Simulation method of river regime under influence of water conservancy project with lack of data

Abstract

This invention relates to a method for simulating river hydrological conditions in areas lacking data due to the influence of water conservancy projects. It is applicable to the field of hydrology, particularly to scenarios where underwater topographic data is lacking and river hydrological conditions (water level, flow rate) are affected by water conservancy project scheduling. The method includes: acquiring river boundary condition data for the study section under the influence of water conservancy project operation; preprocessing the upstream section boundary conditions based on downstream boundary conditions and undetermined coefficients; employing multiple machine learning algorithms based on one-dimensional river control equations to simulate and calculate the water level and flow rate at the stations to be simulated within the study section; measuring the accuracy of the simulation results of each machine learning algorithm, using the accuracy as a weight to iteratively update the upstream section boundary conditions; selecting the combination of undetermined coefficient values ​​with the best simulation accuracy within a period of time; and then selecting the simulation result with the highest accuracy within the period of time from the simulation results of multiple iterations to determine the optimal river hydrological simulation model.

Description

Technical Field

[0001] This invention relates to a method for simulating river conditions in the absence of data under the influence of water conservancy projects. It is applicable to the field of hydrology, and is particularly applicable to simulation scenarios of river conditions (water level, flow) in the absence of underwater topographic data due to the influence of water conservancy project scheduling. Background Technology

[0002] Driven by global climate change, extreme hydrological events, including flash floods, sudden droughts, and rapid shifts between drought and flood, have become significantly more severe, posing a major threat to human life and infrastructure. Furthermore, the recent surge in water conservancy construction has also had a significant impact on river water conditions. Effectively understanding the patterns of river water condition changes under the dual influence of climate change and water conservancy projects urgently requires efficient and accurate hydrodynamic simulation methods.

[0003] Accurately understanding the characteristics of underwater topography is a prerequisite for establishing hydrodynamic models and conducting effective numerical simulations of river hydrodynamic processes. However, due to limitations in funding, technology, manpower, and resources for underwater topographic surveying, current underwater topographic surveys of rivers suffer from problems such as low update frequency, narrow coverage, and large measurement errors, which are not conducive to conducting numerical simulations of hydrodynamic processes through the establishment of hydrodynamic models. To address this issue, employing machine learning to conduct numerical simulations of hydrodynamic processes in rivers with incomplete underwater topographic data has become a popular solution.

[0004] According to literature review, currently used machine learning methods mainly establish complex neural networks to fit complex nonlinear functions and use implicit expressions to characterize the response of river water conditions to major influencing factors such as upstream inflow, tributary confluence, inter-regional confluence, and water conservancy project scheduling. This approach mainly lacks a connection to physical laws. First, neural networks are extremely complex, making it difficult to understand the processing procedures of the neural network on the input data. Second, the normalization of the input data by the neural network removes the original physical dimensions of the data, making it impossible to analyze the physical meaning of the complex calculations of the neural network. Third, the establishment of neural networks is difficult to utilize existing experience in river hydrodynamic simulation, requiring a high level of expertise. Summary of the Invention

[0005] The technical problem to be solved by this invention is to provide a method for simulating river conditions in the absence of data under the influence of water conservancy projects, in view of the above-mentioned problems.

[0006] The technical solution adopted in this invention is: a method for simulating river hydrological conditions in the absence of data under the influence of water conservancy projects, comprising: Data on river boundary conditions of the study section under the influence of water conservancy project operation are obtained. Based on the downstream boundary conditions and undetermined coefficients, the boundary conditions of the upstream section are preprocessed, and multiple sets of undetermined coefficients are set. Based on the pre-processed upstream cross-section boundary conditions, various machine learning algorithms based on one-dimensional river control equations are used to simulate and calculate the water level and flow rate of the stations to be simulated in the research river section. The simulation results of each machine learning algorithm are measured. The simulation results accuracy is used as the weight to iteratively update the boundary conditions of the upstream section. The iteratively updated boundary conditions of the upstream section are then re-substituted into the various machine learning algorithms for simulation. The optimal combination of undetermined coefficient values ​​with the best simulation results within the periodic screening is selected. Then, the simulation results with the highest accuracy within the periodic screening are selected from the simulation results of multiple iterations. The simulation results are then optimized a second time by combining the water level-discharge relationship of the station to be simulated within the periodic screening, thereby determining the optimal river hydrological simulation calculation model.

[0007] By employing the aforementioned technical means and adopting a machine learning algorithm based on physical equations, the black-box model of traditional neural networks is broken away from, ensuring the physical interpretability of the simulation process and results. Through a multi-layered design involving the selection of undetermined coefficients, boundary condition iteration, and secondary optimization of results, the simulation accuracy is significantly improved. The final optimal simulation model can be directly used for hydrological simulation during the forecast period, achieving efficient reuse. It does not rely on underwater topographic data and can be simulated using only easily accessible river boundary condition data, thus solving the core pain point of lacking topographic data.

[0008] As a preferred embodiment, the preprocessing of the upstream section boundary conditions based on downstream boundary conditions and undetermined coefficients includes: ; ; in, , These are the upstream section water level and flow rate after pretreatment; , These are the measured water level and flow rate at the upstream section; , These are the measured water level and flow rate at the downstream section; , These are the upstream section of the station to be simulated, and the tributary flow between the station to be simulated and the downstream section; , These are the flow rates between the upstream section and the station to be simulated, and between the station to be simulated and the downstream section, respectively. , These are undetermined coefficients.

[0009] By employing the aforementioned technical methods, the interference of tributary confluence and inter-regional confluence on upstream water flow is eliminated. By segmenting the confluence, the actual hydrological characteristics of the river channel are accurately depicted, avoiding simulation errors caused by the general superposition of confluence data. A linear adjustment method is adopted instead of traditional normalization processing to fully preserve the original physical dimensions of water level and flow, ensuring the physical interpretability of subsequent simulations.

[0010] As a preferred embodiment, the one-dimensional river channel control equation is the difference form of the momentum conservation equation in the Saint-Venant equations.

[0011] As a preferred embodiment, one of the various machine learning algorithms based on one-dimensional river control equations includes: ; A water level calculation formula is constructed based on the difference form of the momentum conservation equation in the Saint-Venant equations, and a time step is introduced. Spatial distance from upstream section to the site to be simulated Changes in upstream water level after pretreatment and undetermined coefficients , , ,Regulation , After determining the range of values, the optimal value is determined by sliding learning combined with least squares fitting within the pass rate period. Solve for the water level change at the simulated site. Then, the water level Z at the site to be simulated is calculated; ; Based on the water level Z at the station to be simulated, an identity matrix E and a column vector of coefficients to be solved are introduced. Construct a flow recursive formula and estimate the parameters using known parameters within a periodic period to obtain the utilization rate. The traffic Q of the site to be simulated is calculated recursively.

[0012] Through the aforementioned technical means, the water level and flow rate of the simulated station can be accurately calculated step by step using recursive methods. With the change in water level as the core variable to be solved, incremental calculations are used to reduce simulation errors and achieve continuous recursion of water level, which conforms to the physical laws of dynamic changes in river water conditions. The flow rate recursion formula is based on the inherent physical relationship between water level and flow rate, and a linear equation is constructed by combining the identity matrix to achieve rapid recursion of flow rate. This complements the water level solution and completes the continuous simulation of "water level-flow rate". All parameters are easily obtainable hydrological and geographical parameters (Δt, Δx, etc.), and no topographic data is required.

[0013] As a preferred embodiment, one of the various machine learning algorithms based on one-dimensional river control equations includes: ; ; A linear approximation of water level and flow rate is made based on the difference form of the momentum conservation equation in the Saint-Venant equations, introducing the identity matrix E and the column vector of coefficients to be solved. Construct a flow recursive formula and estimate the parameters using known parameters within a periodic period to obtain the utilization rate. ； Introducing the identity matrix E and the column vector of coefficients to be solved Construct a water level estimation formula. The parameters are obtained by estimating the known parameters within the rate period;

[0014] By combining the flow rate recursion formula and the water level estimation formula, and taking the water level at time n+1... ,flow The unknowns are the water level and flow rate of the station to be simulated.

[0015] Through the above-mentioned technical means, the water level and flow rate of the simulated station can be solved simultaneously and quickly. An independent water level estimation formula is constructed and combined with the flow rate recursion formula to form a system of two linear equations, realizing the simultaneous solution of water level and flow rate. Compared with step-by-step recursion, the calculation speed is greatly improved, which is suitable for the needs of rapid simulation. The two algorithms simulate water conditions from different perspectives (step-by-step recursion and simultaneous solution), providing a basis for subsequent error iteration and result selection, and further improving the simulation accuracy.

[0016] As a preferred embodiment, the method for measuring the accuracy of the simulation results of each machine learning algorithm includes: using the Nash efficiency coefficient as an error index to measure the accuracy of the simulation results of each machine learning algorithm based on the simulation results of each machine learning algorithm.

[0017] As a preferred embodiment, the iterative update of the upstream section boundary conditions, weighted by the accuracy of the simulation results, includes: ; ; in, , This represents the upstream cross-section water level and flow rate after iteration; , This indicates the upstream cross-sectional water level and flow rate before the iteration; , The s-th machine learning algorithm represents the simulated water level and flow rate of the station to be simulated; n represents the current time step. Let represent the Nash efficiency coefficient of the simulation results obtained by the s-th machine learning algorithm.

[0018] By employing the aforementioned technical means and using the Nash efficiency coefficient as a weight, the algorithm achieves the goal of "the higher the accuracy of the algorithm, the greater its contribution to the iterative boundary conditions." This ensures that the iterative process has a clear accuracy orientation rather than blind adjustments, thereby guaranteeing that the boundary conditions after iteration are more closely aligned with the actual water conditions.

[0019] As a preferred embodiment, the secondary optimization of the simulation results by combining the water level-flow relationship within the period of the simulated site includes: Based on the water level-flow relationship, another simulated value is re-estimated using a more accurate simulated water level or flow rate. The estimated result is compared with the original simulation result, and the more accurate result is selected as the final simulation result.

[0020] By employing the aforementioned technical means and utilizing the inherent physical correlation between water level and flow rate at river cross-sections, the simulation results are corrected to further reduce simulation errors and improve simulation accuracy. By comparing the estimated results with the original simulation results, it is ensured that the final output simulation results are the most accurate values, providing a high-quality foundation for subsequent model reuse and prediction period simulations.

[0021] A device for simulating river hydrological conditions in areas lacking data due to the impact of water conservancy projects, comprising: The data acquisition and preprocessing module is used to acquire river boundary condition data of the study river section under the influence of water conservancy project operation. Based on the downstream boundary conditions and undetermined coefficients, the upstream section boundary conditions are preprocessed, where multiple sets of undetermined coefficients are set to select values. Multiple algorithm simulation modules are used to simulate and calculate the water level and flow rate of the stations to be simulated in the study section based on the pre-processed upstream cross-section boundary conditions and various machine learning algorithms based on one-dimensional river control equations. The boundary condition iteration module is used to measure the accuracy of the simulation results of each machine learning algorithm. Using the accuracy of the simulation results as the weight, the boundary conditions of the upstream section are iteratively updated, and the iterated boundary conditions of the upstream section are resubmitted into the various machine learning algorithms for simulation. The results screening and optimization module is used to screen the combination of undetermined coefficient values ​​with the best accuracy of the simulation results within the periodicity period. Then, it selects the simulation results with the highest accuracy within the periodicity period from the simulation results of multiple iterations. Combined with the water level-flow relationship of the station to be simulated within the periodicity period, the simulation results are optimized a second time to determine the optimal river hydrological simulation calculation model.

[0022] A storage medium on which a computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the method for simulating river hydrological conditions in the absence of data under the influence of a water conservancy project.

[0023] The beneficial effects of this invention are: This invention does not rely on underwater topographic data of the river channel. It can realize hydrodynamic numerical simulation only through easily obtainable river boundary condition data (upstream / downstream cross-sectional water level and flow, tributary / interval confluence flow), one-dimensional river control equations and machine learning algorithms. It completely gets rid of the dependence on underwater topographic data, adapts to the simulation needs of rivers lacking data, and at the same time reduces the financial, technical and human costs brought about by underwater topographic surveying.

[0024] The various machine learning algorithms used in this invention are all based on one-dimensional river channel control equations, and all formulas and parameters have clear physical meanings. The boundary condition preprocessing adopts a linear adjustment method, which preserves the original physical dimensions of the data and does not perform normalization processing. The physical meaning of the calculation can be analyzed through hydrodynamic expertise. The simulation process (data preprocessing, dual-algorithm simulation, iterative optimization, and result screening) can be broken down step by step and is fully traceable, completely abandoning the black-box fitting of traditional neural networks and greatly improving the physical interpretability of the simulation process and results.

[0025] The core of the algorithm in this invention reuses the Saint-Venant equations, which are widely recognized in the field of hydrodynamic simulation. The optimization stage utilizes the basic physical relationship between river level and flow rate, directly incorporating existing hydrodynamic simulation experience. Parameter fitting (least squares method), equation solving (linear equations), and iterative updates (linear combination) all adopt simple and easy-to-operate mathematical methods, without the need for complex neural network structure design and hyperparameter tuning. The simulation process follows the conventional research approach of hydrodynamic numerical simulation, which can be quickly mastered by researchers in the field of hydrodynamics, greatly reducing the professional threshold for cross-disciplinary collaboration.

[0026] To address the characteristics of river water conditions under the influence of water conservancy projects, this invention achieves precise adaptation to the impact of water conservancy project scheduling by accurately characterizing the boundary conditions of downstream water conservancy projects and segmented confluence data. Through a multi-layered design of complementary dual algorithms, boundary condition iteration, parameter selection, and secondary optimization of results, the simulation accuracy is significantly improved. The lightweight iterative mechanism and linearized algorithm design significantly improve computational efficiency compared to traditional neural networks, enabling rapid completion of water condition simulation and model optimization to meet the needs of actual engineering applications. Attached Figure Description

[0027] Figure 1 This is an overall flowchart of the method for simulating river hydrological conditions in the absence of data under the influence of water conservancy projects, as shown in the example.

[0028] Figure 2 This is a schematic diagram of the river section studied in the example.

[0029] Figure 3 The water level and flow rate simulation results of station 1 (Cuntan Station) in the embodiment are shown.

[0030] Figure 4The simulation results of water level and flow rate at station 2 (Qingxi Station) in the example are shown.

[0031] Figure 5 The water level and flow rate simulation results of station 3 (Wanxian Station) in the embodiment are shown. Detailed Implementation

[0032] The present invention will be further described in detail below with reference to specific embodiments. The following embodiments are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0033] This embodiment takes the Three Gorges Reservoir area (significantly affected by water conservancy projects and lacking underwater topographic data for some river sections) as the research area. The stations to be simulated are Cuntan Station, Qingxichang Station, and Wanxian Station. The downstream water conservancy project is the Three Gorges Dam, and the upstream section is Zhutuo Station.

[0034] The method for simulating river hydrological conditions in the absence of data under the influence of water conservancy projects in this embodiment includes the following steps:

[0035] S100. Obtain river boundary condition data for the study section (from Zhutuo Station to the Three Gorges Dam) under the influence of water conservancy project operation. Based on the downstream boundary conditions and undetermined coefficients, perform preprocessing on the upstream section boundary conditions, where multiple sets of undetermined coefficients are set.

[0036] This example obtains river boundary condition data for the section from Zhutuo Station (upstream section) to the Three Gorges Dam (downstream section) under the influence of the Three Gorges Dam's operation. Specifically, this includes: the measured water level at Zhutuo Station (upstream section). Actual flow rate Measured water level at the downstream section of the Three Gorges Dam Actual flow rate ; Segmented tributary flow (From Zhutuo Station to the station to be simulated) (Simulated station to Three Gorges Dam); Segmented interval confluence flow (From Zhutuo Station to the station to be simulated) (From the simulated site to the Three Gorges Dam), among which , The data were obtained through simulation using a distributed watershed hydrological model. Only the water level Z and flow rate Q data were collected for the stations to be simulated (Cuntan Station, Qingxichang Station, and Wanxian Station) during the periodic period. The remaining boundary condition data were collected simultaneously during the periodic period and the forecast period (the periodic period was used for model fitting, parameter estimation, and iterative optimization, while the forecast period was used to verify the model's predictive ability).

[0037] In this embodiment, the measured water level at Zhutuo Station was used. Actual flow rate Linear adjustment was performed to obtain the upstream cross-sectional water level after pretreatment. ,flow The specific formula is as follows: ; ; The preprocessing process did not normalize the data, thus preserving the water level (m) and flow rate intact. The original physical dimensions provide a guarantee for the physical interpretability of subsequent simulations.

[0038] This embodiment closely follows the hydrodynamic physical laws of the river channel, and sets multiple combinations of undetermined coefficients a1 and b1. Each combination is a linear adjustment coefficient, which is used to eliminate the interference of tributary confluence and inter-regional confluence on upstream water.

[0039] Based on the pre-processed upstream cross-section boundary conditions, two machine learning algorithms based on one-dimensional river control equations are used to simulate and calculate the water level and flow rate of the stations to be simulated in the research river section.

[0040] This step employs two machine learning algorithms based on the difference form of the momentum conservation equation in the Saint-Venant equations to simulate the water level and flow rate at the station to be simulated. The two algorithms run independently and complement each other, as detailed below: S210, Algorithm 1: Step-by-step recursive simulation S211. Water Level Calculation: Based on the difference form of the momentum conservation equation in the Saint-Venant equations, a formula for calculating the water level is constructed: ; Where Δt is the hydrological simulation time step, and Δx is the spatial distance from Zhutuo Station (upstream section) to each station to be simulated. The change in upstream water level after pretreatment within the time step. ΔZ represents the water level change at the simulated station within the time step. b′, a2, and b2 are undetermined coefficients.

[0041] First, the ranges of values ​​for a2 and b2 are defined. Then, using sliding learning within the rate period combined with least squares fitting, the possible values ​​of b′ are iterated to determine the optimal value that minimizes the simulation error of the water level at the simulated site. Then, ΔZ is obtained by solving for it, and combined with the initial water level, the water level Z at each time of the station to be simulated is calculated.

[0042] S212. Flow Rate Recursion: Based on the water level Z obtained from the station to be simulated, the identity matrix E and the column vector b3 of the coefficients to be solved are introduced to construct the flow rate recursion formula: ; Where the superscripts n and n+1 of the parameters represent the parameters at time n and n+1, respectively; E is the identity matrix; It is a column vector of coefficients to be solved, obtained by estimating the known parameters within the calibration period. Based on this formula, starting from the initial flow rate, the flow rate Q at each time point of the station to be simulated is calculated recursively to complete the continuous simulation of "water level-flow rate".

[0043] S220, Algorithm 2: Simulation of linear simultaneous solution

[0044] S221. Based on the difference form of the momentum conservation equation in the Saint-Venant equations, a linear approximation of water level and flow rate is made. The flow rate recursion formula in Algorithm 1 is reused, and the identity matrix E and the column vector b4 of the coefficients to be solved are introduced to construct the water level estimation formula: ; in, To solve for the column vector of coefficients to be solved for water level, and... They are independent of each other and are obtained by estimating parameters using known water level and flow rate data within a certain period.

[0045] S222, Solve the system of equations simultaneously: Combine the flow rate recursion formula and the water level estimation formula into a system of two linear equations, with the water level at time n+l as the equation. ,flow The unknown quantity is the boundary data after preprocessing, the historical water level / flow rate, the identity matrix E, and the coefficients. , All of these are known quantities. By solving the linear equation system, the water level Z and flow rate Q at each time point of the station to be simulated can be obtained directly, thus achieving synchronous and rapid solution of water level and flow rate.

[0046] S300. Measure the accuracy of the simulation results of each machine learning algorithm, use the accuracy of the simulation results as the weight, iteratively update the boundary conditions of the upstream section, and re-substitute the iterated boundary conditions of the upstream section into the various machine learning algorithms for simulation.

[0047] S310. The Nash efficiency coefficient (NSE) is used as an error index to measure the accuracy of the simulation results of Algorithm 1 and Algorithm 2. The Nash efficiency coefficient is calculated as follows: based on the observed water level / flow rate M at the station to be simulated and the average value of the observed values... Simulated value S, average simulated value Quantitative calculations were performed, and the NSE value range was [value missing]. The closer the value is to 1, the higher the simulation accuracy, which can objectively reflect the degree of fit between the simulated value and the measured value.

[0048] ; S320. Using the Nash efficiency coefficients of the two algorithms as weights, the boundary conditions of the upstream section ( , Perform linear combination iterations separately, with the following iteration formula: ; ; in, , This represents the upstream cross-section water level and flow rate after iteration; , This indicates the upstream cross-sectional water level and flow rate before the iteration; , The s-th machine learning algorithm represents the simulated water level and flow rate of the station to be simulated; n represents the current time step. Let represent the Nash efficiency coefficient of the simulation results obtained by the s-th machine learning algorithm.

[0049] S330, Apply the iterated upstream section boundary conditions ( , Substitute the results back into Algorithm 1 and Algorithm 2 for simulation, repeat the iteration 10-20 times, and record the Nash efficiency coefficient after each iteration to achieve dynamic optimization of boundary conditions and gradually reduce simulation errors.

[0050] S400, select the combination of undetermined coefficients with the best accuracy of the simulation results within the periodic screening, and then select the simulation results with the highest accuracy within the periodic screening from the simulation results of multiple iterations. Combine the water level-flow relationship of the station to be simulated within the periodic screening to perform secondary optimization of the simulation results, and then determine the optimal river hydrological simulation calculation model.

[0051] S410: Traverse the multiple sets of a1 and b1 value combinations set in step S100, combine the simulation accuracy (Nash efficiency coefficient) within the periodic rate, and select the a1 and b1 combination that can make the simulation accuracy of water level and flow rate of the station to be simulated optimal, and determine the optimal parameters for preprocessing.

[0052] S420. From the simulation results of multiple iterations, select the set of water level and flow simulation results with the highest Nash efficiency coefficient (optimal accuracy) within the rate period as the basic results.

[0053] S430. Utilizing the inherent correlation between water level and flow rate within the period of the simulated station, and based on the optimization logic of weight 8, if the water level simulation accuracy is higher than the flow rate simulation accuracy, then the flow rate simulation value is re-estimated using the water level simulation value with higher accuracy; if the flow rate simulation accuracy is higher than the water level simulation accuracy, then the water level simulation value is re-estimated using the flow rate simulation value with higher accuracy. By comparing the estimation results with the original simulation results, the result with higher accuracy is selected as the final simulation result, thus achieving secondary optimization of the simulation results.

[0054] S440 integrates the optimal combination of a1 and b1, all parameters such as b′, a2, b2, b3, and b4 that have been fitted, as well as the complete logic of dual-algorithm simulation, boundary iteration, and result optimization, to determine the optimal river hydrological simulation calculation model. This model can be directly used for hydrological simulation during the forecast period.

[0055] In this embodiment, when predicting the hydrological conditions during the forecast period, the optimal simulation model determined in step S400 is reused. There is no need to refit any parameters or repeat the iterations; only the river boundary condition data (Zhutuo Station) for the forecast period is used. , , , , , Three Gorges Dam , Substituting the values ​​into the preprocessing formula of step S100 (using the optimal combination of a1 and b1), we obtain the preprocessed values ​​for the prediction period. , Then, by substituting the values ​​into the dual algorithms in the optimal model, the water level and flow rate of the expected simulation station are calculated, the hydrological simulation during the prediction period is completed, and the predictive ability of the model is verified.

[0056] Taking the simulation case from 2012 to 2018 as an example, the collected time series is divided into a rate-setting period and a prediction period, namely 2012-2015 and 2016-2018. The simulation results of water level and flow at three stations—Cuntan, Qingxichang, and Wanxian—are shown below. Figures 3-5 As shown.

[0057] Overall, the flow simulation results effectively capture the daily flow trends of various river channels, with Nash efficiency coefficients (NSE) generally ranging from 0.97 to 1.00. This indicates that the present invention can accurately capture the flow process under the influence of the Three Gorges Dam's operation. Meanwhile, the accuracy of the water level simulation results at Cuntan station is slightly lower, with a Nash efficiency coefficient (NSE) of 0.93, while the Nash efficiency coefficients (NSE) at Qingxi and Wanxian stations are 0.97 and 1.00, respectively. This demonstrates that the present invention can essentially reproduce the impact of the Three Gorges Dam's operation on the water level process.

[0058] In the above simulation examples, underwater topographic data from Zhutuo Station to the Three Gorges Dam was not provided, demonstrating that the method proposed in this invention can accurately simulate the hydrodynamic processes in the reservoir area affected by the Three Gorges Dam even in the absence of underwater topographic data. Furthermore, the example above achieved good simulation results with only 10 iterations of the boundary conditions at Zhutuo Station, indicating that the machine learning hybrid learning method proposed in this invention is more computationally efficient than machine learning methods using neural networks. In conclusion, this invention can effectively conduct effective hydrodynamic numerical simulations in river sections with insufficient topographic data and affected by hydraulic engineering projects.

[0059] Example 2: This example is a device for simulating river conditions in areas lacking data due to the influence of water conservancy projects, comprising: The data acquisition and preprocessing module is used to acquire river boundary condition data of the study river section under the influence of water conservancy project operation. Based on the downstream boundary conditions and undetermined coefficients, the upstream section boundary conditions are preprocessed, where multiple sets of undetermined coefficients are set to select values. Multiple algorithm simulation modules are used to simulate and calculate the water level and flow rate of the stations to be simulated in the study section based on the pre-processed upstream cross-section boundary conditions and various machine learning algorithms based on one-dimensional river control equations. The boundary condition iteration module is used to measure the accuracy of the simulation results of each machine learning algorithm. Using the accuracy of the simulation results as the weight, the boundary conditions of the upstream section are iteratively updated, and the iterated boundary conditions of the upstream section are resubmitted into the various machine learning algorithms for simulation. The results screening and optimization module is used to screen the combination of undetermined coefficient values ​​with the best accuracy of the simulation results within the periodicity period. Then, it selects the simulation results with the highest accuracy within the periodicity period from the simulation results of multiple iterations. Combined with the water level-flow relationship of the station to be simulated within the periodicity period, the simulation results are optimized a second time to determine the optimal river hydrological simulation calculation model.

[0060] Example 3: This example is a storage medium on which a computer-readable storage medium stores a computer program. When the computer program is executed by a processor, it implements the steps of the method for simulating river hydrological conditions in the absence of data under the influence of water conservancy projects described in Example 1.

Claims

1. A method for simulating river hydrological conditions in areas lacking data due to the influence of water conservancy projects, characterized in that, include: Data on river boundary conditions of the study section under the influence of water conservancy project operation are obtained. Based on the downstream boundary conditions and undetermined coefficients, the boundary conditions of the upstream section are preprocessed, and multiple sets of undetermined coefficients are set. Based on the pre-processed upstream cross-section boundary conditions, various machine learning algorithms based on one-dimensional river control equations are used to simulate and calculate the water level and flow rate of the stations to be simulated in the research river section. The simulation results of each machine learning algorithm are measured. The simulation results accuracy is used as the weight to iteratively update the boundary conditions of the upstream section. The iteratively updated boundary conditions of the upstream section are then re-substituted into the various machine learning algorithms for simulation. The optimal combination of undetermined coefficient values ​​with the best simulation results within the periodic screening is selected. Then, the simulation results with the highest accuracy within the periodic screening are selected from the simulation results of multiple iterations. The simulation results are then optimized a second time by combining the water level-discharge relationship of the station to be simulated within the periodic screening, thereby determining the optimal river hydrological simulation calculation model.

2. The method for simulating river hydrological conditions in the absence of data under the influence of water conservancy projects according to claim 1, characterized in that, The preprocessing of the upstream section boundary conditions based on downstream boundary conditions and undetermined coefficients includes: ； ； in, , These are the upstream section water level and flow rate after pretreatment; , These are the measured water level and flow rate at the upstream section; , These are the measured water level and flow rate at the downstream section; , These are the upstream section of the station to be simulated, and the tributary flow between the station to be simulated and the downstream section; , These are the flow rates between the upstream section and the station to be simulated, and between the station to be simulated and the downstream section, respectively. , These are undetermined coefficients.

3. The method for simulating river hydrological conditions in the absence of data under the influence of water conservancy projects according to claim 1, characterized in that, The one-dimensional river channel control equation is the difference form of the momentum conservation equation in the Saint-Venant equations.

4. The method for simulating river hydrological conditions in the absence of data under the influence of water conservancy projects according to claim 3, characterized in that, One of the various machine learning algorithms based on one-dimensional river control equations includes: ； A water level calculation formula is constructed based on the difference form of the momentum conservation equation in the Saint-Venant equations, and a time step is introduced. Spatial distance from upstream section to the site to be simulated Changes in upstream water level after pretreatment and undetermined coefficients , , ,Regulation , After determining the range of values, the optimal value is determined by sliding learning combined with least squares fitting within the pass rate period. Solve for the water level change at the simulated site. Then, the water level Z at the site to be simulated is calculated; ； Based on the water level Z at the station to be simulated, an identity matrix E and a column vector of coefficients to be solved are introduced. Construct a flow recursive formula and estimate the parameters using known parameters within a periodic period to obtain the utilization rate. The traffic Q of the site to be simulated is calculated recursively.

5. The method for simulating river hydrological conditions in the absence of data under the influence of water conservancy projects according to claim 3, characterized in that, One of the various machine learning algorithms based on one-dimensional river control equations includes: ； ； A linear approximation of water level and flow rate is made based on the difference form of the momentum conservation equation in the Saint-Venant equations, introducing the identity matrix E and the column vector of coefficients to be solved. Construct a flow recursive formula and estimate the parameters using known parameters within a periodic period to obtain the utilization rate. ； Introducing the identity matrix E and the column vector of coefficients to be solved Construct a water level estimation formula. The parameters are obtained by estimating the known parameters within the rate period; By combining the flow rate recursion formula and the water level estimation formula, and taking the water level at time n+1... ,flow The unknowns are the water level and flow rate of the station to be simulated.

6. The method for simulating river hydrological conditions in the absence of data under the influence of water conservancy projects according to claim 1, characterized in that, The method for measuring the accuracy of the simulation results of each machine learning algorithm includes: using the Nash efficiency coefficient as an error index to measure the accuracy of the simulation results of each machine learning algorithm based on the simulation results of each machine learning algorithm.

7. The method for simulating river hydrological conditions in the absence of data under the influence of water conservancy projects according to claim 1, characterized in that, The iterative update of the upstream section boundary conditions, weighted by the accuracy of the simulation results, includes: ； ； in, , This represents the upstream cross-section water level and flow rate after iteration; , This indicates the upstream cross-sectional water level and flow rate before the iteration; , The s-th machine learning algorithm represents the simulated water level and flow rate of the station to be simulated; n represents the current time step. Let represent the Nash efficiency coefficient of the simulation results obtained by the s-th machine learning algorithm.

8. The method for simulating river hydrological conditions in the absence of data under the influence of water conservancy projects according to claim 1, characterized in that, The secondary optimization of the simulation results by combining the water level-flow relationship within the calibration period of the simulated site includes: Based on the water level-flow relationship, another simulated value is re-estimated using a more accurate simulated water level or flow rate. The estimated result is compared with the original simulation result, and the more accurate result is selected as the final simulation result.

9. A device for simulating river hydrological conditions in areas lacking data due to the influence of water conservancy projects, characterized in that, include: The data acquisition and preprocessing module is used to acquire river boundary condition data of the study river section under the influence of water conservancy project operation. Based on the downstream boundary conditions and undetermined coefficients, the upstream section boundary conditions are preprocessed, where multiple sets of undetermined coefficients are set to select values. Multiple algorithm simulation modules are used to simulate and calculate the water level and flow rate of the stations to be simulated in the study section based on the pre-processed upstream cross-section boundary conditions and various machine learning algorithms based on one-dimensional river control equations. The boundary condition iteration module is used to measure the accuracy of the simulation results of each machine learning algorithm. Using the accuracy of the simulation results as the weight, the boundary conditions of the upstream section are iteratively updated, and the iterated boundary conditions of the upstream section are resubmitted into the various machine learning algorithms for simulation. The results screening and optimization module is used to screen the combination of undetermined coefficient values ​​with the best accuracy of the simulation results within the periodicity period. Then, it selects the simulation results with the highest accuracy within the periodicity period from the simulation results of multiple iterations. Combined with the water level-flow relationship of the station to be simulated within the periodicity period, the simulation results are optimized a second time to determine the optimal river hydrological simulation calculation model.

10. A storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the method for simulating river hydrological conditions in the absence of data under the influence of water conservancy projects as described in any one of claims 1-8.

Images