q(t);

[0034] Using machine learning, a mapping relationship between input and output is established: F(X) = n, G(Y) = ξ, where X and Y represent the input parameters required to calculate the diversion ratio n and the head loss coefficient ξ, respectively. The steps are as follows:

[0035] The first step is to use Spearman correlation analysis and chi-square test to process the training set and find out the variables related to the diversion ratio n and the head loss coefficient ξ;

[0036] The second step is to use the random forest classification method to establish the mapping relationship between input and output:

[0037] The third step is to use the test set to calibrate and verify the model, and at the same time add disturbance data to verify whether the model is sensitive to measurement errors; if the verification fails, return to the first step, optimize the neural network and recalculate; if the verification passes, output a complete river network branch point hydrodynamic prediction model.

[0038] The specific implementation method of step 7 includes:

[0039] Obtain the boundary conditions of the water level-time function relationship and flow-time function relationship of the upstream river section, and input the information into the hydrodynamic prediction model of the river network branch point;

[0040] The rapid and accurate prediction model of the hydrodynamics of the river network forks is used to predict the hydrodynamics of the downstream river channel, and the diversion ratio n and the head loss coefficient ξ are obtained.

[0041] The obtained results are input as boundary conditions into the Saint-Venant equation of the downstream channel, and the explicit finite volume method is used to solve the one-dimensional channel hydrodynamic conditions.

[0042] According to the river network branch point hydrodynamic prediction model and one-dimensional river channel explicit finite volume method, the hydraulic elements of the predicted area are updated to the time t+△t.

[0043] The present invention also provides a branch point processing system driven by digital twin water network data, comprising:

[0044] The fork morphological data acquisition module collects the terrain data of river width and river angle through various methods, and divides the forks into three-way and multi-way forks according to the shape of the forks, splits the multi-way forks into multiple three-way forks, and divides the three-way forks into diverging forks and converging forks according to the flow direction of the river. The diverging forks have one upstream river channel and two downstream river channels, and the converging forks have two upstream river channels and one downstream river channel. The cross section of the compound river channel is generalized into an isosceles trapezoid, and the morphological data of each fork is dimensionless.

[0045] The fork point sample condition library establishment module combines the river channel width and river channel angle morphological data of the fork point x obtained by the fork point morphological data acquisition module with the dimensionless parameters and stores them in the vector A x ; Summarize all the sample working conditions of the river network branches and establish the branch sample working condition library matrix A;

[0046] The branch point boundary condition acquisition module, the hydrodynamic data includes the upstream section water level-time function relationship H(t) and the flow-time function relationship Q(t), determines the flow characteristic quantity and sets the dimensionless parameters of the upstream water level and flow, calculates the Froude number Fr, and determines the river flow coefficient;

[0047] The branch point fine hydrodynamic data acquisition module uses computational fluid dynamics (CFD) simulation to obtain the downstream water level and flow rate according to the process parameters of the upstream section and non-dimensionalizes them, and calculates the diversion ratio n and the head loss coefficient ξ;

[0048] The module for establishing the hydrodynamic information database of the fork point is to combine the hydrodynamic characteristics of the fork point x obtained by the fork point boundary condition acquisition module and the fork point fine hydrodynamic data acquisition module with the dimensionless parameters and store them in the vector B x In , the information vectors of different branch points are combined into a matrix B to form a branch point hydrodynamic information database;

[0049] The module for establishing the hydrodynamic prediction model of the river network forks extracts data from the sample operating condition database of the forks and the hydrodynamic information database of the forks, divides the data set into a training set and a test set, uses the training set data to train the machine learning model, uses the test set data to perform parameter calibration and verification, and establishes the hydrodynamic prediction model of the river network forks;

[0050] The hydrodynamic condition calculation module inputs the upstream one-dimensional river hydrodynamic data and the river network branch point topographic data into the river network branch point hydrodynamic prediction model, uses the model to output the downstream river hydrodynamic prediction value, and inputs the obtained result as the boundary condition into the Saint-Venant equation of the downstream river, and uses the explicit finite volume method to solve the one-dimensional river hydrodynamic conditions.

[0051] The present invention also provides a computer device, comprising: one or more processors; a memory; and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and when the programs are executed by the processors, the steps of the branch point processing method driven by digital twin water network data as described above are implemented.

[0052] The present invention also provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the branch point processing method driven by digital twin water network data as described above are implemented.

[0053] Beneficial effects:

[0054] (1) The present invention decomposes river network calculation into branch point calculation and river channel calculation, adopts data-driven calculation of river network branch points, and adopts explicit method to solve one-dimensional river channels, thereby improving solution efficiency.

[0055] (2) The hydrodynamic structure of the branch points in the river network calculation is complex. The traditional river network model cannot consider the conservation relationship of mass, momentum and energy at the same time. However, the present invention uses fluid dynamics simulation to consider the conservation relationship of the three, more accurately characterize the hydrodynamic characteristics, and thus make the calculation more precise.

[0056] (3) The present invention calculates the hydrodynamic data in various situations in advance and stores them in the database. When needed, the data is matched with the hydrodynamic data under actual conditions, and the hydrodynamic conditions such as the diversion ratio and the head loss coefficient are calculated by difference. Compared with the three-level solution, the calculation of the water balance of the branch point is replaced by database interpolation calculation, which avoids iteration, reduces the amount of calculation, and improves the calculation efficiency; compared with the traditional three-dimensional CFD simulation, the massive calculation is completed in advance, avoiding real-time large-scale calculation, and greatly improving the calculation speed.

[0057] (4) Since the present invention has the advantages of higher accuracy, computational efficiency, and fast computational speed, the present invention is adapted to the current needs of smart water conservancy construction and facilitates the realization of digital twin technology for complex river networks.

[0058] (5) The present invention takes into account the similar structure of bifurcations in nature and uses dimensionless parameters to establish a model, thereby reducing the size of the sample library and improving the universality of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 This is a schematic diagram of a river network in an embodiment of the present invention;

[0060] Figure 2 It is a splitting diagram of branch point No. 2 of the river network in an embodiment of the present invention;

[0061] Figure 3 Schematic diagram of the branch point structure in an embodiment of the present invention;

[0062] Figure 4 is a schematic diagram of model training established in an embodiment of the present invention;

[0063] Figure 5 Schematic diagram of a technical framework for establishing a prediction model in an embodiment of the present invention;

[0064] Figure 6 Schematic diagram of application of branch point calculation in a river network according to an embodiment of the present invention;

[0065] Figure 7 It is a schematic diagram of the overall process of establishing a prediction model and performing practical application in an embodiment of the present invention. DETAILED DESCRIPTION

[0066] The technical solution of the present invention is further described below in conjunction with the accompanying drawings.

[0067] The present invention provides a method for processing forks driven by digital twin water network data. The method considers the actual morphological structure of river network forks, establishes a dimensionless generalized model based on the gravity similarity criterion, and forms a fork sample condition library. The fork hydrodynamic data is obtained through laboratory river model experiments, field measurements, and computational fluid dynamics simulation (CFD) to establish a fork hydrodynamic information library. The data in the fork sample condition library and the fork hydrodynamic information library are extracted, and a large number of trainings are performed through machine learning models such as random forest algorithms. For any fork hydrodynamic conditions, the three-dimensional hydrodynamic information of the fork at the next time step can be predicted and input into the one-dimensional river channel explicit calculation to realize data-driven fork hydrodynamic calculation.

[0068] Step 1: Collect and process branch point morphological data.

[0069] S101, river network branch point classification:

[0070] Although the shapes of river network forks in nature are diverse, they mostly have similar morphological structures, which can be mainly divided into three-way forks and multi-way forks (with 4 or more forks). Among them, since the multi-way fork structure can be decomposed into multiple three-way forks, the present invention establishes a generalized model for the three-way forks.

[0071] Taking a complex river network basin as an example, refer to Figure 1 There are 9 rivers and 4 forks in the basin. The rivers numbered ①-③ intersect at fork 1. Fork 2 is the intersection of rivers ②, ⑤, ⑦, and ⑧. Rivers ③, ④, and ⑥ intersect at fork 3. Rivers ⑥, ⑧, and ⑨ intersect at fork 4. In particular, fork 2, it can be split into two three-way structures in the calculation. The splitting method is as follows: Figure 2 In practical applications, other multi-branch structures can be similarly split into multiple three-branch branches.

[0072] According to the flow direction of the river, the three-way fork is divided into a diverging fork and a converging fork. The diverging fork has one upstream channel and two downstream channels, while the converging fork has two upstream channels and one downstream channel. Figure 1 As shown, assuming that the water flow direction in the basin is from left to right (geographically from west to east), fork 1 is a diversion type fork, the upstream is channel ①, and the downstream are channels ② and ③; fork 3 and 4 are confluence type forks, the upstream channels are ③ and ④, ⑥ and ⑧ respectively; the downstream channels are channels ⑥ and ⑨ respectively.

[0073] S102, collecting branch point morphological data:

[0074] By extracting river morphology from remote sensing images, we can collect branch point morphology data. Figure 3 , the branch point morphological data include the river channel width B and the river channel angle θ. In the following calculations, for each diverging three-way branch point, the three associated rivers are represented by ①②③ respectively, and the width of the river channel ① at the branch point x is recorded as The width of the river channel ② is recorded as The width of the river channel ③ is recorded as The angle between river channel ① and ② is recorded as The angle between river channel ① and ③ is recorded as like Figure 3 As shown in (a) in the figure. The three channels associated with each confluence-type three-way fork are represented by ①'②'③' respectively, and the width of the channel ①' at the fork x is recorded as The width of the river channel ②' is recorded as The width of the river channel ③' is recorded as The angle between the river channel ①'③' is recorded as The angle between the river channel ②'③' is recorded as like Figure 3 As shown in (b) in the figure. However, for the convenience of description, when the calculation does not involve divergence or confluence, the description is expressed by the parameters of the divergence type branch point, that is, the calculation of the divergence type branch point is also applicable to the confluence type branch point. In the case where it is necessary to distinguish (that is, involving upstream and downstream distinction), make a clear explanation.

[0075] S103, generalization of river section:

[0076] Considering that most rivers in nature are compound rivers with shoals on both sides of the main channel, the shape of this compound river is close to an isosceles trapezoid. Therefore, existing studies have generally generalized the cross section of the river into an isosceles trapezoid. The present invention also makes the same treatment and generalizes the cross section of the river into an isosceles trapezoid. Figure 3 , take the width of the median line of the trapezoid as B, the angle between the two sides of the trapezoid and the bottom of the riverbed as α, and the height of the trapezoid as h.

[0077] So there is the following relationship:

[0078]

[0079]

[0080] The cross-sectional area A is:

[0081] A=Bh #(3)

[0082] S104, dimensionless terrain data:

[0083] For the width of the river channel, the average width of the median line of the isosceles trapezoid corresponding to each river channel section is taken. For the characteristic quantity, set the dimensionless parameter:

[0084]

[0085]

[0086] are the dimensionless parameters of the three river channel widths at the branch point x.

[0087] The river angle is It can be regarded as a dimensionless number, so no processing is done.

[0088] Step 2: Establish a branch point sample condition library.

[0089] S201, combine the feature quantity and dimensionless parameter and store them in vector A x :

[0090]

[0091] S202, summarizing all N river network branch point sample conditions to obtain a branch point sample condition library:

[0092]

[0093] Step 3: Collect branch point boundary conditions.

[0094] S301, boundary conditions include the upstream section water level-time function relationship H(t) and flow-time function relationship Q(t). For the diverging branch point x, channel ① is the upstream channel of the branch point x, and channels ② and ③ are the downstream channels. For the water level, dimensionless parameters are set:

[0095]

[0096] in, is the average width of the river channel, and h1 is the water level of the river channel①.

[0097] For flow, take the average flow of upstream and downstream rivers For the characteristic quantity, set the dimensionless parameter:

[0098]

[0099] Among them, Q1 is the flow rate of river channel ①.

[0100] For the confluence type branch point x, the river channels ①' and ②' are the upstream river channels of the branch point x, and the river channel ③' is the downstream river channel. For the water level, the dimensionless parameters are set:

[0101]

[0102] in, is the average width of the river channel, h 1′ and h 2′ are the water levels of river channels ①' and ②' respectively.

[0103] For flow, take the average flow of upstream and downstream rivers For the characteristic quantity, set the dimensionless parameter:

[0104]

[0105] Among them, Q 1′ and Q 2′ are the flow rates of channels ①' and ②' respectively.

[0106] S302, for the upstream river channel, calculate the Froude number Fr and determine the river flow pattern:

[0107]

[0108] Wherein, the flow rate U u The flow rate Qu and cross-sectional area A u Work out:

[0109]

[0110] The subscript u represents the upstream river channel. For the diverging type fork, u is ①, and for the converging type fork, u is ①' and ②', that is, two Fr are calculated, and there are two flow coefficients accordingly.

[0111] Fr has the following relationship with river flow pattern:

[0112]

[0113] So define K Fr is the flow coefficient, there are:

[0114]

[0115] Step 4: Obtain detailed hydrodynamic data of branch points.

[0116] The detailed hydrodynamic data of the fork can be obtained through experimental measurement of river engineering models, hydrological station measurement, etc. However, due to the limitations of experimental equipment and measurement equipment, the boundary conditions of the fork obtained by the above two methods may have errors and omissions. Therefore, the present invention mainly adopts CFD detailed simulation calculation to obtain the hydrodynamic data of the fork.

[0117] S401, based on the flow parameters of the upstream section, taking into account the conservation of mass, momentum and energy, the downstream flow parameters are calculated using CFD fine simulation:

[0118] The simulation is performed using relevant computational fluid dynamics software, such as OpenFOAM, Ansys Fluent, Flow-3D, etc. Considering that actual rivers are mostly turbulent, the specific calculation mainly uses the Reynolds-Averaged Navier-Stokes Equation (RANS).

[0119] The calculation mainly uses the RANS equation:

[0120] Continuity equation:

[0121]

[0122] Momentum equation:

[0123]

[0124] Among them, u i ,u j represents the Reynolds mean speed, ν is the dynamic viscosity coefficient, ν tis the eddy viscosity coefficient, ρ is the density, p is the pressure, g j is the volume force, and g1=g2=0, g3=-g (only considering the gravity field), i, j=1, 2, 3 represent the components on the x, y, z axes respectively.

[0125] The RANS equations can be closed by turbulence models, such as the k-ε model, k-ω model, etc. The downstream water depth h and flow rate Q are calculated by the RANS equations.

[0126] S402, for the diverging branch point x, calculate the dimensionless number of the downstream water level according to the following formula:

[0127]

[0128] in, is the average width of the river channel, h2 and h3 represent the water levels of channel ② and channel ③ respectively.

[0129] For the confluence point x, the dimensionless number of the downstream water level is calculated as follows:

[0130]

[0131] Among them, h 3, Indicates the water level of river channel ③'.

[0132] S403, for the diverging point x, calculate the dimensionless number of downstream flow according to the following formula:

[0133]

[0134] in, is the average flow rate of the river, Q2 and Q3 represent the flow rates of river channel ② and river channel ③ respectively.

[0135] For the confluence point x, the dimensionless number of downstream flow is calculated as follows:

[0136]

[0137] Among them, Q 3′ Indicates the flow rate of river channel ③'.

[0138] S404, the split ratio n and the head loss coefficient ξ are calculated below:

[0139] For diverging points, the calculation method is as follows:

[0140]

[0141]

[0142] Among them, the subscripts ①, ②, and ③ are the river channel numbers, indicating the calculation for the river channel; Q is the flow rate; h is the river water level; P is the pressure; ρ is the water density; g is the gravitational acceleration, g = 9.81 m / s 2 ; U is the flow velocity of the river section, all of which adopt international standard units.

[0143] For the confluence type branch point, the diversion ratio is empty, which is expressed as diversion ratio n = NaN; let the upstream river channel be ①' and ②', and the downstream river channel be ③', then the head loss coefficient is:

[0144]

[0145] Step 5: Establish a branch point hydrodynamic information database.

[0146] S501, combine the feature value of branch point x with the dimensionless parameters obtained in steps 3 and 4, and store them in vector B x middle.

[0147]

[0148] S502, for information vectors B of different branch points x , merged into matrix B, we have:

[0149]

[0150] Step 6: Extract data and use machine learning methods such as correlation analysis and random forest classification to establish a rapid and accurate prediction model for the hydrodynamics of river network branches.

[0151] The specific steps include:

[0152] S601, extract and segment data from the branch point sample condition library and the branch point hydrodynamic information library, randomly select 80% of the original data as a training set; and the remaining 20% ​​of the original data as a test set.

[0153] S602, input training set branch point samples: river width B, river angle θ; input training set flow conditions: cross-section water level-time function relationship H(t) and flow-time function relationship Q(t).

[0154] S603, using machine learning to establish a mapping relationship between input and output:

[0155] F(X)=n#(27)

[0156] G(Y)=ξ#(28)

[0157] Among them, X and Y represent the input parameters needed to calculate the diversion ratio n and the head loss coefficient ξ respectively. Figure 4As shown in the figure, the model input parameters include section width, channel angle, section water level, section flow, Froude number, etc., and the output hydrodynamic characteristic parameters are: diversion ratio n and head loss coefficient ξ.

[0158] The specific process is as follows:

[0159] 1. Use Spearman correlation analysis and chi-square test to process the training set and find out the variables related to the diversion ratio n and the head loss coefficient ξ.

[0160] 2. Use random forest classification method to establish the mapping relationship between input and output.

[0161] 3. Use the test set to calibrate the model, that is, to verify the accuracy of the model. At the same time, add "disturbance" data to demonstrate whether the model is sensitive to measurement errors. If the verification fails, return to the first step, optimize the neural network and recalculate; if the verification passes, output a complete river network branch point hydrodynamic rapid and accurate prediction model.

[0162] The technical framework for establishing the prediction model in the present invention is as follows: Figure 5 The prediction model inputs the branch point morphological data and branch point fine hydrodynamic data; and outputs the diversion ratio and head loss coefficient.

[0163] Step 7: Couple the rapid and accurate prediction model of river network branch hydrodynamics with the one-dimensional river channel calculation.

[0164] For a river network branch point in a certain river basin, there are river channels ①②③ connected, and the width of river channel ① is B1, the width of river channel ② is B2, the width of river channel ③ is B3, the angle between river channels ①② is θ1, and the angle between river channels ①③ is θ2.

[0165] Reference Figure 6 In practical applications, the calculation of branch points of a river network includes the following steps:

[0166] S701, obtain boundary conditions such as the water level-time function relationship and flow-time function relationship of the river section ①, and input the information into the rapid and accurate prediction model of the hydrodynamics of the river network branches.

[0167] S702, using the river network branch point hydrodynamic fast and accurate prediction model to output the hydrodynamic prediction values ​​of river channels ②③, namely the diversion ratio n and the head loss coefficient ξ.

[0168] S703, input the obtained results into the Saint-Venant equation of the downstream river channel as boundary conditions.

[0169]

[0170]

[0171] Where A is the cross-sectional area, Q is the flow rate, q is the lateral inflow per unit length, h is the water level, and S f For the slope.

[0172] S704, since the explicit finite volume method has a fast calculation speed and decouples the calculation of the river channel and the branch point, it is compatible with the rapid and accurate prediction model of the hydrodynamics of the river network branch point. Therefore, the explicit finite volume method is used to solve the one-dimensional river channel hydrodynamic conditions and calculate the water level and flow of the river channel.

[0173] S705, based on the rapid and accurate prediction model of the hydrodynamics of the river network branches and the one-dimensional river channel explicit finite volume method calculation, the hydraulic elements of the predicted area are updated to the time t+△t.

[0174] Figure 7 The overall process diagram of establishing a prediction model and carrying out practical application in the present invention is shown.

[0175] In summary, the present invention discloses a data-driven branch point processing method, establishes a branch point sample operating condition library and a branch point hydrodynamic information library, adopts machine learning methods such as correlation analysis and random forest classification, and establishes a rapid and accurate prediction model for the hydrodynamics of river network branches, which is suitable for application in digital twin water networks.

[0176] Based on the same technical concept as the method, the present invention also provides a branch point processing system driven by digital twin water network data, including:

[0177] The fork morphological data acquisition module collects the terrain data of river width and river angle through various methods, and divides the forks into three-way and multi-way forks according to the shape of the forks, splits the multi-way forks into multiple three-way forks, and divides the three-way forks into diverging forks and converging forks according to the flow direction of the river. The diverging forks have one upstream river channel and two downstream river channels, and the converging forks have two upstream river channels and one downstream river channel. The cross section of the compound river channel is generalized into an isosceles trapezoid, and the morphological data of each fork is dimensionless.

[0178] The fork point sample condition library establishment module combines the river channel width and river channel angle morphological data of the fork point x obtained by the fork point morphological data acquisition module with the dimensionless parameters and stores them in the vector A x ; Summarize all the sample working conditions of the river network branches and establish the branch sample working condition library matrix A;

[0179] The branch point boundary condition acquisition module, the hydrodynamic data includes the upstream section water level-time function relationship H(t) and the flow-time function relationship Q(t), determines the flow characteristic quantity and sets the dimensionless parameters of the upstream water level and flow, calculates the Froude number Fr, and determines the river flow coefficient;

[0180] The branch point fine hydrodynamic data acquisition module uses computational fluid dynamics (CFD) simulation to obtain the downstream water level and flow rate according to the process parameters of the upstream section and non-dimensionalizes them, and calculates the diversion ratio n and the head loss coefficient ξ;

[0181] The module for establishing the hydrodynamic information database of the fork point is to combine the hydrodynamic characteristics of the fork point x obtained by the fork point boundary condition acquisition module and the fork point fine hydrodynamic data acquisition module with the dimensionless parameters and store them in the vector B x In , the information vectors of different branch points are combined into a matrix B to form a branch point hydrodynamic information database;

[0182] The module for establishing the hydrodynamic prediction model of the river network forks extracts data from the sample operating condition database of the forks and the hydrodynamic information database of the forks, divides the data set into a training set and a test set, uses the training set data to train the machine learning model, uses the test set data to perform parameter calibration and verification, and establishes the hydrodynamic prediction model of the river network forks;

[0183] The hydrodynamic condition calculation module inputs the upstream one-dimensional river channel hydrodynamic data and the river network branch point topographic data into the river network branch point hydrodynamic prediction model, uses the model to output the downstream river channel hydrodynamic prediction value, and inputs the obtained results as boundary conditions into the Saint-Venant equation of the downstream river channel, and uses the explicit finite volume method to solve the one-dimensional river channel hydrodynamic conditions.

[0184] It should be understood that the branch point processing system driven by digital twin water network data in the embodiment of the present invention can implement all the technical solutions in the above-mentioned method embodiment, and the functions of its various functional modules can be specifically implemented according to the methods in the above-mentioned method embodiments. The specific implementation process can refer to the relevant description in the above-mentioned embodiments, which will not be repeated here.

[0185] The present invention also provides a computer device, comprising: one or more processors; a memory; and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and when the programs are executed by the processors, the steps of the branch point processing method driven by digital twin water network data as described above are implemented.

[0186] The present invention also provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the branch point processing method driven by digital twin water network data as described above are implemented.

[0187] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, devices (systems), computer equipment or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0188] The present invention is described with reference to a flowchart of a method according to an embodiment of the present invention. It should be understood that each process in the flowchart and a combination of processes in the flowchart can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the process in the flowchart. Figure 1 A device that specifies functions in a process or multiple processes.

[0189] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A function specified in a process or multiple processes.

[0190] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 The steps of a specified function in a process or multiple processes.

Claims

1. A branch point processing method driven by digital twin water network data, characterized in that: The following steps are involved: Step 1, collect and process the morphological data of the forks: collect the terrain data of the river width and the river angle through various methods, and divide the forks into three-way and multi-way forks according to the shape of the forks, split the multi-way forks into multiple three-way forks, and divide the three-way forks into diverging forks and converging forks according to the flow direction of the river. The diverging forks have one upstream river channel and two downstream river channels, and the converging forks have two upstream river channels and one downstream river channel. The cross section of the compound river channel is generalized into an isosceles trapezoid, and the morphological data of each fork is dimensionless; Step 2: Establish a sample working condition library of the branch point: Combine the river channel width and river channel angle morphological data of the branch point x obtained in step 1 with the dimensionless parameters and store them in vector A x ; Summarize all the sample working conditions of the river network branches and establish the branch sample working condition library matrix A; Step 3: Collect the boundary conditions of the branch point: The hydrodynamic data include the upstream section water level-time function relationship H(t) and the flow-time function relationship Q(t), determine the flow characteristic and set the dimensionless parameters of the upstream water level and flow, calculate the Froude number Fr, and determine the river flow coefficient. For the upstream water level, set the dimensionless parameters: is the average width of the river channel, h u is the water level of the upstream river u; for the flow rate, take the average flow rate of the upstream and downstream rivers For the characteristic quantity, set the dimensionless parameter: Q u is the flow rate of the upstream river u; calculate the Froude number Flow rate u The upstream river flow Q u and cross-sectional area A u Divide and calculate, g is the acceleration due to gravity; Determining the river flow coefficient includes: Based on the relationship between the Froude number Fr and the river flow pattern Flow coefficient Step 4: Obtain detailed hydrodynamic data of the branch point: According to the process parameters of the upstream section, the computational fluid dynamics (CFD) simulation is used to obtain the downstream water level and flow and non-dimensionalize them, and calculate the diversion ratio n and the head loss coefficient ξ; among which, the dimensionless number of the downstream water level is calculated: is the average width of the river channel, h d is the water level of the downstream river channel d; Calculate the dimensionless number for downstream flow: is the average flow rate of the river, Q d represents the flow rate of the downstream channel d; For the diversion type branch point, the upstream river channel is ①, the downstream river channel is ② and ③, and the diversion ratio n and head loss coefficient are calculated according to the following two formulas respectively: Among them, the subscripts ①, ②, and ③ are the river channel numbers, indicating the calculation for the river channel; Q is the flow rate; h is the river water level; P is the pressure; ρ is the water density; g is the gravitational acceleration; U is the flow velocity of the river section; K Q2 and K Q3 are dimensionless numbers of downstream flow of channels ② and ③ respectively; For a confluence type branch point, the diversion ratio n is empty, n = NaN; let the upstream river channel be ①' and ②', and the downstream river channel be ③', then the head loss coefficient Step 5: Establish a branch point hydrodynamic information database: Combine the hydrodynamic characteristics of branch point x obtained in steps 3 and 4 with the dimensionless parameters and store them in vector B x In , the information vectors of different branch points are combined into a matrix B to form a branch point hydrodynamic information database; Step 6: extract data from the branch point sample operating condition database and the branch point hydrodynamic information database, divide the data set into a training set and a test set, use the training set data to train the machine learning model, use the test set data to perform parameter calibration verification, and establish a river network branch point hydrodynamic prediction model; Step 7: Input the upstream one-dimensional river channel hydrodynamic data and river network branch point topographic data into the river network branch point hydrodynamic prediction model, use the model to output the downstream river channel hydrodynamic prediction value, input the obtained result as the boundary condition into the Saint-Venant equation of the downstream river channel, and use the explicit finite volume method to solve the one-dimensional river channel hydrodynamic conditions.

2. The method according to claim 1, characterized in that: In step 1, the dimensionless transformation of branch point morphological data includes: For the width of the river, the average width of the river is taken For the characteristic quantity, set the dimensionless parameter: is the width of the river channel i at the branch point x, and the value range of i is {1,2,3}; For the channel angle, the channel angle itself is taken as a dimensionless number.

3. The method according to claim 1, characterized in that In step 4, the Reynolds time-averaged stress RANS equation is used to calculate the downstream water depth and flow rate. The RANS equation is closed by the turbulence model, where: Continuity equation: Momentum equation: Among them, u i ,u j represents the Reynolds mean speed, ν is the dynamic viscosity coefficient, ν t is the eddy viscosity coefficient, ρ is the water density, p is the pressure, g j is the volume force, and g1=g2=0, g3=-g, i, j=1, 2, 3 represent the components on the x, y, z axes respectively.

4. The method according to claim 1, characterized in that: In step 6, training the machine learning model using the training set data includes: Input training set branch point samples: river width B, river angle θ; Input training set flow conditions: section water level-time function relationship H(t) and flow-time function relationship Q(t); Using machine learning, a mapping relationship between input and output is established: F(X) = n, G(Y) = ξ, where X and Y represent the input parameters required to calculate the diversion ratio n and the head loss coefficient ξ, respectively. The steps are as follows: The first step is to use Spearman correlation analysis and chi-square test to process the training set and find out the variables related to the diversion ratio n and the head loss coefficient ξ; The second step is to use the random forest classification method to establish the mapping relationship between input and output: The third step is to use the test set to calibrate and verify the model, and at the same time add disturbance data to verify whether the model is sensitive to measurement errors; if the verification fails, return to the first step, optimize the neural network and recalculate; if the verification passes, output a complete river network branch point hydrodynamic prediction model.

5. The method according to claim 1, characterized in that Step 7 includes: Obtain the boundary conditions of the water level-time function relationship and flow-time function relationship of the upstream river section, and input the information into the hydrodynamic prediction model of the river network branch point; The rapid and accurate prediction model of the hydrodynamics of the river network forks is used to predict the hydrodynamics of the downstream river channel, and the diversion ratio n and the head loss coefficient ξ are obtained. The obtained results are input as boundary conditions into the Saint-Venant equation of the downstream channel, and the explicit finite volume method is used to solve the one-dimensional channel hydrodynamic conditions. According to the river network branch point hydrodynamic prediction model and one-dimensional river channel explicit finite volume method, the hydraulic elements of the predicted area are updated to the time t+△t.

6. A branch point processing system driven by digital twin water network data, characterized in that: include: The fork morphological data acquisition module collects the terrain data of river width and river angle through various methods, and divides the forks into three-way and multi-way forks according to the shape of the forks, splits the multi-way forks into multiple three-way forks, and divides the three-way forks into diverging forks and converging forks according to the flow direction of the river. The diverging forks have one upstream river channel and two downstream river channels, and the converging forks have two upstream river channels and one downstream river channel. The cross section of the compound river channel is generalized into an isosceles trapezoid, and the morphological data of each fork is dimensionless. The fork point sample condition library establishment module combines the river channel width and river channel angle morphological data of the fork point x obtained by the fork point morphological data acquisition module with the dimensionless parameters and stores them in the vector A x ; Summarize all the sample working conditions of the river network branches and establish the branch sample working condition library matrix A; The branch point boundary condition acquisition module, the hydrodynamic data includes the upstream section water level-time function relationship H(t) and the flow-time function relationship Q(t), determines the flow characteristic quantity and sets the dimensionless parameters of the upstream water level and flow, calculates the Froude number Fr, and determines the river flow coefficient. For the upstream water level, the dimensionless parameters are set: is the average width of the river channel, h u is the water level of the upstream river u; for the flow rate, take the average flow rate of the upstream and downstream rivers For the characteristic quantity, set the dimensionless parameter: Q u is the flow rate of the upstream river u; calculate the Froude number Flow rate u The upstream river flow Q u and cross-sectional area A u Divide and calculate, g is the acceleration due to gravity; Determining the river flow coefficient includes: Based on the relationship between the Froude number Fr and the river flow pattern Determine the flow pattern coefficient The branch point fine hydrodynamic data acquisition module uses computational fluid dynamics (CFD) simulation to obtain the downstream water level and flow rate according to the process parameters of the upstream section and non-dimensionalizes them, and calculates the diversion ratio n and the head loss coefficient ξ; among them, the dimensionless number of the downstream water level is calculated: is the average width of the river channel, h d is the water level of the downstream river channel d; Calculate the dimensionless number for downstream flow: is the average flow rate of the river, Q d represents the flow rate of the downstream channel d; For the diversion type branch point, the upstream river channel is ①, the downstream river channel is ② and ③, and the diversion ratio n and head loss coefficient are calculated according to the following two formulas respectively: Among them, the subscripts ①, ②, and ③ are the river channel numbers, indicating the calculation for the river channel; Q is the flow rate; h is the river water level; P is the pressure; ρ is the water density; g is the gravitational acceleration; U is the flow velocity of the river section; K Q2 and K Q3 are dimensionless numbers of downstream flow of channels ② and ③ respectively; For a confluence type branch point, the diversion ratio n is empty, n = NaN; let the upstream river channel be ①' and ②', and the downstream river channel be ③', then the head loss coefficient The module for establishing the hydrodynamic information database of the fork point is to combine the hydrodynamic characteristics of the fork point x obtained by the fork point boundary condition acquisition module and the fork point fine hydrodynamic data acquisition module with the dimensionless parameters and store them in the vector B x In , the information vectors of different branch points are combined into a matrix B to form a branch point hydrodynamic information database; The module for establishing the hydrodynamic prediction model of the river network forks extracts data from the sample operating condition database of the forks and the hydrodynamic information database of the forks, divides the data set into a training set and a test set, uses the training set data to train the machine learning model, uses the test set data to perform parameter calibration and verification, and establishes the hydrodynamic prediction model of the river network forks; The hydrodynamic condition calculation module inputs the upstream one-dimensional river hydrodynamic data and the river network branch point topographic data into the river network branch point hydrodynamic prediction model, uses the model to output the downstream river hydrodynamic prediction value, and inputs the obtained result as the boundary condition into the Saint-Venant equation of the downstream river, and uses the explicit finite volume method to solve the one-dimensional river hydrodynamic conditions.

7. A computer device, characterized in that: include: one or more processors; Memory; And one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and when the programs are executed by the processors, the steps of the branch point processing method driven by digital twin water network data as described in any one of claims 1-5 are implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the branch point processing method driven by digital twin water network data as described in any one of claims 1 to 5 are implemented.

Citation Information

Patent Citations

  • A general river network water flow numerical simulation method considering a branch point area

    CN109885931A

  • River network water flow numerical simulation method using bifurcated point influence area water surface gradient

    CN111597732A

  • Method for researching split ratio and sewage separation ratio and evolution rule thereof of bifurcated channel

    CN114049013A

  • River network hydrodynamic process simulation method based on branch water conservation

    CN115796075A

  • Research method for diversion ratio and pollutant transport flux ratio of braided river and evolution rule of both

    WO2023087995A1