A flow measurement method and system under complex water flow conditions

Abstract

The present invention discloses a flow measurement method and system under complex water flow conditions, the method comprising the following steps: according to the flow measurement section position, determine the three-dimensional hydrodynamic model modeling area and the model inlet and outlet positions and verification positions, and establish a three-dimensional hydrodynamic model; use image recognition technology to obtain the water level, surface velocity and boundary section flow data of the boundary section and the verification position; perform parameter calibration and verification on the established three-dimensional hydrodynamic model; use the three-dimensional hydrodynamic model for simulation to obtain the flow of the measurement section; use the measurement section flow as output to construct an LSTM artificial intelligence model and a model training set and a test set, train the LSTM model algorithm, use the trained LSTM model for rapid calculation, and obtain real-time flow measurement data. The present invention solves the problems of no stable corresponding relationship between water level and flow under complex water flow conditions, inaccurate estimation of cross-sectional area using a single water level value for non-planar water surfaces, and provides a rapid calculation method for flow.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a flow measurement method and system under complex water flow conditions. Background Art

[0002] Flow is the basic data reflecting the changes in water resources and water volume in rivers, lakes, reservoirs and other water bodies. It is also the most important hydrological characteristic value of rivers. Real-time flow monitoring can be used to grasp the changes in water volume in natural rivers, which is very important for river health protection, flood and drought prevention, and river and lake water resources management.

[0003] For the real-time flow monitoring of open channel water flow with average cross-section and single straight water flow, the velocity area method can be used to divide the test section into multiple parts, and vertical lines of velocity measurement are arranged on each part. The velocity of the measuring points on each vertical line is measured by current meters and other equipment to obtain the average velocity of the vertical line and calculate the flow of this part, and then the flow of the entire section is accumulated. However, due to factors such as high cost of measurement equipment and complex measurement process, the real-time flow is usually estimated based on the water level-flow ratio curve proposed based on temporary flow measurement results. However, for complex water flow conditions such as tidal rivers, dam-controlled rivers, and confluences of main and tributary rivers, there is no stable correspondence between water level and flow, and the water level-flow ratio curve is not applicable.

[0004] In view of this situation, the prior art often uses an acoustic Doppler current meter to obtain the flow velocity of a water depth unit or the vertical average flow velocity of a monitoring section as an index flow velocity, establishes a functional relationship between the index flow velocity and the average flow velocity of the section, and then establishes a functional relationship between the water level and the cross-sectional area, and obtains the cross-sectional flow by multiplying the average flow velocity by the area. However, this method has the following shortcomings: 1) It is difficult to construct a functional relationship that reflects the average flow velocity of the entire section for a single index flow velocity of a section with a large lateral difference in flow velocity; 2) Due to the influence of Coriolis force and river channel curvature, the water surface of some waters is not a plane, but a slope, and the cross-sectional area estimated by a single water level value is not accurate; 3) The acoustic Doppler current meter equipment is expensive, complex to operate, and has a long measurement cycle. It cannot provide flow data in a timely manner, and is affected by water flow conditions, and the measurement accuracy cannot be guaranteed. There is also a method of simulating the relationship between water level and flow by establishing a two-dimensional hydrodynamic model. However, the model calculation is complex and the parameter calibration requires a large amount of real-time upstream and downstream flow and water level monitoring data to ensure it. Due to the limitations of monitoring costs and the timeliness of monitoring data, the model accuracy and timeliness of flow calculation results are difficult to meet actual measurement requirements. Summary of the invention

[0005] In order to solve the problems existing in the prior art, the present invention provides a flow measurement method and system under complex water flow conditions, which solves the problems of no stable correspondence between water level and flow under complex water flow conditions, inaccurate estimation of cross-sectional area using a single water level value for non-planar water surfaces, and high cost of measuring equipment, complex operation, and inability to provide flow data in a timely manner. By coupling an artificial intelligence model, rapid calculation of flow is achieved.

[0006] The present invention provides a flow measurement method under complex water flow conditions, the method comprising the following steps:

[0007] According to the flow measurement section location, determine the three-dimensional hydrodynamic model modeling area and the model inlet and outlet locations and verification locations, and establish a three-dimensional hydrodynamic model;

[0008] Using image recognition technology, obtain water level, surface velocity and boundary section flow data of boundary sections and verification locations;

[0009] The established three-dimensional hydrodynamic model is calibrated and verified using the water level, surface velocity and boundary section flow data at the boundary section and verification location;

[0010] Simulate the calibrated three-dimensional hydrodynamic model to obtain the cross-sectional flow rate;

[0011] With the measured cross-section flow as output, the LSTM artificial intelligence model and training and test sets are constructed to train the LSTM model algorithm;

[0012] Use the trained LSTM model to perform fast calculations to obtain real-time traffic measurement data.

[0013] Preferably, the step of determining the modeling area of ​​the three-dimensional hydrodynamic model and determining the model inlet and outlet positions and the verification position further includes:

[0014] Collect underwater topography and shoreline data of the modeling area, divide the modeling area into three-dimensional grids, and set the boundary conditions and initial conditions of the three-dimensional hydrodynamic model.

[0015] Preferably, the determined model inlet and outlet positions are implemented by the following steps:

[0016] The model inlet section is selected upstream of the flow measurement section and where the flow pattern is stable, including the main inflow section and the tributary inflow section;

[0017] The model outlet section is selected downstream of the flow measurement section and where the water flow is stable, including the main stream outflow section and the tributary outflow section.

[0018] Preferably, the three-dimensional hydrodynamic model uses the NS equation as the control equation, adopts the σ coordinate system to fit the irregular bottom terrain vertically, uses non-overlapping unstructured triangular grids to spatially discretize the calculation area to fit the shoreline horizontally, and adopts the finite volume discrete method to solve it.

[0019] Preferably, the acquisition of water level, surface velocity and boundary section flow data of the boundary section and the verification position comprises the following steps:

[0020] The shoreline images at the inlet and outlet are collected by cameras to construct a water surface line pixel-elevation relationship dataset. The water level values ​​corresponding to the pixels of the water level line are obtained by the water area shoreline segmentation algorithm, and the average water level value of the cross section is obtained by the spatiotemporal averaging method.

[0021] The surface image of the cross-section water flow is collected by the camera, and the ripples, floating objects, and bubble features in the water surface image are extracted. The speed measurement line is drawn along the direction of the cross-section water flow. The length of the speed measurement line is used as the horizontal axis and the video acquisition time is used as the vertical axis to synthesize the space-time image and calculate the cross-section surface flow monitoring data.

[0022] Preferably, the parameter calibration of the established three-dimensional hydrodynamic model includes: calibrating the model parameters by trial and error method until the calculation results of the three-dimensional hydrodynamic model meet the accuracy requirements.

[0023] Preferably, simulating the calibrated three-dimensional hydrodynamic model to obtain the measured cross-sectional flow includes the following steps:

[0024] Based on the established three-dimensional hydrodynamic model and the parameters obtained by calibration, the real-time monitoring water level data of the inlet section and the outlet section are used to drive the three-dimensional hydrodynamic model calculation, and the water level and three-dimensional flow velocity calculation results at any position in the modeling area are obtained;

[0025] According to the three-dimensional flow velocity calculation results at the flow measurement section, the flow velocity of the entire section is integrated to obtain the flow of the measurement section.

[0026] Preferably, the LSTM artificial intelligence model, training set and test set are constructed with the measured cross-sectional flow as output, and the training LSTM model algorithm includes:

[0027] An LSTM artificial intelligence model is established for the flow measurement section; the output layer of the LSTM artificial intelligence model is the flow of the measurement section, and the model input layer is the water level and flow of the inlet and outlet sections and the water level and surface flow velocity monitoring process at the verification section;

[0028] The image recognition method is used to obtain the water level, flow rate, and surface velocity monitoring process at the inlet and outlet sections and the verification section, as well as the cross-sectional flow rate calculated by the three-dimensional hydrodynamic model, to construct the artificial intelligence model training set and test set;

[0029] The artificial intelligence model is trained, tested and verified using the training set and the test set, and the model accuracy is evaluated using indicators such as root mean square error, mean absolute error and Nash coefficient until the model accuracy meets the flow measurement accuracy requirements.

[0030] Preferably, the use of the trained LSTM model for rapid calculation to obtain real-time flow measurement data includes: using the water level, flow rate, and surface velocity monitoring process data at the inlet and outlet sections and the verification section, and using an artificial intelligence model that has passed training and testing verification to calculate the real-time flow of the flow measurement section, thereby realizing real-time flow measurement under complex water flow conditions.

[0031] The present invention also provides a flow measurement system under complex water flow conditions, the system is applied to any of the above-mentioned flow measurement methods under complex water flow conditions, the system comprising:

[0032] Model building module: used to determine the modeling area of ​​the three-dimensional hydrodynamic model and the model inlet and outlet positions and verification positions according to the flow measurement section position, and establish the three-dimensional hydrodynamic model;

[0033] Model processing module: used to obtain water level, surface velocity and cross-sectional flow data of boundary sections and verification positions using image recognition technology;

[0034] Model verification module: used to calibrate and verify the parameters of the established three-dimensional hydrodynamic model using the water level, surface velocity and cross-sectional flow data of the boundary section and verification position;

[0035] Model simulation module: used to simulate the calibrated three-dimensional hydrodynamic model to obtain the measured cross-sectional flow;

[0036] Model training module: used to build LSTM artificial intelligence model and training set and test set with the measured cross-section flow as output, and train LSTM model algorithm;

[0037] Model calculation module: used to use the trained LSTM model for fast calculation to obtain real-time traffic measurement data.

[0038] Technical effects and advantages of the present invention:

[0039] The present invention establishes a three-dimensional hydrodynamic model using measured topography and hydrological data to perform numerical simulation of the water level and flow field of the test river section, thereby solving the problems of no stable correspondence between water level and flow under complex water flow conditions, and inaccurate estimation of cross-sectional area using a single water level value for non-planar water surfaces. Based on video recognition, a more convenient real-time monitoring method for water level, flow velocity and flow is proposed to solve the problems of high cost of measuring equipment, complex operation and inability to provide water level and flow data in a timely manner, while providing data guarantee for parameter calibration of the three-dimensional hydrodynamic model. By coupling an artificial intelligence model, rapid calculation of flow is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 A schematic diagram of a flow measurement process in a flow measurement method under complex water flow conditions provided by an embodiment of the present invention;

[0041] Figure 2 A schematic diagram of modeling area and cross-section configuration of a flow measurement method under complex water flow conditions provided by an embodiment of the present invention;

[0042] Figure 3 A schematic diagram of surface velocity calculation of a flow measurement method under complex water flow conditions provided by an embodiment of the present invention;

[0043] Figure 4 A schematic diagram of flow calculation of a flow measurement section of a flow measurement method under complex water flow conditions provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0044] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0045] To solve the problems of the prior art, the present invention provides a flow measurement method under complex water flow conditions. Figure 1 As shown, the method comprises the following steps:

[0046] Step 1: According to the flow measurement section position, determine the three-dimensional hydrodynamic model modeling area and the model inlet and outlet positions and verification positions, and establish a three-dimensional hydrodynamic model;

[0047] Specifically, the flow measurement section position includes the following steps: according to the water flow conditions of the river section where the flow measurement section is located, a numerical model modeling area is selected, and the model inlet position, outlet position and verification position are determined.

[0048] Among them, specific references are Figure 2 As shown, the determined model inlet and outlet positions are implemented by the following steps:

[0049] The model inlet section is selected upstream of the flow measurement section and where the flow state is stable, including the main inflow section, tributary inflow section, etc.

[0050] The model outlet section is selected downstream of the flow measurement section and where the water flow is stable, including the main stream outflow section, tributary outflow section, etc.

[0051] according to Figure 2 In the figure, a verification section is set between the inlet section and the outlet section. There are two verification sections. A flow measurement section is set between the two verification sections. In the figure, the direction of the arrow indicates the direction of water flow. To ensure the accuracy of the measurement, model outlet sections and model inlet sections must be set at the inlets and outlets of the tributaries and mainstream of the river.

[0052] According to another embodiment of the present invention, the three-dimensional hydrodynamic modeling area and the determined model inlet and outlet positions and verification positions are determined by numerical simulation, and then also include: collecting underwater terrain and coastline data of the modeling area, and performing three-dimensional grid division on the modeling area; setting model boundary conditions and initial conditions.

[0053] Specifically, according to the modeling scope and the determined import and export locations, the underwater terrain and shoreline data of the modeling area are collected, and the modeling area is divided into three-dimensional grids.

[0054] The setting of the model boundary conditions includes the setting of the inlet boundary conditions and the setting of the outlet boundary conditions, wherein:

[0055] Inlet boundary condition: given the flow monitoring process Q at the inlet m,t Or water level monitoring process Z m,t .

[0056] Outlet boundary condition: given the flow monitoring process Q at the outlet n,t Or water level monitoring process Z n,t .

[0057] Wherein, m = 1, 2, ..., M; n = 1, 2, ..., N; t = 1, 2, ..., T; M is the total number of import positions; N is the total number of export positions; T is the total calculation time of the model, the same below.

[0058] The boundary initial conditions are set by cold start or hot start, wherein:

[0059] 1) Cold start: outlet water level at t=0 and zero flow rate, where

[0060] 2) Hot start: The model results after cold start operation is stabilized are used as the initial conditions.

[0061] The three-dimensional hydrodynamic model is established by selecting a three-dimensional hydrodynamic numerical model for hydrodynamic calculation, the three-dimensional hydrodynamic model uses the NS equation as the control equation, a σ coordinate system is used to fit the irregular bottom terrain in the vertical direction, a widely used Mellor-Yamada 2.5-order mode turbulence model is used for closure, a non-overlapping unstructured triangular grid is used horizontally to spatially discretize the calculation area to fit the shoreline, and a finite volume discretization method is used for solving.

[0062] Step 2: Using image recognition technology, obtain water level, surface velocity and boundary section flow data of boundary sections and verification locations;

[0063] It should be noted that, before the step 2, the step further includes: obtaining boundary condition data; the boundary condition data obtaining specifically includes configuring the import and export boundary sections; further,

[0064] The boundary section configuration is manifested as configuring the water level and flow measurement section corresponding to the river section inlet; configuring the water level and flow measurement section corresponding to the river section outlet.

[0065] The method of obtaining the water level, surface velocity and cross-sectional flow data of the boundary section and the verification position comprises the following steps:

[0066] The shoreline images at the inlet and outlet are collected by cameras to construct a water surface line pixel-elevation relationship dataset. The water level values ​​corresponding to the pixels of the water level line are obtained by the water area shoreline segmentation algorithm, and the average water level value of the cross section is obtained by the spatiotemporal averaging method.

[0067] The surface image of the cross-section water flow is collected by the camera, and the ripples, floating objects, and bubble features in the water surface image are extracted. The speed measurement line is drawn along the direction of the cross-section water flow. The length of the speed measurement line is used as the horizontal axis and the video acquisition time is used as the vertical axis to synthesize the space-time image and calculate the cross-section surface flow monitoring data.

[0068] Obtaining surface velocity includes: For details, please refer to Figure 3 As shown in the figure, the water surface image is collected by the camera, and the features such as ripples, floating objects, bubbles, etc. in the water surface image are extracted. The spatiotemporal image is synthesized in the image space and sequence time. The spatiotemporal image has significant texture features. The angle between the main direction of the texture and the time axis represents the surface flow velocity information. In river velocity measurement, a velocity line is drawn along the downstream direction. The distance that the characteristic quantity of movement moves along the velocity line within time T is S. At the same time, the pixel point moves j pixels within l frames. The relationship between the size v of the pixel velocity vector along this direction and the texture features in the spatiotemporal image can be expressed as:

[0069]

[0070] Among them, v i is the surface velocity of the ith velocity measurement line, S x is the spatial scale represented by pixels, S t is the time interval between each frame of the image, and tanθ is the tangent value of the slope of the space-time image.

[0071] 2) Obtaining boundary section flow data includes: selecting the surface velocity coefficient η s , the surface velocity v i Converted to vertical average velocity

[0072]

[0073] The cross-sectional flow rate Q is calculated using the velocity area method. m,t and Q n,t ;

[0074]

[0075] In the formula, represents the average velocity of the i-th measuring line, A i Represents the cross-sectional area of ​​interval i.

[0076] It should be noted that the boundary section flow data includes Q m,t , Q n,t , where Q m,t represents the flow monitoring process at the inlet, Q n,t represents the flow monitoring process at the outlet, Q m,t , Q n,t It can be obtained by the above formula (2) and formula (3).

[0077] Step 3: calibrate and verify the parameters of the established three-dimensional hydrodynamic model using the water level, surface velocity and boundary section flow data of the boundary section and the verification position;

[0078] In this embodiment, model verification and parameter calibration are performed through the established three-dimensional hydrodynamic model, and three-dimensional hydrodynamic simulation calculations are performed using boundary section flow and water level data as boundaries. The water level and surface flow velocity monitoring data of the verification section are used as verification data to calibrate and verify the three-dimensional hydrodynamic model parameters.

[0079] The three-dimensional hydrodynamic model verification includes: verifying the section configuration, obtaining the verification section water level monitoring process, obtaining the verification section surface flow velocity monitoring process, performing three-dimensional hydrodynamic model calculation, extracting the three-dimensional hydrodynamic model results and verifying the section calculation results and comparing them with the monitoring data; wherein,

[0080] Verification section configuration: its specific performance is to configure the water flow verification section in the river section. Model verification sections should be selected from important sections within the modeling range, including the confluence of main and tributary rivers, upstream and downstream of hydraulic structures, etc., and should be evenly distributed.

[0081] Acquire the verification section water level monitoring process; collect shoreline images at the inlet and outlet through the camera, construct a water surface line pixel-elevation relationship dataset, obtain the water level value corresponding to the pixel of the water level line through the water area shoreline segmentation algorithm, and use the spatiotemporal averaging method to obtain the average water level value of the section, and obtain Z k,t (k=1,2,…,K, K is the number of verification sections).

[0082] Acquire the surface velocity monitoring process of the verification section; collect water surface images through the camera, extract ripples, floating objects, bubbles and other features in the water surface image, and synthesize the space-time image in the image space and sequence time. The space-time image has significant texture features. The angle between the main direction of the texture and the time axis represents the surface velocity information, such as Figure 4 As shown in the figure, in river speed measurement, a speed measurement line is drawn along the downstream direction. The distance that the moving feature moves along the speed measurement line within time T is S. At the same time, the pixel point moves j pixels within l frames. The relationship between the size v of the pixel velocity vector along this direction and the texture feature in the spatiotemporal image can be expressed as

[0083]

[0084] Where S x is the spatial scale represented by pixels, S t is the time interval of each frame image, tanθ is the tangent value of the slope of the space-time image, which can be obtained by formula (4): Among them, i=1,2,…,I k , I k is the number of velocity measurement lines of the k-verification section.

[0085] Carry out three-dimensional hydrodynamic model verification calculation; its specific performance is based on the established three-dimensional hydrodynamic model, using the water level and flow monitoring data Q of the river section inlet and outlet sections m,t , Q n，t , Z m,t , Z n,t Drive the 3D hydrodynamic model calculation to obtain the water level and surface velocity at any location within the modeling area.

[0086] Extraction of three-dimensional hydrodynamic model results; Extraction of model verification section based on verification section Model calculation of water level process Z′ k,t , flow rate process v′ k,t , according to the velocity line position extraction model corresponding to the velocity line flow velocity calculation process

[0087] Compare the calculation results of the verification section with the monitoring data; the water level monitoring process Z at the water flow verification section k,t , Surface velocity monitoring process The water level process Z′ at the same location as calculated by the model k,t and surface velocity process For comparison, the model accuracy was calculated using indicators such as root mean square error, mean absolute error and Nash coefficient.

[0088] Root mean square error formula:

[0089]

[0090]

[0091] Mean absolute error:

[0092]

[0093]

[0094] Nash coefficient:

[0095]

[0096]

[0097]

[0098]

[0099] In the formula, RMSE z,k is the root mean square error of the water level in section k, RMSE v,k is the root mean square error of the surface velocity at section k, MAE z,k is the mean absolute error of the water level at section k, MAE v,k is the mean absolute error of flow velocity in section k, E Z,k is the Nash coefficient of the water level at section k, E v,k is the Nash coefficient of flow velocity in section k, Z k,t is the water level monitoring process of section k, Z′ k,t Calculate the water level process for the k-section model. is the surface velocity monitoring process of the velocity measuring line i at section k, is the process of calculating the surface velocity of the velocity line model of section k i, T is the total calculation time of the model (t=1,2,…,T), is the mean value of the water level monitoring process of section k, is the mean value of the surface velocity monitoring process of velocity measuring line i at section k.

[0100] The three-dimensional hydrodynamic model parameter calibration includes: calibrating the model parameters using a trial and error method until the three-dimensional hydrodynamic model calculation results meet the accuracy requirements.

[0101] Step 4: simulate the calibrated three-dimensional hydrodynamic model to obtain the flow rate of the measured section;

[0102] It should be noted that the simulation of the calibrated three-dimensional hydrodynamic model to obtain the measured cross-sectional flow includes the following steps:

[0103] 1) Model calculation: Based on the established three-dimensional hydrodynamic model and the parameters obtained by calibration, the model calculation is driven by the real-time monitoring water level data of the inlet section and the outlet section to obtain the water level and three-dimensional flow velocity calculation results at any position in the modeling area.

[0104] 2) Flow calculation: According to the flow velocity calculation results of the flow measurement section, the flow velocity of the entire section is integrated. For details, please refer to Figure 4 As shown, the coordinate axes are established with the y-axis and the z-axis. In the figure, Z b Indicates the riverbed bottom elevation of the flow measurement section; Z s Indicates the water surface elevation at time t of the flow measurement section, Q t Indicates the instantaneous flow at the flow measurement section at time t, u yz，t Indicates the flow velocity of the fluid at the cell (y,z) of the river section at time t; calculates the flow Q of the flow measurement section t , the calculation formula is as follows:

[0105] Q t =∫∫u yz,t dydz (13)

[0106] Step 5: Using the measured section flow as output, build an LSTM artificial intelligence model and training and test sets, and train the LSTM model algorithm.

[0107] In this embodiment, the LSTM artificial intelligence model, training set, and test set are constructed with the measured cross-sectional flow as output, and the training LSTM model algorithm specifically includes the following steps:

[0108] First, an LSTM model is established for the flow measurement section. The output layer of the model is the flow Q of the measurement section. t The model input layer is the water level and flow rate at the inlet and outlet sections and the water level and surface velocity monitoring process at the verification section, that is, Z m,t , Z n,t , Q m,t , Q n,t , Z k,t ,

[0109] Secondly, the image recognition method is used to obtain the water level, flow rate, and surface velocity monitoring process Z at the inlet and outlet sections and the verification section. m,t , Z n,t , Q m,t , Q n,t , Z l,t , And the cross-sectional flow Q calculated by the three-dimensional hydrodynamic model t , build artificial intelligence model training and test sets;

[0110] Then, the training set and test set are used to train, test and verify the artificial intelligence model, and the model accuracy is evaluated using indicators such as root mean square error, mean absolute error and Nash coefficient until the model accuracy meets the flow measurement accuracy requirements.

[0111] Step 6: Use the trained LSTM model to perform fast calculations to obtain real-time traffic measurement data.

[0112] Specifically, the use of the trained LSTM model for rapid calculation to obtain real-time flow measurement data specifically includes: using the water level, flow, and surface flow velocity monitoring process data Z at the inlet and outlet sections and the verification section m,t , Z n,t , Q m,t , Q n,t , Z k,t , The artificial intelligence model that has passed training and testing is used to calculate the real-time flow of the flow measurement section, realizing real-time flow measurement under complex water flow conditions.

[0113] At the same time, an embodiment of the present invention further provides a flow measurement system under complex water flow conditions, the system is applied to the flow measurement method under complex water flow conditions described above, wherein the system includes:

[0114] Model building module: used to determine the modeling area of ​​the three-dimensional hydrodynamic model and the model inlet and outlet positions and verification positions according to the flow measurement section position, and establish the three-dimensional hydrodynamic model;

[0115] Model processing module: used to obtain water level, surface velocity and cross-sectional flow data of boundary sections and verification positions using image recognition technology;

[0116] Model verification module: used to calibrate and verify the parameters of the established three-dimensional hydrodynamic model using the water level, surface velocity and boundary section flow data of the boundary section and verification position;

[0117] Model simulation module: used to simulate the calibrated three-dimensional hydrodynamic model to obtain the measured cross-sectional flow;

[0118] Model training module: used to build an LSTM artificial intelligence model for measuring cross-section flow, build artificial intelligence model training sets and test sets, and train the LSTM model algorithm;

[0119] Model calculation module: used to use the trained LSTM model for fast calculation to obtain real-time traffic measurement data.

[0120] It can be understood that the flow measurement system under complex water flow conditions provided by the present invention corresponds to the flow measurement method under complex water flow conditions provided by the aforementioned embodiments. The relevant technical features of the flow measurement system under complex water flow conditions can refer to the relevant technical features of the flow measurement method under complex water flow conditions, which will not be repeated here.

[0121] In summary, the present invention establishes a three-dimensional hydrodynamic model to numerically simulate the water level and flow field of the test river section, thereby solving the problems of no stable correspondence between water level and flow under complex water flow conditions, inaccurate estimation of cross-sectional area using a single water level value for non-planar water surfaces, etc.; a more convenient real-time monitoring method for water level, flow velocity, and flow is proposed based on video recognition to solve the problems of high cost of measuring equipment, complex operation, and inability to provide flow data in a timely manner; and rapid calculation of flow is achieved by coupling an artificial intelligence model.

[0122] The protection scope of the present invention is not limited to the above examples. Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the scope and spirit of the present invention. If these changes and modifications fall within the scope of the claims of the present invention and their equivalents, the present invention is intended to include these changes and modifications.

Claims

1. A flow measurement method under complex water flow conditions, the method comprising the following steps: According to the flow measurement section location, determine the three-dimensional hydrodynamic modeling area and the model inlet and outlet locations and verification locations, collect underwater topography and shoreline data in the modeling area, and perform three-dimensional meshing of the modeling area; The boundary conditions and initial conditions of the three-dimensional hydrodynamic model are set to establish a three-dimensional hydrodynamic model; the three-dimensional hydrodynamic model uses the NS equation as the control equation and adopts The coordinate system is fitted to the irregular bottom terrain, and the calculation area is spatially discretized to fit the shoreline using non-overlapping unstructured triangular grids horizontally, and the finite volume discretization method is used to solve it; the determination of the model inlet and outlet positions is implemented by the following steps: The model inlet section is selected upstream of the flow measurement section and where the flow pattern is stable, including the main inflow section and the tributary inflow section; The model outlet section is selected downstream of the flow measurement section and where the flow pattern is stable, including the main stream outflow section and the tributary outflow section; Use image recognition technology to obtain water level, surface velocity and flow data of boundary sections and verification positions; including: collect shoreline images at the inlet and outlet through cameras, construct a water surface line pixel-elevation relationship data set, obtain the water level value corresponding to the pixel of the water level line through the water area shoreline segmentation algorithm, and use the spatiotemporal averaging method to obtain the average water level value of the boundary section; collect surface images of cross-section water flow through cameras, extract ripples, floating objects, and bubble features in the water surface image, draw speed measurement lines along the cross-section water flow direction, use the length of the speed measurement line as the horizontal axis and the video acquisition time as the vertical axis, synthesize the spatiotemporal image, and calculate the cross-section surface velocity monitoring data; The established three-dimensional hydrodynamic model is calibrated and verified using the acquired boundary section and verification position water level, surface velocity and boundary section flow data; including: calibrating the model parameters using a trial and error method until the three-dimensional hydrodynamic model calculation results meet the accuracy requirements; the three-dimensional hydrodynamic model verification includes: verifying section configuration, obtaining the verification section water level monitoring process, obtaining the verification section surface velocity monitoring process, performing three-dimensional hydrodynamic model calculation, extracting the three-dimensional hydrodynamic model results and comparing the verification section calculation results with the monitoring data; The calibrated three-dimensional hydrodynamic model is simulated to obtain the flow of the measured section; including: based on the established three-dimensional hydrodynamic model and the parameters obtained by calibration, the three-dimensional hydrodynamic model calculation is driven by the real-time monitoring of the water level and flow data of the inlet section and the outlet section to obtain the water level and three-dimensional flow velocity calculation results at any position in the modeling area; according to the three-dimensional flow velocity calculation results at the flow measurement section position, the flow velocity of the entire section is integrated to obtain the flow of the measured section; Taking the flow rate of the measured section as the output, constructing the LSTM artificial intelligence model and the training set and the test set, and training the LSTM model algorithm; including: establishing the LSTM artificial intelligence model for the flow measurement section; the output layer of the LSTM artificial intelligence model is the flow rate of the measured section, and the model input layer is the water level and flow rate of the inlet and outlet sections and the water level and surface flow velocity monitoring process at the verification section; Use image recognition technology to obtain the water level, flow rate, and surface velocity monitoring process at the inlet and outlet sections and the verification section, as well as the cross-sectional flow rate calculated by the three-dimensional hydrodynamic model, to build the artificial intelligence model training set and test set; Use the constructed training set and test set to train, test and verify the artificial intelligence model, and use the root mean square error, mean absolute error and Nash coefficient indicators to evaluate the model accuracy until the model accuracy meets the flow measurement accuracy requirements; The trained LSTM model is used for fast calculation to obtain real-time flow measurement data; including: using the water level, flow, and surface velocity monitoring process data at the inlet and outlet sections and the verification section, and using the artificial intelligence model that has passed the training test verification to calculate the real-time flow of the flow measurement section, so as to realize real-time measurement of flow under complex water flow conditions.

2. A flow measurement system under complex water flow conditions, the system is applied to the flow measurement method under complex water flow conditions as claimed in claim 1, characterized in that: include: Model building module: used to determine the modeling area of ​​the three-dimensional hydrodynamic model and the model inlet and outlet locations and verification locations based on the flow measurement section location, collect underwater terrain and shoreline data in the modeling area, and perform three-dimensional grid division on the modeling area; The boundary conditions and initial conditions of the three-dimensional hydrodynamic model are set to establish a three-dimensional hydrodynamic model; the three-dimensional hydrodynamic model uses the NS equation as the control equation and adopts The coordinate system is fitted to the irregular bottom terrain, and the calculation area is spatially discretized to fit the shoreline using non-overlapping unstructured triangular grids horizontally, and the finite volume discretization method is used to solve it; the determination of the model inlet and outlet positions is implemented by the following steps: The model inlet section is selected upstream of the flow measurement section and where the flow pattern is stable, including the main inflow section and the tributary inflow section; The model outlet section is selected downstream of the flow measurement section and where the flow pattern is stable, including the main stream outflow section and the tributary outflow section; Model processing module: used to use image recognition technology to obtain water level, surface flow velocity and cross-sectional flow data of boundary sections and verification positions; including: collecting shoreline images at the inlet and outlet through cameras, constructing a water surface line pixel-elevation relationship data set, obtaining the water level value corresponding to the pixel of the water level line through the water area shoreline segmentation algorithm, and using the spatiotemporal averaging method to obtain the average water level value of the boundary section; collecting cross-sectional water flow surface images through cameras, extracting ripples, floating objects, and bubble features in the water surface images, drawing speed measurement lines along the cross-sectional water flow direction, using the speed measurement line length as the horizontal axis and the video acquisition time as the vertical axis, synthesizing the spatiotemporal image, and calculating the cross-sectional surface flow velocity monitoring data; Model verification module: used to calibrate and verify the parameters of the established three-dimensional hydrodynamic model using the water level, surface velocity and section monitoring data of the boundary section and the verification position; including: calibrating the model parameters using the trial and error method until the calculation results of the three-dimensional hydrodynamic model meet the accuracy requirements; the three-dimensional hydrodynamic model verification includes: verifying the section configuration, obtaining the verification section water level monitoring process, obtaining the verification section surface velocity monitoring process, performing three-dimensional hydrodynamic model calculation, extracting the three-dimensional hydrodynamic model results and comparing the verification section calculation results with the monitoring data; Model simulation module: used to simulate the calibrated three-dimensional hydrodynamic model to obtain the flow rate of the measured section; including: based on the established three-dimensional hydrodynamic model and the parameters obtained by calibration, using the real-time monitoring water level and flow data of the inlet section and outlet section to drive the three-dimensional hydrodynamic model calculation, and obtain the water level and three-dimensional flow velocity calculation results at any position in the modeling area; according to the three-dimensional flow velocity calculation results at the flow measurement section position, the flow velocity of the entire section is integrated to obtain the flow rate of the measurement section; Model training module: used to build LSTM artificial intelligence model and training set and test set with the flow rate of the measured section as output, and train the LSTM model algorithm; including: establishing LSTM artificial intelligence model for the flow measurement section; the output layer of the LSTM artificial intelligence model is the flow rate of the measured section, and the model input layer is the water level and flow rate of the inlet and outlet sections and the water level and surface flow velocity monitoring process at the verification section; Use image recognition technology to obtain the water level, flow rate, and surface velocity monitoring process at the inlet and outlet sections and the verification section, as well as the cross-sectional flow rate calculated by the three-dimensional hydrodynamic model, to build the artificial intelligence model training set and test set; Use the constructed training set and test set to train, test and verify the artificial intelligence model, and use the root mean square error, mean absolute error and Nash coefficient indicators to evaluate the model accuracy until the model accuracy meets the flow measurement accuracy requirements; Model calculation module: used to quickly calculate the trained LSTM model to obtain real-time flow measurement data; including: using the water level, flow, and surface velocity monitoring process data at the inlet and outlet sections and the verification section, and using the artificial intelligence model that has passed the training test verification to calculate the real-time flow of the flow measurement section, so as to realize real-time measurement of flow under complex water flow conditions.

Images