Parameter screening method, model screening method, prediction method and related devices

Abstract

The invention discloses a parameter screening method, a model screening method, a prediction method and a related device, and adopts a partial least square structure equation model to reduce the influence of a nonlinear relationship among parameters on a prediction model, and provides a cleaner and orthogonality-corrected parameter set for a lasso algorithm, so that the parameter selection of the lasso algorithm is more reliable and robust, and the prediction efficiency is improved. According to the double-layer screening structure, the advantages of the two methods are fully utilized, complementation is achieved, the screened parameter set is scientific and efficient, and the accuracy of the prediction model can be improved.

Description

Technical Field

[0001] This invention relates to a parameter selection method, a model selection method, a prediction method, and related devices, belonging to the field of peak flow prediction for breaches. Background Technology

[0002] The suddenness and unpredictable development of dam breaches greatly increase the difficulty of emergency rescue and disaster relief. Rapid prediction of peak breach flow is crucial for disaster relief and downstream flood assessment, and its prediction methods have long been a major research focus. Existing peak breach flow prediction methods input parameters affecting the peak breach flow into machine learning models. These parameters are diverse (such as dam height, water depth above the breach bottom, reservoir capacity, and breach width), and complex nonlinear relationships often exist between them, leading to decreased accuracy in model predictions. Summary of the Invention

[0003] This invention provides a parameter selection method, a model selection method, a prediction method, and related apparatus, which solve the problems disclosed in the background art.

[0004] According to one aspect of this application, a parameter filtering method is provided, comprising: Obtain all raw parameters that affect the peak flow rate of the breach; The partial least squares structural equation model and the lasso algorithm were used sequentially to filter the original parameters and obtain the parameters used to predict the peak flow of the breach. The partial least squares structural equation model analyzed the influence of each original parameter on the peak flow of the breach, and the original parameters were filtered according to the magnitude of the influence.

[0005] Furthermore, the raw parameters include conventional parameters and categorical parameters. Conventional parameters include dam height, water depth above the breach bottom, final water depth of the breach, reservoir capacity, reservoir capacity above the breach bottom, average breach width, and breach duration. Categorical parameters include breach mode and dam type.

[0006] Furthermore, in the partial least squares structural equation model, reservoir potential energy, breach discharge capacity, material erodibility, and time evolution characteristics are used as latent variables. Path coefficients represent the direct effects of reservoir potential energy, material erodibility, and time evolution characteristics on the breach discharge capacity, the goodness of fit, and the goodness of fit, respectively. Specifically, reservoir potential energy is the gravitational potential energy quantified by dam height, water depth above the breach bottom, reservoir capacity above the breach bottom, and reservoir capacity; breach discharge capacity is characterized by the average breach width and final breach depth; material erodibility is characterized by classification parameters reflecting soil-water interaction characteristics; and time evolution characteristics are characterized by the breach duration. The relationship between the latent variables is as follows: ; In the formula, The first j The and the first i One latent variable, i Not equal to j , I The number of latent variables. For path coefficients, The variation that the partial least squares structural equation model could not explain; The latent variables are linear combinations of the original parameters: ; In the formula, For the first j The th latent variable k One explicit variable, For the first k Each manifest variable and its corresponding latent variable The strength of the relationship, This is for measurement error.

[0007] According to one aspect of this application, a parameter filtering device is provided, comprising: The first acquisition module acquires all raw parameters that affect the peak flow rate of the breach; The first screening module sequentially uses the partial least squares structural equation model and the lasso algorithm to screen the original parameters and obtain the parameters used to predict the peak flow of the breach. Among them, the partial least squares structural equation model analyzes the influence of each original parameter on the peak flow of the breach and screens the original parameters according to the magnitude of the influence.

[0008] According to one aspect of this application, a model screening method is provided, comprising: Using the above parameter filtering method, parameters for predicting peak flow rate of the breach were selected; Using parameters selected from historical moments, various machine learning models were trained and tested. The machine learning model with the best test results was selected as the model for predicting the peak flow of the breach.

[0009] According to one aspect of this application, a model screening apparatus is provided, characterized in that it comprises: The second filtering module uses the above parameter filtering method to filter out the parameters used to predict the peak flow rate of the breach. The module was determined, and the selected parameters were used to train and test various machine learning models. The machine learning model with the best test results was selected as the model for predicting the peak flow of the breach.

[0010] According to one aspect of this application, a prediction method is provided, characterized in that it includes: Retrieve the parameters selected at the current moment; where the selected parameters are those selected using the parameter selection method described above. Input the selected parameters at the current moment into the prediction model to obtain the peak flow prediction result of the breach; wherein, the prediction model is the model selected using the above model selection method.

[0011] According to one aspect of this application, a prediction apparatus is provided, characterized in that it comprises: The second acquisition module acquires the filtered parameters at the current moment; wherein, the filtered parameters are those selected using the parameter filtering method described above; The prediction module inputs the selected parameters at the current moment into the prediction model to obtain the predicted peak flow rate of the breach; wherein, the prediction model is the model selected using the above model selection method.

[0012] According to another aspect of this application, a computer-readable storage medium is provided that stores one or more programs, the one or more programs including instructions that, when executed by a computing device, cause the computing device to perform a parameter selection method, a model selection method, or a prediction method.

[0013] According to another aspect of this application, a computer device is provided, including one or more processors and one or more memories, wherein one or more programs are stored in the one or more memories and configured to be executed by the one or more processors, and the one or more programs include instructions for performing a parameter screening method, a model screening method, or a prediction method.

[0014] The beneficial effects achieved by this invention are as follows: This invention adopts a partial least squares structural equation model, which reduces the impact of nonlinear relationships between parameters on the prediction model, and provides a cleaner and more orthogonal parameter set for the lasso algorithm, making the parameter selection of the lasso algorithm more reliable and robust, and achieving fine screening. This two-layer screening structure makes full use of the advantages of the two methods to achieve complementarity. The selected parameter set is both scientific and efficient, which can improve the accuracy of the prediction model. Attached Figure Description

[0015] Figure 1 A flowchart for the parameter filtering method; Figure 2 This is a schematic diagram of the cross-section of the breach; Figure 3 This is a schematic diagram of the longitudinal section of the breach; Figure 4 This is a schematic diagram of a partial least squares structure equation model (PLS-SEM). Figure 5 The image shows the PLS-SEM path analysis results for a small dam. Figure 6 This is a diagram showing the PLS-SEM path analysis results for a medium-sized dam. Figure 7 The image shows the PLS-SEM path analysis results for a large dam. Figure 8 A diagram showing the impact of each latent variable on the peak flow rate at the breach; Figure 9 For Lasso path graphs; Figure 10 This is a graph of the regression coefficients under the optimal penalty weight condition. Figure 11 A block diagram of the parameter filtering device; Figure 12 A flowchart of the model selection method; Figure 13 The image shows the test results for the XGBoost model without parameter filtering. Figure 14 The image shows the test results of the XGBoost model after parameter filtering. Figure 15 The image shows the test results of the SVR model without parameter filtering. Figure 16 The image shows the test results of the SVR model after parameter filtering. Figure 17 The graph shows the test results for the Ridge Regression model without parameter filtering. Figure 18 The graph shows the test results of the Ridge Regression model after parameter filtering. Figure 19 A block diagram of the model screening device; Figure 20 Here is a flowchart of the prediction method; Figure 21 This is a block diagram of the prediction device. Detailed Implementation

[0016] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit this application or its application or use. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0017] Unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps described in these embodiments do not limit the scope of this application.

[0018] At the same time, it should be understood that, for ease of description, the dimensions of the various parts shown in the accompanying drawings are not drawn according to actual scale.

[0019] Techniques, methods, and equipment known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and equipment should be considered part of the specification.

[0020] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values.

[0021] It should be noted that similar symbols and letters in the following figures represent similar items; therefore, once an item is defined in one figure, it does not need to be discussed further in subsequent figures.

[0022] Furthermore, in the description of the embodiments of this application, the terms "first," "second," etc., are used only for distinguishing descriptions and should not be construed as indicating or implying relative importance. Therefore, features defined with "first" or "second" may explicitly or implicitly include one or more features.

[0023] See Figure 1 , Figure 1 This is a flowchart of a parameter filtering method provided in an embodiment of this application. The parameter filtering method can be executed by a parameter filtering device, which can be a terminal device or a server. The parameter filtering method can include at least the following steps: Step 1: Obtain all raw parameters that affect the peak flow rate of the breach.

[0024] Studies show that the dam-breaking flow first forms an initial scour pit on the downstream embankment slope and then gradually erodes upstream. When the erosion reaches the upstream edge of the embankment crest, the breach process becomes extremely rapid and violent, accompanied by intermittent instability and collapse of the breach slope, eventually forming a relatively stable breach. Typical breach cross-sections, longitudinal sections, and related parameters can be seen as follows: Figure 2 and 3 As shown.

[0025] By collecting and organizing data on various types of earth-rock dam failures from both domestic and international sources, as well as relevant information from the Dam Safety Management Center of the Ministry of Water Resources, a basic database containing 192 dam failure cases has been established. To fully consider the impact of upstream reservoirs and the shape characteristics of the breach, in some embodiments, the original parameters are divided into conventional parameters and categorized parameters. Specifically, seven important conventional parameters may include dam height, water depth above the breach bottom, final water depth of the breach, reservoir capacity, reservoir capacity above the breach bottom, average breach width, and breach duration (see Table 1 for details). Categorized parameters may include breach mode and dam type. Breach modes mainly include whether overtopping or piping occurred, while dam types mainly include homogeneous dams, core dams, concrete dams, and partitioned dams.

[0026] Table 1 Description of Common Parameters

[0027] It should be noted that the original parameters need to be preprocessed before filtering, which may involve removing outliers and filling in missing values ​​using various data imputation methods.

[0028] Step 2: The partial least squares structural equation model and the lasso algorithm are used sequentially to filter the original parameters and obtain the parameters used to predict the peak flow of the breach. The partial least squares structural equation model analyzes the influence of each original parameter on the peak flow of the breach and filters the original parameters according to the magnitude of the influence.

[0029] It's important to note that Partial Least Squares Structural Equation Modeling (PLS-SEM) is a powerful tool for revealing complex causal relationships between variables. In this model, latent variables are abstract concepts that cannot be directly observed but can be indirectly represented by multiple measurable variables. Principal component analysis and multiple regression are combined, employing iterative estimation methods to maximize the explanation of variations in latent variables. PLS-SEM does not require data to follow a normal distribution and can effectively assess interactions between variables. Here, SmartPLS4 software can be used to construct a PLS-SEM model and assess its goodness of fit. R ² is used to measure its prediction accuracy.

[0030] In some embodiments, the structure of PLS-SEM can be found in [reference needed]. Figure 4 In the partial least squares structural equation model, reservoir potential energy (RPE), breach discharge capacity (BDC), material erodibility (ME), and temporal evolution characteristics (TEC) are used as latent variables, and path coefficients are used to determine the latent variables.β The direct effects of reservoir potential energy, material corrosivity, and temporal evolution characteristics on the breach discharge capacity, respectively, represent the effects of these four factors: reservoir potential energy, water depth above the breach bottom, reservoir capacity above the breach bottom, and reservoir capacity. The breach discharge capacity is characterized by the average breach width and final breach depth, conforming to the broad-crested weir principle. Material corrosivity is characterized using classification parameters to reflect soil-water interaction characteristics. Temporal evolution characteristics are characterized by the breach duration.

[0031] The entire model can be divided into two models: the structural model and the measurement model. The structural model, also known as the internal model, is used to estimate the relationships between latent variables and can be defined as follows: ; In the formula, The first j The and the first i One latent variable, i Not equal to j , I The number of latent variables. Path coefficients are used to determine the strength and direction of the relationship between explanatory variables. The variation that the partial least squares structural equation model could not explain represents The portion of the model that cannot be explained by other latent variables in the model.

[0032] The measurement model is used to estimate the relationship between the original parameters and their corresponding latent variables, where the latent variables are linear combinations of the original parameters. ; In the formula, For the first j The th latent variable k One explicit variable, For load, the first k Each manifest variable and its corresponding latent variable The strength of the relationship between the manifest variable and the latent variable is indicated by a higher value. For measurement error, representing the first k The portion of a manifest variable that cannot be explained by its latent variables includes random errors and unmeasured influencing factors.

[0033] PLS-SEM analysis of the original parameters for peak dam break flow ( Q p The impact of the influence is considered, and screening is conducted based on the magnitude of the impact.

[0034] Here, dams are classified into three categories based on their height: small dams (<15 meters), medium dams (15–30 meters), and large dams (>30 meters). The corresponding path analysis results are presented in the sub-sections. Figures 5-7 middle. Figures 5-7 The R² value within the middle box diagram represents the proportion of variance that the influencing factor can explain, as indicated by the arrow. β The values ​​represent the degree of correlation between parameters. PLS-SEM effectively quantifies the causal relationships of key original parameters in dam failure processes of dams of different sizes (small, medium, and large). Validation results show that the model is effective in... Q p The model demonstrates strong explanatory power, with R² values ​​of 0.743, 0.858, and 0.832 for the three dam types, indicating that the model can explain the vast majority of the variance in the outcome variables. Path analysis reveals the evolution of the driving mechanism dominated by dam size: for small dams, RPE ( β =0.876) is Q p The main direct driving factor; while the breach discharge capacity of large dams ( β =0.981) had the most significant influence; medium-sized dams showed a balanced contribution from both variables. A stable and significant positive relationship between RPE and BDC was observed in all dam types. β (Range: 0.661-0.802), indicating that greater potential energy directly promotes breach development, thereby amplifying peak flow.

[0035] Figure 8 Showing the pairs of each latent variable Q p The decomposition results of the direct, indirect, and total effects confirmed the dynamic changes in the effects of each latent variable: RPE and BDC on Q p The total effect of ME was significantly higher across all dam types, while the total effects of TEC were relatively weak (both below 0.3). Therefore, the original parameters included in the latent variables of ME and TEC (including...) T f The failure mode and dam type were excluded from subsequent analysis.

[0036] PLS-SEM excels at handling multicollinearity by extracting latent variables to model the relationships between variables, thus reducing the impact of multicollinearity. Using PLS-SEM here can effectively address highly correlated parameters commonly found in dam-break data, providing a more stable subset of parameters for subsequent analysis.

[0037] It should be noted that the basic principle of the Lasso algorithm is to introduce a penalty weight based on the least squares method. LambdaA penalty term (i.e., the degree of compression of the regression coefficients) is used to constrain the regression coefficients on each parameter of the machine learning model, thereby achieving the purpose of parameter selection. This is achieved by increasing... Lambda The value is used to gradually compress the regression coefficients between insignificantly related parameters to 0, thereby ultimately obtaining a model with fewer parameters and greater efficiency.

[0038] The formula for calculating the objective function is as follows: ; In the formula, J ( w ) is the objective function (or loss function) of the Lasso algorithm. M This represents the total number of samples in the dam failure case database. N This refers to the number of parameters (parameters selected after partial least squares structural equation modeling). w For the regression coefficient vector, w n Indicates the first n Each feature pair Q p The extent of the impact x mn It is the first m The first sample n The specific numerical value of each feature y m It is the first m Actual observations in individual dam failure cases Q p .

[0039] The optimal performance estimate of the Lasso model is obtained by using the K-fold cross-validation method. K-fold cross-validation is a commonly used model evaluation technique to evaluate the performance of machine learning models on training data. It divides the original dataset into K subsets, called folds. The model is trained and validated K times on K different training sets. In each training, one fold is used as the validation set, and the remaining K-1 folds are used as the training set. Finally, the average of K evaluation metrics is calculated as the final model performance evaluation metric.

[0040] Different penalty weights were calculated using 10-fold cross-validation. Lambda The optimal penalty weight is obtained by considering the mean squared error (MSE) under the given conditions. Lambda And the Lasso path diagram for the remaining 6 parameters was plotted, as shown below. Figure 9 As shown in the figure, this diagram gradually eliminates unimportant parameters to obtain those that significantly affect peak flow, making the model sparser and easier to interpret. The horizontal solid lines in the lasso path represent the regression coefficients of each parameter as a function of the penalty weight. LambdaThe trajectory of change. With... Lambda As ln increases, the regression coefficients of each parameter gradually approach zero. The slower the change, the more important the parameter's influence. Lambda The value is 0.1, which is the optimal penalty weight. Lambda When it is -1, h b , S , V w as well as B ave The corresponding regression coefficients are greater than 0, indicating a high correlation among the four parameters. Therefore, the Lasso algorithm is used to extract four highly correlated parameters from the six parameters. h b , S , V w as well as B ave ), eliminating two low-correlation parameters whose regression coefficients were compressed to zero ( h w , h d Thus, parameter selection is achieved, and the regression coefficients of each parameter are as follows: Figure 10 As shown.

[0041] While the Lasso algorithm possesses parameter selection capabilities, it may randomly select one parameter while ignoring others when parameters are highly correlated, leading to unstable or uninterpretable results. Therefore, PLS-SEM preprocessing is used here to reduce parameter correlation, providing a cleaner and more orthogonal parameter set for Lasso. This avoids the instability of the Lasso algorithm in such cases, making its parameter selection more reliable and robust.

[0042] The above method employs a partial least squares structural equation model to reduce the impact of nonlinear relationships between parameters on the prediction model, providing a cleaner and more orthogonal parameter set for the lasso algorithm. This makes the parameter selection of the lasso algorithm more reliable and robust, achieving fine screening. This two-layer screening structure fully utilizes the advantages of the two methods to achieve complementarity. The selected parameter set is both scientific and efficient, which can improve the accuracy of the prediction model.

[0043] See Figure 11 , Figure 11 This is a block diagram of a parameter filtering device provided in an embodiment of this application. It is a virtual device that can be loaded and executed by a computer device, which may include the aforementioned parameter filtering device. Figure 11The apparatus may include a first acquisition module and a first filtering module, which, when executing the above parameter filtering method, can: The first acquisition module obtains all raw parameters that affect the peak flow rate of the breach.

[0044] The first screening module sequentially uses the partial least squares structural equation model and the lasso algorithm to screen the original parameters and obtain the parameters used to predict the peak flow of the breach. Among them, the partial least squares structural equation model analyzes the influence of each original parameter on the peak flow of the breach and screens the original parameters according to the magnitude of the influence.

[0045] The device described above employs a partial least squares structural equation model, which reduces the impact of nonlinear relationships between parameters on the prediction model. This provides a cleaner and more orthogonal parameter set for the lasso algorithm, making parameter selection more reliable and robust, and enabling fine-grained screening. This two-layer screening structure fully utilizes the advantages of both methods to achieve complementarity. The selected parameter set is both scientific and efficient, which can improve the accuracy of the prediction model.

[0046] See Figure 12 , Figure 12 This is a flowchart of a model selection method provided in an embodiment of this application. The model selection method can be executed by a model selection device, which can be a terminal device or a server. The model selection method can include at least the following steps: S1) Using the above parameter screening method, parameters for predicting peak flow rate of the breach are selected.

[0047] S2) Using parameters selected from historical moments, various machine learning models are trained and tested. The machine learning model with the best test results is used as the model for predicting the peak flow of the breach.

[0048] It should be noted that the machine learning model can arbitrarily choose from several known models, such as XGBoost, Support Vector Regression (SVR), and Ridge Regression, and collect historical data. h b , S , V w , B ave Construct a training database and divide it into training and testing sets in an 8:2 ratio. Then train and test the model. The results can be found in [link to relevant documentation]. Figures 13-18 And Table 2.

[0049] Table 2. Prediction accuracy of each model on the test set

[0050] As can be seen, among the XGBoost, SVR, and Ridge Regression models, the XGBoost model's training set and test set fitted lines are closest to the y=x line, and its RMSE, MAE, and MSE on the test set are all the smallest. R 2 The maximum accuracy demonstrates the superior performance of the XGBoost model. After feature selection, all three models showed improved prediction accuracy, with the XGBoost model showing the most significant improvement. This proves that using parameter selection methods can effectively improve model performance, and that the XGBoost model is the most suitable among the three models for predicting peak breach flow.

[0051] The above method first filters the parameters, and then filters the models based on the filtered parameters, which can obtain the model with the highest accuracy in predicting the peak flow of the breach.

[0052] See Figure 19 , Figure 19 This is a block diagram of a model screening device provided in an embodiment of this application. It is a virtual device that can be loaded and executed by a computer device, which may include the aforementioned model screening device. Figure 19 The apparatus may include a second screening module and a determining module, which, when executing the above-described model screening method, can: The second filtering module uses the above parameter filtering method to filter out the parameters used to predict the peak flow rate of the breach.

[0053] The module was determined, and the selected parameters were used to train and test various machine learning models. The machine learning model with the best test results was selected as the model for predicting the peak flow of the breach.

[0054] The aforementioned device first filters the parameters, and then filters the models based on the filtered parameters to obtain the model with the highest accuracy in predicting the peak flow of the breach.

[0055] See Figure 20 , Figure 20 This is a flowchart of a prediction method provided in an embodiment of this application. The prediction method can be executed by a prediction device, which can be a terminal device or a server. The prediction method can include at least the following steps: A1) Obtain the filtered parameters at the current moment; where the filtered parameters are those filtered by the parameter filtering method described above.

[0056] A2) Input the selected parameters at the current moment into the prediction model to obtain the peak flow prediction result of the breach; wherein, the prediction model is the model selected by the above model selection method.

[0057] The peak flow rate at the breach is influenced by a variety of parameters. Too many parameters complicate the model, increasing the difficulty of prediction. The above method reduces the number of model input parameters through parameter screening and reduces the nonlinear relationship between parameters. By selecting the most suitable model, the highest accuracy in predicting peak flow at the breach can be achieved.

[0058] See Figure 21 , Figure 21 This is a block diagram of a prediction device provided in an embodiment of this application. It is a virtual device that can be loaded and executed by a computer device, which may include the aforementioned prediction device. Figure 21 The apparatus may include a second acquisition module and a prediction module, which, when performing the above prediction method, can: The second acquisition module acquires the filtered parameters at the current moment; wherein the filtered parameters are those selected by the parameter filtering method described above.

[0059] The prediction module inputs the selected parameters at the current moment into the prediction model to obtain the predicted peak flow rate of the breach; wherein, the prediction model is the model selected using the above model selection method.

[0060] The aforementioned device reduces the number of model input parameters through parameter screening and also reduces the nonlinear relationships between parameters. By selecting the most suitable model through model screening, it can achieve the most accurate prediction of peak flow rate at the breach.

[0061] This application also relates to a computer-readable storage medium that stores one or more programs, the one or more programs including instructions that, when executed by a computing device, cause the computing device to perform a parameter selection method, a model selection method, or a prediction method.

[0062] This application also relates to a computer device including one or more processors and one or more memories, wherein one or more programs are stored in the one or more memories and configured to be executed by the one or more processors, and the one or more programs include instructions for performing a parameter selection method, a model selection method, or a prediction method.

[0063] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0064] 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.

[0065] 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.

[0066] 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.

[0067] The above are merely embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention are included within the scope of the claims of the present invention pending approval.

Claims

1. A parameter selection method, characterized in that, include: Obtain all raw parameters that affect the peak flow rate of the breach; The partial least squares structural equation model and the lasso algorithm were used sequentially to filter the original parameters and obtain the parameters used to predict the peak flow of the breach. The partial least squares structural equation model analyzed the influence of each original parameter on the peak flow of the breach, and the original parameters were filtered according to the magnitude of the influence.

2. The method according to claim 1, characterized in that, The raw parameters include conventional parameters and categorical parameters. Conventional parameters include dam height, water depth above the breach bottom, final water depth of the breach, reservoir capacity, reservoir capacity above the breach bottom, average breach width, and breach duration. Categorical parameters include breach mode and dam type.

3. The method according to claim 2, characterized in that, In the partial least squares structural equation model, reservoir potential energy, breach discharge capacity, material erodibility, and time evolution characteristics are used as latent variables. Path coefficients represent the direct effects of reservoir potential energy, material erodibility, and time evolution characteristics on the breach discharge capacity, breach discharge capacity, and goodness of fit, respectively. Specifically, reservoir potential energy is the gravitational potential energy quantified by dam height, water depth above the breach bottom, reservoir capacity above the breach bottom, and reservoir capacity; breach discharge capacity is characterized by the average breach width and final breach depth; material erodibility is characterized by classification parameters reflecting soil-water interaction characteristics; and time evolution characteristics are represented by the breach duration. The relationship between the latent variables is as follows: ； In the formula, , The first j The and the first i One latent variable, i Not equal to j , I The number of latent variables. For path coefficients, The variation that the partial least squares structural equation model could not explain; The latent variables are linear combinations of the original parameters: ； In the formula, For the first j The th latent variable k One explicit variable, For the first k Each manifest variable and its corresponding latent variable The strength of the relationship, This is for measurement error.

4. A parameter screening device, characterized in that, include: The first acquisition module acquires all raw parameters that affect the peak flow rate of the breach; The first screening module sequentially uses the partial least squares structural equation model and the lasso algorithm to screen the original parameters and obtain the parameters used to predict the peak flow of the breach. Among them, the partial least squares structural equation model analyzes the influence of each original parameter on the peak flow of the breach and screens the original parameters according to the magnitude of the influence.

5. A model selection method, characterized in that, include: Using the method described in any one of claims 1 to 3, parameters for predicting peak flow rate of the breach are selected; Using parameters selected from historical moments, various machine learning models were trained and tested. The machine learning model with the best test results was selected as the model for predicting the peak flow of the breach.

6. A model screening device, characterized in that, include: The second screening module uses the method described in any one of claims 1 to 3 to screen out parameters used to predict the peak flow rate of the breach. The module was determined, and the selected parameters were used to train and test various machine learning models. The machine learning model with the best test results was selected as the model for predicting the peak flow of the breach.

7. A prediction method, characterized in that, include: Obtain the filtered parameters at the current moment; wherein the filtered parameters are those filtered using the method described in any one of claims 1 to 3; The parameters selected at the current moment are input into the prediction model to obtain the peak flow prediction result of the breach; wherein, the prediction model is the model selected by the method described in claim 5.

8. A prediction device, characterized in that, include: The second acquisition module acquires the filtered parameters at the current moment; wherein the filtered parameters are those filtered using the method described in any one of claims 1 to 3; The prediction module inputs the selected parameters at the current moment into the prediction model to obtain the peak flow prediction result of the breach; wherein the prediction model is the model selected by the method described in claim 5.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores one or more programs, the one or more programs including instructions that, when executed by a computing device, cause the computing device to perform the method of any one of claims 1 to 3, 5, and 7.

10. A computer device, characterized in that, include: One or more processors and one or more memories, one or more programs stored in one or more memories and configured to be executed by one or more processors, the one or more programs including instructions for performing the method of any one of claims 1 to 3, 5, and 7.

Images