Meteorological data screening method and device for runoff forecasting model

Abstract

The invention relates to the technical field of hydrology, and discloses a meteorological data screening method and device for a runoff forecasting model, and the method comprises the steps: obtaining the actual runoff data and actual meteorological data of a target runoff region at historical time, and the fake meteorological data; inputting the actual meteorological data and the imitated meteorological data into a neural network model, and iteratively updating a first weight and a second weight of a filter layer of the neural network model and a third weight of a multi-layer sensor of the neural network model until the neural network model outputs actual runoff data, obtaining a newest first weight, a newest second weight and a newest third weight; generating a counterfeit statistic based on the latest first weight, the latest second weight and the latest third weight; generating a counterfeit threshold value based on a preset error rate threshold value and the counterfeit statistical magnitude; and selecting the actual meteorological data of which the mapped counterfeit statistical magnitude is greater than a counterfeit threshold value from the actual meteorological data as the meteorological data of the runoff forecasting model.

Description

Technical Field

[0001] The invention relates to the field of hydrological technology, and in particular to a meteorological data screening method and device for a runoff forecast model. Background Art

[0002] At present, runoff forecasting is crucial for water resource management and disaster prevention and mitigation. Physically driven models and data-driven models are common forecasting methods. Physically driven models rely on the physical mechanism of hydrological processes, but are limited by accurate meteorological data. Data-driven models establish mathematical models through historical data analysis. They do not rely on physical processes, but rely on the selection of appropriate meteorological data.

[0003] The selection of meteorological data has a huge impact on the accuracy of the runoff forecast model. Traditional methods mostly rely on expert experience and data correlation analysis, which have certain subjectivity and limitations. Therefore, how to accurately screen meteorological data for runoff forecast models is a technical problem that needs to be solved urgently. Summary of the invention

[0004] In view of this, the present invention provides a method and device for screening meteorological data of a runoff forecast model to solve the problem of how to accurately screen the meteorological data of a runoff forecast model.

[0005] In a first aspect, the present invention provides a method for screening meteorological data of a runoff forecast model, the method comprising: Obtaining actual runoff data and N actual meteorological data of the target runoff area in historical time, and N simulated meteorological data mapped one-to-one with the N actual meteorological data, where N is a positive integer; Inputting N actual meteorological data and N simulated meteorological data into a neural network model, and iteratively updating a first weight and a second weight of a filter layer of the neural network model, and a third weight of a multilayer perceptron of the neural network model, until the neural network model outputs actual runoff data, and obtaining the latest first weight, the latest second weight, and the latest third weight, wherein the first weight is used to weight the N actual meteorological data, the second weight is used to weight the N simulated meteorological data, and the third weight is determined based on the weight of each layer of the multilayer perceptron; Based on the latest first weight, the latest second weight and the latest third weight, N simulated statistics are generated, wherein one simulated statistic indicates the difference between one actual meteorological data and one mapped simulated meteorological data, and one simulated statistic maps one actual meteorological data; Generate a counterfeit threshold based on a preset error rate threshold and N counterfeit statistics; Actual meteorological data whose mapped imitation statistics are greater than the imitation threshold are selected from N actual meteorological data as meteorological data for the runoff forecast model.

[0006] In an optional implementation, after selecting actual meteorological data whose mapped counterfeit statistics are greater than the counterfeit threshold from N actual meteorological data as meteorological data for the runoff forecast model, the method further includes: Based on the preset initial time series length and the initial prediction step number, M target actual meteorological data are selected from the meteorological data of the runoff forecast model, where M≤N and is a positive integer; The actual meteorological data of M targets and the actual runoff data of the corresponding future time are divided into a training set and a validation set according to a preset ratio; The runoff forecast model is trained based on the training set and the validation set, and the hyper parameters of the runoff forecast model are obtained after optimization; Based on the optimized hyperparameters, an updated runoff forecast model is established; Merge the training set and the validation set to obtain a combined training set; Based on the combined training set, an updated runoff forecast model is trained to obtain a trained runoff forecast model.

[0007] In an optional implementation, after training the updated runoff forecast model based on the combined training set to obtain the trained runoff forecast model, the method further includes: The test set is divided into M target actual meteorological data and actual runoff data corresponding to the future time according to the preset ratio; Based on the test set, the generalization ability and performance of the trained runoff forecast model are evaluated.

[0008] In an optional implementation, obtaining actual runoff data and N actual meteorological data of a target runoff area in historical time, and N simulated meteorological data mapped one-to-one with the N actual meteorological data, includes: Obtain the actual runoff data and N actual meteorological data of the target runoff area in historical time; Based on the adaptive moving smoothing method, the actual runoff data is optimized; Based on N actual meteorological data and approximate semi-positive definite programming method, N simulated meteorological data are generated.

[0009] In an optional implementation, after inputting N actual meteorological data and N simulated meteorological data into the neural network model, the method further includes: Get the pre-weights of the multi-layer perceptron connected to the filter layer and each layer-by-layer weight of the output layer of the multi-layer perceptron connected to the neural network model; Based on the previous weight and each layer-by-layer weight, a third weight is determined.

[0010] In an optional implementation, N actual meteorological data and N simulated meteorological data are input into a neural network model, including: Inputting N simulated meteorological data and N actual meteorological data into the filtering layer, and using the first weight and the second weight to generate a weighted sum of the N simulated meteorological data and the N actual meteorological data as N fused meteorological features; Input N fused meteorological features into the first hidden layer of the multilayer perceptron to obtain the output result of the first hidden layer; The output results of the first hidden layer are input into the middle hidden layer of the multilayer perceptron layer by layer, and the output results of the middle hidden layer of the multilayer perceptron are obtained layer by layer; Input the output result of the last middle hidden layer of the multilayer perceptron into the output layer of the multilayer perceptron; obtain the output result of the multilayer perceptron; The output results of the multilayer perceptron are input into the output layer of the neural network model to obtain the output results of the neural network model.

[0011] In a second aspect, the present invention provides a meteorological data screening device for a runoff forecast model, the device comprising: An acquisition module is used to acquire actual runoff data and N actual meteorological data of the target runoff area in historical time, and N simulated meteorological data mapped one-to-one with the N actual meteorological data, where N is a positive integer; An updating module, used for inputting N actual meteorological data and N simulated meteorological data into the neural network model, and iteratively updating the first weight and the second weight of the filter layer of the neural network model, and the third weight of the multilayer perceptron of the neural network model, until the neural network model outputs the actual runoff data, and obtaining the latest first weight, the latest second weight and the latest third weight, wherein the first weight is used to weight the N actual meteorological data, the second weight is used to weight the N simulated meteorological data, and the third weight is determined based on the weight of each layer of the multilayer perceptron; A first generating module is used to generate N simulated statistics based on the latest first weight, the latest second weight and the latest third weight, wherein one simulated statistic indicates the difference between one actual meteorological data and one mapped simulated meteorological data, and one simulated statistic maps one actual meteorological data; A second generating module is used to generate a counterfeit threshold based on a preset error rate threshold and N counterfeit statistics; The selection module is used to select actual meteorological data whose mapped counterfeit statistics are greater than a counterfeit threshold from N actual meteorological data as meteorological data for the runoff forecast model.

[0012] In a third aspect, the present invention provides a computer device, comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the meteorological data screening method for the runoff forecast model of the first aspect or any corresponding embodiment thereof by executing the computer instructions.

[0013] In a fourth aspect, the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the meteorological data screening method for a runoff forecast model of the first aspect or any corresponding embodiment thereof.

[0014] In a fifth aspect, the present invention provides a computer program product, comprising computer instructions for causing a computer to execute the meteorological data screening method for a runoff forecast model of the first aspect or any corresponding embodiment thereof.

[0015] The embodiment provided by the present invention has the following effects: the weight of the filter layer is automatically iterated and optimized through the neural network model, and a simulated statistic reflecting the difference of the meteorological data is generated, thereby avoiding the manual screening of meteorological data in the traditional method, thereby greatly improving the data processing efficiency and accuracy. The model can assign appropriate weights to different data according to the different characteristics of the actual meteorological data and the simulated meteorological data by automatically optimizing the first, second, and third weights. The model can more accurately extract valuable information from multi-source data, thereby improving the reliability of runoff forecasting. By setting an error rate threshold and generating a simulated statistic, the difference between each group of actual meteorological data and its mapped simulated meteorological data can be automatically evaluated, thereby screening out actual meteorological data that has a higher contribution to the forecast, so that the model can better process complex meteorological data and improve the accuracy of runoff prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0017] Figure 1 is a flow chart of a meteorological data screening method for a runoff forecast model according to an embodiment of the present invention; Figure 2 is a schematic diagram of a meteorological data screening device for a runoff forecast model according to an embodiment of the present invention; Figure 3 It is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0019] According to an embodiment of the present invention, an embodiment of a method for screening meteorological data for a runoff forecast model is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in an order different from that shown here.

[0020] In this embodiment, a meteorological data screening method for a runoff forecast model is provided, which can be used for a computer. Figure 1 is a flow chart of a meteorological data screening method for a runoff forecast model according to an embodiment of the present invention. Figure 1 The method comprises: Step S101, obtaining actual runoff data and N actual meteorological data of a target runoff area in historical time, and N simulated meteorological data mapped one-to-one with the N actual meteorological data.

[0021] Where N is a positive integer. Actual runoff data refers to runoff data related to the target area obtained based on historical records or observations. Actual meteorological data includes multiple meteorological elements such as temperature, precipitation, humidity, wind speed, etc., reflecting historical weather conditions. Simulated meteorological data is hypothetical data corresponding to actual meteorological data generated in some way, which can be obtained through simulation, algorithm generation or some inference method.

[0022] One way to obtain it is to obtain meteorological data through satellite remote sensing technology, or directly obtain data from meteorological stations in the target area, or directly collect measured actual runoff data from hydrological stations.

[0023] Step S102, input N actual meteorological data and N simulated meteorological data into the neural network model, and iteratively update the first weight and the second weight of the filter layer of the neural network model, and the third weight of the multilayer perceptron of the neural network model, until the neural network model outputs the actual runoff data, and obtains the latest first weight, the latest second weight and the latest third weight.

[0024] Among them, the first weight is used to weight N actual meteorological data, the second weight is used to weight N simulated meteorological data, and the third weight is determined based on the weight of each layer of the multilayer perceptron. Specifically, the first weight adjusts the influence of the actual meteorological data, and the second weight adjusts the influence of the simulated meteorological data. The third weight is the weight of the multilayer perceptron used to process and generate the final output.

[0025] One update method is: each round of iteration uses the training data for forward propagation and back propagation, calculates the loss and updates the weights. After each iteration, the model gradually approaches the actual runoff data and optimizes its prediction results. Through multiple iterations, the neural network model continuously adjusts its internal weights until the model's output is close to the actual runoff data.

[0026] Step S103, generating N simulated statistics based on the latest first weight, the latest second weight and the latest third weight.

[0027] Among them, a simulated statistic indicates the difference between an actual meteorological data and a mapped simulated meteorological data, and a simulated statistic maps an actual meteorological data. The simulated statistic is a measure of the difference between the actual meteorological data and its corresponding simulated meteorological data, indicating the deviation between the simulated meteorological data and the actual meteorological data in certain characteristics (such as temperature, precipitation, etc.). Each actual meteorological data will correspond to a simulated statistic, which is used to measure the difference between it and the simulated meteorological data.

[0028] One way to generate is: the simulated statistic can be defined as some error measure, such as mean square error or correlation coefficient. For each actual meteorological data and its corresponding simulated meteorological data, the mean square error or other statistics between them can be calculated.

[0029] Step S104: generating a counterfeit threshold based on a preset error rate threshold and N counterfeit statistics.

[0030] Among them, by setting a preset error rate threshold, those fake data that are significantly different from the actual meteorological data can be screened out.

[0031] One way to generate is to calculate the statistical distribution of the differences between all actual meteorological data and simulated meteorological data, and use quantiles (such as the 90% quantile) to set the threshold. If the simulated statistic exceeds the threshold, it is considered that there is a significant difference between the data. The choice of error threshold can be based on the error rate of the model and the required accuracy. The best error threshold can be selected through cross-validation, so that the screened data can improve the prediction accuracy of the model without losing too much information.

[0032] Step S105 , selecting actual meteorological data whose mapped counterfeit statistics are greater than a counterfeit threshold from the N actual meteorological data as meteorological data for the runoff forecast model.

[0033] For each actual meteorological data, if the corresponding counterfeit statistic is greater than the counterfeit threshold, the actual meteorological data is considered to be data with a large difference from the counterfeit meteorological data, so that the actual meteorological data with a large difference from the counterfeit data can be screened out.

[0034] One selection method is to select the actual meteorological data whose counterfeit statistics are greater than the counterfeit threshold, and believe that these data are more different from the counterfeit data in meteorological characteristics and therefore have higher credibility. By screening out these data, the runoff forecast model can accurately learn the more complex relationship between meteorology and runoff.

[0035] The meteorological data screening method of the runoff forecast model provided by the present invention automatically iterates and optimizes the weight of the filter layer through a neural network model, and generates a simulated statistic reflecting the difference of the meteorological data, thereby avoiding the manual screening of meteorological data in the traditional method, thereby greatly improving the data processing efficiency and accuracy. The model can assign appropriate weights to different data according to the different characteristics of the actual meteorological data and the simulated meteorological data by automatically optimizing the first, second, and third weights. The model is enabled to more accurately extract valuable information from multi-source data, thereby improving the reliability of runoff forecasting. By setting an error rate threshold and generating a simulated statistic, the difference between each group of actual meteorological data and its mapped simulated meteorological data can be automatically evaluated, thereby screening out actual meteorological data that has a higher contribution to the forecast. The model is enabled to better process complex meteorological data and improve the accuracy of runoff prediction.

[0036] In an optional embodiment, in order to further train the runoff forecast model, after selecting actual meteorological data whose mapped imitation statistics are greater than the imitation threshold from N actual meteorological data as the meteorological data of the runoff forecast model, it also includes: based on the preset initial time series length and the initial prediction step number, M target actual meteorological data are screened out from the meteorological data of the runoff forecast model, where M≤N is a positive integer. A training set and a validation set are divided according to a preset ratio from the M target actual meteorological data and the actual runoff data corresponding to the future time. The runoff forecast model is trained based on the training set and the validation set to obtain the hyperparameters of the runoff forecast model after optimization. Based on the optimized hyperparameters, an updated runoff forecast model is established. The training set and the validation set are merged to obtain a merged training set. Based on the merged training set, an updated runoff forecast model is trained to obtain a trained runoff forecast model.

[0037] In this embodiment, first, a part of the actual meteorological data set is selected, and this part of the data is greater than the counterfeit threshold in the mapped counterfeit statistics. Based on a preset initial time series length (for example, actual meteorological data for a period of time in the past) and the corresponding initial prediction steps (i.e., the length of the future time period to be predicted. For example, predicting the runoff in the next 1 day, 3 days, 7 days or longer), M target actual meteorological data are screened out from these actual meteorological data. From the screened M target actual meteorological data and the actual runoff data of the corresponding historical time, they are divided into training sets and verification sets according to a certain ratio. The training set is used for actual model training, and the verification set is used to evaluate the generalization ability of the model. The runoff forecast model is trained based on the training set and the verification set, and the optimized hyperparameters are obtained by model optimization. The optimization of hyperparameters can be achieved using methods such as grid search and Bayesian optimization. Through the optimized hyperparameters, an updated runoff forecast model is established. This process is the key to model optimization, which ensures that the model can adapt to the characteristics of the training data, thereby improving the prediction accuracy.

[0038] The method provided in this embodiment, by screening appropriate meteorological data and historical runoff data corresponding to the future time, and optimizing hyperparameters, the model can better adapt to the actual data, thereby improving the accuracy of runoff prediction. Selecting data whose mapped imitation statistics are greater than the imitation threshold can effectively remove irrelevant or low-quality meteorological data and improve the training efficiency and quality of the model. By merging the training set and the validation set and training on the merged data set, the data can be more fully utilized, the information lost in the validation set can be reduced, and the model can learn more complete data features. Through hyperparameter optimization, it is ensured that the model can be trained under better parameter settings, avoiding overfitting or underfitting problems during the training process. Establishing an updated runoff forecast model based on the optimized hyperparameters and further training the model using the merged training set can make the model more adaptable and stable in practical applications.

[0039] In an optional implementation, in order to further evaluate the runoff forecast model, after training the updated runoff forecast model based on the merged training set to obtain the trained runoff forecast model, it also includes: dividing a test set according to a preset ratio from the M target actual meteorological data and the actual runoff data corresponding to the future time. Based on the test set, the generalization ability and performance of the trained runoff forecast model are evaluated.

[0040] In this embodiment, after the training is completed and the updated runoff forecast model is obtained, a test set is divided out according to a preset ratio based on M target actual meteorological data and the actual runoff data corresponding to the future time. The test set should be data that has not been seen during the training process, and can effectively evaluate the generalization ability of the model, that is, the performance of the model in practical applications. The generalization ability of the model refers to its performance on new data that has not been seen. By evaluating the model through the test set, we can understand whether the model can accurately predict new runoff data, not just perform well on the training data. Use standard evaluation indicators to quantitatively evaluate the performance of the model. By comparing with real actual runoff data, we can further understand the accuracy of the model in practical applications.

[0041] The method provided in this embodiment is effectively verified by the use of the test set in the practical application ability of the model. The model should not only optimize the performance on the training set, but also be able to accurately predict unknown data to ensure its wide application potential. Through the evaluation of the test set, it can be found that the model may have overfitting or underfitting problems. If the performance of the model on the test set is poor, the stability and reliability of the model can be improved by adjusting the training process, reselecting data or optimizing hyperparameters. Based on the evaluation results of the test set, the model is further optimized. For example, if it is found that certain specific types of meteorological data or historical time periods have a greater impact on the model prediction results, it can be considered to enhance the training of this part of the data or make corresponding adjustments.

[0042] In an optional implementation, in order to further accurately obtain data, the actual runoff data and N actual meteorological data of the target runoff area in historical time, and N simulated meteorological data mapped one-to-one with the N actual meteorological data are obtained, including: obtaining the actual runoff data and N actual meteorological data of the target runoff area in historical time. Based on the adaptive moving smoothing method, the actual runoff data is optimized. Based on the N actual meteorological data and the approximate semi-positive definite programming method, N simulated meteorological data are generated.

[0043] In this embodiment, the window size is adjusted by an adaptive algorithm, and the weight in the smoothing process is dynamically determined, so that smoothing is performed according to the local change trend of the data. The outliers or interference information in the actual runoff data are removed to make the data more accurate and convenient for subsequent analysis and modeling. When the actual meteorological data is irrelevant to the runoff, the simulated data generated by the approximate semi-positive programming method are consistent in distribution with the actual meteorological data in a certain sense; when the actual meteorological data is related to the runoff, the simulated data constructed are quite different from the meteorological data. These simulated data can further expand the training set, provide more diverse meteorological conditions for the model, and thus improve the generalization ability of the model.

[0044] The method provided in this embodiment optimizes the actual runoff data through an adaptive moving smoothing method, making it smoother and more accurate, thereby improving the prediction accuracy of the model. By generating simulated meteorological data, the characteristic dimension of the training data set of the model is expanded. Through more real and rich input data (including optimized actual runoff data and generated simulated meteorological data), the model can better capture the complex relationship between meteorological factors and runoff, further improving the accuracy of runoff forecasting.

[0045] In an optional implementation, in order to further accurately determine the third weight, after inputting the N actual meteorological data and the N simulated meteorological data into the neural network model, the method further includes: obtaining the pre-weight of the multi-layer perceptron connected to the filter layer and each layer-by-layer weight of the output layer of the multi-layer perceptron connected to the neural network model. Based on the pre-weight and each layer-by-layer weight, the third weight is determined.

[0046] In this embodiment, the pre-weight is the pre-weight of the connection filter layer in the multi-layer perceptron, that is, the influence of the pre-weight on the data when the data enters the neural network through the filter layer. These weights are crucial for feature extraction and subsequent processing of the data. The calculation of the pre-weight can be trained based on gradient descent or other optimization algorithms to ensure that the features are effectively extracted. The layer-by-layer weights are located between the hidden layers of the neural network. The layer-by-layer weights of each hidden layer reflect the connection strength of the neurons in that layer. The layer-by-layer weights are parameters that are continuously optimized during the training process of the neural network. Through multiple iterations, the layer-by-layer weights are gradually adjusted to minimize the fitting error of the model.

[0047] The method provided in this embodiment can effectively improve the learning ability of the neural network model for complex data relationships through accurate third weights, so that the neural network model can fit the input actual meteorological data and simulated meteorological numbers, and then obtain weights for screening actual meteorological data, so as to accurately screen the meteorological data of the runoff forecast model in the future, thereby improving the accuracy of runoff forecast. By optimizing the weight structure in the multi-layer perceptron, the model can better cope with different data inputs and improve its performance on unknown data. By determining the appropriate third weight, the convergence process of the model can be accelerated, the training time can be reduced, and the model efficiency can be improved.

[0048] In an optional embodiment, in order to further accurately use the neural network model, N actual meteorological data and N simulated meteorological data are input into the neural network model, including: inputting the N simulated meteorological data and the N actual meteorological data into the filter layer, and using the first weight and the second weight to generate the weighted sum of the N simulated meteorological data and the N actual meteorological data as N fused meteorological features. The N fused meteorological features are input into the first hidden layer of the multilayer perceptron to obtain the output result of the first hidden layer. The output result of the first hidden layer is input into the middle hidden layer of the multilayer perceptron layer by layer, and the output result of the middle hidden layer of the multilayer perceptron is obtained layer by layer. The output result of the last middle hidden layer of the multilayer perceptron is input into the multilayer perceptron and the final output result of the neural network model is output by the output layer.

[0049] In this embodiment, a weighted sum operation is performed on N actual meteorological data and N simulated meteorological data through a first weight and a second weight. The first weight and the second weight represent the relative importance of the actual meteorological data and the simulated meteorological data in the model. The result of the weighted sum is N fused meteorological features, which will be used as input for subsequent calculations. The N fused meteorological features are input to the first hidden layer of the multilayer perceptron. The output results of the first hidden layer are input to the middle hidden layers of the multilayer perceptron layer by layer. In each layer, the data is processed by the weighted summation and activation function of the neurons, and more complex features are gradually learned. In each layer, the adjustment and optimization process of the weights will affect the output results. The multilayer perceptron is processed by several middle hidden layers, and finally the output results of the middle hidden layers are input to the output layer of the multilayer perceptron. The result generated by the output layer is the final output of the neural network model.

[0050] The method provided in this embodiment enables the model to cope with different types of input data by fusing actual and simulated meteorological data, thereby improving the forecast accuracy in a changeable meteorological environment. Through layer-by-layer processing of a multi-layer perceptron, the model can extract more complex and abstract features, thereby improving the screening ability. Through weighted fusion and layer-by-layer transmission, the model can better learn the nonlinear relationship between meteorological data, thereby obtaining a more accurate forecast result.

[0051] The present invention provides a meteorological data screening device for a runoff forecast model. Figure 2 Schematic diagram of a meteorological data screening device for a runoff forecast model according to an embodiment of the present invention, the device comprising: The acquisition module 201 is used to acquire the actual runoff data and N actual meteorological data of the target runoff area in historical time, and N simulated meteorological data mapped one-to-one with the N actual meteorological data, where N is a positive integer.

[0052] The updating module 202 is used to input N actual meteorological data and N simulated meteorological data into the neural network model, and iteratively update the first weight and the second weight of the filter layer of the neural network model, and the third weight of the multi-layer perceptron of the neural network model, until the neural network model outputs the actual runoff data, and obtains the latest first weight, the latest second weight and the latest third weight, wherein the first weight is used to weight the N actual meteorological data, the second weight is used to weight the N simulated meteorological data, and the third weight is determined based on the weight of each layer of the multi-layer perceptron.

[0053] The first generation module 203 is used to generate N simulated statistics based on the latest first weight, the latest second weight and the latest third weight, wherein one simulated statistic indicates the difference between an actual meteorological data and a mapped simulated meteorological data, and one simulated statistic maps an actual meteorological data.

[0054] The second generating module 204 is used to generate a counterfeit threshold based on a preset error rate threshold and N counterfeit statistics.

[0055] The selection module 205 is used to select actual meteorological data whose mapped counterfeit statistics are greater than the counterfeit threshold from the N actual meteorological data as meteorological data for the runoff forecast model.

[0056] The meteorological data screening device of the runoff forecast model provided by the present invention automatically iterates and optimizes the weight of the filter layer through a neural network model, and generates a simulated statistic reflecting the difference of the meteorological data, thereby avoiding the manual screening of meteorological data in the traditional method, thereby greatly improving the data processing efficiency and accuracy. The model can assign appropriate weights to different data according to the different characteristics of the actual meteorological data and the simulated meteorological data by automatically optimizing the first, second, and third weights. The model is enabled to more accurately extract valuable information from multi-source data, thereby improving the reliability of runoff forecasting. By setting an error rate threshold and generating a simulated statistic, the difference between each group of actual meteorological data and its mapped simulated meteorological data can be automatically evaluated, thereby screening out data with a higher contribution to the forecast. The model is enabled to better process complex meteorological data and improve the accuracy of runoff prediction.

[0057] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.

[0058] The meteorological data screening device of the runoff forecast model in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

[0059] The embodiment of the present invention also provides a computer device having the above Figure 2 The meteorological data screening device of the runoff forecast model is shown.

[0060] See also Figure 3 , Figure 3 is a schematic diagram of the structure of a computer device provided by an optional embodiment of the present invention, such as Figure 3 As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components are connected to each other using different buses for communication, and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in or on the memory to display the graphical information of the GUI on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 3 A processor 10 is taken as an example.

[0061] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include an integrated circuit. The integrated circuit may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic or any combination thereof.

[0062] The memory 20 stores instructions executable by at least one processor 10, so that the at least one processor 10 executes the method shown in the above embodiment.

[0063] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0064] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive; the memory 20 may also include a combination of the above types of memory.

[0065] The computer device further comprises a communication interface 30 for the computer device to communicate with other devices or a communication network.

[0066] The embodiment of the present invention also provides a computer-readable storage medium. The method according to the embodiment of the present invention can be implemented in hardware, firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium through a network download, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state hard disk, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor, or hardware, the method shown in the above embodiment is implemented.

[0067] A part of the present invention may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present invention through the operation of the computer. Those skilled in the art should understand that the existence of the computer program instruction in a computer-readable medium includes, but is not limited to, a source file, an executable file, an installation package file, etc., and accordingly, the way in which the computer program instruction is executed by the computer includes, but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium may be any available computer-readable storage medium or communication medium accessible to the computer.

[0068] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A meteorological data screening method for a runoff forecast model, characterized in that: The method comprises: Obtaining actual runoff data and N actual meteorological data of the target runoff area in historical time, and N simulated meteorological data mapped one-to-one with the N actual meteorological data, where N is a positive integer; Input the N actual meteorological data and the N simulated meteorological data into a neural network model, and iteratively update a first weight and a second weight of a filter layer of the neural network model, and a third weight of a multilayer perceptron of the neural network model, until the neural network model outputs the actual runoff data, and obtains the latest first weight, the latest second weight, and the latest third weight, wherein the first weight is used to weight the N actual meteorological data, the second weight is used to weight the N simulated meteorological data, and the third weight is determined based on the weight of each layer of the multilayer perceptron; Based on the latest first weight, the latest second weight and the latest third weight, N simulated statistics are generated, wherein one simulated statistic indicates the difference between one actual meteorological data and one mapped simulated meteorological data, and one simulated statistic maps one actual meteorological data; Generate a counterfeit threshold based on a preset error rate threshold and the N counterfeit statistics; Actual meteorological data whose mapped simulated statistics are greater than the simulated threshold value are selected from the N actual meteorological data as meteorological data for the runoff forecast model.

2. The method according to claim 1, characterized in that After selecting the actual meteorological data whose mapped simulated statistics are greater than the simulated threshold value from the N actual meteorological data as the meteorological data of the runoff forecast model, the method further includes: Based on the preset initial time series length and the initial prediction step number, M target actual meteorological data are selected from the meteorological data of the runoff forecast model, where M≤N is a positive integer; Dividing the M target actual meteorological data and the actual runoff data corresponding to the future time into a training set and a validation set according to a preset ratio; Training the runoff prediction model based on the training set and the validation set to obtain optimized hyperparameters of the runoff prediction model; Based on the optimized hyperparameters, an updated runoff forecast model is established; Merging the training set and the validation set to obtain a merged training set; Based on the combined training set, the updated runoff forecast model is trained to obtain a trained runoff forecast model.

3. The method according to claim 2, characterized in that After training the updated runoff forecast model based on the combined training set to obtain the trained runoff forecast model, the method further includes: Dividing the M target actual meteorological data and the actual runoff data corresponding to the future time into a test set according to a preset ratio; Based on the test set, the generalization ability and performance of the trained runoff forecast model are evaluated.

4. The method according to any one of claims 1 to 3, characterized in that: The method of obtaining the actual runoff data of the target runoff area in historical time and N actual meteorological data, as well as N simulated meteorological data mapped one-to-one with the N actual meteorological data, includes: Obtain the actual runoff data and N actual meteorological data of the target runoff area in historical time; Optimizing the actual runoff data based on an adaptive moving smoothing method; Based on the N actual meteorological data and an approximate semi-positive definite programming method, the N simulated meteorological data are generated.

5. The method according to any one of claims 1 to 3, characterized in that: After the N actual meteorological data and the N simulated meteorological data are input into the neural network model, the method further includes: Obtaining the pre-weight of the multi-layer perceptron connected to the filter layer and each layer-by-layer weight of the multi-layer perceptron connected to the output layer of the neural network model; The third weight is determined based on the pre-weight and each of the layer-by-layer weights.

6. The method according to any one of claims 1 to 3, characterized in that: The step of inputting the N actual meteorological data and the N simulated meteorological data into a neural network model comprises: Inputting the N simulated meteorological data and the N actual meteorological data into the filtering layer, and using the first weight and the second weight to generate a weighted sum of the N simulated meteorological data and the N actual meteorological data as N fused meteorological features; Inputting the N fused meteorological features into the first hidden layer of the multilayer perceptron to obtain an output result of the first hidden layer; Input the output results of the first hidden layer into the middle hidden layers of the multilayer perceptron layer by layer, and obtain the output results of the middle hidden layers of the multilayer perceptron layer by layer; Input the output result of the last intermediate hidden layer of the multilayer perceptron into the output layer of the multilayer perceptron to obtain the output result of the multilayer perceptron; The output result of the multilayer perceptron is input into the output layer of the neural network model to obtain the output result of the neural network model.

7. A meteorological data screening device for a runoff forecast model, characterized in that: The device comprises: An acquisition module is used to acquire actual runoff data and N actual meteorological data of the target runoff area in historical time, and N simulated meteorological data mapped one-to-one with the N actual meteorological data, where N is a positive integer; An updating module, used for inputting the N actual meteorological data and the N simulated meteorological data into a neural network model, and iteratively updating a first weight and a second weight of a filter layer of the neural network model, and a third weight of a multilayer perceptron of the neural network model, until the neural network model outputs the actual runoff data, and obtaining the latest first weight, the latest second weight, and the latest third weight, wherein the first weight is used to weight the N actual meteorological data, the second weight is used to weight the N simulated meteorological data, and the third weight is determined based on the weight of each layer of the multilayer perceptron; A first generating module is used to generate N simulated statistics based on the latest first weight, the latest second weight and the latest third weight, wherein one simulated statistic indicates the difference between one actual meteorological data and one mapped simulated meteorological data, and one simulated statistic maps one actual meteorological data; A second generating module, configured to generate a counterfeit threshold based on a preset error rate threshold and the N counterfeit statistics; The selection module is used to select the actual meteorological data whose mapped simulated statistics are greater than the simulated threshold from the N actual meteorological data as the meteorological data of the runoff forecast model.

8. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the meteorological data screening method for the runoff forecast model according to any one of claims 1 to 6 by executing the computer instructions.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the meteorological data screening method for the runoff forecast model according to any one of claims 1 to 6.

10. A computer program product, characterized in that The method comprises computer instructions for causing a computer to execute the meteorological data screening method for a runoff forecast model according to any one of claims 1 to 6.

Images