River channel evolution model construction method and device, electronic equipment and readable storage medium

Abstract

The invention provides a river channel evolution model construction method and device, electronic equipment and a readable storage medium, and relates to the technical field of intelligent water conservancy. By constructing normal prior distribution of riverbed roughness parameters and combining a customized likelihood function and a Bayesian inference framework, accurate estimation of posterior distribution of the parameters is realized, the defects that a traditional method only outputs a single optimal solution, is easy to sink into local optimum and cannot quantify uncertainty are overcome, and the parameter calibration precision is remarkably improved. A self-adaptive and multi-chain parallel dynamic simulation algorithm is adopted, parallel sampling is carried out, the step length is dynamically adjusted, and the sampling efficiency and the convergence speed in a complex posterior space are greatly improved. Through deep coupling of a hydrodynamic error model and a likelihood function, the consistency of the model and an actual hydraulic process is enhanced. Parameter uncertainty is propagated to simulation results, mean values, variances and confidence intervals of water level, flow and other prediction results are constructed, and a scientific and reliable uncertainty quantification basis is provided for flood early warning and scheduling decision making.

Description

Technical Field

[0001] This invention relates to the field of intelligent water conservancy technology, and in particular to a method, apparatus, electronic device, and readable storage medium for constructing a river evolution model. Background Technology

[0002] In the field of water conservancy engineering, river evolution models are widely used in decision-making scenarios such as flood forecasting and water resource allocation. Riverbed roughness, as a key hydrodynamic parameter, has a decisive impact on the accuracy of model output, but it is difficult to measure directly and usually needs to be calibrated using historical data.

[0003] Existing parameter calibration methods mostly employ deterministic optimization techniques (such as least squares), which can only output a single optimal parameter value. This presents several technical problems: First, the accuracy and uncertainty characterization of parameter estimation are insufficient. When faced with complex posterior distributions that are multimodal, nonlinear, and highly correlated, they are prone to getting trapped in local optima and cannot provide confidence intervals for prediction results, making it difficult to provide a reliable basis for high-risk decisions such as flood warnings. Second, the error modeling method is not sufficiently coupled with hydraulic characteristics. Traditional methods use the assumption of fixed variance noise, neglecting level-discharge conversion errors, response hysteresis effects, and heteroscedasticity, leading to a disconnect between the error model and the hydrodynamic equations. Third, the sampling efficiency and convergence performance are insufficient to meet engineering requirements. The lack of multi-chain parallelism and adaptive adjustment mechanisms results in low computational efficiency and slow convergence speed in complex scenarios such as large watersheds and long sequences.

[0004] Therefore, there is an urgent need for a method to construct a river evolution model that can couple hydrodynamic characteristics, quantify parameter uncertainties, and improve sampling efficiency. Summary of the Invention

[0005] The purpose of this invention is to provide a method, apparatus, electronic device, and readable storage medium for constructing a river channel evolution model. By constructing a normal prior distribution of the riverbed roughness parameter and combining it with a customized likelihood function and a Bayesian inference framework, accurate estimation of the parameter posterior distribution is achieved. This overcomes the shortcomings of traditional methods, which only output a single optimal solution, are prone to local optima, and cannot quantify uncertainties, thus significantly improving the accuracy of parameter calibration. An adaptive and multi-chain parallel dynamic simulation algorithm is employed, with parallel sampling and dynamic adjustment of the step size, greatly improving sampling efficiency and convergence speed in complex posterior spaces. Deep coupling between the hydrodynamic error model and the likelihood function enhances the consistency between the model and actual hydraulic processes. Parameter uncertainties are propagated to the simulation results, constructing the mean, variance, and confidence intervals of predicted results such as water level and flow rate, providing a scientifically reliable basis for uncertainty quantification for flood warning and dispatching decisions.

[0006] In a first aspect, the present invention provides a method for constructing a river channel evolution model, comprising: The raw data within the target area is acquired and standardized, missing values ​​are imputed, and outliers are detected to form standardized time series data. The raw data includes at least historical water level monitoring data and historical flow monitoring data. Based on the basic topographic data of river cross-section scanning, riverbed soil type, bank slope structure and vegetation coverage, combined with the external hydraulic driving field of water level and flow, a river evolution model is constructed, and the driving field is used to simulate the river evolution model to obtain the river flow response process under multiple rainfall scenarios. Based on riverbed soil type, particle size composition, vegetation cover and preset hydraulic empirical formula, a prior distribution of riverbed roughness parameter is constructed. The prior distribution adopts a normal distribution. The mean and variance of the normal distribution are determined by consulting empirical parameter intervals, historical calibration results and literature statistics to describe the reasonable range of riverbed roughness parameter under the condition of no observation constraints. Based on the river channel evolution model and prior distribution, observational data are introduced into the Bayesian inference framework. By constructing a hydrodynamic error model and combining it with hydrodynamic characteristics, a pre-defined customized likelihood function is used to reflect the actual observation error. The posterior distribution of the riverbed roughness parameter is updated by sampling through an adaptive and multi-chain parallel dynamic simulation algorithm. The uncertainty of the posterior distribution of the riverbed roughness parameter is propagated to the hydrodynamic simulation results of the river channel. Parameter samples are generated by sampling from the posterior distribution, and each parameter sample is input into the river evolution model to obtain multiple sets of simulation results for water level, flow rate, and velocity. The simulation results are then aggregated and statistically analyzed to construct the mean prediction, variance, and confidence intervals with pre-set confidence levels, thereby quantifying the uncertainty of the model prediction.

[0007] In some preferred embodiments of the present invention, the steps of standardizing the original data, imputing missing values, and detecting outliers include: Based on the Min-Max standardization method, the original data is scaled to the range [0, 1] using the following formula to achieve dimension unification: ; in, This represents the original value of the i-th variable at time t; Let be the standardized value of the i-th variable at time t; For variables Minimum value during the entire acquisition period; For variables The maximum value during the entire acquisition period; The missing values ​​in the standardized data are imputed using the following linear interpolation formula to obtain complete time series data. The calculation formula is as follows: ; in, It is the value after filling in the blanks; It is time The known standard value at that location; It is time The known standard value at that location; For missing moments; Given time a; Given time b; Outlier detection based on Z-Score statistics is used to identify data points that significantly deviate from the data distribution in the complete time series data. Outliers exceeding a preset threshold are replaced with the neighborhood mean. The formula for calculating the Z-Score is as follows: ; ; ; in, A sliding window with a preset fixed time interval; The mean value within the sliding window; The standard deviation within the sliding window; This is the Z-Score value.

[0008] In some preferred embodiments of the present invention, the river evolution model is constrained by the following Saint-Venant equations: ; ; Where Q is the instantaneous cross-sectional flow rate of the river section, in m³. 3 / s; A is the cross-sectional area of ​​the river channel, in m². 2 x represents the length of the river segment in meters (m); t represents time in seconds (s). B represents the lateral inflow of water; B is the width of the water surface in meters; g is the acceleration due to gravity. For riverbed slope; The coefficient of friction; This represents the lateral inflow velocity.

[0009] In some preferred embodiments of the present invention, the prior distribution of the riverbed roughness parameter is constrained by the following formula: ; in, denoted as , where μ is the mean value of the riverbed roughness parameter. The variance of the riverbed roughness parameter; Let be the prior probability density of the riverbed roughness parameter.

[0010] In some preferred embodiments of the present invention, the customized likelihood function is constrained by the following formula: ; in, For customized likelihood functions; This refers to the error variance; Let i be the i-th observation; is the i-th output of the river evolution model; N is the number of samples.

[0011] In some preferred embodiments of the present invention, the step of sampling and updating the posterior distribution of the riverbed roughness parameter using an adaptive and multi-chain parallel dynamic simulation algorithm includes: The posterior distribution of the unnormalized riverbed roughness parameter is determined based on the prior distribution of the riverbed roughness parameter and the customized likelihood function. An adaptive Markov chain Monte Carlo algorithm based on multi-chain parallelism is used for sampling. All sampling chains are run in parallel, and the step size is gradually adjusted based on the residual value and the preset acceptance rate during the sampling process until all sampling chains converge. The initial parameters of each sampling chain are sampled from the prior distribution. After all sampling chains converge, the posterior distribution of the riverbed roughness parameter is constructed based on the sampling results of all sampling chains.

[0012] In some preferred embodiments of the present invention, the uncertainty of the posterior distribution of the riverbed roughness parameter is propagated to the river hydrodynamic simulation results. This is achieved by generating parameter samples from the posterior distribution, inputting each parameter sample into the river evolution model, obtaining multiple sets of simulation results for water level, flow rate, and velocity, and performing ensemble statistics on the simulation results to construct mean prediction, variance, and confidence intervals with pre-set confidence levels. This process quantifies the uncertainty of the model prediction. The steps include: A parameter sample set is generated by sampling from the posterior distribution of the riverbed roughness parameter; Each sample is input into the river evolution model and run to obtain hydrodynamic simulation results; Statistical analysis was performed on all sample results at each time point and spatial location; the analysis results included: mean prediction, variance, and confidence intervals at the confidence level.

[0013] Secondly, the present invention provides a river channel evolution model construction device, comprising: The observation data processing module is used to acquire raw data within the target area and perform standardization, missing value imputation, and outlier detection on the raw data to form standardized time series data. The raw data includes at least historical water level monitoring data and historical flow monitoring data. The river flow evolution module is used to construct a river evolution model based on basic topographic data such as river cross-section scanning data, riverbed soil type, bank slope structure and vegetation coverage, combined with external hydraulic driving fields of water level and flow. The driving field is then used to simulate the river evolution model to obtain the river flow response process under multiple rainfall scenarios. The prior information configuration module is used to construct the prior distribution of the riverbed roughness parameter based on the riverbed soil type, particle size composition, vegetation cover and preset hydraulic empirical formulas. The prior distribution adopts a normal distribution. The mean and variance of the normal distribution are determined by consulting empirical parameter intervals, historical calibration results and literature statistics to describe the reasonable range of values ​​for the riverbed roughness parameter under the condition of no observation constraints. The multi-chain parallel sampling module is used to introduce observation data into the Bayesian inference framework based on the river evolution model and prior distribution. By constructing a hydrodynamic error model and combining hydrodynamic characteristics, it adopts a preset customized likelihood function to reflect the actual observation error. It then uses an adaptive and multi-chain parallel dynamic simulation algorithm to sample and update the posterior distribution of the riverbed roughness parameter. The uncertainty analysis module is used to propagate the uncertainty of the posterior distribution of the riverbed roughness parameter to the hydrodynamic simulation results of the river channel. It generates parameter samples by sampling from the posterior distribution and inputs each parameter sample into the river evolution model to obtain multiple sets of simulation results for water level, flow rate and velocity. It also performs ensemble statistics on the simulation results to construct the mean prediction, variance and confidence intervals with preset confidence levels, so as to quantify the uncertainty of the model prediction.

[0014] Thirdly, the present invention provides an electronic device including a processor and a memory, the memory storing computer-executable instructions that can be executed by the processor, and the processor executing the computer-executable instructions to implement the river evolution model construction method provided in the first aspect above.

[0015] Fourthly, the present invention provides a readable storage medium storing computer-executable instructions, which, when called and executed by a processor, cause the processor to implement the river evolution model construction method provided in the first aspect above.

[0016] This invention brings the following beneficial effects: This invention provides a method, apparatus, electronic device, and readable storage medium for constructing a river channel evolution model. The method includes: acquiring raw data within a target area, and performing standardization, missing value imputation, and outlier detection on the raw data to form standardized time-series data; wherein the raw data includes at least: historical water level monitoring data and historical flow monitoring data; based on basic topographic data such as river cross-section scanning data, riverbed soil type, bank slope structure, and vegetation coverage, combined with external hydraulic driving fields of water level and flow, constructing a river channel evolution model, and simulating the river channel evolution model using the driving field to obtain the river flow response process under multiple rainfall scenarios; constructing a prior distribution of the riverbed roughness parameter based on riverbed soil type, particle size composition, vegetation coverage, and a preset hydraulic empirical formula; wherein the prior distribution adopts a normal distribution, and the mean and variance of the normal distribution are determined by consulting empirical parameter intervals, historical calibration results, and literature statistics to describe the reasonable range of values ​​for the riverbed roughness parameter under no observation constraints; in the river Based on the channel evolution model and prior distribution, observational data is introduced into a Bayesian inference framework. By constructing a hydrodynamic error model and combining it with hydrodynamic characteristics, a pre-defined customized likelihood function is used to reflect the actual observation error. Sampling is performed through an adaptive and multi-chain parallel dynamic simulation algorithm to update the posterior distribution of the riverbed roughness parameter. The uncertainty of the posterior distribution of the riverbed roughness parameter is propagated to the channel hydrodynamic simulation results. Parameter samples are generated by sampling from the posterior distribution, and each parameter sample is input into the channel evolution model to obtain multiple sets of simulation results for water level, flow rate, and velocity. The simulation results are then aggregated and statistically analyzed to construct mean prediction, variance, and confidence intervals with pre-set confidence levels, thereby quantifying the uncertainty of model prediction. By constructing a normal prior distribution of the riverbed roughness parameter and combining it with a customized likelihood function and a Bayesian inference framework, accurate estimation of the parameter posterior distribution is achieved. This overcomes the shortcomings of traditional methods, which only output a single optimal solution, are prone to local optima, and cannot quantify uncertainty, significantly improving the accuracy of parameter calibration. An adaptive and multi-chain parallel dynamic simulation algorithm is employed, with parallel sampling and dynamic step size adjustment, significantly improving sampling efficiency and convergence speed in complex posterior spaces. Deep coupling of the hydrodynamic error model and the likelihood function enhances the consistency between the model and actual hydraulic processes. Parameter uncertainties are propagated to the simulation results, constructing the mean, variance, and confidence intervals of predicted results such as water level and flow rate, providing a scientifically reliable quantitative basis for flood warning and dispatching decisions. Attached Figure Description

[0017] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 A flowchart of a method for constructing a river channel evolution model is provided in this embodiment of the invention; Figure 2 This is a diagram of the simulated flood discharge process at Station A provided in an embodiment of the present invention. Figure 3 This is a simulated flood flow process diagram based on actual measurements from Bilibili provided in an embodiment of the present invention. Figure 4 This is a schematic diagram of a river evolution model construction device provided in an embodiment of the present invention; Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention.

[0019] Icons: 310 - Observation data processing module; 320 - River flow evolution module; 330 - Prior information configuration module; 340 - Multi-chain parallel sampling module; 350 - Uncertainty analysis module; 400 - Memory; 401 - Processor; 402 - Bus; 403 - Communication interface. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0021] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0022] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0023] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship commonly used when the product of this invention is in use. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this invention. In addition, the terms "first," "second," "third," etc., are only used to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0024] Furthermore, terms such as "horizontal," "vertical," and "sag" do not imply that components must be absolutely horizontal or suspended, but rather that they can be slightly tilted. For example, "horizontal" simply means that its direction is more horizontal relative to "vertical," and does not mean that the structure must be completely horizontal, but can be slightly tilted.

[0025] In the description of this invention, it should also be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," and "link" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0026] In the field of hydraulic engineering and hydrodynamic modeling, river evolution models are widely used in decision support scenarios such as flood forecasting, optimal water resource allocation, and river management. These models typically describe the flow patterns based on numerical methods such as the Saint-Venant equations, simulating the spatiotemporal variations of key hydrological elements like water level and flow rate by inputting boundary conditions, topographic data, and hydrodynamic parameters (such as riverbed roughness). Riverbed roughness, as a key parameter characterizing river resistance, has a decisive impact on the accuracy of the model output. However, this parameter is difficult to obtain through direct measurement and usually needs to be estimated using parameter calibration methods based on historical observation data. Existing parameter calibration methods mostly employ deterministic optimization techniques, such as least squares or maximum likelihood estimation, by searching for a single optimal parameter value that minimizes the simulation error.

[0027] However, the above-mentioned traditional methods have the following technical problems in practical applications: First, the accuracy and uncertainty characterization of parameter estimation are insufficient. Traditional deterministic optimization methods can only output a single optimal parameter value and cannot quantitatively describe the uncertainty of the parameter estimation results. In hydrodynamic models, due to observational data noise, model structure errors, and complex coupling relationships between parameters, the posterior distribution of parameters often exhibits complex characteristics such as multimodality, nonlinearity, and high-dimensional correlation. Single optimization methods are insufficient to fully explore such complex distribution spaces, easily getting trapped in local optima, causing the parameter estimation results to deviate from the actual physical process, thus affecting the accuracy of model predictions. Especially in high-risk decision-making scenarios such as flood warnings, the lack of a quantitative description of the confidence interval of the prediction results makes it impossible for decision-makers to assess the credibility of the forecast results and to formulate scientific and reasonable scheduling plans and emergency response plans.

[0028] Second, the error modeling approach is insufficiently coupled with hydraulic characteristics. Existing methods typically employ a fixed-variance Gaussian noise assumption to handle observation errors, neglecting the unique error structure characteristics inherent in hydrodynamic observations. For example, hydrodynamic characteristics such as the conversion error between water level and flow rate, the hysteresis effect of river flow response to rainfall events, and the heterogeneity of observation error variance at different water level levels are difficult to effectively reflect in traditional error models. This simplified error handling leads to a loose coupling between the observation error model and the hydrodynamic equations, failing to fully characterize the spatiotemporal structure of model biases and reducing the physical consistency and reliability of parameter calibration results.

[0029] Third, the parameter sampling efficiency and convergence performance are insufficient to meet the needs of engineering applications. In complex scenarios such as large watersheds, multiple cross-sections, long sequences, and multiple operating conditions, the computational cost of a single run of a river evolution model is high. Traditional parameter calibration methods typically require a large number of repetitive model simulations, resulting in low overall computational efficiency. Furthermore, existing methods generally lack multi-chain parallel mechanisms and adaptive sampling strategies, failing to dynamically adjust the search step size and proposal distribution based on residual changes and acceptance rates. This leads to slow convergence speed and low sampling efficiency in complex posterior spaces, making it difficult to meet the timeliness requirements of engineering applications while ensuring calibration accuracy.

[0030] This invention provides a method, apparatus, electronic device, and readable storage medium for constructing a river channel evolution model. It quantifies the uncertainty of parameter calibration and prediction in the river channel evolution model based on a customized likelihood function. Compared with existing technologies, it has the following technical advantages: Customized Likelihood Function Coupled with Hydrodynamic Constraints: By introducing a customized likelihood function oriented towards hydrodynamic characteristics, factors such as water level-discharge conversion errors, observation lag effects, and river hydraulic transport characteristics are explicitly incorporated into the Bayesian inference framework, enhancing the model's sensitivity to changes in hydrodynamic characteristics. Compared to traditional methods such as least squares, this approach not only provides optimal parameter values ​​but also quantifies parameter uncertainties, significantly improving parameter estimation accuracy under complex posterior distributions (such as multimodal and nonlinear distributions).

[0031] Multi-chain parallel and adaptive MCMC sampling: A multi-chain parallel MCMC mechanism is introduced in the parameter sampling process, which enables multiple chains to search the posterior space synchronously under different initial values, avoiding getting trapped in local optima and significantly improving the take-off and take-off speed. The adaptive step size and covariance matrix adjustment strategy dynamically optimizes the proposal distribution according to the error structure and parameter changes during the sampling process, so that the sampling process automatically adapts to complex, nonlinear or multimodal posterior distributions, improving sampling efficiency and stability.

[0032] Parameter uncertainty quantification and output propagation: Sampled parameters from the posterior distribution are input one by one into the river evolution model to generate multiple sets of simulation results for water level, flow rate, and velocity. Aggregate statistics of the simulation results yield the mean, variance, and confidence intervals, thus clarifying the impact of parameter uncertainty on hydrodynamic processes. This function provides quantitative evidence for flood warning, joint scheduling, and risk management, making decision-making more reliable and scientific.

[0033] Overall, the above approach significantly improves the accuracy and reliability of river evolution model parameter calibration while maintaining physical consistency. Furthermore, by incorporating uncertainty quantification results, it provides a scientific basis for decision-making in engineering applications, including flood warning, scheduling optimization, and risk management, balancing accuracy and computational efficiency.

[0034] The following detailed description of some embodiments of the present invention is provided in conjunction with the accompanying drawings. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0035] Example 1 This invention provides a method for constructing a river channel evolution model, see [link to relevant documentation]. Figure 1 The illustrated embodiment of the present invention provides a flowchart of a method for constructing a river channel evolution model, the method comprising: Step S102: Obtain the raw data within the target area, and perform standardization processing, missing value imputation, and outlier detection on the raw data to form standardized time series data; wherein, the raw data includes at least: historical water level monitoring data and historical flow monitoring data.

[0036] Specifically, sensors are deployed within the target area using IoT devices and remote sensing technology to collect historical monitoring data such as water level and flow rate. The collected raw data is transmitted to a central platform via wireless or wired networks. To ensure data accuracy, data cleaning and preprocessing techniques are used to standardize the raw data, fill in missing data, and detect outliers. The processed data is then transmitted to a data storage platform, forming standardized time-series data, providing a high-quality data foundation for subsequent analysis and decision-making. Through systematic data collection and preprocessing, noise and anomalies in the raw data are eliminated, ensuring the integrity and consistency of the time series, and providing reliable data support for the subsequent construction of river evolution models and parameter calibration.

[0037] Furthermore, in some preferred embodiments of the present invention, the steps of standardizing the original data, imputing missing values, and detecting outliers include: scaling the original data to the range [0, 1] using the Min-Max standardization method according to the following formula to obtain standardized data: ;in, This represents the original value of the i-th variable at time t; Let be the standardized value of the i-th variable at time t; For variables Minimum value during the entire acquisition period; For variables The maximum value during the entire acquisition period; missing values ​​in the standardized data are filled using the following linear interpolation formula to obtain complete time series data. The calculation formula is: ;in, It is the value after filling in the blanks; It is time The known standard value at that location; It is time The known standard value at that location; For missing moments; Given time a; Given time b, outlier detection is performed on the complete time series data using Z-Score statistics. Data points that significantly deviate from the data distribution are identified, and outliers exceeding a preset threshold are replaced with the neighborhood mean. The Z-Score value is calculated using the following formula: ; ; ;in, A sliding window with a preset fixed time interval; The mean value within the sliding window; The standard deviation within the sliding window; This is the Z-Score value.

[0038] Specifically, if Then determine These are outliers, where This is a preset outlier threshold. For data points identified as outliers, the mean or median within their time neighborhood can be used to replace them, thus completing the outlier correction.

[0039] For example, a one-dimensional river evolution model is constructed using a typical river section in a certain region, and parameter calibration and uncertainty quantification are performed. The river section extends from section A (the starting point of the Nanliao River) to section B (the ending point of the Liao River), with a river length of 20 km, 579 sections, and 6 monitoring stations. The observation data uses the flow series from these monitoring stations, with a time step of 1 hour. In the boundary conditions, the upstream boundary uses the Nanliao River inflow process, and the downstream boundary uses the Liao River water level process. The initial condition uses the initial water level field obtained from steady-state preheating calculations. Based on riverbed soil type, particle size distribution, vegetation cover, and bank slope structure, the river is divided into 8 representative river sections, and a Manning roughness parameter is set for each representative section to form a parameter vector. ,in Its prior distribution is determined by fusing normative / literature intervals, historical calibration statistics, and scene feature mapping. A customized heteroscedasticity likelihood function is employed. The parameters were updated using Bayesian methods, and posterior sampling was performed using adaptive MCMC with 10 parallel chains. The total number of valid posterior samples was 400. The sampling convergence criterion was adopted. And effective sample size Substituting the posterior samples into the hydrodynamic solver yields the output sets of the water level and flow processes. The 5%–95% confidence interval is calculated using the quantile method, and confidence bands are plotted. Simultaneously, uncertainty evaluation indicators such as observation coverage and interval width are calculated.

[0040] Step S104: Based on the basic topographic data of river cross-section scanning data, riverbed soil type, bank slope structure and vegetation coverage, combined with the external hydraulic driving field of water level and flow, a river evolution model is constructed, and the driving field is used to simulate the river evolution model to obtain the river flow response process under multiple rainfall scenarios.

[0041] Specifically, based on basic topographic data such as river cross-section survey data, riverbed soil type, bank slope structure, and vegetation cover, and combined with external hydraulic driving fields such as water level and flow rate, a river evolution model is constructed. The river evolution model is then simulated using these driving fields to obtain the river flow response process under various rainfall scenarios. The river flow process is numerically calculated by solving the continuity equation and momentum equation at the nodes, i.e., the Saint-Venant equations, to obtain the calculated results of hydrodynamic parameters such as river water level and flow rate. In the one-dimensional hydrodynamic model, the river is typically discretized into several calculation cross-sections, each described by variables such as water level and flow rate. The most accurate values ​​are obtained by solving the Saint-Venant equations. The constructed river evolution model accurately reflects the flow movement patterns under complex topographic and hydrological conditions, providing a physical basis for subsequent parameter calibration. Furthermore, the simulation results from multiple scenarios can be used to verify the model's generalization ability.

[0042] Furthermore, in some preferred embodiments of the present invention, the river evolution model is constrained by the following Saint-Venant equations: ; Where Q is the instantaneous cross-sectional flow rate of the river section, in m³ / s. 3 / s; A is the cross-sectional area of ​​the river channel, in m². 2 x represents the length of the river segment in meters (m); t represents time in seconds (s). B represents the lateral inflow of water; B is the width of the water surface in meters; g is the acceleration due to gravity. For riverbed slope; The coefficient of friction; This represents the lateral inflow velocity.

[0043] The boundary conditions of the river flow evolution model can be flexibly configured based on regional hydrological and topographical features, including using upstream inflow processes as upstream boundary conditions, downstream water level processes as downstream boundary conditions, and lateral inflow processes as lateral boundary conditions. At the implementation level, the model sets up two typical boundary condition configuration methods: flow process lines and water level process lines, to adapt to different types of regional characteristics and more realistically simulate the river flow evolution process under actual operating conditions.

[0044] Under the constraints of the aforementioned governing equations, the river flow evolution model, driven by historical boundary conditions, performs a reproducible simulation of typical historical processes, using historical water levels as a basis. Compared with historical traffic As the primary state variable, output any cross-section of the river channel. Time-varying sequences within historical processes .in, Time step The time resolution can be as fine as the minute.

[0045] Specifically, the Manning formula is used to calculate the coefficient of friction. As shown in the following formula: ; Where n is the channel roughness coefficient; R is the hydraulic radius, in meters.

[0046] Step S106: Based on the riverbed soil type, particle size composition, vegetation cover, and preset hydraulic empirical formula, construct the prior distribution of the riverbed roughness parameter. The prior distribution adopts a normal distribution. The mean and variance of the normal distribution are determined by consulting empirical parameter intervals, historical calibration results, and literature statistics to describe the reasonable range of riverbed roughness parameter under the condition of no observation constraints.

[0047] Specifically, a prior distribution of the riverbed roughness parameter is constructed based on riverbed soil type, particle size composition, vegetation cover, and relevant hydraulic empirical formulas. The prior distribution adopts a normal distribution. The mean (μ) and variance (σ²) of the normal distribution are determined by consulting empirical parameter intervals, historical calibration results, and literature statistics. This normal distribution describes the reasonable range of riverbed roughness values ​​under unobserved constraints. The normal distribution is used to represent the potential range of variation in riverbed roughness, and its mean and variance are adjusted using these historical data and empirical parameter intervals, thereby quantifying the uncertainty of the roughness. By integrating multi-source empirical information to construct the prior distribution, the uncertainty range of the parameter is reasonably quantified, providing prior knowledge consistent with engineering experience for Bayesian inference, which contributes to the stability and accuracy of subsequent posterior updates.

[0048] Furthermore, in some preferred embodiments of the present invention, the prior distribution of the riverbed roughness parameter is constrained by the following formula: ;in, is the riverbed roughness parameter; μ is the mean value of the riverbed roughness parameter, representing the average value provided by historical data and empirical formulas; The variance of the riverbed roughness parameter reflects its possible range of variation or uncertainty. Let be the prior probability density of the riverbed roughness parameter.

[0049] Specifically, the river channel is divided based on the acquired information. First, based on measured cross-sectional data and riverbed sediment sampling, the soil type and particle size composition of each cross-section are determined, and the median particle size (D50) and the proportion of coarse particles are used to characterize the roughness of the substrate. Second, the vegetation cover and vegetation type of the riverbanks and beaches are obtained based on remote sensing imagery and field surveys, serving as the basis for segmentation of the resistance addition. Third, the river segments are morphologically classified according to the bank slope structure and slope gradient. Finally, the target river channel is divided into several representative segments using consistency rules of similar substrate, similar vegetation cover, and similar bank slope structure, ensuring that the substrate friction mechanism remains consistent within the same segment, thus providing clear spatial support for the subsequent construction of roughness a priori distribution and calibration parameter vectors for segment division.

[0050] Subsequently, based on the riverbed soil type and particle size composition, and referring to hydraulic design specifications (Hydraulics Handbook, River Engineering Model Test Procedures) and publicly available literature, a recommended range for Manning roughness consistent with the riverbed type was selected. Table 1 shows the recommended ranges for Manning roughness corresponding to common riverbed types.

[0051] Table 1

[0052] Next, historical calibration results of existing river evolution models for the same river system or similar river sections are collected to form a roughness sample set, and the sample mean and sample variance are calculated. Then, vegetation cover and bank slope structure are used as correction factors to perform consistency verification and fine-tuning on the base interval. The higher and denser the vegetation, the greater the roughness. As the vegetation cover increases from 10% to 50%, the roughness increases by approximately 0.01-0.03. Finally, the prior central value of each representative river section is obtained ( ) and fluctuation scale ( ).

[0053] Reasonable ranges for riverbed roughness parameters are determined for various river sections. For each type of river section, precision parameters can be obtained. experience range This includes the sample mean and fluctuation range obtained from historical calibration. This information is used to provide a basis for the mean and variance of the prior distribution.

[0054] In this embodiment, with Let represent the riverbed roughness parameter (e.g., Manning roughness coefficient), assuming it follows a one-dimensional normal distribution under no observational constraints. Let the prior distribution be: ; In this embodiment, the median of the empirical interval is used as... The initial estimate, and determined based on the interval width. .

[0055] For each representative river segment, the corresponding rough prior distribution parameter pair is finally obtained. The parameters are stored in a parameterized form in the model configuration library, forming a prior information set of riverbed roughness parameters. In subsequent Bayesian inference, the above prior distribution is combined with the likelihood function formed by observed water level, flow rate, and other data to update the posterior distribution of the roughness parameters. This allows for a roughness parameter estimate that better reflects the actual hydraulic characteristics of the river channel, while taking into account both historical experience and real-time observation constraints.

[0056] Step S108: Based on the river channel evolution model and prior distribution, the observation data is introduced into the Bayesian inference framework. By constructing a hydrodynamic error model and combining hydrodynamic characteristics, a preset customized likelihood function is used to reflect the actual observation error. The posterior distribution of the riverbed roughness parameter is updated by sampling through an adaptive and multi-chain parallel dynamic simulation algorithm.

[0057] Specifically, based on the river evolution model and prior distribution, observational data is incorporated into a Bayesian inference framework. Within this framework, a hydrodynamic error model is first constructed, incorporating hydrodynamic characteristics (such as level-discharge conversion errors and hysteresis effects), and employing a customized likelihood function to reflect actual observational errors. Then, an adaptive and multi-chain parallel MCMC algorithm is used for sampling to update the posterior distribution of parameters. The adaptive step size and proposal distribution adjustment mechanism enable the sampling process to automatically optimize based on changes in the error structure and observational data, thereby improving sampling efficiency and ensuring the accuracy of the final posterior distribution. The Bayesian framework organically combines prior information with observational data, achieving quantitative updates to parameter uncertainties; the adaptive multi-chain parallel algorithm significantly improves sampling efficiency and convergence speed, effectively handling complex posterior distributions.

[0058] Furthermore, in some preferred embodiments of the present invention, the customized likelihood function is constrained by the following formula: ;in, For customized likelihood functions; This refers to the error variance; Let i be the i-th observation; is the i-th output of the river evolution model; N is the number of samples.

[0059] Specifically, an observation error model based on hydrodynamic characteristics is constructed: based on the statistical characteristics of monitoring data and the hydraulic response law of the river channel, an observation error model that can reflect the changes in observation error with water level and flow rate is established.

[0060] The riverbed roughness parameter vector is: ; in, Let represent the roughness of the i-th river segment or the i-th type of riverbed, and the parameter vector at this point. It is the parameter vector of the state to be calibrated.

[0061] In this embodiment, the observed quantities and simulated quantities correspond to the isomorphic observation mapping values ​​output by the monitoring system and the river evolution model, respectively. The simulated quantities are solved using the river evolution model: as parameter vectors... Using historical operating conditions as input, the hydrodynamic solution outputs the entire field's water level and flow rate changes, and then maps the observation stations to the model cross-sections. Record the observations at the station corresponding to the i-th river segment. For the i-th observation Output of the river evolution model Define residual: ; Assume the residuals follow a distribution with zero mean: ; in, It can be set as a function that changes with water level or flow rate based on data characteristics.

[0062] ; In the formula, a and b are the first and second non-negative coefficients set through statistical analysis or experience, respectively.

[0063] This error model can more realistically reflect the impact of measurement noise, systematic errors, and river channel complexity on observation data, laying the foundation for the subsequent construction of the likelihood function.

[0064] For river evolution model based on parameters The simulated output. The residual scale RMS is defined based on the residuals. ; in, The number of time steps or the total number of observed samples This represents the residual at time t. Residual scale. It can reflect the magnitude of the simulation error caused by the current parameters, and serves as a direct basis for subsequent adaptive step size and covariance scaling.

[0065] Constructing the likelihood function under hydrodynamic constraints: Based on the error model, parameters are established under the constraints of the channel evolution model. The likelihood function is used to measure the consistency between the model simulation results and the observation data. The likelihood function is directly coupled with the results of the river hydraulic calculation. It can utilize the internal structural information of the river evolution model during the inference process to improve the sensitivity of parameter calibration to the river process.

[0066] Simulated values ​​are obtained by combining river evolution models with hydrological data. The likelihood function uses the accumulation of the squared observation error to measure the parameters. The degree of fit to the observed data. By embedding the hydrodynamic solver into the likelihood function, the likelihood value is ensured to strictly follow the channel dynamics structure during the calculation process. Any parameter combination that does not conform to the hydrodynamic equation will produce abnormal results, which will then be automatically eliminated through likelihood, thereby significantly improving the sensitivity of parameter calibration to the spatial variation of riverbed roughness.

[0067] In this embodiment, the simulation operator is defined as a "combined operator that maps parameters to historical conditions and then to the observation space." It consists of a continuity equation and momentum equation solving module from the river evolution model, and an observation mapping module (the mapping relationship between observation stations and model cross sections). The hydrodynamic solution operator takes as input a parameter vector, boundary conditions BC (upstream inflow process, lateral inflow process, and downstream water level process), and initial conditions IC (initial river water level), and outputs a spatiotemporal state field of hydrodynamic elements. ;set up The observation mapping operator represents the state field in the set of observation spatiotemporal points. Upsample and output the observed isomorphic vector. Construct the solution operator based on the channel evolution model: ; in, This represents a river evolution model based on the continuity and momentum equations, i.e., a hydrodynamic solver, with roughness parameters as input. The output is the corresponding analog quantity such as water level and flow rate. The observation mapping operator represents the correspondence between observation stations and model sections; BC represents the boundary conditions; and IC represents the initial conditions. The input variables, internal processing logic, and output variables of the simulation operator form clear module boundaries in engineering implementation, enabling the likelihood function to directly call this combined operator to complete the closed-loop calculation of "parameters → simulation → residual → likelihood".

[0068] The likelihood function is constructed based on the error model: ; in, For parameters are The probability density of the observed data, i.e., the likelihood function; For parameters The simulation results for the i-th observation are as follows: The aforementioned error variance; The roughness parameter to be calibrated.

[0069] Step S110: The uncertainty of the posterior distribution of the riverbed roughness parameter is propagated to the hydrodynamic simulation results of the river channel. Parameter samples are generated by sampling from the posterior distribution, and each parameter sample is input into the river evolution model to obtain multiple sets of simulation results for water level, flow rate, and velocity. The simulation results are then aggregated and statistically analyzed to construct the mean prediction, variance, and confidence intervals with preset confidence levels, thereby quantifying the uncertainty of the model prediction.

[0070] Specifically, the uncertainty of the posterior distribution of the riverbed roughness parameter is propagated to the hydrodynamic simulation results, quantifying the uncertainty of the model prediction. By sampling from the posterior distribution to generate parameter samples and inputting each sample into the river evolution model, multiple sets of simulation results for water level, flow rate, and velocity can be obtained. Ensemble statistics of these results can construct mean predictions, variances, and confidence intervals, thereby clarifying the impact of parameter uncertainty on the hydrodynamic process. This step provides a quantitative basis for flood warning, scheduling optimization, and risk management, making the decision-making process more reliable and scientific when considering model uncertainty. By propagating parameter uncertainty to the model output, the probability distribution of the prediction results is obtained, enabling a quantitative assessment of the prediction's credibility range and providing a quantitative basis for risk assessment in flood warning and scheduling decisions.

[0071] Furthermore, in some preferred embodiments of the present invention, the step of sampling and updating the posterior distribution of the riverbed roughness parameter through an adaptive and multi-chain parallel dynamic simulation algorithm includes: determining the posterior distribution of the unnormalized riverbed roughness parameter based on the prior distribution of the riverbed roughness parameter and a customized likelihood function; sampling using an adaptive Markov chain Monte Carlo algorithm based on multi-chain parallelism, running all sampling chains in parallel, and gradually adjusting the step size based on the residual value and a preset acceptance rate during the sampling process until all sampling chains converge; wherein, the initial parameters of each sampling chain are sampled from the prior distribution; after all sampling chains converge, the posterior distribution of the riverbed roughness parameter is constructed based on the sampling results of all sampling chains.

[0072] Specifically, the likelihood and the prior knowledge of construction are compared. Combining these yields the unnormalized posterior: .

[0073] Multi-chain initialization: Suppose K Markov chains are running in parallel, each running independently with initial parameters sampled from a prior distribution. ; The proposed distribution is obtained by using an adaptive Metropolis (Monte Carlo) algorithm based on random walks: ; ; in, The candidate parameter vector is generated by the proposal distribution; Let be the step size factor of the k-th chain in the t-th iteration; It is the covariance matrix that is dynamically updated based on chain history sampling.

[0074] In this formula, the parameter vector is directly used as the core variable of "current state → candidate state" and enters the hydrodynamic solution and likelihood calculation through the candidate state, realizing a complete closed loop from proposal distribution to acceptance-rejection judgment.

[0075] Covariance is updated using the following formula: ; in, Let be the current parameter vector of the k-th chain in the t-th iteration; Let be the random perturbation vector of the t-th iteration, whose distribution is controlled by the adaptive covariance matrix and the step size factor. is a small constant with positive definiteness; I is the identity matrix.

[0076] The step size factor is adaptively adjusted based on the deviation between the current acceptance rate and the target acceptance rate, so that the sampling process can achieve a balance between exploration and convergence.

[0077] The step size is automatically adjusted based on the acceptance rate and the magnitude of the residual; that is, the increase or decrease of the step size is determined by whether the acceptance rate deviates from the target value. ; in: Let be the actual acceptance rate of the k-th chain in the t-th iteration; The target acceptance rate is typically set to 0.2-0.4 in some preferred embodiments of the present invention. The learning rate decreases with the number of iterations.

[0078] Combining the aforementioned hydrodynamic constraint likelihood function, the Metropolis-Hastings criterion is constructed: ; During the sampling process, K chains explore the posterior parameter space in parallel from different initial points, and the convergence status is determined using conventional inter-chain diagnostic metrics. This embodiment of the invention employs multi-chain parallel sampling and sets a convergence criterion to determine the sampling termination time. Specifically, for each dimension of the parameter vector θ, the inter-chain consistency diagnostic metric R is calculated. When all parameters are satisfied At this point, it is assumed that different Markov chains have been sampled within the same stable region. Simultaneously, the effective sample size (ESS) is calculated. When the ESS of the key parameter is not less than a preset threshold of 400, the sample size is considered sufficient to support a stable estimate of the posterior statistic. Based on the above statistical criteria, the stability of the mean and variance of the flow fit evaluation index (RMSE) within the most recent iteration windows is further checked. When its fluctuation amplitude is less than a preset threshold of 5, sampling is determined to have entered a steady-state stage and sampling is terminated. Finally, multi-chain samples are merged as a posterior sample set for uncertainty quantification output.

[0079] Simultaneously, convergence diagnosis is performed on the K chains during the sampling process, allowing multiple chains to explore the posterior space simultaneously from different initial points, avoiding getting trapped in local optima and improving convergence speed. After all chains converge, the multi-chain samples are merged to construct the posterior distribution: ; in, The parameter sample set obtained by parallel MCMC sampling is used to analyze the posterior distribution. Numerical approximations are performed, and the estimated parameter values ​​and their uncertainty intervals are calculated accordingly.

[0080] By obtaining the parameter mean, confidence interval, and uncertainty range, parameter calibration and uncertainty quantification are achieved, thereby enabling the calibration and uncertainty quantification of the riverbed roughness parameter, providing more reliable hydrodynamic parameter support for subsequent river flow evolution simulation and scheduling optimization.

[0081] Furthermore, in some preferred embodiments of the present invention, the uncertainty of the posterior distribution of the riverbed roughness parameter is propagated to the hydrodynamic simulation results. This is achieved by sampling from the posterior distribution to generate parameter samples, inputting each parameter sample into the river evolution model to obtain multiple sets of simulation results for water level, flow rate, and velocity. The simulation results are then aggregated and statistically analyzed to construct mean prediction, variance, and confidence intervals at preset confidence levels, thereby quantifying the uncertainty of the model prediction. This includes: generating a parameter sample set by sampling from the posterior distribution of the riverbed roughness parameter; inputting each sample into the river evolution model to obtain hydrodynamic simulation results; and performing statistical analysis on all sample results at each time point and spatial location. The analysis results include mean prediction, variance, and confidence intervals at preset confidence levels. This provides quantitative evidence for flood warning, scheduling optimization, and risk management, making the decision-making process more reliable and scientific when considering model uncertainties.

[0082] Specifically, from the posterior distribution of the above parameters In this process, Monte Carlo sampling is used to generate the parameter sample set: ; Wherein, N is the number of samples, which can be set based on the model accuracy requirements. In some preferred embodiments of the present invention, N=1000-5000.

[0083] For each sample Input river evolution model The simulation was run, and the hydrodynamic results in time and space were obtained: ; Output variables may include water level h, flow rate Q, etc.

[0084] Statistical analysis was performed on all sample results at each time t and spatial location x: Calculate the mean: ; Calculate the variance: ; Confidence interval: ; in, This is the 0.05N value; It is the 0.95N value.

[0085] Furthermore, the confidence interval is calculated using the "posterior sample-driven model output set" as input, and the output is the quantile interval at the specified confidence level. Let's assume that the parameter sample set is obtained through posterior sampling. , where M is the sample size used for statistical calculations. For any The output is obtained by inputting it into the hydrodynamic solver and mapping it through observation. Therefore, an output sample sequence is formed for any fixed point (t, x). Sort it in ascending order to get Then the 5% and 95% quantiles are respectively taken as , Thus, an uncertainty range of 5%–95% is obtained. The confidence band can be obtained by repeating the process for all time points and locations.

[0086] Based on the uncertainty propagation results driven by posterior samples, the system outputs the mean predictions, variances, and confidence intervals of key hydrological quantities such as water level processes, flow processes, and peak flood characteristics. This quantitatively characterizes the transmission relationship from the uncertainty of hydrodynamic parameters such as riverbed roughness to the uncertainty of hydrodynamic response. Specifically, the mean prediction reflects the most likely hydrodynamic response level under current observation constraints, while the variance and confidence interval reflect the dispersion and reliability range of the prediction results. This allows for the identification of key river sections and critical time periods with significant uncertainties, and the calculation of the risk level (such as exceedance probability or risk classification) of peak flood levels and flow rates exceeding warning thresholds or engineering control levels. Based on this quantitative information, in flood warning scenarios, the upper bound of the confidence interval can be used as a conservative prediction basis to formulate more robust warning thresholds and response triggering conditions. In scheduling optimization scenarios, the confidence interval or exceedance probability can be introduced into scheduling constraints and objective functions to form risk-oriented scheduling rules. In watershed risk management scenarios, indicators such as uncertainty bandwidth and coverage can be used to assess the weaknesses of the monitoring system and model parameters, guiding further monitoring enhancement and model updates. By explicitly considering model uncertainties during the decision-making process, embodiments of the present invention can support decision-makers in formulating more robust early warning thresholds, scheduling rules, and emergency plans under different risk preferences, thereby improving the reliability and scientific rigor of flood control and water resource allocation decisions. Specifically, see [link to relevant documentation]. Figure 2 The diagram shown in the embodiment of the present invention illustrates the measured simulated flood discharge process at station A. Figure 3 The simulated flood flow process diagram of station B provided in the embodiment of the present invention shows that, from June 27, 2024 to July 8, 2024, the comparison results of the calculated flow and the measured flow of station A and station B after calibration are relatively close, and the error is less than the preset value.

[0087] This invention provides a method for constructing a river channel evolution model, comprising: acquiring raw data within a target area, and performing standardization, missing value imputation, and outlier detection on the raw data to form standardized time series data; wherein the raw data includes at least: historical water level monitoring data and historical flow monitoring data; based on basic topographic data such as river cross-section scanning data, riverbed soil type, bank slope structure, and vegetation coverage, combined with external hydraulic driving fields of water level and flow, constructing a river channel evolution model, and simulating the river channel evolution model using the driving field to obtain the river flow response process under multiple rainfall scenarios; constructing a prior distribution of the riverbed roughness parameter based on riverbed soil type, particle size composition, vegetation coverage, and a preset hydraulic empirical formula; wherein the prior distribution adopts a normal distribution, and the mean and variance of the normal distribution are determined by consulting empirical parameter intervals, historical calibration results, and literature statistics to describe the reasonable range of values ​​for the riverbed roughness parameter under no observation constraints; and in the river channel evolution model and prior... Based on the distribution, observational data are introduced into a Bayesian inference framework. By constructing a hydrodynamic error model and combining it with hydrodynamic characteristics, a pre-defined customized likelihood function is used to reflect the actual observation error. Sampling is performed through an adaptive and multi-chain parallel dynamic simulation algorithm to update the posterior distribution of the riverbed roughness parameter. The uncertainty of the posterior distribution of the riverbed roughness parameter is propagated to the river hydrodynamic simulation results. Parameter samples are generated by sampling from the posterior distribution, and each parameter sample is input into the river evolution model to obtain multiple sets of simulation results for water level, flow rate, and velocity. The simulation results are then aggregated and statistically analyzed to construct mean prediction, variance, and confidence intervals with pre-set confidence levels, thereby quantifying the uncertainty of the model prediction. By constructing a normal prior distribution of the riverbed roughness parameter and combining it with a customized likelihood function and a Bayesian inference framework, accurate estimation of the parameter posterior distribution is achieved. This overcomes the shortcomings of traditional methods, which only output a single optimal solution, are prone to local optima, and cannot quantify uncertainty, significantly improving the accuracy of parameter calibration. An adaptive and multi-chain parallel dynamic simulation algorithm is employed, with parallel sampling and dynamic step size adjustment, significantly improving sampling efficiency and convergence speed in complex posterior spaces. Deep coupling of the hydrodynamic error model and the likelihood function enhances the consistency between the model and actual hydraulic processes. Parameter uncertainties are propagated to the simulation results, constructing the mean, variance, and confidence intervals of predicted results such as water level and flow rate, providing a scientifically reliable quantitative basis for flood warning and dispatching decisions.

[0088] Example 2 Based on the above embodiments, this invention provides a river channel evolution model construction device, see [link to previous embodiment]. Figure 4 The diagram shown is a structural schematic of a river evolution model construction device provided in an embodiment of the present invention. The device includes: The observation data processing module 310 is used to acquire raw data within the target area and perform standardization, missing value imputation, and outlier detection on the raw data to form standardized time series data; wherein, the raw data includes at least: historical water level monitoring data and historical flow monitoring data; The river flow evolution module 320 is used to construct a river evolution model based on basic topographic data such as river cross-section scanning data, riverbed soil type, bank slope structure and vegetation coverage, combined with external hydraulic driving fields of water level and flow. The driving field is used to simulate the river evolution model to obtain the river flow response process under multiple rainfall scenarios. The prior information configuration module 330 is used to construct the prior distribution of the riverbed roughness parameter based on the riverbed soil type, particle size composition, vegetation cover and preset hydraulic empirical formulas. The prior distribution adopts a normal distribution. The mean and variance of the normal distribution are determined by consulting empirical parameter intervals, historical calibration results and literature statistics to describe the reasonable range of values ​​for the riverbed roughness parameter under the condition of no observation constraints. The multi-chain parallel sampling module 340 is used to introduce observation data into a Bayesian inference framework based on the river evolution model and prior distribution. By constructing a hydrodynamic error model and combining hydrodynamic characteristics, it adopts a preset customized likelihood function to reflect the actual observation error. It performs sampling through an adaptive and multi-chain parallel dynamic simulation algorithm to update the posterior distribution of the riverbed roughness parameter. The uncertainty analysis module 350 is used to propagate the uncertainty of the posterior distribution of the riverbed roughness parameter to the hydrodynamic simulation results of the river channel. It generates parameter samples by sampling from the posterior distribution and inputs each parameter sample into the river evolution model to obtain multiple sets of simulation results for water level, flow rate and velocity. It also performs ensemble statistics on the simulation results to construct the mean prediction, variance and confidence intervals with preset confidence levels, so as to realize the quantification of the uncertainty of the model prediction.

[0089] Furthermore, in some preferred embodiments of the present invention, the observation data processing module 310 is used for the steps of standardizing, imputing missing values, and detecting outliers in the original data, including: scaling the original data to the range [0, 1] using the Min-Max standardization method according to the following formula to obtain standardized data: ;in, This represents the original value of the i-th variable at time t; Let be the standardized value of the i-th variable at time t; For variables Minimum value during the entire acquisition period; For variables The maximum value during the entire acquisition period; missing values ​​in the standardized data are filled using the following linear interpolation formula to obtain complete time series data. The calculation formula is: ;in, It is the value after filling in the blanks; It is time The known standard value at that location; It is time The known standard value at that location; For missing moments; Given time a; Given time b, outlier detection is performed on the complete time series data using Z-Score statistics. Data points that significantly deviate from the data distribution are identified, and outliers exceeding a preset threshold are replaced with the neighborhood mean. The Z-Score value is calculated using the following formula: ; ; ;in, A sliding window with a preset fixed time interval; The mean value within the sliding window; The standard deviation within the sliding window; This is the Z-Score value.

[0090] Furthermore, in some preferred embodiments of the present invention, the river flow evolution module 320 is used to constrain the river evolution model by the following Saint-Venant equations: ; Where Q is the instantaneous cross-sectional flow rate of the river section, in m³ / s. 3 / s; A is the cross-sectional area of ​​the river channel, in m². 2 x represents the length of the river segment in meters (m); t represents time in seconds (s). B represents the lateral inflow of water; B is the width of the water surface in meters; g is the acceleration due to gravity. For riverbed slope; The coefficient of friction; This represents the lateral inflow velocity.

[0091] Furthermore, in some preferred embodiments of the present invention, the prior information configuration module 330 is used to constrain the prior distribution of the riverbed roughness parameter based on the following formula: ;in, denoted as , where μ is the mean value of the riverbed roughness parameter. The variance of the riverbed roughness parameter; Let be the prior probability density of the riverbed roughness parameter.

[0092] Furthermore, in some preferred embodiments of the present invention, the multi-chain parallel sampling module 340 is used to constrain the customized likelihood function by the following formula: ;in, For customized likelihood functions; This refers to the error variance; Let i be the i-th observation; is the i-th output of the river evolution model; N is the number of samples.

[0093] Furthermore, in some preferred embodiments of the present invention, the uncertainty analysis module 350 is used to update the posterior distribution of the riverbed roughness parameter by sampling through an adaptive and multi-chain parallel dynamic simulation algorithm, including: determining the posterior distribution of the unnormalized riverbed roughness parameter based on the prior distribution of the riverbed roughness parameter and a customized likelihood function; sampling using an adaptive Markov chain Monte Carlo algorithm based on multi-chain parallelism, running all sampling chains in parallel, and gradually adjusting the step size based on the residual value and a preset acceptance rate during the sampling process until all sampling chains converge; wherein, the initial parameters of each sampling chain are sampled from the prior distribution; after all sampling chains converge, the posterior distribution of the riverbed roughness parameter is constructed based on the sampling results of all sampling chains.

[0094] Furthermore, in some preferred embodiments of the present invention, the uncertainty analysis module 350 is used to propagate the uncertainty of the posterior distribution of the riverbed roughness parameter to the hydrodynamic simulation results of the river channel. This is achieved by sampling from the posterior distribution to generate parameter samples, inputting each parameter sample into the river channel evolution model for execution, obtaining multiple sets of simulation results for water level, flow rate, and velocity, and performing aggregate statistics on the simulation results to construct mean prediction, variance, and confidence intervals at preset confidence levels. The steps include: sampling from the posterior distribution of the riverbed roughness parameter to generate a parameter sample set; inputting each sample into the river channel evolution model for execution to obtain hydrodynamic simulation results; and performing statistical analysis on all sample results at each time point and spatial location. The analysis results include: mean prediction, variance, and confidence intervals at preset confidence levels.

[0095] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the river evolution model construction device described above can be referred to the corresponding process in the aforementioned embodiments of the river evolution model construction method, and will not be repeated here.

[0096] Example 3 This invention also provides an electronic device for running a method for constructing a river channel evolution model; see [link to related documentation]. Figure 5The schematic diagram of an electronic device provided by the embodiment of the present invention shown below includes a memory 400 and a processor 401. The memory 400 is used to store one or more computer instructions, which are executed by the processor 401 to implement the above-mentioned method for constructing a river evolution model.

[0097] Furthermore, Figure 5 The electronic device shown also includes a bus 402 and a communication interface 403. The processor 401, the communication interface 403 and the memory 400 are connected via the bus 402.

[0098] The memory 400 may include high-speed random access memory (RAM) and may also include non-volatile memory, such as at least one disk storage device. Communication between this system network element and at least one other network element is achieved through at least one communication interface 403 (which can be wired or wireless), such as the Internet, wide area network, local area network, metropolitan area network, etc. The bus 402 can be an ISA bus, PCI bus, or EISA bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 5 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.

[0099] Processor 401 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of processor 401 or by instructions in software form. Processor 401 can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this invention. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this invention can be directly manifested as execution by a hardware decoding processor, or execution by a combination of hardware and software modules in the decoding processor. The software module can reside in a readily available storage medium in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory 400, and processor 401 reads information from memory 400 and, in conjunction with its hardware, completes the steps of the method described in the foregoing embodiments.

[0100] This invention also provides a readable storage medium storing computer-executable instructions. When these computer-executable instructions are called and executed by a processor, they cause the processor to implement the above-described method for constructing a river channel evolution model. For specific implementation details, please refer to the method embodiments, which will not be repeated here.

[0101] The computer program products of the river evolution model construction method, apparatus and electronic equipment provided in the embodiments of the present invention include a readable storage medium storing program code. The instructions included in the program code can be used to execute the methods in the preceding method embodiments. For specific implementation, please refer to the method embodiments, which will not be repeated here.

[0102] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the system and / or device described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0103] Furthermore, in the description of the embodiments of the present invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in the present invention based on the specific circumstances.

[0104] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0105] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for constructing a river channel evolution model, characterized in that, include: The raw data within the target area is acquired, and the raw data is standardized, missing values ​​are imputed, and outliers are detected to form standardized time series data; wherein, the raw data includes at least: historical water level monitoring data and historical flow monitoring data; Based on the basic topographic data of river cross-section scanning, riverbed soil type, bank slope structure and vegetation coverage, combined with the external hydraulic driving field of water level and flow, a river evolution model is constructed, and the driving field is used to simulate the river evolution model to obtain the river flow response process under multiple rainfall scenarios. Based on riverbed soil type, particle size composition, vegetation cover, and a preset hydraulic empirical formula, a prior distribution of riverbed roughness parameter is constructed. The prior distribution adopts a normal distribution, and the mean and variance of the normal distribution are determined by consulting empirical parameter intervals, historical calibration results, and literature statistics to describe the reasonable range of values ​​for the riverbed roughness parameter under the condition of no observation constraints. Based on the river channel evolution model and the prior distribution, the observation data is introduced into the Bayesian inference framework. By constructing a hydrodynamic error model and combining it with hydrodynamic characteristics, a preset customized likelihood function is used to reflect the actual observation error. The posterior distribution of the riverbed roughness parameter is updated by sampling through an adaptive and multi-chain parallel dynamic simulation algorithm. The uncertainty of the posterior distribution of the riverbed roughness parameter is propagated to the hydrodynamic simulation results of the river channel. Parameter samples are generated by sampling from the posterior distribution, and each parameter sample is input into the river channel evolution model for operation. Multiple sets of simulation results for water level, flow rate, and velocity are obtained. The simulation results are then aggregated and statistically analyzed to construct the mean prediction, variance, and confidence intervals with preset confidence levels, thereby quantifying the uncertainty of the model prediction.

2. The method for constructing a river channel evolution model according to claim 1, characterized in that, The steps of standardizing, imputing missing values, and detecting outliers in the raw data include: Based on the Min-Max standardization method, the original data is scaled to the range [0, 1] using the following formula to achieve dimension unification: ； in, This represents the original value of the i-th variable at time t; Let be the standardized value of the i-th variable at time t; For variables Minimum value during the entire acquisition period; For variables The maximum value during the entire acquisition period; The missing values ​​in the standardized data are imputed using the following linear interpolation formula to obtain complete time series data. The calculation formula is as follows: ； in, It is the value after filling in the blanks; It is time The known standard value at that location; It is time The known standard value at that location; For missing moments; Given time a; Given time b; Outlier detection is performed on the complete time series data using Z-Score statistics to identify data points that significantly deviate from the data distribution. Outliers exceeding a preset threshold are replaced with the neighborhood mean. The Z-Score value is calculated using the following formula: ； ； ； in, A sliding window with a preset fixed time interval; The mean value within the sliding window; The standard deviation within the sliding window; This is the Z-Score value.

3. The method for constructing a river channel evolution model according to claim 1, characterized in that, The river evolution model is constrained by the following set of Saint-Venant equations: ； ； Where Q is the instantaneous cross-sectional flow rate of the river section, in m³. 3 / s; A is the cross-sectional area of ​​the river channel, in m². 2 x represents the length of the river segment in meters (m); t represents time in seconds (s). B represents the lateral inflow of water; B is the width of the water surface in meters; g is the acceleration due to gravity. For riverbed slope; The coefficient of friction; This represents the lateral inflow velocity.

4. The method for constructing a river channel evolution model according to claim 1, characterized in that, The prior distribution of the riverbed roughness parameter is constrained by the following formula: ； in, denoted as , where μ is the mean value of the riverbed roughness parameter. The variance of the riverbed roughness parameter; Let be the prior probability density of the riverbed roughness parameter.

5. The method for constructing a river channel evolution model according to claim 1, characterized in that, The customized likelihood function is constrained by the following formula: ； in, For the customized likelihood function; This refers to the error variance; Let i be the i-th observation; is the i-th output of the river evolution model; N is the number of samples.

6. The method for constructing a river channel evolution model according to claim 1, characterized in that, The step of sampling and updating the posterior distribution of the riverbed roughness parameter using an adaptive and multi-chain parallel dynamic simulation algorithm includes: The posterior distribution of the unnormalized riverbed roughness parameter is determined based on the prior distribution of the riverbed roughness parameter and the customized likelihood function. An adaptive Markov chain Monte Carlo algorithm based on multi-chain parallelism is used for sampling. All sampling chains are run in parallel, and the step size is gradually adjusted based on the residual value and the preset acceptance rate during the sampling process until all sampling chains converge. The initial parameters of each sampling chain are sampled from the prior distribution. After all the sampling chains converge, the posterior distribution of the riverbed roughness parameter is constructed based on the sampling results of all the sampling chains.

7. The method for constructing a river channel evolution model according to claim 1, characterized in that, The uncertainty of the posterior distribution of the riverbed roughness parameter is propagated to the river hydrodynamic simulation results. This is achieved by generating parameter samples from the posterior distribution and inputting each parameter sample into the river evolution model to obtain multiple sets of simulation results for water level, flow rate, and velocity. The simulation results are then aggregated statistically analyzed to construct mean prediction, variance, and confidence intervals with pre-set confidence levels, thereby quantifying the uncertainty of the model prediction. The steps include: A parameter sample set is generated by sampling from the posterior distribution of the riverbed roughness parameter; Each sample is input into the river channel evolution model and run to obtain hydrodynamic simulation results; Statistical analysis was performed on all sample results at each time point and spatial location; the analysis results included: mean prediction, variance, and confidence intervals for the stated confidence levels.

8. A device for constructing a river channel evolution model, characterized in that, include: The observation data processing module is used to acquire raw data within the target area and perform standardization, missing value imputation, and outlier detection on the raw data to form standardized time series data; wherein, the raw data includes at least: historical water level monitoring data and historical flow monitoring data; The river flow evolution module is used to construct a river evolution model based on basic topographic data such as river cross-section scanning data, riverbed soil type, bank slope structure and vegetation coverage, combined with external hydraulic driving fields of water level and flow. The driving field is then used to simulate the river evolution model to obtain the river flow response process under multiple rainfall scenarios. The prior information configuration module is used to construct the prior distribution of the riverbed roughness parameter based on the riverbed soil type, particle size composition, vegetation cover and preset hydraulic empirical formulas. The prior distribution adopts a normal distribution, and the mean and variance of the normal distribution are determined by consulting empirical parameter intervals, historical calibration results and literature statistics to describe the reasonable range of values ​​of the riverbed roughness parameter under the condition of no observation constraints. The multi-chain parallel sampling module is used to introduce observation data into a Bayesian inference framework based on the river evolution model and the prior distribution. By constructing a hydrodynamic error model and combining hydrodynamic characteristics, a preset customized likelihood function is used to reflect the actual observation error. The module performs sampling through an adaptive and multi-chain parallel dynamic simulation algorithm to update the posterior distribution of the riverbed roughness parameter. The uncertainty analysis module is used to propagate the uncertainty of the posterior distribution of the riverbed roughness parameter to the river hydrodynamic simulation results. It generates parameter samples by sampling from the posterior distribution and inputs each parameter sample into the river evolution model to obtain multiple sets of simulation results for water level, flow rate and velocity. The simulation results are then aggregated and statistically analyzed to construct the mean prediction, variance and confidence intervals with preset confidence levels, so as to quantify the uncertainty of the model prediction.

9. An electronic device, characterized in that, The system includes a processor and a memory, the memory storing computer-executable instructions that can be executed by the processor, the processor executing the computer-executable instructions to implement the river evolution model construction method according to any one of claims 1 to 7.

10. A readable storage medium, characterized in that, The readable storage medium stores computer-executable instructions, which, when invoked and executed by a processor, cause the processor to implement the river evolution model construction method according to any one of claims 1 to 7.

Images