A method and device for real-time and rapid inversion of roughness parameters, and a storage medium

Abstract

The present application provides a kind of roughness parameter real-time fast inversion method, device and storage medium, it is related to hydrological simulation technical field.The water power simulation model is constructed, and waterway is dispersed into multiple discrete units;Add disturbance to the roughness parameter of each discrete unit and generate roughness parameter set.Roughness parameter training set is obtained and substituted into water power simulation model, and the water flow flow simulation data of each first target discrete unit is obtained.Response surface surrogate model of one-dimensional water power simulation model is generated, roughness parameter set is input into response surface surrogate model, and the water flow flow simulation data of second target discrete unit is obtained, roughness parameter set is updated by ensemble Kalman filtering method, and the optimal roughness parameter of each discrete unit at each time is obtained based on updated roughness parameter set.The present application considers the influence of roughness parameter with flow stage etc., to improve the accuracy of flow simulation in roughness parameter inversion process.The calculation is small, and the parameter inversion speed is improved.

Description

Technical Field

[0001] This invention relates to the field of hydrological simulation technology, and in particular to a method, apparatus, and storage medium for real-time and rapid inversion of roughness parameters. Background Technology

[0002] Roughness parameters are very important in hydraulic calculations because they directly affect the accuracy of hydraulic models, which in turn affects many aspects such as engineering design, flood simulation, riverbed evolution analysis, and water conveyance efficiency.

[0003] The value of roughness parameter is influenced by various factors such as riverbed material, riverbed morphology, and flow characteristics. Currently, there is no universal method to accurately calculate roughness. In practical applications, the determination of roughness parameter usually relies on empirical formulas, field measurements, or model tests. Manual adjustment methods are subjective and uncertain, affecting the accuracy of roughness parameter inversion.

[0004] Existing roughness parameter inversion methods typically ignore the influence of roughness parameter variations with flow rate and water level during the inversion process. Since the inverted roughness parameter value is a fixed value, this affects the accuracy of flow simulation. Furthermore, traditional parameter inversion methods require solving numerous hydrodynamic equations, resulting in a huge computational burden and slow inversion speed. Summary of the Invention

[0005] In view of the above problems, the present invention proposes a method, apparatus and storage medium for real-time and rapid inversion of roughness parameters.

[0006] This invention provides a method for real-time and rapid inversion of roughness parameters, the method comprising:

[0007] Acquire waterway data and construct a corresponding one-dimensional hydrodynamic simulation model based on the waterway data, wherein the waterway includes: river or canal;

[0008] Based on the one-dimensional hydrodynamic simulation model, the waterway is discretized into multiple discrete units;

[0009] When there is water flow rate observation data, a perturbation is added to the roughness parameter of each discrete unit, and a roughness parameter set is generated by random sampling method;

[0010] The roughness parameter training set of the response surface substitution model is obtained by sampling from the roughness parameter set using the random sampling method and then substituted into the one-dimensional hydrodynamic simulation model. The flow rates of the inflow and outflow sections of the waterway are used as boundary conditions to solve the one-dimensional hydrodynamic simulation model and obtain the water flow simulation data of each of the first target discrete units.

[0011] Based on the roughness parameter training set and the water flow simulation data of each of the first target discrete units, a response surface replacement model of the one-dimensional hydrodynamic simulation model is generated, wherein the first target discrete unit is the discrete unit corresponding to each element in the roughness parameter training set.

[0012] The roughness parameter set is input into the response surface substitution model to obtain the water flow simulation data of the second target discrete unit, where the second target discrete unit is the discrete unit corresponding to each element of the roughness parameter set.

[0013] The roughness parameter set is updated using the observed water flow rate and the simulated water flow rate of the second target discrete unit through the ensemble Kalman filtering method.

[0014] Based on the updated set of roughness parameters, the optimal roughness parameters for each discrete unit at each time step are obtained.

[0015] Optionally, waterway data is acquired, and a corresponding one-dimensional hydrodynamic simulation model is constructed based on the waterway data, including:

[0016] Obtain the dimensional data of the river channel or the canal, the dimensional data including: cross-sectional dimensions and slope;

[0017] Based on the cross-sectional dimensions and the slope, the one-dimensional hydrodynamic simulation model is constructed.

[0018] Optionally, based on the one-dimensional hydrodynamic simulation model, the waterway is discretized into multiple discrete units, including:

[0019] The one-dimensional hydrodynamic simulation model is calculated using the finite volume method, and the waterway is discretized into multiple discrete units.

[0020] Optionally, at times when water flow rate observation data is available, a perturbation is added to the roughness parameter of each discrete unit, and a set of roughness parameters is generated using a random sampling method, including:

[0021] At the moment when water flow rate observation data is available, the roughness parameter of each discrete unit is assigned a value that fluctuates up and down according to a normal distribution.

[0022] The roughness parameter set is generated by the Latin hypercube sampling method after the assignment operation.

[0023] Optionally, the roughness parameter training set of the response surface substitution model is obtained by sampling from the roughness parameter set using the random sampling method, including:

[0024] The roughness parameter training set of the response surface substitution model is obtained by sampling from the roughness parameter set using the Latin hypercube sampling method.

[0025] Optionally, based on the roughness parameter training set and the simulated flow rate data of each of the first target discrete units, a response surface replacement model for the one-dimensional hydrodynamic simulation model is generated, including:

[0026] Using the roughness parameter training set as input and the simulated water flow rate data of each of the first target discrete units as output, the response surface substitution model is trained.

[0027] Optionally, the response surface replacement model is trained to include:

[0028] An initial response surface replacement model is generated using the Gaussian process algorithm;

[0029] The roughness parameter training set is used as the input of the Gaussian process algorithm, and the water flow rate simulation data of each of the first target discrete units is used as the output of the Gaussian process algorithm. The hyperparameters to be optimized in the Gaussian process algorithm are optimized using the particle swarm optimization algorithm.

[0030] The response surface replacement model is obtained based on the optimized hyperparameters.

[0031] Optionally, the roughness parameter set is updated using the observed water flow rate and the simulated water flow rate of the second target discrete unit through an ensemble Kalman filter method, including:

[0032] Using the ensemble Kalman analysis formula, the observed water flow rate and the simulated water flow rate of the second target discrete unit are calculated to obtain the updated value of the roughness parameter set;

[0033] The roughness parameter set is updated using the updated values.

[0034] Optionally, based on the updated set of roughness parameters, the optimal roughness parameters for each discrete unit at each time step are obtained, including:

[0035] The updated roughness parameter set is averaged, and the average value is used as the optimal roughness parameter for each discrete unit at each time step.

[0036] Optionally, constructing the one-dimensional hydrodynamic simulation model includes:

[0037] The constructed one-dimensional hydrodynamic simulation model is expressed using the Saint-Venant equations as follows:

[0038]

[0039] In the above two equations, Q is the flow rate; x is the direction of water flow; B is the channel width; Z is the water level; t is the time; A is the cross-sectional area of ​​the water passage; g is the gravitational acceleration; n is the roughness parameter; R is the hydraulic radius; V is the average velocity of the cross section; ql and ul are the lateral inflow rate and the component of the lateral inflow in the x direction along the single field length, respectively.

[0040] Optionally, the Gaussian process algorithm is represented by a mean function μ(m) and a covariance function k(m, m'), wherein the mean function μ(m) is chosen to be 0;

[0041] The diagonal squared exponential covariance function is chosen as the covariance function k(m, m'), and its expression is: In this formula, σf and λ are the hyperparameters to be optimized in the Gaussian process algorithm.

[0042] Optionally, using the ensemble Kalman analysis formula, the observed water flow rate and the simulated water flow rate of the second target discrete unit are calculated to obtain the expression for the updated value of the roughness parameter set:

[0043]

[0044] In the above formula, X is the set of roughness parameters, j is the element number, f is the prediction, a is the analysis, i is the time, H is the observation operator, O is the observed water flow rate, and K i+1 Kalman gain;

[0045] The observation operator is determined by the simulated water flow rate of the second target discrete unit.

[0046] Optionally, the Kalman gain K i+1 The expression is:

[0047] K i+1 =P i+1 H T HP i+1 H T +R i+1 ) -1

[0048] In the above formula, R i+1 For the observation error covariance, P i+1 Let P be the covariance of the observation error of the predicted value. i+1 The expression is as follows:

[0049]

[0050] In the above formula: for The mean.

[0051] This invention provides an apparatus for real-time and rapid inversion of roughness parameters, the apparatus comprising:

[0052] A model building module is used to acquire waterway data and build a corresponding one-dimensional hydrodynamic simulation model based on the waterway data. The waterway includes: river or canal.

[0053] The discrete module is used to discretize the waterway into multiple discrete units based on the one-dimensional hydrodynamic simulation model.

[0054] The perturbation generation set module is used to add perturbations to the roughness parameters of each discrete unit when there is water flow observation data, and generate a roughness parameter set through a random sampling method;

[0055] The solution module is used to sample the roughness parameter training set of the response surface substitution model from the roughness parameter set using the random sampling method and substitute it into the one-dimensional hydrodynamic simulation model. The flow rates of the inflow and outflow sections of the waterway are used as boundary conditions to solve the one-dimensional hydrodynamic simulation model and obtain the water flow simulation data of each of the first target discrete units.

[0056] A response surface replacement model module is used to generate a response surface replacement model for the one-dimensional hydrodynamic simulation model based on the roughness parameter training set and the flow rate simulation data of the first target discrete unit. The first target discrete unit is the discrete unit corresponding to each element in the roughness parameter training set.

[0057] The flow simulation data module is used to input the roughness parameter set into the response surface substitution model to obtain the water flow simulation data of the second target discrete unit, where the second target discrete unit is the discrete unit corresponding to each element in the roughness parameter set.

[0058] The update module is used to update the roughness parameter set using the observed water flow rate and the simulated water flow rate of the second target discrete unit through a set Kalman filter method.

[0059] The optimal roughness parameter module is used to obtain the optimal roughness parameter for each discrete unit at each time step based on the updated roughness parameter set.

[0060] This invention also provides a computer-readable storage medium storing a computer program that enables a processor to execute the method for real-time fast inversion of roughness parameters as described in any of the preceding claims.

[0061] The method for real-time and rapid inversion of roughness parameters provided by this invention first acquires waterway data and constructs a corresponding one-dimensional hydrodynamic simulation model based on the waterway data. Then, based on the one-dimensional hydrodynamic simulation model, the waterway is discretized into multiple discrete units. When there is water flow observation data, a perturbation is added to the roughness parameter of each discrete unit, and a roughness parameter set is generated through a random sampling method.

[0062] Then, by random sampling, the roughness parameter training set of the response surface substitution model is obtained from the roughness parameter set and substituted into the one-dimensional hydrodynamic simulation model. The flow rates of the inflow and outflow sections of the waterway are used as boundary conditions to solve the one-dimensional hydrodynamic simulation model and obtain the water flow simulation data of each of the first target discrete units.

[0063] Based on the roughness parameter training set and the simulated flow rate data of the first target discrete unit, a response surface replacement model for the one-dimensional hydrodynamic simulation model is generated. The roughness parameter set is input into the response surface replacement model to obtain the simulated flow rate data of the second target discrete unit. The roughness parameter set is updated using the observed flow rate and the simulated flow rate of the second target discrete unit through the ensemble Kalman filter method. Finally, based on the updated roughness parameter set, the optimal roughness parameter for each discrete unit at each time step is obtained.

[0064] This invention addresses the shortcomings of traditional roughness parameter inversion methods, which neglect the influence of roughness parameter changes with flow rate and water level. Furthermore, it addresses the issues of large computational load and long processing time associated with hydrodynamic model simulations. The invention proposes discretizing the waterway into multiple discrete units and adding a perturbation to the roughness parameter of each discrete unit when flow rate observation data is available. This incorporates the influence of roughness parameter changes with flow rate and water level into the roughness parameter inversion process, thereby improving the accuracy of water flow simulation.

[0065] Simultaneously, the response surface model is trained and used to replace the solution of the traditional one-dimensional hydrodynamic river model. Only a few-order one-dimensional hydrodynamic river model needs to be solved, instead of solving a one-dimensional hydrodynamic river model for each element in the roughness parameter set. While accurately obtaining the optimal roughness parameters, the computational load is small and the time consumption is short, which greatly improves the parameter inversion speed and allows for faster estimation of the optimal roughness parameters. Attached Figure Description

[0066] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0067] Figure 1This is a flowchart of a method for real-time and rapid inversion of roughness parameters according to an embodiment of the present invention;

[0068] Figure 2 This is a block diagram of a device for real-time and rapid inversion of roughness parameters according to an embodiment of the present invention. Detailed Implementation

[0069] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present invention, and are only some, not all, embodiments of the present invention, and are not intended to limit the present invention.

[0070] Reference Figure 1 The flowchart illustrates a method for real-time fast inversion of roughness parameters according to an embodiment of the present invention. The method includes:

[0071] Step 101: Obtain waterway data and construct a corresponding one-dimensional hydrodynamic simulation model based on the waterway data. Waterways include: rivers or canals.

[0072] The method for real-time and rapid inversion of roughness parameters proposed in this invention first acquires waterway data and then constructs a corresponding one-dimensional hydrodynamic simulation model based on the waterway data. Here, waterways include rivers or canals and other waterways that involve hydrological simulation and roughness parameters.

[0073] A preferred method for constructing a one-dimensional hydrodynamic simulation model includes: first obtaining the dimensional data of the river or channel, including cross-sectional dimensions and slope; and then constructing a one-dimensional hydrodynamic simulation model based on the cross-sectional dimensions and slope.

[0074] To better understand the constructed one-dimensional hydrodynamic simulation model, it is preferable to express the model using the Saint-Venant equations, as follows:

[0075]

[0076] In the above two equations, Q is the flow rate; x is the direction of water flow; B is the channel width; Z is the water level; t is the time; A is the cross-sectional area of ​​the water passage; g is the gravitational acceleration; n is the roughness parameter; R is the hydraulic radius; V is the average velocity of the cross section; ql and ul are the lateral inflow rate and the component of the lateral inflow in the x direction along the single field length, respectively.

[0077] Step 102: Based on the one-dimensional hydrodynamic simulation model, the waterway is discretized into multiple discrete units.

[0078] After constructing a one-dimensional hydrodynamic simulation model, the waterway can be discretized into multiple discrete elements based on this model. A preferred method for discretizing the waterway into multiple discrete elements includes performing calculations on the one-dimensional hydrodynamic simulation model using the finite volume method to discretize the waterway into multiple discrete elements. For example, the channel can be discretized into 50 discrete elements using the finite volume method.

[0079] Step 103: When there is water flow observation data, add perturbation to the roughness parameter of each discrete unit, and generate a roughness parameter set by random sampling method.

[0080] After the discrete elements are completed, perturbations are added to the roughness parameters of each discrete element when water flow observation data is available, and a set of roughness parameters is generated through random sampling. This process of adding perturbations incorporates the effects of changes in flow rate and water level on the roughness parameters into the roughness parameter inversion process, thereby improving the accuracy of subsequent water flow simulations.

[0081] A better approach is to assign values ​​to the roughness parameters of each discrete unit that follow a normal distribution, ensuring the perturbation error of the roughness parameters conforms to a normal distribution with zero mean and variance S. Then, the roughness parameters after this assignment are used to generate a roughness parameter set using the Latin hypercube sampling method. For example, after adding a perturbation to the roughness parameters of each discrete unit, 100 values ​​are randomly sampled using the Latin hypercube sampling method as 100 elements; these 100 elements form the roughness parameter set.

[0082] Step 104: Using a random sampling method, sample the roughness parameter training set of the response surface substitution model from the roughness parameter set and substitute it into the one-dimensional hydrodynamic simulation model. Using the inflow and outflow cross-sections of the waterway as boundary conditions, solve the one-dimensional hydrodynamic simulation model to obtain the water flow simulation data of each of the first target discrete units.

[0083] After the roughness parameter set is generated, the roughness parameter training set of the response surface substitution model is obtained by sampling from the roughness parameter set through random sampling method and substituted into the one-dimensional hydrodynamic simulation model. The flow rates of the inflow and outflow sections of the waterway are used as boundary conditions to solve the one-dimensional hydrodynamic simulation model and obtain the water flow simulation data of each of the first objective discrete units.

[0084] A preferred approach is to use the Latin hypercube sampling method to sample the roughness parameters of the response surface surrogate model from the roughness parameter set. For example, the Latin hypercube sampling method can be used to sample 30 elements from 100 elements in the roughness parameter set to obtain the roughness parameter training set, and the remaining 70 elements are used to form the roughness parameter validation set. Generally, the number of elements in the roughness parameter training set does not exceed the number of elements in the roughness parameter validation set. That is, taking 100 roughness parameters as an example, the roughness parameter training set has a maximum of 50 elements, and the remaining 50 elements are the elements in the roughness parameter validation set.

[0085] Step 105: Based on the roughness parameter training set and the water flow simulation data of the first target discrete unit, generate the response surface replacement model of the one-dimensional hydrodynamic simulation model. The first target discrete unit is the discrete unit corresponding to each element in the roughness parameter training set.

[0086] After obtaining the roughness parameter training set and the simulated flow rate data for each of the first target discrete units, a response surface substitution model corresponding to the one-dimensional hydrodynamic simulation model can be generated. Here, the first target discrete unit is the discrete unit corresponding to each element in the roughness parameter training set.

[0087] A superior method for generating response surface replacement models includes: using a roughness parameter training set as input and simulated flow rate data for each of the first objective discrete units as output to train the response surface replacement model. Specifically:

[0088] The Gaussian process algorithm is used to generate an initial response surface replacement model. The roughness parameter training set is used as the input of the Gaussian process algorithm, and the simulated water flow rate data of each of the first target discrete units is used as the output of the Gaussian process algorithm. The hyperparameters to be optimized in the Gaussian process algorithm are optimized using the particle swarm optimization algorithm (PSO algorithm). The response surface replacement model is obtained based on the optimized hyperparameters.

[0089] Preferably, the Gaussian process algorithm is represented by the mean function μ(m) and the covariance function k(m, m'). In this embodiment of the invention, the mean function μ(m) is preferably chosen to be 0.

[0090] A preferred approach is to choose the diagonal squared exponential covariance function as the covariance function k(m, m'), whose expression is: In this formula, σf and λ are the hyperparameters to be optimized in the Gaussian process algorithm.

[0091] After the response surface substitution model is generated, it still needs to be validated. Therefore, based on the roughness parameter set, a roughness parameter validation set for the response surface substitution model needs to be generated, and the simulation accuracy of the response substitution model is verified by using this roughness parameter validation set to meet the accuracy requirements.

[0092] Similarly, the Latin hypercube sampling method can be used to sample from the target roughness parameter set to obtain the roughness parameter validation set. The target roughness parameter set is the set of elements remaining after removing elements from the roughness parameter training set. Continuing the previous example: sampling 30 elements from 100 elements in the roughness parameter set using the Latin hypercube sampling method yields the roughness parameter training set, leaving 70 elements. From these remaining 70 elements, another 50 elements are sampled using the Latin hypercube sampling method to obtain the roughness parameter validation set. Naturally, it's understandable that the number of elements in the roughness parameter set, the number of elements in the roughness parameter training set, and the number of elements in the roughness parameter validation set are determined by domain experts based on experience. A larger number of elements results in higher accuracy for the response surface substitution model, but a longer time to generate and validate the model; conversely, a smaller number of elements results in lower accuracy for the response surface substitution model, but a shorter time to generate and validate it. After extensive testing and research, the inventors found that an optimal number of elements is 100 in the roughness parameter set, 30 in the roughness parameter training set, and 50 in the roughness parameter validation set. This ensures the accuracy of the response surface replacement model while also optimizing the time required to generate and validate it.

[0093] To verify that the simulation accuracy of the response substitution model meets the accuracy requirements, if the simulation accuracy of the response substitution model does not meet the accuracy requirements, the number of elements in the roughness parameter training set is increased based on the roughness parameter to form a new training set, and the steps of generating and optimizing the response surface and verifying the simulation accuracy of the response substitution model are iteratively executed until the simulation accuracy of the response substitution model meets the accuracy requirements.

[0094] For example, if a response substitution model is generated using a roughness parameter training set of 30 elements, but the simulation accuracy of this model does not meet the accuracy requirements, then the number of elements in the roughness parameter training set is increased to 40, forming a new training set. Then, using these 40 elements as input, the response surface substitution model is optimized again, and a new roughness parameter validation set of 50 elements is generated. This new validation set is used to test the simulation accuracy of the response substitution model and determine if it meets the accuracy requirements. This process is repeated iteratively until the simulation accuracy of the response substitution model meets the accuracy requirements.

[0095] Step 106: Input the roughness parameter set into the response surface substitution model to obtain the water flow simulation data of the second target discrete unit. The second target discrete unit is the discrete unit corresponding to each element in the roughness parameter set.

[0096] After the response surface substitution model is generated and validated, the roughness parameter set is input into the response surface substitution model to obtain the water flow simulation data of the second target discrete unit. The second target discrete unit is the discrete unit corresponding to each element in the roughness parameter set. For example, if there are 100 elements in the roughness parameter set, then inputting these 100 elements into the response surface substitution model will yield the water flow simulation data of the discrete units corresponding to these 100 elements.

[0097] Step 107: Update the roughness parameter set using the observed water flow rate and the simulated water flow rate of the second objective discrete unit through the ensemble Kalman filter method.

[0098] After obtaining the simulated flow rate data, the roughness parameter set can be updated using the observed flow rate values ​​and the simulated flow rate values ​​of the second objective discrete unit through ensemble Kalman filtering. A preferred method for updating the roughness parameter set includes:

[0099] Using the ensemble Kalman analysis formula, the observed water flow rate and the simulated water flow rate of the second objective discrete unit are calculated to obtain the updated value of the roughness parameter set; the updated value is then used to update the roughness parameter set.

[0100] Specifically, the expression for updating the roughness parameter set is:

[0101]

[0102] In the above formula, X is the set of roughness parameters, j is the element number, f is the prediction, a is the analysis, i is the time, H is the observation operator, O is the observed water flow rate, and K... i+1 The Kalman gain is used; the observation operator is determined by the simulated flow rate of the water in the second objective discrete unit.

[0103] Among these, the Kalman gain K i+1 The expression is:

[0104] K i+1 =P i+1 H T HP i+1 H T +R i+1 ) -1

[0105] In the above formula, R i+1 For the observation error covariance, P i+1 Let P be the covariance of the observation error of the predicted value. i+1 The expression is as follows:

[0106]

[0107] In the above formula: for The mean.

[0108] Step 108: Based on the updated set of roughness parameters, obtain the optimal roughness parameters for each discrete unit at each time step.

[0109] After updating the roughness parameter set, the optimal roughness parameters for each discrete unit at each time step can be obtained based on the updated roughness parameter set. Better methods for obtaining the optimal roughness parameters include:

[0110] The updated roughness parameter set is averaged, and the average value is used as the optimal roughness parameter for each discrete unit at each time step.

[0111] The above-described rapid roughness parameter estimation method can be better understood by taking the simulation of the water flow propagation and receding process in a water conveyance channel as an example. Assume the channel length is 350m, with an observation section set up every 50m, and the water flow rate at each section measured every 1 minute. After the gate at the head of the channel is opened, water is released at a flow rate of 3.5L / s for 110 minutes, with free flow at the tail of the channel.

[0112] Based on the data from the aforementioned channel, assuming a trapezoidal cross-sectional shape and a gradient of 0.0025, a one-dimensional hydrodynamic simulation model of the channel is constructed, using the Saint-Venant equations as follows:

[0113]

[0114] In the formula: Q is the flow rate; x is the direction of water flow; B is the channel width; Z is the water level; t is time; A is the cross-sectional area of ​​the water passage; g is the gravitational acceleration; n is the roughness parameter; R is the hydraulic radius; V is the average velocity of the cross section; ql and ul are the lateral inflow rate and the x-component of the lateral inflow along the single field length, respectively. The channel is then discretized into 50 discrete elements using the finite volume method.

[0115] When observational data is available, a perturbation is added to the roughness parameter of each discrete unit. It is assumed that the perturbation error follows a normal distribution with zero mean and variance S, and a roughness parameter set is generated using the Latin hypercube sampling method. This perturbation process incorporates the effects of changes in flow rate and water level on the roughness parameter into the roughness parameter inversion process, thus improving the accuracy of subsequent flow simulations.

[0116] The roughness parameter training set of the response surface substitution model is obtained by sampling the roughness parameter set using the Latin hypercube sampling method. The roughness parameter training set is then substituted into the one-dimensional hydrodynamic simulation model. The flow rates at the inflow and outflow sections of the channel are used as boundary conditions to solve the Saint-Venant equations, thereby obtaining the flow rate simulation of the discrete unit corresponding to the elements in the roughness parameter training set.

[0117] Using the roughness parameter training set as input and the simulated flow rate data of each discrete unit corresponding to the elements in the roughness parameter training set as output, a response surface replacement model for the one-dimensional hydrodynamic simulation model is trained. The response surface replacement model can employ the Gaussian process method. The Gaussian process algorithm is represented by the mean function μ(m) and the covariance function k(m, m'). The mean function can be chosen as 0, and the diagonal squared exponential covariance function can be chosen as the covariance function, expressed as: In the formula, σf and λ are the hyperparameters of the Gaussian process algorithm, and their values ​​are obtained through training.

[0118] To further verify the accuracy of the response surface substitution model, after its accuracy meets the requirements, the roughness parameter set is input into the response surface substitution model to obtain the water flow simulation data of each discrete unit corresponding to each element in the roughness parameter set.

[0119] Using the ensemble Kalman analysis formula, the observed flow rate and the simulated flow rate of the second objective discrete element are calculated to obtain the updated values ​​of the roughness parameter set; the expression for the updated values ​​of the roughness parameter set is:

[0120]

[0121] In the above formula, X is the set of roughness parameters, j is the element number, f is the prediction, a is the analysis, i is the time, H is the observation operator, O is the observed water flow rate, and K... i+1 The Kalman gain is used; the observation operator is determined by the simulated flow rate of the water in the second objective discrete unit.

[0122] Among these, the Kalman gain K i+1 The expression is:

[0123] K i+1 =P i+1 H T HP i+1 H T +R i+1 ) -1

[0124] In the above formula, R i+1 For the observation error covariance, P i+1 Let P be the covariance of the observation error of the predicted value. i+1 The expression is as follows:

[0125]

[0126] In the above formula: for The mean.

[0127] Finally, the average value of the updated roughness parameter set is taken as the optimal roughness parameter for each discrete unit at each time step.

[0128] In this embodiment of the invention, based on the above-described method for real-time and rapid inversion of roughness parameters, a device for real-time and rapid inversion of roughness parameters is also proposed, with reference to... Figure 2 The block diagram shown is of a device for real-time and rapid inversion of roughness parameters, which includes:

[0129] Model building module 210 is used to acquire waterway data and build a corresponding one-dimensional hydrodynamic simulation model based on the waterway data. The waterway includes: river or canal.

[0130] Discretization module 220 is used to discretize the waterway into multiple discrete units based on the one-dimensional hydrodynamic simulation model.

[0131] The perturbation generation set module 230 is used to add perturbations to the roughness parameters of each discrete unit when there is water flow observation data, and generate a roughness parameter set by random sampling method;

[0132] The solution module 240 is used to sample the roughness parameter training set of the response surface substitution model from the roughness parameter set using the random sampling method and substitute it into the one-dimensional hydrodynamic simulation model. The flow rates of the inflow and outflow sections of the waterway are used as boundary conditions to solve the one-dimensional hydrodynamic simulation model and obtain the water flow simulation data of each of the first target discrete units.

[0133] The response surface replacement model generation module 250 is used to generate the response surface replacement model of the one-dimensional hydrodynamic simulation model based on the roughness parameter training set and the water flow simulation data of the first target discrete unit. The first target discrete unit is the discrete unit corresponding to each element in the roughness parameter training set.

[0134] The flow simulation data module 260 is used to input the roughness parameter set into the response surface substitution model to obtain the water flow simulation data of the second target discrete unit, where the second target discrete unit is the discrete unit corresponding to each element in the roughness parameter set.

[0135] Update module 270 is used to update the roughness parameter set using the observed water flow rate and the simulated water flow rate of the second target discrete unit through a set Kalman filter method;

[0136] The optimal roughness parameter module 280 is used to obtain the optimal roughness parameter for each discrete unit at each time step based on the updated roughness parameter set.

[0137] Optionally, the model building module 210 is specifically used for:

[0138] Obtain the dimensional data of the river channel or the canal, the dimensional data including: cross-sectional dimensions and slope;

[0139] Based on the cross-sectional dimensions and the slope, the one-dimensional hydrodynamic simulation model is constructed.

[0140] The construction of the one-dimensional hydrodynamic simulation model includes:

[0141] The constructed one-dimensional hydrodynamic simulation model is expressed using the Saint-Venant equations as follows:

[0142]

[0143] In the above two equations, Q is the flow rate; x is the direction of water flow; B is the channel width; Z is the water level; t is the time; A is the cross-sectional area of ​​the water passage; g is the gravitational acceleration; n is the roughness parameter; R is the hydraulic radius; V is the average velocity of the cross section; ql and ul are the lateral inflow rate and the component of the lateral inflow in the x direction along the single field length, respectively.

[0144] Optionally, the discrete module 220 is specifically used for:

[0145] The one-dimensional hydrodynamic simulation model is calculated using the finite volume method, and the waterway is discretized into multiple discrete units.

[0146] Optionally, the disturbance generation set module 230 is specifically used for:

[0147] At the moment when water flow rate observation data is available, the roughness parameter of each discrete unit is assigned a value that fluctuates up and down according to a normal distribution.

[0148] The roughness parameter set is generated by the Latin hypercube sampling method after the assignment operation.

[0149] Optionally, the solution module 240 is specifically used for:

[0150] The roughness parameter training set of the response surface substitution model is obtained by sampling from the roughness parameter set using the Latin hypercube sampling method.

[0151] Optionally, the module 250 for generating response surface substitution models is specifically used for:

[0152] Using the roughness parameter training set as input and the simulated water flow rate data of each of the first target discrete units as output, the response surface substitution model is trained.

[0153] Optionally, the module 250 for generating response surface substitution models is further specifically used for:

[0154] An initial response surface replacement model is generated using the Gaussian process algorithm;

[0155] The roughness parameter training set is used as the input of the Gaussian process algorithm, and the water flow rate simulation data of each of the first target discrete units is used as the output of the Gaussian process algorithm. The hyperparameters to be optimized in the Gaussian process algorithm are optimized using the particle swarm optimization algorithm.

[0156] The response surface replacement model is obtained based on the optimized hyperparameters;

[0157] The Gaussian process algorithm is represented by a mean function μ(m) and a covariance function k(m, m'), wherein the mean function μ(m) is chosen to be 0;

[0158] The diagonal squared exponential covariance function is chosen as the covariance function k(m, m'), and its expression is: In this formula, σf and λ are the hyperparameters to be optimized in the Gaussian process algorithm.

[0159] Optionally, the update module 270 is specifically used for:

[0160] Using the ensemble Kalman analysis formula, the observed water flow rate and the simulated water flow rate of the second target discrete unit are calculated to obtain the updated value of the roughness parameter set;

[0161] The roughness parameter set is updated using the updated values;

[0162] Specifically, using the ensemble Kalman analysis formula, the observed water flow rate and the simulated water flow rate of the second target discrete unit are calculated to obtain the expression for the updated value of the roughness parameter set:

[0163]

[0164] In the above formula, X is the set of roughness parameters, j is the element number, f is the prediction, a is the analysis, i is the time, H is the observation operator, O is the observed water flow rate, and K i+1 Kalman gain;

[0165] The observation operator is determined by the simulated water flow rate of the second target discrete unit;

[0166] The Kalman gain K i+1 The expression is:

[0167] K i+1 =P i+1 H T HP i+1 H T +R i+1 ) -1

[0168] In the above formula, R i+1 For the observation error covariance, P i+1 Let P be the covariance of the observation error of the predicted value. i+1 The expression is as follows:

[0169]

[0170] In the above formula: for The mean.

[0171] Optionally, the optimal roughness parameter module 280 is specifically used for:

[0172] The updated roughness parameter set is averaged, and the average value is used as the optimal roughness parameter for each discrete unit at each time step.

[0173] In this embodiment of the invention, based on the above-described method for real-time and rapid inversion of roughness parameters, a computer-readable storage medium is also proposed, wherein the computer program stored therein enables a processor to execute the method for real-time and rapid inversion of roughness parameters as described in any one of steps 101 to 108.

[0174] In summary, the method for real-time and rapid inversion of roughness parameters of the present invention first acquires waterway data, constructs a corresponding one-dimensional hydrodynamic simulation model based on the waterway data, and then discretizes the waterway into multiple discrete units based on the one-dimensional hydrodynamic simulation model; when there is water flow observation data, a perturbation is added to the roughness parameter of each discrete unit, and a roughness parameter set is generated by random sampling.

[0175] Then, by random sampling, the roughness parameter training set of the response surface substitution model is obtained from the roughness parameter set and substituted into the one-dimensional hydrodynamic simulation model. The flow rates of the inflow and outflow sections of the waterway are used as boundary conditions to solve the one-dimensional hydrodynamic simulation model and obtain the water flow simulation data of each of the first target discrete units.

[0176] Based on the roughness parameter training set and the simulated flow rate data of the first target discrete unit, a response surface replacement model for the one-dimensional hydrodynamic simulation model is generated. The roughness parameter set is input into the response surface replacement model to obtain the simulated flow rate data of the second target discrete unit. The roughness parameter set is updated using the observed flow rate and the simulated flow rate of the second target discrete unit through the ensemble Kalman filter method. Finally, based on the updated roughness parameter set, the optimal roughness parameter for each discrete unit at each time step is obtained.

[0177] This invention addresses the shortcomings of traditional roughness parameter inversion methods, which neglect the influence of roughness parameter changes with flow rate and water level. Furthermore, it addresses the issues of large computational load and long processing time associated with hydrodynamic model simulations. The invention proposes discretizing the waterway into multiple discrete units and adding a perturbation to the roughness parameter of each discrete unit when flow rate observation data is available. This incorporates the influence of roughness parameter changes with flow rate and water level into the roughness parameter inversion process, thereby improving the accuracy of water flow simulation.

[0178] Simultaneously, the response surface model is trained and used to replace the solution of the traditional one-dimensional hydrodynamic river model. Only a few-order one-dimensional hydrodynamic river model needs to be solved, instead of solving a one-dimensional hydrodynamic river model for each element in the roughness parameter set. While accurately obtaining the optimal roughness parameters, the computational load is small and the time consumption is short, which greatly improves the parameter inversion speed and allows for faster estimation of the optimal roughness parameters.

[0179] Although preferred embodiments of the present invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present invention.

[0180] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0181] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims. All of these forms are within the protection scope of the present invention.

Claims

1. A method for real-time and rapid inversion of roughness parameters, characterized in that, The method for real-time and rapid inversion of the roughness parameter includes: Acquire waterway data and construct a corresponding one-dimensional hydrodynamic simulation model based on the waterway data, wherein the waterway includes: river or canal; Based on the one-dimensional hydrodynamic simulation model, the waterway is discretized into multiple discrete units; When there is water flow rate observation data, a perturbation is added to the roughness parameter of each discrete unit, and a roughness parameter set is generated by random sampling method; The roughness parameter training set of the response surface substitution model is obtained by sampling from the roughness parameter set using the random sampling method and then substituted into the one-dimensional hydrodynamic simulation model. The flow rates of the inflow and outflow sections of the waterway are used as boundary conditions to solve the one-dimensional hydrodynamic simulation model and obtain the water flow simulation data of each of the first target discrete units. Based on the roughness parameter training set and the water flow simulation data of each of the first target discrete units, a response surface replacement model of the one-dimensional hydrodynamic simulation model is generated, wherein the first target discrete unit is the discrete unit corresponding to each element in the roughness parameter training set. The roughness parameter set is input into the response surface substitution model to obtain the water flow simulation data of the second target discrete unit, where the second target discrete unit is the discrete unit corresponding to each element of the roughness parameter set. The roughness parameter set is updated using the observed water flow rate and the simulated water flow rate of the second target discrete unit through the ensemble Kalman filtering method. Based on the updated set of roughness parameters, the optimal roughness parameters for each discrete unit at each time step are obtained.

2. The method for real-time and rapid inversion of roughness parameters according to claim 1, characterized in that, Acquire waterway data and construct a corresponding one-dimensional hydrodynamic simulation model based on the waterway data, including: Obtain the dimensional data of the river channel or the canal, the dimensional data including: cross-sectional dimensions and slope; Based on the cross-sectional dimensions and the slope, the one-dimensional hydrodynamic simulation model is constructed.

3. The method for real-time and rapid inversion of roughness parameters according to claim 1, characterized in that, Based on the aforementioned one-dimensional hydrodynamic simulation model, the waterway is discretized into multiple discrete units, including: The one-dimensional hydrodynamic simulation model is calculated using the finite volume method, and the waterway is discretized into multiple discrete units.

4. The method for real-time and rapid inversion of roughness parameters according to claim 1, characterized in that, At times when water flow rate observation data is available, a perturbation is added to the roughness parameter of each discrete unit, and a set of roughness parameters is generated using a random sampling method, including: At the moment when water flow rate observation data is available, the roughness parameter of each discrete unit is assigned a value that fluctuates up and down according to a normal distribution. The roughness parameter set is generated by the Latin hypercube sampling method after the assignment operation.

5. The method for real-time and rapid inversion of roughness parameters according to claim 1, characterized in that, The roughness parameter training set of the response surface substitution model is obtained by sampling from the roughness parameter set using the random sampling method, including: The roughness parameter training set of the response surface substitution model is obtained by sampling from the roughness parameter set using the Latin hypercube sampling method.

6. The method for real-time and rapid inversion of roughness parameters according to claim 1, characterized in that, Based on the roughness parameter training set and the simulated flow rate data of each of the first target discrete units, a response surface replacement model for the one-dimensional hydrodynamic simulation model is generated, including: Using the roughness parameter training set as input and the simulated water flow rate data of each of the first target discrete units as output, the response surface substitution model is trained.

7. The method for real-time and rapid inversion of roughness parameters according to claim 6, characterized in that, The trained response surface replacement model includes: An initial response surface replacement model is generated using the Gaussian process algorithm; The roughness parameter training set is used as the input of the Gaussian process algorithm, and the water flow rate simulation data of each of the first target discrete units is used as the output of the Gaussian process algorithm. The hyperparameters to be optimized in the Gaussian process algorithm are optimized using the particle swarm optimization algorithm. The response surface replacement model is obtained based on the optimized hyperparameters.

8. The method for real-time and rapid inversion of roughness parameters according to claim 1, characterized in that, The roughness parameter set is updated using the ensemble Kalman filtering method, utilizing the observed water flow rate and the simulated water flow rate of the second target discrete unit, including: Using the ensemble Kalman analysis formula, the observed water flow rate and the simulated water flow rate of the second target discrete unit are calculated to obtain the updated value of the roughness parameter set; The roughness parameter set is updated using the updated values.

9. The method for real-time and rapid inversion of roughness parameters according to claim 1, characterized in that, Based on the updated set of roughness parameters, the optimal roughness parameters for each discrete unit at each time step are obtained, including: The updated roughness parameter set is averaged, and the average value is used as the optimal roughness parameter for each discrete unit at each time step.

10. The method for real-time and rapid inversion of roughness parameters according to claim 2, characterized in that, Constructing the one-dimensional hydrodynamic simulation model includes: The constructed one-dimensional hydrodynamic simulation model is expressed using the Saint-Venant equations as follows: In the above two equations, Q is the flow rate; x is the direction of water flow; B is the channel width; Z is the water level; t is the time; A is the cross-sectional area of ​​the water passage; g is the gravitational acceleration; n is the roughness parameter; R is the hydraulic radius; V is the average velocity of the cross section; ql and ul are the lateral inflow rate and the component of the lateral inflow in the x direction along the single field length, respectively.

11. The method for real-time and rapid inversion of roughness parameters according to claim 7, characterized in that, The Gaussian process algorithm is represented by the mean function μ(m) and the covariance function k(m, m'), wherein the mean function μ(m) is chosen to be 0; The diagonal squared exponential covariance function is chosen as the covariance function k(m, m'), and its expression is: In this formula, σf and λ are the hyperparameters to be optimized in the Gaussian process algorithm.

12. The method for real-time and rapid inversion of roughness parameters according to claim 8, characterized in that, Using the ensemble Kalman analysis formula, the observed water flow rate and the simulated water flow rate of the second target discrete unit are calculated to obtain the expression for the updated value of the roughness parameter set: In the above formula, X is the set of roughness parameters, j is the element number, f is the prediction, a is the analysis, i is the time, H is the observation operator, O is the observed water flow rate, and K i+1 Kalman gain; The observation operator is determined by the simulated water flow rate of the second target discrete unit.

13. The method for real-time and rapid inversion of roughness parameters according to claim 12, characterized in that, The Kalman gain K i+1 The expression is: K i+1 =P i+1 H T (HP i+1 H T +R i+1 ) -1 In the above formula, R i+1 For the observation error covariance, P i+1 P represents the covariance of the observation error of the predicted value. i+1 The expression is as follows: In the above formula: for The mean.

14. A device for real-time and rapid inversion of roughness parameters, characterized in that, The device for real-time and rapid inversion of the roughness parameter includes: A model building module is used to acquire waterway data and build a corresponding one-dimensional hydrodynamic simulation model based on the waterway data. The waterway includes: river or canal. The discrete module is used to discretize the waterway into multiple discrete units based on the one-dimensional hydrodynamic simulation model. The perturbation generation set module is used to add perturbations to the roughness parameters of each discrete unit when there is water flow observation data, and generate a roughness parameter set through a random sampling method; The solution module is used to sample the roughness parameter training set of the response surface substitution model from the roughness parameter set using the random sampling method and substitute it into the one-dimensional hydrodynamic simulation model. The flow rates of the inflow and outflow sections of the waterway are used as boundary conditions to solve the one-dimensional hydrodynamic simulation model and obtain the water flow simulation data of each of the first target discrete units. A response surface replacement model module is used to generate a response surface replacement model for the one-dimensional hydrodynamic simulation model based on the roughness parameter training set and the flow rate simulation data of the first target discrete unit. The first target discrete unit is the discrete unit corresponding to each element in the roughness parameter training set. The flow simulation data module is used to input the roughness parameter set into the response surface substitution model to obtain the water flow simulation data of the second target discrete unit, where the second target discrete unit is the discrete unit corresponding to each element in the roughness parameter set. The update module is used to update the roughness parameter set using the observed water flow rate and the simulated water flow rate of the second target discrete unit through a set Kalman filter method. The optimal roughness parameter module is used to obtain the optimal roughness parameter for each discrete unit at each time step based on the updated roughness parameter set.

15. A computer-readable storage medium, characterized in that, The stored computer program enables the processor to execute the method for real-time rapid inversion of roughness parameters as described in any one of claims 1 to 13.

Images