Runoff model parameter inversion method and device based on multi-response variable correlation prediction

Abstract

The present invention discloses a method and device for inverse parameter estimation of a runoff model based on multi-response variable correlation prediction, including: obtaining hydrological observation data for parameter inversion; preprocessing the hydrological observation data; inputting the preprocessed hydrological observation data, external environmental data, and initial values of model parameters into a runoff model to generate hydrological simulation data; calculating an observation correlation matrix of the hydrological observation data and a simulation correlation matrix of the simulation data respectively; calculating individual fitting errors and correlation structure errors respectively; constructing a composite objective function by using the individual fitting errors and correlation structure errors; performing iterative optimization adjustment of model parameters with the minimization of the composite objective function as the goal, and finally obtaining optimal model parameters. The estimation accuracy and robustness are improved. It is applicable to scenarios with limited or complex data such as mountain flood prediction, and can effectively improve the model prediction accuracy and reliability.

Description

Technical Field

[0001] The present application relates to the technical field of parameter inversion, and particularly to a runoff model parameter inversion method and device based on multi-response variable correlation prediction. Background Art

[0002] In the current field of runoff model parameter inversion, the technical status quo mainly relies on optimization techniques to adjust model parameters to minimize the difference between observed data and simulated data. Common methods include:

[0003] 1. Least squares method: Estimate parameters by minimizing the sum of squared errors between observed values and simulated values.

[0004] 2. Genetic algorithm: A global optimization method based on evolutionary computation that searches for optimal parameters by simulating the natural selection process.

[0005] 3. Bayesian method: Combine prior knowledge and observed data to estimate parameters through posterior probability distribution.

[0006] These methods have been widely applied in the field of hydrology, such as in flood forecasting, water resource management, etc. However, the implementation of existing technologies usually assumes that each response variable is independent, or only integrates the errors of multiple response variables through a simple weighting method, without fully considering the correlation between variables.

[0007] In addition, the existing technologies in runoff model parameter inversion also have the following defects and deficiencies: Traditional methods require a large amount of high-quality observed data to ensure the accuracy of parameter estimation. When data is scarce or there is significant noise, the estimation results are often unreliable. Noise in the observed data (such as measurement errors) will significantly interfere with the optimization process, resulting in large parameter estimation deviations. Optimization algorithms (such as gradient descent or genetic algorithms) may fall into local optimal solutions rather than global optimal solutions, especially when the parameter space is complex. When dealing with large-scale data or multi-variable models, the computational complexity of traditional methods is high, with many iterations and long time consumption. Summary of the Invention

[0008] In view of this, in view of the deficiencies of traditional parameter inversion technologies in data scarcity, noise interference, and local optimal problems, the embodiments of the present application provide a runoff model parameter inversion method and device based on multi-response variable correlation prediction, aiming to improve the accuracy and robustness of runoff model parameter estimation by utilizing the correlation between multiple hydrological variables, especially applicable to hydrological scenarios with scarce data or noise interference.

[0009] According to the first aspect of the embodiments of the present application, a runoff model parameter inversion method based on multi-response variable correlation prediction is provided, including:

[0010] Obtain hydrological observation data for parameter inversion. The hydrological observation data consists of several hydrological variables, and each hydrological variable contains hydrological data;

[0011] Preprocess the hydrological observation data;

[0012] Input the preprocessed hydrological observation data, external environmental data, and initial values of model parameters into a runoff model to generate hydrological simulation data. The model parameters include soil permeability and surface roughness coefficient;

[0013] Calculate the observation correlation matrix of the hydrological observation data and the simulation correlation matrix of the simulation data respectively;

[0014] For each hydrological variable, calculate the sum of squared errors between its hydrological observation data and hydrological simulation data respectively, and then sum up the sum of squared errors of all hydrological variables to obtain an individual fitting error; use the Frobenius norm to measure the difference between the observation correlation matrix and the simulation correlation matrix, multiply the Frobenius norm by a weight factor to obtain a correlation structure error; use the individual fitting error and the correlation structure error to construct a composite objective function;

[0015] Iteratively optimize and adjust the model parameters with the goal of minimizing the composite objective function, and finally obtain the optimal model parameters.

[0016] According to the second aspect of the embodiments of the present application, there is provided a device for inverse runoff model parameters based on multi-response variable correlation prediction, including:

[0017] An acquisition module, configured to acquire hydrological observation data for parameter inversion. The hydrological observation data consists of several hydrological variables, and each hydrological variable contains hydrological data;

[0018] A preprocessing module, configured to preprocess the hydrological observation data;

[0019] A data generation module, configured to input the preprocessed hydrological observation data, external environmental data, and initial values of model parameters into a runoff model to generate hydrological simulation data. The model parameters include soil permeability and surface roughness coefficient;

[0020] A calculation module, configured to calculate the observation correlation matrix of the hydrological observation data and the simulation correlation matrix of the simulation data respectively;

[0021] A function construction module is used to calculate the sum of squared errors between the hydrological observation data and the hydrological simulation data for each hydrological variable respectively, and then sum up the sum of squared errors of all hydrological variables to obtain the individual fitting error; use the Frobenius norm to measure the difference between the observed correlation matrix and the simulated correlation matrix, multiply the Frobenius norm by a weight factor to obtain the correlation structure error; use the individual fitting error and the correlation structure error to construct a composite objective function.

[0022] An optimization iteration module is used to perform iterative optimization and adjustment of the model parameters with the goal of minimizing the composite objective function, and finally obtain the optimal model parameters.

[0023] According to the third aspect of the embodiments of the present application, there is provided an electronic device, including:

[0024] One or more processors;

[0025] A memory for storing one or more programs;

[0026] When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in the first aspect.

[0027] According to the fourth aspect of the embodiments of the present application, there is provided a computer-readable storage medium, on which computer instructions are stored, and when the instructions are executed by a processor, the steps of the method as described in the first aspect are implemented.

[0028] The technical solutions provided by the embodiments of the present application may include the following beneficial effects:

[0029] As can be seen from the above embodiments, the present application makes full use of the relationships between multiple hydrological variables, calculates the observed correlation matrix of the hydrological observation data and the simulated correlation matrix of the simulated data respectively, uses these two correlation matrices to construct the individual fitting error and the correlation structure error respectively, and then constructs a composite objective function including the individual fitting error and the correlation structure error. Finally, the composite objective function is iteratively optimized, and the accuracy and robustness of parameter estimation are ultimately improved, especially showing excellent performance in scenarios of scarce data or noise interference.

[0030] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The accompanying drawings herein are incorporated into the specification and form a part of the specification, showing embodiments consistent with the present application, and are used together with the specification to explain the principles of the present application.

[0032] Figure 1It is a flowchart of a runoff model parameter inversion method based on multi-response variable correlation prediction shown according to an exemplary embodiment.

[0033] Figure 2 It is a block diagram of a runoff model parameter inversion device based on multi-response variable correlation prediction shown according to an exemplary embodiment.

[0034] Figure 3 It is a schematic structural diagram of an electronic device shown according to an exemplary embodiment. Detailed implementation manners

[0035] Here, the exemplary embodiments will be described in detail, and the examples are shown in the drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementation manners described in the following exemplary embodiments do not represent all implementation manners consistent with the present application. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0036] The terms used in the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application. The singular forms "a", "the", and "said" used in the present application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0037] It should be understood that although the terms first, second, third, etc. may be used in the present application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "when" or "while" or "in response to determining".

[0038] Key technical terms and their explanations:

[0039] Runoff model, a mathematical model for simulating terrestrial water flow, used to describe how input conditions such as rainfall and terrain are converted into output quantities such as water flow and water level.

[0040] Parameter inversion is the process of estimating model parameters from observed data, usually by an optimization method to make the model simulation results match the observed data as much as possible.

[0041] Hydrological variables (Response variables), which are the outputs or measurable quantities of the runoff model, such as rainfall, water flow velocity, and water level, etc. These variables are usually jointly affected by the model parameters.

[0042] Figure 1 is a flowchart of a method for inverse inference of runoff model parameters based on multi-response variable association prediction shown according to an exemplary embodiment, as Figure 1 shown. This method is applied to a terminal and may include the following steps:

[0043] S1: Obtain hydrological observation data for parameter inversion. The hydrological observation data is composed of several hydrological variables, and each hydrological variable has hydrological data. This step may include the following sub-steps:

[0044] S11: Real-time monitor hydrological variables through multiple data collection sites distributed in the target basin. The hydrological variables include rainfall, water flow velocity, and water level;

[0045] Specifically, multiple data collection sites (such as hydrological sensors or measurement stations) distributed in the target basin real-time monitor hydrological variables. These variables include rainfall, water flow velocity, and water level, etc.

[0046] S12: For each hydrological variable, generate hydrological observation data y obs =[y1 obs ,y2 obs ,...,yi obs ,...,yn obs , where yi obs represents the measurement time series of the i-th site within a predetermined time period, and n is the total number of hydrological variables.

[0047] Specifically, for each hydrological variable, generate hydrological observation data y obs =[y1 obs ,y2 obs ,...,yi obs ,...,yn obs , where yi obs represents the measurement time series of the i-th site within a predetermined time period, and n is the total number of hydrological variables. For example, in a flash flood monitoring scenario, assume there are three sites A, B, and C located at the upstream, middle, and downstream of a river respectively. Site A may be equipped with a rain gauge and a water level gauge, recording rainfall and water level data every 10 minutes, generating time series such as [10mm, 12mm, 15mm,...] and [0.5m, 0.6m, 0.7m,...]. These data collection sites are usually installed at key positions in the basin and can automatically collect data by built-in sensors and temporarily store it through an internal storage unit.

[0048] The data acquisition station transmits the collected hydrological observation data to the data processing server. The transmission method can be a wireless network (such as 4G / 5G network) or a wired connection (such as a data cable), depending on the geographical location of the station and the communication conditions. For example, Station A packages the data into JSON format (such as {"station":"A","time":"2023-10-01 08:00","rainfall":10,"water_level":0.5}) through a 4G module and uploads it to the server in batches every hour. To ensure data reliability, the data acquisition station attaches a timestamp and a checksum to each data packet, and the server will verify whether the checksum is correct after receiving it.

[0049] The data processing server conducts a preliminary check and storage on the received data. The server first checks the integrity of the data packet, for example, to confirm whether there is data loss (discontinuous timestamps) or format errors. If an anomaly is found (such as the data for a certain hour is missing), the server will record a log and mark the data for that period as pending processing; if the data is normal, it will be stored in the database for subsequent steps. For example, the server may store the data of Stations A, B, and C in tables station_A_data, station_B_data, and station_C_data respectively, and each table contains fields such as timestamp, rainfall, and water level.

[0050] The execution entities of this step are the data acquisition station and the data processing server. The data acquisition station is responsible for monitoring and sending data, and the data processing server is responsible for receiving and preliminary sorting. The finally output observed data y_obs is a vector containing time series of multiple stations and serves as the input for the next preprocessing step.

[0051] S2: Preprocess the hydrological observation data;

[0052] Specifically, the data processing server cleans and normalizes the received hydrological observation data to ensure that the data quality is suitable for subsequent model simulation and analysis. This process needs to be completed step by step in multiple sub-steps.

[0053] The preprocessing may include outlier detection, missing value imputation, normalization, and unified format processing, as follows:

[0054] (1) The data processing server checks for outliers in the observed data. Outliers may be caused by sensor failures or extreme weather conditions, such as water levels below the riverbed height (e.g., -0.1 m) or negative water flow velocities (-0.5 m / s). The server scans each data point against preset physical thresholds (e.g., water level range [0, 10] m, water flow velocity range [0, 5] m / s). If a data point exceeds the range, it is marked as an outlier. For example, if a record at site B shows a water level of -0.2 m, the server will replace it with a missing value marker (e.g., NaN) and record an anomaly log.

[0055] (2) The data processing server processes missing values in the data. For short-term missing data (e.g., no data for two consecutive hours at a site), the server uses linear interpolation to fill in the missing values. The calculation formula is y t =y t-1 +(y t+1 -y t-1 )*(t - t t-1 ) / (t t+1 -t t-1 ), where t is the missing time point. For example, if the water levels at site A at 08:00 and 10:00 are 0.5 m and 0.7 m respectively, the interpolated value at 09:00 is 0.6 m. However, if the missing data exceeds a certain duration (e.g., 6 hours), the server marks that period as unavailable and excludes it from subsequent analysis.

[0056] (3) The data processing server standardizes the data. Since different hydrological variables have different dimensions (e.g., rainfall is measured in millimeters, water level is measured in meters), direct comparison would affect the calculation of correlations. Therefore, the server normalizes each time series to the [0, 1] interval. The calculation formula is (x - x min ) / (x max -x min ), where x min and x max are the minimum and maximum values of the hydrological variable in the entire dataset. For example, for the water level series [0.5, 0.6, 0.7] at site A, the minimum and maximum values are 0.5 and 0.7 respectively. After normalization, it becomes [0, 0.5, 1]. This operation ensures that all variable value ranges are consistent.

[0057] (4) The data processing server organizes the preprocessed data into a unified format, generating the preprocessed hydrological observation data y obsFor example, the water level and water flow velocity data of three stations are integrated into a multi-dimensional array with dimensions (n_stations, n_timesteps, n_variables), where n_stations = 3, n_timesteps is the time step (e.g., 24 hours), and n_variables = 2 (water level and water flow velocity). This step is independently completed by the data processing server, and the preprocessed y obs is used as the input for model simulation.

[0058] S3 inputs the preprocessed hydrological observation data, external environmental data, and initial values of model parameters into the runoff model to generate hydrological simulation data. The model parameters include soil permeability and surface roughness coefficient.

[0059] Specifically, the data processing server prepares the input data for the runoff model. The input data includes the preprocessed hydrological observation data y obs , external environmental data (such as rainfall time series and terrain information represented by digital elevation model), and initial values of model parameters θ. Model parameters θ may include soil permeability (range [0.01, 0.5]), surface roughness coefficient (range [0.01, 0.1]), etc. The initial values can be selected from the literature (e.g., the soil permeability is set to 0.1).

[0060] The hydrological model used can be a distributed hydrological model (such as SWAT or TOPMODEL). The server passes the input data into the hydrological model by calling the model interface. For example, in a flash flood scenario, the model calculates the water level and water flow velocity at each station based on rainfall and terrain to generate hydrological simulation data y sim (θ) = [y1 sim (θ), y2 sim (θ),..., yn sim (θ)]. Assume that the rainfall input at station A is [10mm, 12mm, 15mm], and the model may output a water level sequence [0.6m, 0.65m, 0.8m]. The time step and spatial location of the simulation results are consistent with the hydrological observation data.

[0061] The data processing server can further verify the validity of the simulation data. Check whether the simulation data is within the physically reasonable range (e.g., the water level is non - negative). If an anomaly is found (such as the water level at a certain station is - 0.1 meters), a log is recorded and the parameter configuration is marked as unavailable. However, it is usually assumed that the initial parameters are reasonable, and the simulation data directly enters the next step.

[0062] S4 calculates the observation correlation matrix of the hydrological observation data and the simulation correlation matrix of the simulation data respectively; this step includes the following sub - steps:

[0063] S41: For every two hydrological variables in the hydrological observation data, calculate the Pearson correlation coefficient between them to construct an n×n observation correlation matrix, where n is the total number of hydrological variables;

[0064] Specifically, the data processing server processes the preprocessed hydrological observation data y obs Calculate the correlation matrix R obs . For each pair of response variables (such as the water level at site A and the water flow velocity at site B), the server calculates the Pearson correlation coefficient, and the formula is corr(yi,yj)=cov(yi,yj) / (σi*σj), where cov is the covariance, and σ_i and σ_j are the standard deviations. For example, if the water level sequences at sites A and B are [0, 0.5, 1] and [0.2, 0.6, 0.8] respectively, the calculated correlation coefficient is approximately 0.95. The server traverses all variable pairs to construct an n×n symmetric matrix R obs , where n is the total number of variables (such as 6 variables including the water levels and water flow velocities at 3 sites).

[0065] S42: For every two hydrological variables in the hydrological simulation data, calculate the Pearson correlation coefficient between them to construct an n×n simulation correlation matrix;

[0066] Specifically, similarly, the data processing server uses the same calculation method as in S41 to generate the correlation matrix R sim (θ). Assume the simulated water level sequence is [0.1, 0.55, 0.9], and a similar correlation coefficient matrix is calculated with the observed data. This step ensures that the correlation structure of the simulated data can be compared with the observed data.

[0067] The server checks the numerical stability of the matrix. For example, if the standard deviation of a certain pair of variables is close to zero (the data hardly changes), it will cause the calculation of the correlation coefficient to be unstable. The server will record a warning and fill the corresponding element with a default value (such as 0), but generally assume that there is no such problem after the data is preprocessed.

[0068] S5: For each hydrological variable, calculate the sum of the squared errors between its hydrological observation data and hydrological simulation data respectively, and then sum up the sum of the squared errors of all hydrological variables to obtain the individual fitting error; use the Frobenius norm to measure the difference between the observation correlation matrix and the simulation correlation matrix, multiply the Frobenius norm by a weight factor to obtain the correlation structure error; use the individual fitting error and the correlation structure error to construct a composite objective function.

[0069] Specifically, the data processing server calculates the individual fitting error. For each response variable, calculate the sum of the squared errors between its hydrological observation data and hydrological simulation data respectively: , where ||.||2 is the L2 norm of the time series. Since the data series is in matrix form, the square root of the sum of the squares of all elements of the matrix is in the form of a norm. For example, if the observed water level at site A is [0, 0.5, 1] and the simulated water level is [0.1, 0.55, 0.9], then the error is (0 - 0.1)^2+(0.5 - 0.55)^2+(1 - 0.9)^2 = 0.0225. The server repeats this calculation for all variables to obtain the total error, which is the individual fitting error.

[0070] The data processing server calculates the correlation structure error. The Frobenius norm is used to measure the difference between R obs and R sim (θ): ||R obs -R sim (θ)||^2, which is the square root of the sum of the squares of the differences of the matrix elements. For example, if R obs and R sim (θ) are [[1, 0.95], [0.95, 1]] and [[1, 0.9], [0.9, 1]] respectively, then the difference matrix is [[0, 0.05], [0.05, 0]], and the Frobenius norm is sqrt(0.05^2+0.05^2)≈0.0707. The server multiplies by the weight factor λ (set to 1 for example) to obtain the correlation error term.

[0071] The data processing server combines the two parts of the error and defines the composite objective function:

[0072] ;

[0073] This function requires the model parameter θ to optimize both the numerical fitting and the correlation matching simultaneously. The server calculates the value of F(θ) (such as 0.0225 + 1*0.0707^2≈0.0275) as the basis for the optimization step.

[0074] S6 performs iterative optimization and adjustment of the model parameters with the minimization of the composite objective function as the goal, and finally obtains the optimal model parameters. This step includes the following sub-steps:

[0075] S61: The server initializes the parameter θ. According to the physical meaning of the model parameters, the range of the model parameters is set (such as the soil permeability [0.01, 0.5]), and an initial value of the model parameter is randomly selected from the range (such as 0.1).

[0076] S62: The server updates the parameter using an optimization algorithm (such as the gradient descent method). With the minimization of the composite objective function as the goal, the server updates the model parameters using an optimization algorithm; the server calculates the partial derivative of the composite objective function F(θ) with respect to θ (which can be approximated by numerical differentiation) and updates along the negative gradient direction: θ new= θ - α * ∇F(θ), where α is the learning rate (e.g., 0.01). For example, if ∇F(θ) is 0.2, then θ is updated from 0.1 to 0.098.

[0077] S63: The server iteratively performs optimization. After each update of θ, the model simulation (S3) is rerun to generate a new y sim (θ), and the new R is calculated sim (θ) (S4), F(θ) is updated (S5), and the convergence condition is checked (e.g., the change in F(θ) is less than 0.001). If not converged, continue iterating; if converged, output the optimal parameter θ*, for example, the optimization result may be a soil permeability of 0.15 and a surface roughness coefficient of 0.03.

[0078] As can be seen from the above embodiments, the present application makes full use of the relationships between multiple hydrological variables, calculates the observation correlation matrix of the hydrological observation data and the simulation correlation matrix of the simulation data respectively, constructs the individual fitting error and the correlation structure error using these two correlation matrices, and then constructs a composite objective function containing the individual fitting error and the correlation structure error. Finally, the composite objective function is iteratively optimized to ultimately improve the accuracy and robustness of parameter estimation, especially performing excellently in scenarios of scarce data or noise interference.

[0079] Corresponding to the foregoing embodiments of the runoff model parameter inversion method based on multi-response variable correlation prediction, the present application also provides embodiments of a runoff model parameter inversion device based on multi-response variable correlation prediction.

[0080] Figure 2 is a block diagram of a runoff model parameter inversion device based on multi-response variable correlation prediction shown according to an exemplary embodiment. Referring to Figure 2 , the device includes:

[0081] An acquisition module 1 for acquiring hydrological observation data for parameter inversion, the hydrological observation data being composed of several hydrological variables, and each hydrological variable having hydrological data;

[0082] A preprocessing module 2 for preprocessing the hydrological observation data;

[0083] A data generation module 3 for inputting the preprocessed hydrological observation data, external environment data, and initial values of model parameters into a runoff model to generate hydrological simulation data, the model parameters including soil permeability and surface roughness coefficient;

[0084] A calculation module 4 for calculating the observation correlation matrix of the hydrological observation data and the simulation correlation matrix of the simulation data respectively;

[0085] The function construction module 5 is used to calculate the sum of squared errors between the hydrological observation data and the hydrological simulation data for each hydrological variable respectively, and then sum up the sum of squared errors of all hydrological variables to obtain the individual fitting error; use the Frobenius norm to measure the difference between the observed correlation matrix and the simulated correlation matrix, multiply the Frobenius norm by the weight factor to obtain the correlation structure error; use the individual fitting error and the correlation structure error to construct a composite objective function;

[0086] The optimization iteration module 6 is used to perform iterative optimization adjustment of the model parameters with the goal of minimizing the composite objective function, and finally obtain the optimal model parameters.

[0087] Regarding the device in the above embodiments, the specific ways in which each module performs operations have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0088] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can refer to the descriptions in the method embodiments. The device embodiments described above are only illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of the present application. Those of ordinary skill in the art can understand and implement it without creative efforts.

[0089] Correspondingly, the present application also provides an electronic device, including: one or more processors; a memory for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement the runoff model parameter inversion method based on multi-response variable correlation prediction as described above. As Figure 3 shown, it is a hardware structure diagram of a device for inverting runoff model parameters based on multi-response variable correlation prediction provided by an embodiment of the present invention in any device with data processing capabilities. In addition to Figure 3 the processors and memory shown, any device with data processing capabilities where the device in the embodiment is located usually also includes other hardware according to the actual functions of the device, which will not be elaborated here.

[0090] Correspondingly, the present application further provides a computer-readable storage medium, on which computer instructions are stored. When the instructions are executed by a processor, the runoff model parameter inversion method based on multi-response variable correlation prediction as described above is implemented. The computer-readable storage medium may be an internal storage unit of any device with data processing capabilities described in any of the foregoing embodiments, such as a hard disk or memory. The computer-readable storage medium may also be an external storage device, such as a plug-in hard disk, a Smart Media Card (SMC), an SD card, a Flash Card, etc. equipped on the device. Further, the computer-readable storage medium may also include both an internal storage unit of any device with data processing capabilities and an external storage device. The computer-readable storage medium is used to store the computer program and other programs and data required by any device with data processing capabilities, and may also be used to temporarily store data that has been output or will be output.

[0091] Those skilled in the art will readily conceive of other embodiments of the present application after considering the specification and practicing the content disclosed herein. The present application is intended to cover any variations, uses, or adaptations of the present application, which follow the general principles of the present application and include well-known common general knowledge or conventional technical means in the technical field not disclosed in the present application. The specification and embodiments are only regarded as exemplary, and the true scope and spirit of the present application are pointed out by the claims.

[0092] It should be understood that the present application is not limited to the exact structures described above and shown in the drawings, and various modifications and changes can be made without departing from its scope. The scope of the present application is only limited by the appended claims.

Claims

1. An inversion method for runoff model parameters based on multi-response variable correlation prediction, characterized in that Including: S1: Obtain hydrological observation data for parameter inversion. The hydrological observation data consists of several hydrological variables, and each hydrological variable contains hydrological data; S2: Preprocess the hydrological observation data; S3: Input the preprocessed hydrological observation data, external environmental data, and initial values of model parameters into a runoff model to generate hydrological simulation data. The model parameters include soil permeability and surface roughness coefficient; S4: Calculate the observation correlation matrix of the hydrological observation data and the simulation correlation matrix of the simulation data respectively; S5: For each hydrological variable, calculate the sum of squared errors between its hydrological observation data and hydrological simulation data, and then sum up the sum of squared errors of all hydrological variables to obtain the individual fitting error; Use the Frobenius norm to measure the difference between the observation correlation matrix and the simulation correlation matrix, multiply the Frobenius norm by a weight factor to obtain the correlation structure error; Use the individual fitting error and the correlation structure error to construct a composite objective function; S6: Take the minimization of the composite objective function as the goal to perform iterative optimization adjustment of the model parameters, and finally obtain the optimal model parameters; Among them, calculating the observation correlation matrix of the hydrological observation data and the simulation correlation matrix of the simulation data respectively includes: S41: For every two hydrological variables in the hydrological observation data, calculate the Pearson correlation coefficient between them, and use this to construct an n×n observation correlation matrix, where n is the total number of hydrological variables; S42: For every two hydrological variables in the hydrological simulation data, calculate the Pearson correlation coefficient between them, and use this to construct an n×n simulation correlation matrix.

2. The method according to claim 1, wherein Obtaining hydrological observation data for parameter inversion. The hydrological observation data consists of several hydrological variables, and each hydrological variable contains hydrological data, including: Real-time monitoring of hydrological variables through multiple data acquisition stations distributed in the target basin. The hydrological variables include rainfall, water flow velocity, and water level; For each hydrological variable, generate hydrological observation data y obs =[y1 obs ,y2 obs ,...,yi obs ,...,yn obs , where yi obs represents the measurement time series of the i-th station within a predetermined time period, and n is the total number of hydrological variables.

3. The method according to claim 1, wherein Preprocessing the hydrological observation data includes: Performing outlier detection, missing value imputation, normalization, and unified format processing on the hydrological observation data.

4. The method according to claim 1, wherein The composite objective function is: ； In the formula, F ( θ ) is the objective function, θ is the model parameter, which is the core parameter set that needs to be estimated by optimization in the runoff model, n is the total number of hydrological variables, y i obs is the measurement time series of the i-th station within the predetermined time period, y i sim ( θ ) is the value of the simulated i-th response variable, λ is the weight factor, R obs is the observation correlation matrix, R sim ( θ ) is the simulation correlation matrix.

5. The method according to claim 1, wherein Taking the minimization of the composite objective function as the goal to perform iterative optimization adjustment of the model parameters, and finally obtaining the optimal model parameters, including: S61: According to the physical meaning of the model parameters, set the range of the model parameters, and randomly select the initial values of the model parameters from the range; S62: Taking the minimization of the composite objective function as the goal, use an optimization algorithm to update the model parameters; S63: After each update of the model parameters, repeat S3 - S5 until convergence, and output the optimal model parameters.

6. An apparatus for inverse inversion of runoff model parameters based on multi-response variable correlation prediction, characterized in that, Including: An acquisition module for obtaining hydrological observation data for parameter inversion. The hydrological observation data consists of several hydrological variables, and each hydrological variable contains hydrological data; A preprocessing module for preprocessing the hydrological observation data; A data generation module for inputting preprocessed hydrological observation data, external environmental data, and initial values of model parameters into a runoff model to generate hydrological simulation data, where the model parameters include soil permeability and surface roughness coefficient; A calculation module for calculating an observation correlation matrix of the hydrological observation data and a simulation correlation matrix of the simulation data respectively; A function construction module for calculating the sum of squared errors between the hydrological observation data and the hydrological simulation data for each hydrological variable, and then summing up the sum of squared errors of all hydrological variables to obtain an individual fitting error; using the Frobenius norm to measure the difference between the observation correlation matrix and the simulation correlation matrix, multiplying the Frobenius norm by a weight factor to obtain a correlation structure error; constructing a composite objective function using the individual fitting error and the correlation structure error; An optimization iteration module for iteratively optimizing and adjusting the model parameters with the goal of minimizing the composite objective function to finally obtain the optimal model parameters; Among them, calculating the observation correlation matrix of the hydrological observation data and the simulation correlation matrix of the simulation data respectively includes: S41: Calculating the Pearson correlation coefficient between every two hydrological variables in the hydrological observation data, and constructing an n×n observation correlation matrix with this, where n is the total number of hydrological variables; S42: Calculating the Pearson correlation coefficient between every two hydrological variables in the hydrological simulation data, and constructing an n×n simulation correlation matrix with this.

7. An electronic device, characterized in that, Including: One or more processors; A memory for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1-5.

8. A computer-readable storage medium having computer instructions stored thereon, characterized in that, When the instruction is executed by the processor, it implements the steps of the method according to any one of claims 1-5.

Images