A flood flow prediction method and system based on bias correction bayesian model averaging

Abstract

This invention discloses a method and system for flood discharge prediction based on bias-corrected Bayesian model averaging. The method includes: constructing a hydrodynamic model using acquired river cross-section data, hydrodynamic model parameter data, historical discharge data, or discharge data output from a hydrological model; constructing a model set using the hydrodynamic model; and outputting the discharge prediction results of the model set. For the model set in step 1... K The method of this invention involves considering several candidate models, averaging the Bayesian models based on correction coefficients according to the law of total probability, and constructing a log-likelihood function. The SCE-UA algorithm is then used to maximize the log-likelihood function, calculating the average optimal parameters of the Bayesian models. Based on the calculated average optimal parameters, the final predicted value and confidence interval are calculated. This method can improve the accuracy of flood discharge prediction.

Description

Technical Field

[0001] This invention relates to a flood discharge prediction method based on the average of a bias-corrected Bayesian model, belonging to the field of flood risk analysis. Background Technology

[0002] In flood risk analysis, numerical simulation technology is a key tool for predicting flood flow and analyzing flood propagation characteristics. By simulating the dynamic flow process of floods, the spatiotemporal distribution characteristics of floods within a specific watershed can be revealed, providing important scientific support for flood early warning systems. Based on the results of numerical simulations, flood early warning systems can release flood risk information to the public in real time, providing a basis for decision-making to effectively mitigate flood disasters.

[0003] However, due to uncertainties in model inputs, structural simplification, and parameter estimation errors, numerical simulation results often exhibit certain biases, affecting the accuracy and stability of flood flow prediction. To further improve the reliability of simulation results, prediction methods based on ensemble models have gradually attracted attention. Ensemble models, by combining the prediction results of multiple individual models, can not only reduce the limitations of single-model predictions but also effectively reduce errors, improving the accuracy and robustness of flood flow prediction. However, they cannot reduce the systematic errors of individual models within the model ensemble, thus limiting the improvement in the accuracy of the final prediction results, and the uncertainty of the final estimated confidence interval is too large. Furthermore, traditional Bayesian model averaging methods typically use the EM algorithm to solve for the average parameters of the Bayesian model, but the EM algorithm can only guarantee convergence to a local optimum, not that the parameters are globally optimal.

[0004] Combining bias correction with Bayesian model averaging reduces the systematic errors of individual models within the model ensemble, further optimizing the predictive performance of the integrated model. Bias-corrected Bayesian model averaging corrects the systematic errors between model outputs and observed data, making the predictions of individual models closer to reality. Furthermore, by integrating the corrected multi-model results through probability distribution, it generates a more reliable comprehensive prediction, reducing and quantifying prediction uncertainty. Employing the globally optimal algorithm SCE-UA improves the performance of approximating globally optimal parameter estimates. This method, combining the SCE-UA algorithm and bias-corrected Bayesian model averaging, provides a high-precision and high-reliability technical approach for flood risk analysis. Summary of the Invention

[0005] The technical problem to be solved by this invention is that in the process of post-processing ensemble forecasting using ordinary Bayesian models, when there are systematic errors in the models, the performance of individual models cannot be adjusted, which ultimately leads to insufficient prediction accuracy and excessive uncertainty.

[0006] To address the aforementioned technical problems, this invention proposes a flood discharge prediction method based on the averaging of a bias-corrected Bayesian model, comprising the following steps:

[0007] Step 1: Construct a hydrodynamic model using river cross-section data, hydrodynamic model parameter data, historical flow data, or hydrological model output flow data; then construct a model set using the hydrodynamic model.

[0008] Step 2: For the candidate models in the model set in Step 1, post-process the average of the Bayesian models based on the correction coefficients according to the law of total probability to construct the log-likelihood function.

[0009] Step 3: Maximize the log-likelihood function using the SCE-UA algorithm to calculate the average optimal parameters of the Bayesian model;

[0010] Step 4: Based on the Bayesian model average optimal parameter estimation results calculated in Step 3, calculate the final predicted value and confidence interval.

[0011] The aforementioned flood discharge prediction method based on bias-corrected Bayesian model averaging includes, in step 1:

[0012] (11) Collect river cross-section information, obtain historical flow data recorded by hydrological stations in the basin, or output flow data through hydrological models as upstream basic flow data. Based on basic traffic data Construct a hydrodynamic model;

[0013] (12) Select the appropriate basic roughness coefficient according to the riverbed type and flow characteristics. ;

[0014] (13) Using the obtained flow data, construct a model set using the hydrodynamic model.

[0015] The aforementioned flood discharge prediction method based on bias-corrected Bayesian model averaging includes, in step (13):

[0016] (131) Based on the basic flow determined in step (11) and the basic roughness coefficient determined in step (12) Set the traffic scaling factor and roughness scaling factor Traffic scaling factor Includes 'a' roughness scaling factors. It contains b items;

[0017] (132) The flow input data and the initially set roughness data are scaled and combined to form K=a×b candidate models, which form a model set.

[0018] In the aforementioned method for flood flow prediction based on the average of a bias-corrected Bayesian model, in step (132), the hydrodynamic model includes flow files, geometry files, and operation plan files. The flow files are processed according to the flow scaling factor to obtain a flow files. The roughness in the geometry files is modified to obtain b geometry files. By combining the flow files and geometry files, a total of K = a × b plan files are obtained. By running the K plan files, K model outputs are obtained, and a model set is constructed.

[0019] The aforementioned flood discharge prediction method based on bias-corrected Bayesian model averaging includes, in step 2:

[0020] (21) Using the K candidate models from step 1, establish the posterior distribution of the prediction variable y based on the law of total probability.

[0021] (1)

[0022] in, It is the probability density function predicted by the average probability of the Bayesian model; Given observation data Under the condition that the kth model outputs The posterior probability is the weight. ; Given the output of the k-th model and observation data Under the condition of predictor variables The conditional probability distribution;

[0023] (22) The conditional probability distribution in equation (1) By setting the probability density to Gaussian, the corresponding log-likelihood function is obtained. Simultaneously, the Bayesian model output is corrected using correction coefficients.

[0024] (2)

[0025] in, Let be the log-likelihood function. Where t is the length of the traffic data and t is the time step. The number of candidate models. This represents the model output value at time step t of the k-th model. For the average parameters of the Bayesian model that need to be estimated, This includes the weights of each model. The standard deviation of the probability density for each model Correction coefficients for each model Correction coefficient two ;

[0026] (23) Maximizing equation (2) is equivalent to maximizing the likelihood function, thereby obtaining the estimated average parameters of the Bayesian model. The calculation formula is as follows:

[0027] (3)

[0028] in To maximize the parameter estimates after maximizing the likelihood function.

[0029] The aforementioned flood discharge prediction method based on bias-corrected Bayesian model averaging includes the following steps in step (3):

[0030] (31) Initialize the number of complexes Number of interior points of the complex Sample size The maximum number of iterations of the algorithm, rep;

[0031] (32) Randomly sample within the initially set parameter space For each sample, calculate the negative log-likelihood function value;

[0032] (33) Arrange the negative log-likelihood function values ​​obtained from step (32) in ascending order. Each sample point is stored in buffer D;

[0033] (34) Divide the points in buffer D into... In a complex, each complex contains One point;

[0034] (35) Perform CCE evolution calculations for each complex;

[0035] (36) Arrange the points in the complex in ascending order in buffer D;

[0036] (37) If the convergence condition is met, the process ends and the optimal average parameter of the Bayesian model is output. If the convergence condition is not met, the process returns to step (34) until the convergence condition is met or the maximum number of iterations rep of the algorithm is reached. At this point, the operation stops and the optimal average parameter estimation result is output.

[0037] The aforementioned flood discharge prediction method based on bias-corrected Bayesian model averaging includes, in step (35):

[0038] (351) Initialize CCE evolution, determine the number of points q in the sub-complex, the number of iterations α in the sub-complex, and the number of evolutions β in the complex, with the constraints being 2≤q≤m, α≥1, and β≥1;

[0039] (352) Based on the linear probability distribution, calculate the selection probability of each point within the complex. The calculation formula is:

[0040] (4)

[0041] The points are indexed and sorted from lowest to highest based on their negative log-likelihood function values. The number of interior points of the complex; The normalization factor is the sum of the possibilities for choosing all points, ensuring the selection probability. The sum is 1;

[0042] (353) Based on the selection probability calculated in step (352), random sampling without replacement is performed within the complex, and a total of q points are selected and stored in buffer B. The q points together form a sub-complex, and the position information L at this time is recorded.

[0043] (354) Generate the next iteration population and obtain the final buffer B;

[0044] (355) Replace the sub-complexes within the original complex with the final buffer B, i.e., position L, and sort the complexes in ascending order according to the negative log-likelihood function values;

[0045] (356) Repeat steps (352) to (355) β times.

[0046] The aforementioned flood discharge prediction method based on bias-corrected Bayesian model averaging includes, in step (354):

[0047] (01). Sort buffer B in ascending order of objective function value and find the worst estimate point in B. The centroids of the remaining q-1 points are denoted as ;

[0048] (02). Calculate the reflection point of the worst point. If the reflection point If the parameters are within the feasible space, proceed to step (04); otherwise, proceed to step (03).

[0049] (03). Randomly generated points Replacement reflection point Proceed to step (04);

[0050] (04). Calculate the reflection point The negative log-likelihood function, if the reflection point The negative log-likelihood function is less than the original worst-case scenario. So, taking the reflection point Replace the worst estimate point Otherwise, proceed to step (05);

[0051] (05). Calculation If point The negative log-likelihood function is less than the original worst estimate point. So, with point Replace the worst estimate point Otherwise, randomly generate points. Replace the worst estimate point ;

[0052] (06). Repeat steps (01) to (05) α times to obtain the final buffer B.

[0053] The aforementioned flood discharge prediction method based on bias-corrected Bayesian model averaging includes, in step 4:

[0054] (41) Calculate the final predicted value based on the Bayesian model average parameters and Bayesian model output obtained in step (3):

[0055] (5)

[0056] Here The expected value of the final prediction. , , The optimal parameter values ​​estimated in step (3) are... This is the output of the k-th model in the model set;

[0057] (42) Based on the obtained average parameters and model output of the Bayesian model, calculate the average variance of the Bayesian model, and then obtain the confidence interval of the average final predicted value of the Bayesian model.

[0058] (6)

[0059] This represents the average variance of the Bayesian model. The output of the k-th model in the model set; the weights of each model. Standard deviation of the probability density for each model Correction coefficients for each model Correction coefficient two All of these are the optimal parameter values ​​estimated in step 3;

[0060] (7)

[0061] This represents the confidence interval for the average final predicted value of the Bayesian model. Corresponding confidence level The critical value of the standard normal distribution.

[0062] A computer device / apparatus / system includes a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the above-described method.

[0063] The beneficial effects achieved by this invention are as follows: The method of this invention, employing a Bayesian model averaging method based on correction coefficients, can improve prediction accuracy by integrating the outputs of multiple corrected models while reducing the systematic error of a single model in the model set. Furthermore, since it operates within the framework of Bayesian model averaging, it can also obtain the confidence interval of the prediction results, reflecting the uncertainty of the prediction results. Simultaneously, the SCE-UA algorithm is a global optimization algorithm that avoids convergence to local optima through complex evolution and random search, thus solving the problem of potential convergence to local optima in the high-dimensional and complex problem of solving for Bayesian model averaging parameters. Attached Figure Description

[0064] Figure 1 This is a flowchart of a flood discharge prediction method based on the averaging of a bias-corrected Bayesian model, as described in Embodiment 1 of the present invention.

[0065] Figure 2 The geometric model of the hydrodynamic model established in Embodiment 1 of the present invention, along with the corresponding basic flow data and model parameter diagram;

[0066] Figure 3 This is a graph showing the parameter estimation results obtained using the SCE-UA algorithm in Embodiment 1 of the present invention;

[0067] Figure 4 This is a performance comparison chart of a single model before and after calibration in Embodiment 1 of the present invention;

[0068] Figure 5 This is a comparison chart of the final prediction results and their confidence intervals obtained by averaging the results of the SCE-UA algorithm and the bias-corrected Bayesian model in Embodiment 1 of the present invention, as well as the prediction results of the single model and the uncorrected model. Detailed Implementation

[0069] The method of the present invention will be described in detail below with reference to the accompanying drawings.

[0070] Example 1

[0071] like Figure 1 As shown, this example provides a flood discharge prediction method based on the averaging of a bias-corrected Bayesian model, including the following steps:

[0072] Step 1: Construct a hydrodynamic model using acquired river cross-section data, hydrodynamic model parameter data, historical flow data, or flow data output from a hydrological model. Then, use the hydrodynamic model to build a model set and output the flow prediction results of the model set, including:

[0073] (1.1) Collect river cross-section information through field measurements or high-resolution digital elevation models, obtain historical flow data recorded by hydrological stations within the basin, or output flow data through hydrological models as upstream baseline flow data. Based on basic traffic data Construct a hydrodynamic model;

[0074] (1.2) Select the appropriate basic roughness coefficient according to the riverbed type and flow characteristics. ;

[0075] (1.3) Using the acquired flow data, construct a model set using hydrodynamic models, including:

[0076] (1.3.1) Based on the basic flow determined in step (1.1) and the basic roughness coefficient determined in step (1.2) Set the traffic scaling factor and roughness scaling factor Traffic scaling factor Includes 'a' roughness scaling factors. It contains b items;

[0077] (1.3.2) The flow input data and the initially set roughness data are scaled and combined to form K=a×b candidate models, which form a model set;

[0078] In step (1.3.2), the hydrodynamic model includes flow rate files, geometry files, and operation plan files. The flow rate files are processed according to a flow rate scaling factor, resulting in *a* flow rate files. The roughness in the geometry files is modified, resulting in *b* geometry files. By combining the flow rate files and geometry files, a total of K = *a* × *b* operation plan files are obtained. Running these K operation plan files yields K model outputs, constructing a model set. Using this model set for subsequent predictions can reduce uncertainty and improve the accuracy and robustness of the hydrodynamic model.

[0079] In this embodiment, the error is calculated as 20%, that is... = =[0.8,1.0,1.2] represents the coefficient scaling combination, resulting in a total of 3×3=9 models, such as Figure 2 The figure shows the established hydrodynamic model geometry, as well as the corresponding basic flow data and basic model parameters;

[0080] Step 2: For the K candidate models in the model set from Step 1, post-process the average of the Bayesian models based on the correction coefficients according to the law of total probability to construct the log-likelihood function; integrate the outputs of multiple candidate models to improve the reliability and accuracy of the prediction, including:

[0081] (2.1) Using the K candidate models from step 1, establish the posterior distribution of the prediction variable y based on the law of total probability.

[0082] (1)

[0083] in, It is the probability density function predicted by the average probability of the Bayesian model; Given observation data Under the condition that the kth model outputs The posterior probability is the weight. ; Given the output of the k-th model and observation data Under the condition of predictor variables The conditional probability distribution;

[0084] (2.2) The conditional probability distribution in equation (1) By setting the probability density to Gaussian, the corresponding log-likelihood function is obtained. Simultaneously, the Bayesian model output is corrected using correction coefficients.

[0085] (2)

[0086] in, Let be the log-likelihood function. Where t is the length of the traffic data and t is the time step. The number of candidate models. This represents the model output value at time step t of the k-th model. For the average parameters of the Bayesian model that need to be estimated, This includes the weights of each model. The standard deviation of the probability density for each model Correction coefficients for each model Correction coefficient two ;

[0087] (2.3) Maximizing equation (2) is equivalent to maximizing the likelihood function, thereby obtaining the estimated average parameters of the Bayesian model. The calculation formula is as follows:

[0088] (3)

[0089] in To maximize the parameter estimates after maximizing the likelihood function;

[0090] Step 3: Maximize the log-likelihood function using the SCE-UA algorithm to calculate the average optimal parameters of the Bayesian model. The SCE-UA algorithm is a method for minimizing the objective function; maximizing the log-likelihood function is equivalent to minimizing the negative log-likelihood function, and includes the following steps:

[0091] (3.1) Initialize the number of complexes Number of interior points of the complex Sample size The maximum number of iterations of the algorithm, rep;

[0092] (3.2) Randomly sample within the initially set parameter space For each sample, calculate the negative log-likelihood function value;

[0093] (3.3) Arrange the negative log-likelihood function values ​​obtained from step (3.2) in ascending order. Each sample point is stored in buffer D;

[0094] (3.4) Divide the points in buffer D into... In a complex, each complex contains One point;

[0095] (3.5) Perform CCE evolution calculations for each complex, including:

[0096] (3.5.1) Initialize CCE evolution, determine the number of points q in the sub-complex, the number of sub-complex iterations α, and the number of complex evolutions β, with the constraints 2≤q≤m, α≥1, and β≥1;

[0097] (3.5.2) Based on the linear probability distribution, calculate the selection probability of each point within the complex. The calculation formula is:

[0098] (4)

[0099] The points are indexed and sorted from lowest to highest based on their negative log-likelihood function values. The number of interior points of the complex; The normalization factor is the sum of the possibilities for choosing all points, ensuring the selection probability. The sum is 1;

[0100] (3.5.3) Based on the selection probability calculated in step (3.5.2), random sampling without replacement is performed within the complex, and a total of q points are selected and stored in buffer B. The q points together form a sub-complex, and the position information L at this time is recorded.

[0101] (3.5.4) Generate the next iteration population and obtain the final buffer B, including:

[0102] (01). Sort buffer B in ascending order of objective function value and find the worst estimate point in B. The centroids of the remaining q-1 points are denoted as ;

[0103] (02). Calculate the reflection point of the worst point. If the reflection point If the parameters are within the feasible space, proceed to step (04); otherwise, proceed to step (03).

[0104] (03). Randomly generated points Replacement reflection point Proceed to step (04);

[0105] (04). Calculate the reflection point The negative log-likelihood function, if the reflection point The negative log-likelihood function is less than the original worst-case scenario. So, taking the reflection point Replace the worst estimate point Otherwise, proceed to step (05);

[0106] (05). Calculation If point The negative log-likelihood function is less than the original worst estimate point. So, with point Replace the worst estimate point Otherwise, randomly generate points. Replace the worst estimate point ;

[0107] (06). Repeat steps (01) to (05) α times to obtain the final buffer B.

[0108] (3.5.5) Replace the sub-complexes within the original complex with the final buffer B, i.e., position L, and sort the complexes in ascending order according to the negative log-likelihood function values;

[0109] (3.5.6) Repeat steps (3.5.2) to (3.5.5) β times;

[0110] (3.6) Arrange the points in the complex in ascending order in buffer D;

[0111] (3.7) If the convergence condition is met, the process ends and the optimal average parameter of the Bayesian model is output. If the convergence condition is not met, the process returns to step 3.4 until the convergence condition is met or the maximum number of iterations rep of the algorithm is reached. At this point, the operation stops and the optimal average parameter estimation result is output.

[0112] like Figure 3 The figure shows the average parameter estimates of the Bayesian model after 500,000 iterations (rep=50w). Figure 3 In Figure (a), the relationship between the calculated weight and standard deviation estimates, arranged in descending order of weight, and the RMSE after model correction is shown. It can be seen that models with lower standard deviations and RMSEs are assigned higher weights. Figure 3 In the middle (b), the value of the deviation correction parameter is estimated at this time.

[0113] Step 4: Based on the Bayesian model average optimal parameter estimation results calculated in Step 3, calculate the final predicted value and confidence interval:

[0114] (4.1) Calculate the final predicted value based on the average parameters and output of the Bayesian model obtained in step 3:

[0115] (5)

[0116] Here The expected value of the final prediction. , , These are the optimal parameter values ​​estimated in step 3. This is the output of the k-th model in the model set;

[0117] (4.2) Based on the obtained average parameters and model output of the Bayesian model, calculate the average variance of the Bayesian model, and then obtain the confidence interval of the average final predicted value of the Bayesian model.

[0118] (6)

[0119] This represents the average variance of the Bayesian model. The output of the k-th model in the model set; the weights of each model. Standard deviation of the probability density for each model Correction coefficients for each model Correction coefficient two All of these are the optimal parameter values ​​estimated in step 3;

[0120] (7)

[0121] This represents the confidence interval for the average final predicted value of the Bayesian model. Corresponding confidence level The critical value of the standard normal distribution.

[0122] Figure 4The chart shows the changes in RMSE of individual models before and after correction (arranged from largest to smallest according to model weight) and the average RMSE of the Bayesian model without bias correction. It can be seen that after bias correction, the bias of the models is significantly reduced, and in the final ensemble prediction, the bias-corrected Bayesian model has an average lower RMSE.

[0123] Figure 5 The table shows the predicted values ​​and confidence intervals of the Bayesian model mean (a) without bias correction and the Bayesian model mean (b) with bias correction. It can be seen that the Bayesian model mean with bias correction significantly reduces the size of the confidence interval, indicating that it significantly reduces the uncertainty of the final prediction result of the ensemble model.

[0124] Example 2

[0125] A computer device / apparatus / system includes a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the above-described method.

[0126] Example 3

[0127] A computer-readable storage medium having a computer program / instructions stored thereon, characterized in that the computer program / instructions, when executed by a processor, implement the steps of the above-described method.

[0128] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0129] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0130] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0131] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A flood discharge prediction method based on bias-corrected Bayesian model averaging, characterized in that, Includes the following steps: Step 1: Construct a hydrodynamic model using river cross-section data, hydrodynamic model parameter data, historical flow data, or hydrological model output flow data; then construct a model set using the hydrodynamic model. Step 2: For the candidate models in the model set from Step 1, post-process the Bayesian models based on the correction coefficients according to the law of total probability to construct the log-likelihood function; including: 21) Using the K candidate models from step 1, establish the posterior distribution of the predicted variable y based on the law of total probability. (1) in, It is the probability density function predicted by the average probability of the Bayesian model; Given observation data Under the condition that the kth model outputs The posterior probability is the weight. ; Given the output of the k-th model and observation data Under the condition of predictor variables The conditional probability distribution; 22) The conditional probability distribution in equation (1) By setting the probability density to Gaussian, the corresponding log-likelihood function is obtained. Simultaneously, the Bayesian model output is corrected using correction coefficients. (2) in, Let be the log-likelihood function. Where t is the length of the traffic data and t is the time step. The number of candidate models. This represents the model output value at time step t of the k-th model. For the average parameters of the Bayesian model that need to be estimated, This includes the weights of each model. The standard deviation of the probability density for each model Correction coefficients for each model Correction coefficient two ; (23) Maximizing equation (2) is equivalent to maximizing the likelihood function, thereby obtaining the estimated average parameters of the Bayesian model. The calculation formula is as follows: (3) in To maximize the parameter estimates after maximizing the likelihood function; Step 3: Maximize the log-likelihood function using the SCE-UA algorithm to calculate the average optimal parameters of the Bayesian model; Step 4: Based on the Bayesian model average optimal parameter estimation results calculated in Step 3, calculate the final predicted value and confidence interval.

2. The flood discharge prediction method based on the bias-corrected Bayesian model averaging according to claim 1, characterized in that, Step 1 includes: (11) Collect river cross-section information, obtain historical flow data recorded by hydrological stations in the basin, or output flow data through hydrological models as upstream basic flow data. Based on basic traffic data Construct a hydrodynamic model; (12) Select the appropriate basic roughness coefficient according to the riverbed type and flow characteristics. ; (13) Using the obtained flow data, construct a model set using the hydrodynamic model.

3. The flood discharge prediction method based on the averaging of a bias-corrected Bayesian model according to claim 2, characterized in that, Step (13) includes: (131) Based on the basic flow determined in step (11) and the basic roughness coefficient determined in step (12) Set the traffic scaling factor and roughness scaling factor Traffic scaling factor Includes 'a' roughness scaling factors. It contains b items; (132) The flow input data and the initially set roughness data are scaled and combined to form K=a×b candidate models, which form a model set.

4. The flood discharge prediction method based on the averaging of a bias-corrected Bayesian model according to claim 3, characterized in that, In step (132), the hydrodynamic model includes flow files, geometry files, and operation plan files. The flow files are processed according to the flow scaling factor to obtain a flow files. The roughness in the geometry files is modified to obtain b geometry files. By combining the flow files and geometry files, a total of K = a × b plan files are obtained. By running the K plan files, K model outputs are obtained, and a model set is constructed.

5. The flood discharge prediction method based on the bias-corrected Bayesian model averaging according to claim 1, characterized in that, Step 3 includes the following steps: (31) Initialize the number of complexes Number of interior points of the complex Sample size The maximum number of iterations of the algorithm, rep; (32) Randomly sample within the initially set parameter space For each sample, calculate the negative log-likelihood function value; (33) Arrange the negative log-likelihood function values ​​obtained from step (32) in ascending order. Each sample point is stored in buffer D; (34) Divide the points in buffer D into... In a complex, each complex contains One point; (35) Perform CCE evolution calculations for each complex; (36) Arrange the points in the complex in ascending order in buffer D; (37) If the convergence condition is met, the process ends and the optimal average parameter of the Bayesian model is output. If the convergence condition is not met, the process returns to step (34) until the convergence condition is met or the maximum number of iterations rep of the algorithm is reached. At this point, the operation stops and the optimal average parameter estimation result is output.

6. The flood discharge prediction method based on the averaging of a bias-corrected Bayesian model according to claim 5, characterized in that, Step (35) includes: (351) Initialize CCE evolution, determine the number of points q in the sub-complex, the number of iterations α in the sub-complex, and the number of evolutions β in the complex, with the constraints being 2≤q≤m, α≥1, and β≥1; (352) Based on the linear probability distribution, calculate the selection probability of each point within the complex. The calculation formula is: (4) The points are indexed and sorted from lowest to highest based on their negative log-likelihood function values. The number of interior points of the complex; The normalization factor is the sum of the possibilities for choosing all points, ensuring the selection probability. The sum is 1; (353) Based on the selection probability calculated in step (352), random sampling without replacement is performed within the complex, and a total of q points are selected and stored in buffer B. The q points together form a sub-complex, and the position information L at this time is recorded. (354) Generate the next iteration population and obtain the final buffer B; (355) Replace the sub-complexes within the original complex with the final buffer B, i.e., position L, and sort the complexes in ascending order according to the negative log-likelihood function values; (356) Repeat steps (352) to (355) β times.

7. The flood discharge prediction method based on the averaging of a bias-corrected Bayesian model according to claim 6, characterized in that, Step 354 includes: (01). Sort buffer B in ascending order of objective function value and find the worst estimate point in B. The centroids of the remaining q-1 points are denoted as ; (02). Calculate the reflection point of the worst point. If the reflection point If the parameters are within the feasible space, proceed to step (04); otherwise, proceed to step (03). (03). Randomly generated points Replacement reflection point Proceed to step (04); (04). Calculate the reflection point The negative log-likelihood function, if the reflection point The negative log-likelihood function is less than the original worst-case scenario. So, taking the reflection point Replace the worst estimate point Otherwise, proceed to step (05); (05). Calculation If point The negative log-likelihood function is less than the original worst estimate point. So, with point Replace the worst estimate point Otherwise, randomly generate points. Replace the worst estimate point ; (06). Repeat steps (01) to (05) a times to obtain the final buffer B.

8. The flood discharge prediction method based on the averaging of a bias-corrected Bayesian model according to claim 1, characterized in that, Step 4 includes: (41) Calculate the final predicted value based on the average parameters of the Bayesian model obtained in step 3 and the output of the Bayesian model: (5) Here The expected value of the final prediction. , , These are the optimal parameter values ​​estimated in step 3. This is the output of the k-th model in the model set; (42) Based on the obtained average parameters and model output of the Bayesian model, calculate the average variance of the Bayesian model, and then obtain the confidence interval of the average final predicted value of the Bayesian model. (6) This represents the average variance of the Bayesian model. The output of the k-th model in the model set; the weights of each model. Standard deviation of the probability density for each model Correction coefficients for each model Correction coefficient two All of these are the optimal parameter values ​​estimated in step 3; (7) This represents the confidence interval for the average final predicted value of the Bayesian model. Corresponding confidence level The critical value of the standard normal distribution.

9. A computer system, characterized in that, It includes a memory, a processor, and a computer program stored in the memory, the processor executing the computer program to perform the steps of the method as claimed in any one of claims 1-8.

Images