A two-dimensional hydrodynamic model data assimilation method, system and storage medium

Abstract

The present invention discloses a two-dimensional hydrodynamic model data assimilation method, system and storage medium, and relates to the field of data assimilation processing technology. A forward model is used to obtain fitting data of prediction indicators under the context of measured input parameters and model parameters in historical flood events, and the measured data and fitting data are used to calculate the prediction error rate of the forward model for the prediction indicators. A neural network model is constructed with the normalized values of the input parameters and model parameters as input data and the prediction error rate as output data. The prediction error rate of the neural network model when the normalized values of random combinations within the parameter range are used as input data is obtained, and a data mapping table is constructed. By searching the data mapping table for the same input parameters as the new flood event, the model parameters with the minimum weighted value of the input parameters and the prediction error rate of all prediction indicators are obtained as driving data of the forward model, and the prediction indicator fitting data of the new flood event are obtained.

Description

Technical Field

[0001] The present invention relates to the technical field of data assimilation processing, and in particular to a two-dimensional hydrodynamic model data assimilation method, system and storage medium. Background Art

[0002] A two-dimensional hydrodynamic model is a mathematical model used to simulate the movement of water in two dimensions (usually a horizontal plane). It is widely used in hydrodynamic research and engineering practice in water bodies such as rivers, lakes, estuaries, and nearshore waters. It is a crucial tool in hydrodynamic research and engineering applications. By continuously improving the model's theoretical foundation and numerical methods, combined with actual observational data, the model's accuracy and applicability can be enhanced, providing strong support for water resources management, flood prevention and disaster reduction, and ecological and environmental protection.

[0003] During the simulation process, hydrodynamic models are affected by uncertainties such as model parameters, input data, and model structure, leading to deviations between simulation results and actual observations. Data assimilation combines observed data with model predictions, using assimilation algorithms to continuously adjust model parameters, bringing model outputs closer to actual observations. Furthermore, data assimilation can optimize the model's initial fields and boundary conditions based on observed data to better align with reality, thereby improving model simulation accuracy.

[0004] Currently, existing technologies for assimilation of hydrodynamic models primarily include particle filtering, ensemble Kalman filtering, and variational methods. In practical applications, these methods significantly increase the demand for computing resources when the amount of variable data in the model is large. This not only leads to low computational efficiency and increased computational costs, but also makes it difficult to meet the needs of real-time or near-real-time hydrodynamic simulations. This is particularly true for flood warnings and emergency responses. Existing assimilation methods, due to their high computational complexity, often fail to complete calculations in a timely manner, preventing rapid updates to model states and significantly reducing the timeliness of model fitting, thus limiting their application in actual flood warnings. To address this issue, we propose a two-dimensional hydrodynamic model data assimilation method, system, and storage medium. Summary of the Invention

[0005] The main purpose of the present invention is to provide a two-dimensional hydrodynamic model data assimilation method, system and storage medium, which can effectively solve the problems in the background technology.

[0006] In order to achieve the above purpose, the technical solution adopted by the present invention is:

[0007] A two-dimensional hydrodynamic model data assimilation method includes:

[0008] Collect measured data on prediction indicators of historical flood events in the area to be monitored , substitute the measured data into the forward model of flood events, and use the forward model to obtain the input parameters of , the model parameters are Fitting the predictors to the data in context ,in, It represents the data at time t in the flood event; u, v, and w are the numbers of the prediction index, input parameter, and model parameter, respectively;

[0009] Using measured data and fitting data Calculate the prediction error rate of the forward model for the u-th prediction indicator at time t in the flood event , where t∈T; T is the total fitting time of the forward model for the flood event;

[0010] Build with input parameters , model parameters The normalized value is the input data, with the prediction error rate A neural network model is used to output data and is trained by adjusting the weight parameters, bias parameters, and hyperparameters of the neural network model until its accuracy is not less than the set threshold;

[0011] Set the input parameter value range of the forward model respectively [ , ] and the range of model parameter values[ , ], get a random combination of parameters within the value range ,in, ∈[ , ]; ∈[ , ], get the neural network model to randomly combine The prediction error rate when the normalized value of is used as input data , build a data mapping table → ;

[0012] Search the data mapping table for the input parameters of the new flood event at time t Same input parameters , to input parameters The model parameter that takes the minimum weighted value of the prediction error rate of all prediction indicators As the driving data of the forward model, obtain the corresponding prediction index fitting data .

[0013] A two-dimensional hydrodynamic model data assimilation system, comprising:

[0014] Data acquisition module, used to collect the measured data of the prediction indicators of historical flood events in the monitored area ;

[0015] Data fitting module, used to obtain the measured data of the prediction indicators Substitute it into the forward model of flood events and use the forward model to obtain the input parameters , the model parameters are Fitting the predictors to the data in context ,in, It represents the data at time t in the flood event; u, v, and w are the numbers of the prediction index, input parameter, and model parameter, respectively;

[0016] Fitting data processing module, used to utilize measured data and fitting data Calculate the prediction error rate of the forward model for the u-th prediction indicator at time t in the flood event , where t∈T; T is the total fitting time of the forward model for the flood event;

[0017] Neural network building blocks for constructing neural networks with input parameters , model parameters The normalized value is the input data, with the prediction error rate Build an LSTM neural network model for output data and train it by adjusting the weight parameters, bias parameters, and hyperparameters of the neural network model until its accuracy is not less than the set threshold;

[0018] Mapping data acquisition module, used to set the input parameter value range of the forward model [ , ] and the range of model parameter values[ , ], get a random combination of parameters within the value range ,in, ∈[ , ]; ∈[ , ], get the neural network model to randomly combine The prediction error rate when the normalized value of is used as input data , build a data mapping table → ;

[0019] Drive the data acquisition module to search the data mapping table for the input parameters of the new flood event at time t Same input parameters , to input parameters The model parameter that takes the minimum weighted value of the prediction error rate of all prediction indicators As the driving data of the forward model, obtain the corresponding prediction index fitting data .

[0020] A two-dimensional hydrodynamic model data assimilation storage medium, wherein an electronic program is stored in the storage medium.

[0021] Furthermore, the prediction indicators include water level, flow rate and flooding range;

[0022] The input parameters include rainfall runoff parameters and terrain parameters;

[0023] The model parameters include the Manning roughness coefficient of the monitoring area, boundary condition parameters, calculation unit parameters, time step parameters and model operation parameters.

[0024] Furthermore, the prediction error rate The calculation formula is: - | / ×100%.

[0025] Furthermore, the neural network model is an LSTM neural network model, and the hyperparameters of the neural network model include hidden layer size, learning rate and number of training rounds.

[0026] Furthermore, the process of determining the input parameter value range and the model parameter value range includes the following steps:

[0027] Using the input parameters of historical flood events and model parameters Construct a multidimensional data vector z, z=[ , ] T , and the multidimensional data vector z~N(μ,∑), where μ is the d-dimensional mean vector, representing the expected value of each data dimension; ∑ is the d×d covariance matrix, representing the covariance relationship between the data dimensions; d= ;

[0028] Construct the normal distribution probability density function of the multidimensional data vector z ,in, = ; is the inverse matrix of the covariance matrix; π is the pi;

[0029] Sampling a noise vector from a normal distribution of the constructed multidimensional data vector z;

[0030] Constructing a generative adversarial network model, wherein the generator G of the generative adversarial network model takes the sampled noise vector as input and the generated multidimensional virtual data vector as output; the discriminator D of the generative adversarial network model takes the normalized value of the multidimensional data vector z as input and the probability that the generated multidimensional virtual data vector is real data as output;

[0031] Use the real data in the multidimensional data vector z and the virtual data in the multidimensional virtual data vector generated by the generator to train the discriminator D, calculate the loss of the real data and the loss of the generated data, and update the parameters of the discriminator D; input the virtual data generated by the generator G into the discriminator D, calculate the loss of the generator G, and update the parameters of the generator G so that the generator G can generate multidimensional virtual data that is closer to the real data in the multidimensional data vector z;

[0032] Use the trained generator G to generate multidimensional virtual data with a sample size of Q, and use the minimum value of the input parameter in the generated multidimensional virtual data as the lower limit of the input parameter value range , the maximum value of the input parameter is used as the upper limit of the input parameter value range , the minimum value of the model parameter is used as the lower limit of the model parameter value range , the maximum value of the model parameter is used as the upper limit of the model parameter value range .

[0033] Furthermore, the sampling steps of the noise vector are as follows:

[0034] Define the mean vector μ, where μ=[μ1, μ2, ..., μ d ] T ; where μ d is the mean of the d-th dimension data;

[0035] Define the d×d dimensional covariance matrix ∑, where ∑= , It is expressed as the covariance between the p-th dimension data and the q-th dimension data, and p=1,2,...,d; q=1,2,...,d;

[0036] Numerical methods are used to sample noise vectors from the normal distribution of multidimensional data.

[0037] Furthermore, the sample size Q of the multidimensional virtual data generated by the generator G satisfies the relationship: Q ≥ C / ε 2 ; Where C is the capacity of the generator G; ε is the difference between the data distribution of the virtual data generated by the generator G and the data distribution of the real data.

[0038] The present invention has the following beneficial effects:

[0039] Compared with the existing technology, this technical solution uses the forward model to obtain the fitting data of the prediction indicators under the measured input parameters and model parameters in the historical flood events, uses the measured data and the fitting data to calculate the prediction error rate of the forward model for the prediction indicators, constructs a neural network model with the normalized values of the input parameters and model parameters as input data and the prediction error rate as output data, obtains the prediction error rate of the neural network model when the normalized values of the random combination within the parameter range are used as input data, and constructs a data mapping table. By searching the data mapping table for the same input parameters as the new flood event, the prediction error rate of the input parameters and all prediction indicators is obtained. The model parameters with the minimum weighted value of the difference rate are used as the driving data of the forward model to obtain the fitting data of the prediction indicators of new flood events. By fitting the historical data with the neural network model, the construction of the neural network model and the data fitting prediction process can be completed in the non-flood season, thereby establishing a data mapping table. When a new flood event occurs, the various setting parameters of the model can be quickly determined by directly searching the data mapping table, realizing the cyclic drive of the forward model. When conducting flood warning and emergency response, the calculation can be completed quickly and timely, so that the model status is quickly updated, the timeliness of the model fitting is improved, and the real-time or near real-time hydrodynamic simulation needs are met. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 A schematic flow chart of a two-dimensional hydrodynamic model data assimilation method according to the present invention;

[0041] Figure 2 The figure is a structural diagram of a two-dimensional hydrodynamic model data assimilation system according to the present invention. DETAILED DESCRIPTION

[0042] The present invention will be further described below in conjunction with specific embodiments. The accompanying drawings are for illustrative purposes only and represent only schematic diagrams rather than actual drawings. They should not be understood as limiting the present invention. In order to better illustrate the specific embodiments of the present invention, some parts of the drawings may be omitted, enlarged or reduced, and do not represent the size of the actual product.

[0043] The specific implementation process of the technical solution of the present invention includes the following steps:

[0044] Step 1: Collect the measured data of the prediction indicators of historical flood events in the monitoring area , substitute the measured data into the forward model of flood events, and use the forward model to obtain the input parameters of , the model parameters are Fitting the predictors to the data in context ,in, It represents the data at time t in the flood event; u, v, and w are the numbers of the prediction index, input parameter, and model parameter, respectively;

[0045] Among them, the prediction indicators include water level, flow and inundation range;

[0046] Input parameters include rainfall-runoff parameters and terrain parameters;

[0047] The model parameters include the Manning roughness coefficient of the monitoring area, boundary condition parameters, calculation unit parameters, time step parameters and model operation parameters.

[0048] In this embodiment, the forward model is LISFLOOD-FP model as an example for explanation.

[0049] In the LISFLOOD-FP model, the following parameters need to be set:

[0050] Manning roughness coefficient

[0051] Manning roughness coefficient is an important parameter in hydraulics and hydrology, which is used to describe the friction resistance between water flow and riverbed, river bank or inner wall of pipe.

[0052] Terrain parameters

[0053] DEM data: A digital elevation model (DEM) is one of the fundamental inputs to the model, used to define the terrain of the computational domain. The resolution and accuracy of the DEM directly impact the accuracy of the simulation results.

[0054] Boundary condition parameters

[0055] Boundary condition files: These files include upstream flow boundary (.bci file) and downstream water level boundary (.bdy file). These files define the input and output conditions of the model and are key to simulating flood propagation.

[0056] Point source boundary conditions: For example, the inflow of a rainwater inspection well can be calculated using the weir flow formula and input into the model as a point source boundary condition.

[0057] Calculation unit parameters

[0058] Calculation unit division: The entire monitored area is divided into structural calculation units according to the terrain grid. Each calculation unit contains information such as elevation, roughness coefficient, infiltration parameters, percentage of impervious area and depth of depression.

[0059] Land use type classification: Each calculation unit is divided into permeable areas and impermeable areas according to the land use type or surface cover type, and runoff calculations are performed separately.

[0060] Rainfall runoff parameters

[0061] Rainfall intensity: Rainfall intensity is an important input parameter for simulating urban waterlogging and affects surface runoff.

[0062] Infiltration: In permeable areas, runoff is equal to rainfall intensity minus infiltration losses, so infiltration is a key parameter.

[0063] Time step parameter

[0064] Time step: According to the CFL condition, the time step is linearly proportional to the grid cell size to ensure the stability of the numerical simulation.

[0065] Model running parameters

[0066] Model operation file: includes the settings of model operation parameters, such as total simulation time, time step, etc.

[0067] After setting the LISFLOOD-FP model parameters, prepare the input file:

[0068] Prepare input files, including DEM files, boundary condition files (.bci and .bdy), channel information files (.river), etc. These files need to be organized and named according to the format requirements of the LISFLOOD-FP model.

[0069] Model Execution:

[0070] Run the simulation using the LISFLOOD-FP model executable. The model will simulate the temporal and spatial evolution of flooding based on the input DEM, boundary conditions, and channel information.

[0071] Result Analysis

[0072] Output file interpretation:

[0073] Model outputs include parameters such as water depth and flow velocity in raster format, as well as predicted water levels and hydrographs. Output files are typically provided at hourly resolution, facilitating analysis of the spatiotemporal evolution of flooding.

[0074] Step 2: Using measured data and fitting data Calculate the prediction error rate of the forward model for the u-th prediction indicator at time t in the flood event , the calculation formula is: | | / ×100%; where t∈T; T is the total fitting time of the forward model for the flood event.

[0075] Step 3: Build to input parameters , model parameters The normalized value is the input data, with the prediction error rate An LSTM neural network model is constructed to output data and trained by adjusting the weight parameters, bias parameters, and hyperparameters of the neural network model until its accuracy is not less than the set threshold. The hyperparameters include the hidden layer size, learning rate, and number of training rounds.

[0076] Step 4: Set the range of input parameters of the forward model [ , ] and the range of model parameter values[ , ].

[0077] Specifically, the process of determining the value range is as follows:

[0078] Step 41: Using the input parameters of historical flood events and model parameters Construct a multidimensional data vector z, z=[ , ] T , and the multidimensional data vector z~N(μ,∑), where μ is the d-dimensional mean vector, representing the expected value of each data dimension; ∑ is the d×d covariance matrix, representing the covariance relationship between the data dimensions; d= ;

[0079] Step 42: Construct the normal distribution probability density function of the multidimensional data vector z ,in, = ; is the inverse matrix of the covariance matrix; π is the pi;

[0080] Step 43: Sampling a noise vector from the normal distribution of the constructed multidimensional data vector z. The specific process is as follows: I. Define the mean vector μ, where μ = [μ1, μ2, ..., μ d ] T ; where μ d is the mean of the d-th dimension data; II. Define the d×d-dimensional covariance matrix ∑, where ∑= , Expressed as the covariance between the p-th dimension data and the q-th dimension data, and p = 1, 2, ..., d; q = 1, 2, ..., d; III. Using numerical methods to sample noise vectors from the normal distribution of multidimensional data;

[0081] Step 44: Construct a generative adversarial network model, wherein the generator G of the generative adversarial network model takes the sampled noise vector as input and outputs the generated multidimensional virtual data vector; the discriminator D of the generative adversarial network model takes the normalized value of the multidimensional data vector z as input and outputs the probability that the generated multidimensional virtual data vector is real data;

[0082] Step 45: Use the real data in the multidimensional data vector z and the virtual data in the multidimensional virtual data vector generated by the generator to train the discriminator D, calculate the loss of the real data and the loss of the generated data, and update the parameters of the discriminator D; input the virtual data generated by the generator G into the discriminator D, calculate the loss of the generator G, and update the parameters of the generator G so that the generator G can generate multidimensional virtual data that is closer to the real data in the multidimensional data vector z;

[0083] Step 46: Use the trained generator G to generate multidimensional virtual data with a sample size of Q, and use the minimum value of the input parameter in the generated multidimensional virtual data as the lower limit of the input parameter value range , the maximum value of the input parameter is used as the upper limit of the input parameter value range , the minimum value of the model parameter is used as the lower limit of the model parameter value range , the maximum value of the model parameter is used as the upper limit of the model parameter value range .

[0084] Step 5: Get a random combination of parameters within the range ,in, ∈[ , ]; ∈[ , ], get the neural network model to randomly combine The prediction error rate when the normalized value of is used as input data , build a data mapping table → .

[0085] Step 6: Search the data mapping table for the input parameters of the new flood event at time t Same input parameters , to input parameters The model parameter that takes the minimum weighted value of the prediction error rate of all prediction indicators As the driving data of the forward model, the prediction error rate weighted value = , It is expressed as the weight of the u-th prediction index. In this embodiment, the weight coefficients of water level, flow rate and flooding range can be taken as 1 / 3 of the value when they are equal. Then the weighted value of the prediction error rate = ;

[0086] At this time, enter the parameters For the collected data of the new flood event at time t, when the weighted value of the prediction error rate is minimized, the data mapping table is searched → , we can determine the optimal model parameters of the LISFLOOD-FP model at this time , the input parameters of the new flood event collected at time t and the optimal model parameters As the driving data of the LISFLOOD-FP model at this time, the fitting data of various prediction indicators of the LISFLOOD-FP model can be obtained , and so on, to realize the data assimilation process.

[0087] The technical solution of the present invention utilizes a neural network model to fit historical data, and can complete the construction of the neural network model and the data fitting and prediction process in the non-flood season, thereby establishing a data mapping table. When a new flood event occurs, the various setting parameters of the model can be quickly determined by directly searching the data mapping table, thereby realizing the cyclic drive of the forward model. When conducting flood warnings and emergency responses, the calculation can be completed quickly and timely, so that the model status is quickly updated, the timeliness of the model fitting is improved, and the real-time or near real-time hydrodynamic simulation needs are met.

[0088] The basic principles, main features, and advantages of the present invention are shown and described above. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are merely illustrative of the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and modifications are intended to fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. A two-dimensional hydrodynamic model data assimilation method, characterized in that: include: Collect measured data on prediction indicators of historical flood events in the area to be monitored ; The forward model is used to obtain the input parameters , the model parameters are Fitting the predictors to the data in context ; Using measured data and fitting data Calculate the prediction error rate of the forward model for the u-th prediction indicator at time t in the flood event ; Build with input parameters , model parameters The normalized value is the input data, with the prediction error rate A neural network model for output data; Set the range of input parameters of the forward model [ , ] and the range of model parameter values[ , ], get a random combination of parameters within the value range ; Construct a data mapping table, the data mapping table is a random combination Mapping to a neural network model with random combination The prediction error rate when the normalized value of is used as input data ; Search the data mapping table for the input parameters of the new flood event at time t Same input parameters , to input parameters The model parameter that takes the minimum weighted value of the prediction error rate of all prediction indicators As the driving data of the forward model, obtain the corresponding prediction index fitting data .

2. A two-dimensional hydrodynamic model data assimilation method according to claim 1, characterized in that: Said prediction indicators include water level, flow and inundation extent; The input parameters include rainfall runoff parameters and terrain parameters; The model parameters include the Manning roughness coefficient of the monitoring area, boundary condition parameters, calculation unit parameters, time step parameters and model operation parameters.

3. A two-dimensional hydrodynamic model data assimilation method according to claim 1, characterized in that: Forecast error rate The calculation formula is: - | / ×100%.

4. A two-dimensional hydrodynamic model data assimilation method according to claim 1, characterized in that: The neural network model is an LSTM neural network model, and the hyperparameters of the neural network model include hidden layer size, learning rate and number of training rounds.

5. A two-dimensional hydrodynamic model data assimilation method according to claim 1, characterized in that: The process of determining the value range of input parameters and model parameters includes the following steps: Using the input parameters of historical flood events and model parameters Construct a multidimensional data vector z, z=[ , ] T , and the multidimensional data vector z~N(μ,∑), where μ is the d-dimensional mean vector, representing the expected value of each data dimension; ∑ is the d×d covariance matrix, representing the covariance relationship between the data dimensions; d= ; Construct the normal distribution probability density function of the multidimensional data vector z ,in, = ; is the inverse matrix of the covariance matrix; π is the pi; Sampling a noise vector from a normal distribution of the constructed multidimensional data vector z; Constructing a generative adversarial network model, wherein the generator G of the generative adversarial network model takes the sampled noise vector as input and the generated multidimensional virtual data vector as output; the discriminator D of the generative adversarial network model takes the normalized value of the multidimensional data vector z as input and the probability that the generated multidimensional virtual data vector is real data as output; Use the real data in the multidimensional data vector z and the virtual data in the multidimensional virtual data vector generated by the generator to train the discriminator D, calculate the loss of the real data and the loss of the generated data, and update the parameters of the discriminator D; input the virtual data generated by the generator G into the discriminator D, calculate the loss of the generator G, and update the parameters of the generator G so that the generator G can generate multidimensional virtual data that is closer to the real data in the multidimensional data vector z; Use the trained generator G to generate multidimensional virtual data with a sample size of Q, and use the minimum value of the input parameter in the generated multidimensional virtual data as the lower limit of the input parameter value range , the maximum value of the input parameter is used as the upper limit of the input parameter value range , the minimum value of the model parameter is used as the lower limit of the model parameter value range , the maximum value of the model parameter is used as the upper limit of the model parameter value range .

6. A two-dimensional hydrodynamic model data assimilation method according to claim 5, characterized in that: The sampling steps of the noise vector are as follows: Define the mean vector μ, where μ=[μ1, μ2, ..., μ d ] T ; where μ d is the mean of the d-th dimension data; Define the d×d dimensional covariance matrix ∑, where ∑= , It is expressed as the covariance between the p-th dimension data and the q-th dimension data, and p=1,2,...,d; q=1,2,...,d; Numerical methods are used to sample noise vectors from normal distribution of multidimensional data.

7. A two-dimensional hydrodynamic model data assimilation method according to claim 5, characterized in that: The sample size Q of the multidimensional virtual data generated by the generator G satisfies the relationship: Q ≥ C / ε 2 ; Where C is the capacity of the generator G; ε is the difference between the data distribution of the virtual data generated by the generator G and the data distribution of the real data.

8. A two-dimensional hydrodynamic model data assimilation system, characterized by: include: Data acquisition module, used to collect the measured data of the prediction indicators of historical flood events in the monitored area ; The data fitting module is used to use the forward model to obtain the input parameters , the model parameters are Fitting the predictors to the data in context ; Fitting data processing module, used to utilize measured data and fitting data Calculate the prediction error rate of the forward model for the u-th prediction indicator at time t in the flood event ; Neural network building blocks for constructing neural networks with input parameters , model parameters The normalized value is the input data, with the prediction error rate LSTM neural network model for output data; Mapping data acquisition module for constructing Mapping to a neural network model with random combination The prediction error rate when the normalized value of is used as input data Data mapping table; Drive the data acquisition module to search the data mapping table for the input parameters of the new flood event at time t Same input parameters , to input parameters The model parameter that takes the minimum weighted value of the prediction error rate of all prediction indicators As the driving data of the forward model, obtain the corresponding prediction index fitting data .

9. A two-dimensional hydrodynamic model data assimilation storage medium, characterized in that: The storage medium stores an electronic program, wherein the electronic program, when executed by a processor, can implement the steps of a two-dimensional hydrodynamic model data assimilation method according to any one of claims 1 to 7.

Images