A method and system for extended-range weather forecasting based on multi-model integration
Through multi-model integration and improved convolutional neural network, LSTM/RNN model, combined with meteorological factor field and information sequence for weather forecasting, the problems of uncertainty and error in traditional methods in extended weather forecasting are solved, and the forecasting is improved accuracy.
Patent Information
- Application Number
- CN202411014691.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-26
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2044-07-26
AI Technical Summary
Traditional numerical weather forecasting methods have great uncertainty and error in extended weather forecasts. The single neural network model has great limitations when processing complex meteorological data, making it difficult to provide accurate forecast results.
Using a multi-model integration method, the precipitation prediction model of two branches is constructed, based on meteorological factor field and meteorological information sequence, and feature extraction and prediction are combined with improved convolutional neural network and LSTM/RNN model. Predict results to improve accuracy.
It improves the accuracy and robustness of extended weather forecasts and reduces the uncertainty of forecast results.
Smart Images

Figure CN118550005B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of meteorological forecasting and artificial intelligence technology, and in particular to an extended-range weather forecasting method and system based on multi-model integration. Background Art
[0002] Extended-range weather forecast refers to the weather forecast for the next 10-30 days, which is of great significance to agricultural production, energy scheduling, disaster prevention and other fields. However, due to the complexity and nonlinear characteristics of the atmospheric system, traditional numerical weather forecast methods have large uncertainties and errors in extended-range forecasts.
[0003] With the development of artificial intelligence technology, neural network models have shown great application prospects in the field of weather forecasting. However, a single neural network model has limitations in processing complex meteorological data and patterns. In order to overcome this problem, the present invention proposes an extended-range weather forecast method and system based on multi-model integration, which integrates multiple models to perform extended-range weather forecasting to improve the accuracy and robustness of the forecast. Summary of the invention
[0004] The purpose of the present invention is to solve the problem of inaccurate forecast results of a single model in the above-mentioned prior art, and to provide an extended-term weather forecast method and system based on multi-model integration to achieve the technical effect of improving the accuracy of extended-term weather forecasts.
[0005] The present invention is achieved through the following technical solutions:
[0006] The present invention provides an extended-range weather forecast method based on multi-model integration, comprising:
[0007] Get the meteorological element field at n time points within a time period And the meteorological information sequence at the corresponding time ;
[0008] Constructing a precipitation prediction model, the precipitation prediction model comprising a first branch and a second branch, the first branch predicting based on a meteorological element field to obtain a first predicted rainfall, and the second branch predicting based on a meteorological information sequence to obtain a second predicted rainfall;
[0009] A prediction result is obtained based on the first predicted rainfall amount and the second predicted rainfall amount.
[0010] Further, the first branch predicts based on the meteorological element field to obtain the first predicted rainfall, including:
[0011] The meteorological element field at the n time points Input the improved convolutional neural network respectively to extract the characteristics of the meteorological element field at n time points , the characteristics of the meteorological element field at the n time points After merging, the data are input into the RNN network for prediction to obtain the first predicted rainfall.
[0012] Furthermore, the improved convolutional neural network includes:
[0013] The meteorological element field at the i-th time point Input four sequentially connected convolution modules to obtain features Fc1, Fc2, Fc3, and Fc4, and perform a concatenation operation on the features Fc1, Fc2, Fc3, and Fc4 to obtain a feature Fy';
[0014] Input the feature Fy' into the sequentially connected convolution layer, BN layer, ReLu layer, maximum pooling layer, convolution layer, ReLu layer, maximum pooling layer, and SE attention module to obtain the feature Fb1;
[0015] Input the feature Fy' into the sequentially connected convolution layer, ReLu layer, average pooling layer, convolution layer, ReLu layer, average pooling layer, SE attention module to obtain the feature Fb2;
[0016] Concatenate the feature Fb1 and feature Fb2 to obtain the feature at the i-th time point .
[0017] Furthermore, the convolution module includes sequentially connected: a convolution layer, a BN layer, a maximum pooling layer, and a CBAM layer.
[0018] Furthermore, the second branch predicts the second predicted rainfall based on the meteorological information sequence, including:
[0019] The meteorological information sequence at the n time points Input the LSTM network for prediction and get the second predicted rainfall.
[0020] Furthermore, = ,in They represent temperature, air pressure, humidity, wind speed and wind direction respectively.
[0021] The present invention also provides an extended-range weather forecast system based on multi-model integration, comprising:
[0022] Information collection module: The information collection module is used to obtain the meteorological element field at n time points within a period of time. And the meteorological information sequence at the corresponding time ;
[0023] Prediction module: The prediction module is used to construct a precipitation prediction model, which includes a first branch and a second branch. The first branch predicts based on the meteorological element field to obtain a first predicted rainfall, and the second branch predicts based on the meteorological information sequence to obtain a second predicted rainfall.
[0024] Result acquisition module: The result acquisition module is used to obtain the prediction result based on the first predicted rainfall and the second predicted rainfall.
[0025] Further, the first branch predicts based on the meteorological element field to obtain the first predicted rainfall, including:
[0026] The meteorological element field at the n time points Input the improved convolutional neural network respectively to extract the characteristics of the meteorological element field at n time points , the characteristics of the meteorological element field at the n time points After merging, the data are input into the RNN network for prediction to obtain the first predicted rainfall.
[0027] Furthermore, the improved convolutional neural network includes:
[0028] The meteorological element field at the i-th time point Input four sequentially connected convolution modules to obtain features Fc1, Fc2, Fc3, and Fc4, and perform a concatenation operation on the features Fc1, Fc2, Fc3, and Fc4 to obtain a feature Fy';
[0029] Input the feature Fy' into the sequentially connected convolution layer, BN layer, ReLu layer, maximum pooling layer, convolution layer, ReLu layer, maximum pooling layer, and SE attention module to obtain the feature Fb1;
[0030] Input the feature Fy' into the sequentially connected convolution layer, ReLu layer, average pooling layer, convolution layer, ReLu layer, average pooling layer, SE attention module to obtain the feature Fb2;
[0031] Concatenate the feature Fb1 and feature Fb2 to obtain the feature at the i-th time point .
[0032] Furthermore, the convolution module includes sequentially connected: a convolution layer, a BN layer, a maximum pooling layer, and a CBAM layer.
[0033] Furthermore, the second branch predicts the second predicted rainfall based on the meteorological information sequence, including:
[0034] The meteorological information sequence at the n time points Input the LSTM network for prediction and get the second predicted rainfall.
[0035] Furthermore, = ,in They represent temperature, air pressure, humidity, wind speed and wind direction respectively.
[0036] Compared with the prior art, the advantages of the present invention are as follows:
[0037] The present invention uses two branches to predict the first predicted rainfall based on the meteorological element field and the second predicted rainfall based on the meteorological information sequence. It also uses an improved convolutional neural network model for feature extraction and performs extended-term weather forecasting based on LSTM and RNN models, thereby improving the accuracy of the extended-term weather forecast. BRIEF DESCRIPTION OF THE DRAWINGS
[0038] In order to more clearly illustrate the specific implementation of the present invention or the technical solution in the prior art, the drawings required for describing the specific implementation or the prior art will be briefly introduced below. Obviously, the drawings described below are only one implementation of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0039] Figure 1 This is a flowchart of the extended-range weather forecast method based on multi-model integration;
[0040] Figure 2 This is a schematic diagram of the precipitation prediction model;
[0041] Figure 3 Schematic diagram of the improved convolutional neural network;
[0042] Figure 4 Schematic diagram of the convolution module;
[0043] Figure 5 This is the structural diagram of the extended-range weather forecast system based on multi-model integration. DETAILED DESCRIPTION
[0044] In order to make the technical means, creative features, objectives and effects achieved by the present invention easy to understand, the technical scheme in the specific implementation mode of the present invention is clearly and completely described below to further illustrate the present invention. Obviously, the specific implementation mode described is only a part of the implementation mode of the present invention, rather than all styles.
[0045] The present invention is further described in detail below in conjunction with the accompanying drawings:
[0046] As attached Figure 1As shown, the present invention provides an extended-range weather forecast method based on multi-model integration, comprising:
[0047] Get the meteorological element field at n time points within a time period And the meteorological information sequence at the corresponding time ;
[0048] Construct a precipitation prediction model, as shown in the attached Figure 2 As shown, the precipitation prediction model includes a first branch and a second branch, the first branch predicts based on the meteorological element field to obtain a first predicted rainfall, and the second branch predicts based on the meteorological information sequence to obtain a second predicted rainfall;
[0049] A prediction result is obtained based on the first predicted rainfall amount and the second predicted rainfall amount.
[0050] Further, the first branch predicts based on the meteorological element field to obtain the first predicted rainfall, including:
[0051] The meteorological element field at the n time points Input the improved convolutional neural network respectively to extract the characteristics of the meteorological element field at n time points , the characteristics of the meteorological element field at the n time points After merging, the data are input into the RNN network for prediction to obtain the first predicted rainfall.
[0052] Further, as attached Figure 3 As shown, the improved convolutional neural network includes:
[0053] The meteorological element field at the i-th time point Input four sequentially connected convolution modules to obtain features Fc1, Fc2, Fc3, and Fc4, and perform a concatenation operation on the features Fc1, Fc2, Fc3, and Fc4 to obtain a feature Fy';
[0054] Input the feature Fy' into the sequentially connected convolution layer, BN layer, ReLu layer, maximum pooling layer, convolution layer, ReLu layer, maximum pooling layer, and SE attention module to obtain the feature Fb1;
[0055] Input the feature Fy' into the sequentially connected convolution layer, ReLu layer, average pooling layer, convolution layer, ReLu layer, average pooling layer, SE attention module to obtain the feature Fb2;
[0056] Concatenate the feature Fb1 and feature Fb2 to obtain the feature at the i-th time point .
[0057] Further, as attached Figure 4 As shown, the convolution module includes the following sequentially connected layers: convolution layer, BN layer, maximum pooling layer, and CBAM layer.
[0058] Furthermore, the second branch predicts the second predicted rainfall based on the meteorological information sequence, including:
[0059] The meteorological information sequence at the n time points Input the LSTM network for prediction and get the second predicted rainfall.
[0060] Furthermore, = ,in They represent temperature, air pressure, humidity, wind speed and wind direction respectively.
[0061] As attached Figure 5 As shown, the present invention also provides an extended-range weather forecast system based on multi-model integration, comprising:
[0062] Information collection module: The information collection module is used to obtain the meteorological element field at n time points within a period of time. And the meteorological information sequence at the corresponding time ;
[0063] Prediction module: The prediction module is used to construct a precipitation prediction model, which includes a first branch and a second branch. The first branch predicts based on the meteorological element field to obtain a first predicted rainfall, and the second branch predicts based on the meteorological information sequence to obtain a second predicted rainfall.
[0064] Result acquisition module: The result acquisition module is used to obtain the prediction result based on the first predicted rainfall and the second predicted rainfall.
[0065] Further, the first branch predicts based on the meteorological element field to obtain the first predicted rainfall, including:
[0066] The meteorological element field at the n time points Input the improved convolutional neural network respectively to extract the characteristics of the meteorological element field at n time points , the characteristics of the meteorological element field at the n time points After merging, the data are input into the RNN network for prediction to obtain the first predicted rainfall.
[0067] Furthermore, the improved convolutional neural network includes:
[0068] The meteorological element field at the i-th time point Input four sequentially connected convolution modules to obtain features Fc1, Fc2, Fc3, and Fc4, and perform a concatenation operation on the features Fc1, Fc2, Fc3, and Fc4 to obtain a feature Fy';
[0069] Input the feature Fy' into the sequentially connected convolution layer, BN layer, ReLu layer, maximum pooling layer, convolution layer, ReLu layer, maximum pooling layer, and SE attention module to obtain the feature Fb1;
[0070] Input the feature Fy' into the sequentially connected convolution layer, ReLu layer, average pooling layer, convolution layer, ReLu layer, average pooling layer, SE attention module to obtain the feature Fb2;
[0071] Concatenate the feature Fb1 and feature Fb2 to obtain the feature at the i-th time point .
[0072] Furthermore, the convolution module includes sequentially connected: a convolution layer, a BN layer, a maximum pooling layer, and a CBAM layer.
[0073] Furthermore, the second branch predicts the second predicted rainfall based on the meteorological information sequence, including:
[0074] The meteorological information sequence at the n time points Input the LSTM network for prediction and get the second predicted rainfall.
[0075] Furthermore, = ,in They represent temperature, air pressure, humidity, wind speed and wind direction respectively.
[0076] Compared with the prior art, the advantages of the present invention are as follows:
[0077] The present invention uses two branches to predict the first predicted rainfall based on the meteorological element field and the second predicted rainfall based on the meteorological information sequence. It also uses an improved convolutional neural network model for feature extraction and performs extended-term weather forecasting based on LSTM and RNN models, thereby improving the accuracy of the extended-term weather forecast.
[0078] The above describes the main technical features and basic principles of the present invention and the related advantages. It is obvious to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and the present invention can be implemented in other specific forms without departing from the concept or basic features of the present invention. Therefore, no matter from which point of view, the above specific embodiments should be regarded as exemplary and non-restrictive. The scope of the present invention is defined by the attached claims rather than the above description, and it is intended that all changes that fall within the meaning and scope of the equivalent elements of the claims are included in the present invention.
[0079] In addition, it should be understood that although the present specification is described according to various implementation methods, not every implementation method contains only one independent technical solution. This narrative method of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each implementation method may also be appropriately combined to form other implementation methods that can be understood by those skilled in the art.
Claims
1. An extended-range weather forecast method based on multi-model integration, characterized in that: The method comprises: Get the meteorological element field at n time points within a time period And the meteorological information sequence at the corresponding time ; Constructing a precipitation prediction model, the precipitation prediction model comprising a first branch and a second branch, the first branch predicting based on a meteorological element field to obtain a first predicted rainfall, and the second branch predicting based on a meteorological information sequence to obtain a second predicted rainfall; Obtaining a prediction result based on the first predicted rainfall and the second predicted rainfall; The first branch predicts the first predicted rainfall based on the meteorological element field, including: The meteorological element field at the n time points Input the improved convolutional neural network respectively to extract the characteristics of the meteorological element field at n time points , the characteristics of the meteorological element field at the n time points After merging, the data are input into the RNN network for prediction to obtain the first predicted rainfall; The improved convolutional neural network comprises: The meteorological element field at the i-th time point Input four sequentially connected convolution modules to obtain features Fc1, Fc2, Fc3, and Fc4, and perform a concatenation operation on the features Fc1, Fc2, Fc3, and Fc4 to obtain a feature Fy'; Input the feature Fy' into the sequentially connected convolution layer, BN layer, ReLu layer, maximum pooling layer, convolution layer, ReLu layer, maximum pooling layer, and SE attention module to obtain the feature Fb1; Input the feature Fy' into the sequentially connected convolution layer, ReLu layer, average pooling layer, convolution layer, ReLu layer, average pooling layer, SE attention module to obtain the feature Fb2; Concatenate the feature Fb1 and feature Fb2 to obtain the feature at the i-th time point ; The second branch predicts the second predicted rainfall based on the meteorological information sequence, including: The meteorological information sequence at the n time points Input the LSTM network for prediction and get the second predicted rainfall.
2. The extended-range weather forecast method based on multi-model integration according to claim 1, characterized in that: The convolution module includes sequentially connected: a convolution layer, a BN layer, a maximum pooling layer, and a CBAM layer.
3. The extended-range weather forecast method based on multi-model integration according to claim 1, characterized in that: = ,in They represent temperature, air pressure, humidity, wind speed and wind direction respectively.
4. An extended-range weather forecast system based on multi-model integration, characterized in that: The system comprises: Information collection module: The information collection module is used to obtain the meteorological element field at n time points within a period of time. And the meteorological information sequence at the corresponding time ; Prediction module: The prediction module is used to construct a precipitation prediction model, which includes a first branch and a second branch. The first branch predicts based on the meteorological element field to obtain a first predicted rainfall, and the second branch predicts based on the meteorological information sequence to obtain a second predicted rainfall. Result acquisition module: The result acquisition module is used to obtain a prediction result based on the first predicted rainfall and the second predicted rainfall; The first branch predicts the first predicted rainfall based on the meteorological element field, including: The meteorological element field at the n time points Input the improved convolutional neural network respectively to extract the characteristics of the meteorological element field at n time points , the characteristics of the meteorological element field at the n time points After merging, the data are input into the RNN network for prediction to obtain the first predicted rainfall; The improved convolutional neural network comprises: The meteorological element field at the i-th time point Input four sequentially connected convolution modules to obtain features Fc1, Fc2, Fc3, and Fc4, and perform a concatenation operation on the features Fc1, Fc2, Fc3, and Fc4 to obtain a feature Fy'; Input the feature Fy' into the sequentially connected convolution layer, BN layer, ReLu layer, maximum pooling layer, convolution layer, ReLu layer, maximum pooling layer, and SE attention module to obtain the feature Fb1; Input the feature Fy' into the sequentially connected convolution layer, ReLu layer, average pooling layer, convolution layer, ReLu layer, average pooling layer, SE attention module to obtain the feature Fb2; Concatenate the feature Fb1 and feature Fb2 to obtain the feature at the i-th time point ; The second branch predicts the second predicted rainfall based on the meteorological information sequence, including: The meteorological information sequence at the n time points Input the LSTM network for prediction and get the second predicted rainfall.
5. The extended-range weather forecast system based on multi-model integration according to claim 4, characterized in that: The convolution module includes sequentially connected: a convolution layer, a BN layer, a maximum pooling layer, and a CBAM layer.
6. The extended-range weather forecast system based on multi-model integration according to claim 4, characterized in that: = ,in They represent temperature, air pressure, humidity, wind speed and wind direction respectively.
