Precipitation prediction method, precipitation prediction device, storage medium, and electronic device
By using the random forest algorithm to select key factors and combining it with convolutional neural networks, backpropagation neural networks, and temporal convolutional network models, the problems of large computational load and inaccurate accuracy in precipitation prediction in existing technologies are solved, achieving higher prediction accuracy and stability.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- YUNNAN POWER GRID CO LTD ELECTRIC POWER RES INST
- Filing Date
- 2024-12-04
- Publication Date
- 2026-07-24
Smart Images
Figure CN119620240B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of data processing technology, and in particular to a precipitation prediction method, a precipitation prediction device, a storage medium, and an electronic device. Background Technology
[0002] Currently, meteorological models and statistical methods are commonly used to predict persistent precipitation. Meteorological models use physical principles to simulate atmospheric motion and climate change. These models typically consider many factors, such as temperature, humidity, wind speed, pressure, and the interactions between these factors. By inputting initial and boundary conditions, meteorological models can predict future weather conditions. However, this method has drawbacks: it is computationally intensive, requiring high-performance computers for computation, and is highly sensitive to initial and boundary conditions, which may lead to unstable prediction results. Statistical methods are methods that use historical data to build mathematical models to predict future trends. These methods typically include time series analysis, regression analysis, and machine learning. The advantages of these methods are their simplicity and ease of implementation, and lower computational resource requirements. However, the accuracy of statistical methods can be affected by factors such as data quality and quantity, model selection, and training and testing datasets. These issues contribute to the inaccuracy and risk in persistent precipitation prediction to some extent. Summary of the Invention
[0003] This application provides a precipitation prediction method, a precipitation prediction device, a storage medium, and an electronic device, which can improve the accuracy of precipitation prediction.
[0004] In a first aspect, this application provides a precipitation prediction method, including:
[0005] Acquire precipitation observation data during periods of continuous precipitation in the target forecast area, as well as environmental impact factors for the target forecast area;
[0006] The importance of the environmental impact factors to the precipitation observation data is determined based on random forest, and the target features are determined based on the importance.
[0007] The convolutional neural network model, the backpropagation neural network model, and the temporal convolutional network model are obtained by training the precipitation observation data and the target features.
[0008] The current precipitation data of the target forecast area is input into the convolutional neural network model, the backpropagation neural network model, and the temporal convolutional network model respectively to obtain the first prediction data, the second prediction data, and the third prediction data output by each model.
[0009] The persistent precipitation data for the target forecast area is determined by combining the first forecast data, the second forecast data, and the third forecast data.
[0010] According to the precipitation prediction method in this embodiment, a prediction model is constructed by acquiring precipitation observation data during persistent precipitation in the target forecast area and environmental impact factors of the target forecast area. This makes the model more targeted to the target forecast area and persistent precipitation, improving the accuracy of persistent precipitation prediction. Furthermore, the random forest algorithm is used to extract target features from environmental impact factors, which helps identify the key factors with the greatest impact on precipitation, thereby improving the model's predictive ability. Model ensemble is achieved by training three different models: convolutional neural network, backpropagation neural network, and temporal convolutional network. This leverages the advantages of different models to improve the robustness and accuracy of the prediction. The temporal convolutional network model can capture the time dependence in precipitation data, improving the prediction accuracy for persistent precipitation. Combining the prediction results of the three different models yields a more comprehensive and reliable precipitation prediction, reducing the prediction error of individual models and improving overall prediction performance.
[0011] In one exemplary embodiment, determining the persistent precipitation data for the target forecast area by combining the first forecast data, the second forecast data, and the third forecast data includes:
[0012] Determine the weights of the first predicted data, the second predicted data, and the third predicted data;
[0013] The first, second, and third predicted data are weighted and averaged based on the weights to obtain the continuous precipitation data.
[0014] In one exemplary implementation, determining the weights of the first predicted data, the second predicted data, and the third predicted data includes:
[0015] Obtain a test set, which includes persistent precipitation observation data for the target forecast area;
[0016] The continuous precipitation observation data are predicted using the convolutional neural network model, the backpropagation neural network model, and the temporal convolutional network model, respectively, to obtain the first test result, the second test result, and the third test result.
[0017] Based on the first test result, the second test result, and the third test result, the prediction pass rate of the convolutional neural network model, the backpropagation neural network model, and the temporal convolutional network model are determined respectively.
[0018] The weights of the convolutional neural network model, the backpropagation neural network model, and the temporal convolutional network model are determined based on the predicted pass rate.
[0019] In one exemplary embodiment, it further includes:
[0020] The least squares method is used to solve for the first test result, the second test result, and the third test result to obtain the weights of the convolutional neural network model, the backpropagation neural network model, and the temporal convolutional network model.
[0021] In one exemplary embodiment, acquiring the current precipitation data of the target forecast area and the current environmental impact factors of the target forecast area includes:
[0022] Obtain current precipitation data for the target forecast area from the meteorological information center;
[0023] Obtain meteorological forecast data output by the CFSv2 model, and extract the current environmental impact factors of the target forecast area from the meteorological forecast data.
[0024] In one exemplary implementation, the extraction of target features from the current environmental impact factors based on random forest includes:
[0025] Obtain a training set, which includes daily precipitation observation data and environmental impact factors during historical periods of continuous precipitation;
[0026] The random forest is trained using the training set, enabling it to predict the precipitation data for the following day based on the environmental influencing factors and the previous day's precipitation observation data.
[0027] The importance of the current environmental influencing factors to the current precipitation observation data is determined based on the trained random forest.
[0028] Target features are extracted from the current environmental impact factors based on the aforementioned importance.
[0029] In one exemplary embodiment, it further includes:
[0030] The random forest is used to extract feature variables from the environmental impact factors;
[0031] The convolutional neural network model, backpropagation neural network model, and temporal convolutional network model are trained based on the daily precipitation observation data during the historical continuous precipitation period and the feature variables.
[0032] Secondly, this application provides a precipitation prediction device, comprising:
[0033] The data acquisition module is used to acquire current precipitation observation data and current environmental impact factors of the target forecast area.
[0034] The feature extraction module is used to extract target features from the current environmental impact factors based on random forest;
[0035] The prediction module is used to input the current precipitation observation data and the target features into the convolutional neural network model, the backpropagation neural network model, and the temporal convolutional network model to obtain the first prediction data, the second prediction data, and the third prediction data output by each model.
[0036] The integrated forecasting module is used to combine the first forecast data, the second forecast data, and the third forecast data to determine the persistent precipitation data for the target forecast area.
[0037] Thirdly, this application provides an electronic device including a memory and one or more processors. The memory stores one or more computer programs, each including instructions that, when executed by the processor, cause the electronic device to perform the precipitation prediction method as described in the first aspect.
[0038] Fourthly, this application provides a computer-readable storage medium storing instructions that, when executed on an electronic device, cause the electronic device to perform the precipitation prediction method as described in the first aspect.
[0039] Fifthly, this application provides a computer program product that, when run on an electronic device, causes the electronic device to perform the precipitation prediction method as described in the first aspect.
[0040] Understandably, the beneficial effects achieved by the above-mentioned precipitation forecasting devices, electronic devices, computer-readable storage media, and computer program products can be referred to the beneficial effects in the first aspect, and will not be repeated here. Attached Figure Description
[0041] Figure 1 A schematic flowchart illustrating the precipitation prediction method provided in the embodiments of this application;
[0042] Figure 2 Line graph showing the prediction effect of the precipitation prediction method provided in the embodiments of this application;
[0043] Figure 3 This is a schematic diagram of the precipitation prediction device provided in the embodiments of this application;
[0044] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation
[0045] To facilitate a clear description of the technical solutions in the embodiments of this application, the terms "first" and "second" are used in the embodiments of this application to distinguish identical or similar items with substantially the same function and effect. For example, "first chip" and "second chip" are only used to distinguish different chips and do not limit their order. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and the terms "first" and "second" do not necessarily imply that they are different. It should be noted that in the embodiments of this application, the words "exemplary" or "for example" are used to indicate that they are examples, illustrations, or descriptions. Any embodiment or design scheme described as "exemplary" or "for example" in this application should not be construed as being better or more advantageous than other embodiments or design schemes. Specifically, the use of the words "exemplary" or "for example" is intended to present the relevant concepts in a specific manner. In the embodiments of this application, "at least one" means one or more, and "more than one" means two or more.
[0046] It should be noted that "at the time of..." in the embodiments of this application can be either at the instant when a certain situation occurs, or for a period of time after the occurrence of a certain situation. The embodiments of this application do not make specific limitations on this.
[0047] The implementation of this embodiment will now be described in detail with reference to the accompanying drawings.
[0048] This embodiment provides a precipitation forecasting method that can predict persistent precipitation and improve forecast accuracy. For example, this precipitation forecasting method can be applied to various electronic devices such as computers (PCs), tablets, virtual reality / augmented reality devices, wearable devices, industrial computers, and in-vehicle systems; it can also be applied to servers, the cloud, server clusters, etc., and this embodiment does not impose any special limitations on it.
[0049] Figure 1 A flowchart illustrating the precipitation prediction method provided in this application embodiment is shown.
[0050] like Figure 1 As shown, this precipitation prediction method may include the following steps:
[0051] Step 101: Obtain the current precipitation observation data of the target forecast area and the current environmental impact factors of the target forecast area.
[0052] Current precipitation observation data can include daily precipitation data up to the forecast date, such as daily precipitation amounts. The target forecast area refers to the region where precipitation forecasting is required.
[0053] Specifically, current precipitation data for the target forecast area is obtained from the meteorological information center; meteorological forecast data output by the CFSv2 model is obtained, and current environmental impact factors for the target forecast area are extracted from the meteorological forecast data.
[0054] CFSv2 (Climate Forecast System Version 2) is a global climate model used for climate prediction and seasonal weather forecasting. This application uses data output from the CFSv2 model as its data source.
[0055] First, two sets of data are prepared: one set is precipitation observation data from the National Meteorological Information Center, and the other set is a preliminary screening factor dataset. Daily meteorological observation data, i.e., current precipitation data, can be obtained from the National Meteorological Information Center. Based on the CFSv2 model, precipitation, sea level pressure, geopotential height at 200hPa, 500hPa, 700hPa, 850hPa, and 925hPa, relative humidity, temperature, and UV wind speed data are extracted according to the forecast time. This constructs a factor dataset containing 27 feature variables, i.e., the current environmental impact factors.
[0056] Step 102: Extract target features from the current environmental impact factors based on random forest.
[0057] Random Forest (RF) is an ensemble learning method that improves prediction accuracy by constructing multiple decision trees and voting on their predictions. In this implementation, the random forest is pre-trained and then used to evaluate the importance of features (environmental impact factors). Specifically, this includes: obtaining a training set containing daily precipitation observation data and environmental impact factors during historical periods of continuous precipitation; training the random forest using the training set so that it can predict the next day's precipitation data based on the environmental impact factors and the daily precipitation observation data; determining the importance of the current environmental impact factors to the current precipitation observation data based on the trained random forest; and extracting target features from the current environmental impact factors based on their importance.
[0058] This implementation method can predict continuous precipitation. Therefore, it extracts daily precipitation observation data and environmental impact factors from historical continuous precipitation periods. The data is preprocessed, such as removing missing values, handling outliers, standardizing, and normalizing the data, to obtain a training set. The acquired environmental impact factors are used as input features to construct a decision tree. Each tree is constructed by randomly sampling feature variables. The optimal split point can be determined using bootstrap sampling, information gain, or Gini impurity. Node splitting continues until a preset stopping condition is met, resulting in leaf nodes. Each tree can predict the outcome, and the prediction with the highest probability is ultimately selected. Through training, the difference between the precipitation data predicted by the random forest based on earlier daily precipitation observation data and the actual precipitation data can be reduced to below a threshold.
[0059] In a random forest, the contribution of each feature variable to node splits can be calculated for each tree. This contribution is assessed by measuring the increase in node purity after the feature variable split. Feature purity is calculated using Gini impurity and information gain. The contribution of each feature variable across all trees is then calculated, resulting in an importance score for each feature variable. Based on this importance score, the 10 most important environmental impact factors are selected as the final feature variables.
[0060] In this embodiment, the current environmental impact factors can be input into the trained random forest to obtain the importance score of each current environmental impact factor, and then output a certain number of environmental factors with high importance, such as the top 10, as target features.
[0061] Random forest is a powerful nonlinear model capable of effectively handling complex data relationships and exhibiting good generalization ability. Factor selection using random forest involves calculating the importance of factors and selecting the top 10 factors with the greatest contribution. This method can identify the main factors that have the greatest impact on persistent rainfall in Yunnan Province, thus providing a scientific basis for rainstorm forecasting and disaster prevention and mitigation, and improving the resilience of Yunnan Province under extreme weather conditions.
[0062] Step 103: Input the current precipitation observation data and the target features into the convolutional neural network model, the backpropagation neural network model, and the temporal convolutional network model to obtain the first predicted data, the second predicted data, and the third predicted data output by each model.
[0063] The target features output by the random forest and the current precipitation observation data are used as inputs to the convolutional neural network model, the backpropagation neural network model, and the temporal convolutional network model to obtain the prediction data output by the three models, which are the first prediction data, the second prediction data, and the third prediction data, respectively.
[0064] Among them, the Temporal Convolutional Network (TCN) model is a deep learning model for processing time series data. It can effectively handle long sequences and controls the receptive field of the network by adjusting the inflation factor, making it more suitable for processing time series data with long-term dependencies.
[0065] Convolutional Neural Network (CNN) models can extract and identify complex patterns in data, and due to the characteristics of the convolutional structure, they are data invariant to translation, scaling, and rotation.
[0066] Backpropagation Neural Network (BPNN) models can perform supervised learning. They can adjust the weights and biases in the network based on the training set using the backpropagation algorithm, thereby minimizing prediction errors.
[0067] This implementation also includes: extracting feature variables from the environmental impact factors through the random forest; and training the convolutional neural network model, the backpropagation neural network model, and the temporal convolutional network model based on the daily precipitation observation data during the historical continuous precipitation and the feature variables.
[0068] Obtain a training set, and extract a set of pre-defined feature variables with high importance from the environmental impact factors in the training set using a random forest. Input these feature variables and the precipitation observation data of the previous i days into a convolutional neural network model, a backpropagation neural network model, and a temporal convolutional network model. Train these three models respectively. Adjust the model parameters by the difference between the predicted precipitation data of the i+1th day and the actual precipitation data of the i+1th day until the difference between the predicted precipitation data and the actual precipitation data of the model output meets the preset conditions, and the training is completed.
[0069] The trained convolutional neural network model, backpropagation neural network model, and temporal convolutional network model are used to predict the current precipitation data.
[0070] Step 104: Combine the first forecast data, the second forecast data, and the third forecast data to determine the persistent precipitation data for the target forecast area.
[0071] Persistent precipitation data refers to precipitation data for periods after the current time, such as precipitation amount. For example, current precipitation observation data could be the precipitation data from the previous j days, while persistent precipitation data could be the precipitation data for days j, j+1, j+2, etc.
[0072] In this embodiment, the final persistent precipitation data can be obtained by calculating the average of the three predicted data output by the three models. Alternatively, a weighted average method can be used to obtain the final persistent precipitation data. Specifically, this includes: determining the weights of the first predicted data, the second predicted data, and the third predicted data; and performing a weighted average of the first predicted data, the second predicted data, and the third predicted data based on the weights to obtain the persistent precipitation data.
[0073] For example, the process of determining the weights is as follows: A test set is obtained, which includes persistent precipitation observation data for the target forecast area; the persistent precipitation observation data is predicted using the convolutional neural network model, the backpropagation neural network model, and the temporal convolutional network model to obtain a first test result, a second test result, and a third test result, respectively; the prediction pass rate of the convolutional neural network model, the backpropagation neural network model, and the temporal convolutional network model is determined based on the first test result, the second test result, and the third test result, respectively; and the weights of the convolutional neural network model, the backpropagation neural network model, and the temporal convolutional network model are determined based on the prediction pass rate.
[0074] The prediction success rates of convolutional neural network (CNN), backpropagation neural network (HNN), and temporal convolutional neural network (TCNN) models can be statistically analyzed using a test set beforehand. Weights are then assigned to each model based on these success rates. For example, if the success rates of the three models during the calibration period are 0.6, 0.7, and 0.8 respectively, the weights for each model can be calculated as follows: 0.6 / (0.6+0.7+0.8) = 0.286, 0.7 / (0.6+0.7+0.8) = 0.334, and 0.8 / (0.6+0.7+0.8) = 0.380.
[0075] In one exemplary embodiment, the least squares method is used to solve for the first test result, the second test result, and the third test result to obtain the weights of the convolutional neural network model, the backpropagation neural network model, and the temporal convolutional network model.
[0076] In one exemplary implementation, the weights can be optimized using simulated annealing, or other optimization algorithms can be used, such as the Adam optimizer, genetic algorithm, stochastic gradient descent, etc.
[0077] This embodiment also provides a forecasting model (Ensemble) that can implement steps 102 to 104 above. Specifically, the forecasting model can be constructed using random forest, convolutional neural network, backpropagation neural network, or temporal convolutional network. Specifically, the forecasting model can include three branches: the first branch consists of a random forest and a convolutional neural network forecasting model; the second branch consists of a random forest and a backpropagation neural network forecasting model; and the third branch consists of a random forest and a temporal convolutional network forecasting model. The outputs of the three branches are weighted and averaged by an adaptive weighting module to finally output persistent precipitation data. In each branch, target features are extracted using a random forest, and these features are then processed for prediction using a convolutional neural network model, a backpropagation neural network model, or a temporal convolutional network model to obtain future precipitation data. Then, the adaptive weighting module performs a weighted average of the precipitation data output from each branch to output the final result, i.e., persistent precipitation data. This scheme can accurately predict persistent precipitation in the target forecast area, thereby providing valuable reference information for related fields and contributing to the prevention and response to extreme weather events.
[0078] In this embodiment, the forecasting model constructed based on random forest selection factors can reduce the limitations of a single model in capturing data trends. By combining the advantages of multiple models, it significantly improves the overall forecasting performance, outperforming any single model. By fusing the prediction results of multiple models, the forecast can effectively reduce the bias and uncertainty in individual model predictions, thereby providing more robust and accurate forecast results.
[0079] Based on the above analysis, various models (including the aforementioned forecast model, a standalone convolutional neural network model, a backpropagation neural network model, and a temporal convolutional network model) were constructed using 10 factors selected by the random forest algorithm and observed precipitation data. The model prediction results are as follows: Figure 2 As shown, OBS refers to the actual observed precipitation data. From Figure 2 As can be seen, the overall trend of the predicted values of each model is largely consistent with the measured values. The correlation coefficients of Convolutional Neural Network (CNN), Temporal Convolutional Network (TCN), Backpropagation Neural Network (BPNN), and Ensemble Prediction Model are 0.75, 0.80, 0.78, and 0.89, respectively, indicating that the performance of ensemble prediction is better than that of other single models, and the ensemble prediction effect is better.
[0080] Furthermore, this embodiment also provides a precipitation prediction device, which can be used to execute the above-described precipitation prediction method. For example... Figure 3As shown, the precipitation forecasting device 300 may specifically include: a data acquisition module 301, used to acquire current precipitation observation data and current environmental impact factors of the target forecast area; a feature extraction module 302, used to extract target features from the current environmental impact factors based on random forest; a prediction module 303, used to input the current precipitation observation data and the target features into the convolutional neural network model, the backpropagation neural network model, and the temporal convolutional network model to obtain the respective outputs of first prediction data, second prediction data, and third prediction data; and a comprehensive prediction module 304, used to combine the first prediction data, second prediction data, and third prediction data to determine the persistent precipitation data of the target forecast area.
[0081] In one exemplary embodiment, the comprehensive prediction module 304 specifically includes a weighting module, used to determine the weights of the first prediction data, the second prediction data, and the third prediction data respectively; and to perform a weighted average of the first prediction data, the second prediction data, and the third prediction data based on the weights to obtain persistent precipitation data.
[0082] In one exemplary embodiment, the weighting module is further configured to: acquire a test set, the test set including persistent precipitation observation data of the target forecast area; predict the persistent precipitation observation data using the convolutional neural network model, the backpropagation neural network model, and the temporal convolutional network model, respectively, to obtain a first test result, a second test result, and a third test result; determine the prediction pass rate of the convolutional neural network model, the backpropagation neural network model, and the temporal convolutional network model based on the first test result, the second test result, and the third test result, respectively; and determine the weights of the convolutional neural network model, the backpropagation neural network model, and the temporal convolutional network model based on the prediction pass rate.
[0083] In one exemplary embodiment, the weighting module is further specifically used to: solve the first test result, the second test result, and the third test result using the least squares method to obtain the weights of the convolutional neural network model, the backpropagation neural network model, and the temporal convolutional network model.
[0084] In one exemplary embodiment, the data acquisition module 301 is specifically used to acquire current precipitation data of the target forecast area from the meteorological information center; acquire meteorological forecast data output by the CFSv2 model; and extract the current environmental impact factors of the target forecast area from the meteorological forecast data.
[0085] In one exemplary embodiment, the feature extraction module 302 is specifically used to: acquire a training set, the training set including daily precipitation observation data and environmental impact factors during historical continuous precipitation; train a random forest using the training set, so that the random forest can predict precipitation data for the following day based on the environmental impact factors and the previous day's precipitation observation data; determine the importance of the current environmental impact factors to the current precipitation observation data based on the trained random forest; and extract target features from the current environmental impact factors based on the importance.
[0086] In one exemplary embodiment, the precipitation prediction device 300 further includes a model training module for extracting feature variables from the environmental impact factors through the random forest; and training the convolutional neural network model, the backpropagation neural network model, and the temporal convolutional network model based on the daily precipitation observation data during the historical continuous precipitation and the feature variables.
[0087] The specific details of each module or unit in the above-mentioned precipitation prediction device have been described in detail in the corresponding precipitation prediction method, so they will not be repeated here.
[0088] This application also provides an electronic device. Figure 4 A schematic diagram of the structure of an electronic device suitable for implementing embodiments of the present disclosure is shown. Figure 4 The electronic device 600 shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments disclosed herein.
[0089] like Figure 4 As shown, the electronic device 600 includes a central processing unit (CPU) 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage section 608 into a random access memory (RAM) 603. The RAM 603 also stores various programs and data required for system operation. The CPU 601, ROM 602, and RAM 603 are interconnected via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.
[0090] The following components are connected to I / O interface 605: an input section 606 including a keyboard, mouse, etc.; an output section 607 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 608 including a hard disk, etc.; and a communication section 609 including a network interface card such as a LAN card, modem, etc. The communication section 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to I / O interface 605 as needed. A removable medium 611, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on drive 610 as needed so that computer programs read from it can be installed into storage section 608 as needed.
[0091] In particular, according to embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable storage medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 609, and / or installed from removable medium 611. When the computer program is executed by central processing unit (CPU) 601, it performs the functions defined in the embodiments of this application.
[0092] For example, when the computer program is executed by the central processing unit (CPU) 601, it can perform the following: acquire current precipitation observation data and current environmental impact factors of the target forecast area; extract target features from the current environmental impact factors based on random forest; input the current precipitation observation data and the target features into the convolutional neural network model, the backpropagation neural network model, and the temporal convolutional network model to obtain the first prediction data, the second prediction data, and the third prediction data output by each model; and combine the first prediction data, the second prediction data, and the third prediction data to determine the persistent precipitation data of the target forecast area.
[0093] It should be noted that the computer-readable medium disclosed herein may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium may be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this disclosure, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media can also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.
[0094] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.
[0095] The units described in the embodiments of this disclosure can be implemented in software or hardware, and the described units can also be located in a processor. The names of these units do not necessarily limit the unit itself.
[0096] In another aspect, this application also provides a computer-readable medium, which may be included in the electronic device described in the above embodiments; or it may exist independently and not assembled into the electronic device. The computer-readable medium carries one or more programs, which include instructions that, when executed by the electronic device, cause the electronic device to perform the methods described in the above embodiments.
[0097] It should be noted that although several modules or units for the device used to perform actions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to the embodiments of this application, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.
[0098] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
Claims
1. A precipitation prediction method, characterized in that, include: Obtain current precipitation observation data and current environmental impact factors for the target forecast area; The forecast model was constructed using random forest, convolutional neural network, backpropagation neural network, and temporal convolutional network models. The forecasting model consists of three branches: the first branch is composed of a random forest and a convolutional neural network forecasting model; the second branch is composed of a random forest and a backpropagation neural network forecasting model; and the third branch is composed of a random forest and a temporal convolutional network forecasting model. In each branch, target features are extracted from the current environmental impact factors based on random forest; The current precipitation observation data and the target features are input into a convolutional neural network model, a backpropagation neural network model, and a temporal convolutional network model to obtain the first, second, and third predicted data output by each model. The persistent precipitation data for the target forecast area is determined by combining the first forecast data, the second forecast data, and the third forecast data. The outputs of the three branches are weighted and averaged by an adaptive weighting module to finally output persistent precipitation data.
2. The precipitation prediction method according to claim 1, characterized in that, The step of determining persistent precipitation data for the target forecast area by combining the first forecast data, the second forecast data, and the third forecast data includes: Determine the weights of the first predicted data, the second predicted data, and the third predicted data; The first, second, and third predicted data are weighted and averaged based on the weights to obtain the continuous precipitation data.
3. The precipitation prediction method according to claim 2, characterized in that, Determining the weights of the first predicted data, the second predicted data, and the third predicted data includes: Obtain a test set, which includes persistent precipitation observation data for the target forecast area; The continuous precipitation observation data are predicted using the convolutional neural network model, the backpropagation neural network model, and the temporal convolutional network model, respectively, to obtain the first test result, the second test result, and the third test result. Based on the first test result, the second test result, and the third test result, the prediction pass rate of the convolutional neural network model, the backpropagation neural network model, and the temporal convolutional network model are determined respectively. The weights of the convolutional neural network model, the backpropagation neural network model, and the temporal convolutional network model are determined based on the predicted pass rate.
4. The precipitation prediction method according to claim 3, characterized in that, Also includes: The least squares method is used to solve for the first test result, the second test result, and the third test result to obtain the weights of the convolutional neural network model, the backpropagation neural network model, and the temporal convolutional network model.
5. The precipitation prediction method according to claim 1, characterized in that, The acquisition of current precipitation data and current environmental impact factors for the target forecast area includes: Obtain current precipitation data for the target forecast area from the meteorological information center; Obtain meteorological forecast data output by the CFSv2 model, and extract the current environmental impact factors of the target forecast area from the meteorological forecast data.
6. The precipitation prediction method according to claim 3, characterized in that, The extraction of target features from the current environmental impact factors based on random forest includes: Obtain a training set, which includes daily precipitation observation data and environmental impact factors during historical periods of continuous precipitation; The random forest is trained using the training set, enabling it to predict the precipitation data for the following day based on the environmental influencing factors and the previous day's precipitation observation data. The importance of the current environmental influencing factors to the current precipitation observation data is determined based on the trained random forest. Target features are extracted from the current environmental impact factors based on the aforementioned importance.
7. The precipitation prediction method according to claim 6, characterized in that, Also includes: The random forest is used to extract feature variables from the environmental impact factors; The convolutional neural network model, backpropagation neural network model, and temporal convolutional network model are trained based on the daily precipitation observation data during the historical continuous precipitation period and the feature variables.
8. A precipitation prediction device, characterized in that, include: The data acquisition module is used to acquire current precipitation observation data and current environmental impact factors of the target forecast area. The forecasting model is constructed using a random forest, convolutional neural network, backpropagation neural network, and temporal convolutional network model. The model comprises three branches: the first branch consists of a random forest and a convolutional neural network; the second branch also consists of a random forest and a backpropagation neural network; and the third branch consists of a random forest and a temporal convolutional network. Within each branch... The feature extraction module is used to extract target features from the current environmental impact factors based on random forest; The prediction module is used to input the current precipitation observation data and the target features into a convolutional neural network model, a backpropagation neural network model, and a temporal convolutional network model to obtain the first prediction data, the second prediction data, and the third prediction data output by each model. The integrated forecasting module is used to combine the first forecast data, the second forecast data, and the third forecast data to determine the persistent precipitation data for the target forecast area; The outputs of the three branches are weighted and averaged by an adaptive weighting module to finally output persistent precipitation data.
9. A computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to perform the precipitation prediction method as described in any one of claims 1 to 7.
10. An electronic device, characterized in that, The device includes a processor and a memory, the memory storing one or more computer programs, the one or more computer programs including instructions that, when executed by the electronic device, cause the electronic device to perform the precipitation prediction method according to any one of claims 1 to 7.