CN114139690A - Short-term rainfall prediction method and device - Google Patents
Short-term rainfall prediction method and device Download PDFInfo
- Publication number
- CN114139690A CN114139690A CN202111498466.5A CN202111498466A CN114139690A CN 114139690 A CN114139690 A CN 114139690A CN 202111498466 A CN202111498466 A CN 202111498466A CN 114139690 A CN114139690 A CN 114139690A
- Authority
- CN
- China
- Prior art keywords
- network
- precipitation
- rainfall
- convlstm
- radar
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 43
- 238000001556 precipitation Methods 0.000 claims abstract description 38
- 238000012549 training Methods 0.000 claims abstract description 16
- 230000007246 mechanism Effects 0.000 claims abstract description 12
- 238000007781 pre-processing Methods 0.000 claims abstract description 4
- 238000010586 diagram Methods 0.000 claims description 6
- 230000015654 memory Effects 0.000 claims description 6
- 238000002592 echocardiography Methods 0.000 claims description 5
- 238000004590 computer program Methods 0.000 claims description 4
- 239000003086 colorant Substances 0.000 claims description 3
- 230000007704 transition Effects 0.000 claims description 3
- 230000008569 process Effects 0.000 abstract description 5
- 230000008859 change Effects 0.000 abstract description 4
- 230000000694 effects Effects 0.000 abstract description 4
- 230000000007 visual effect Effects 0.000 abstract description 4
- 238000013528 artificial neural network Methods 0.000 description 6
- 230000004913 activation Effects 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 230000006872 improvement Effects 0.000 description 3
- 230000007787 long-term memory Effects 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 238000011161 development Methods 0.000 description 2
- 230000005284 excitation Effects 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 230000000306 recurrent effect Effects 0.000 description 2
- 230000004044 response Effects 0.000 description 2
- 230000006403 short-term memory Effects 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 230000004931 aggregating effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000013210 evaluation model Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000013277 forecasting method Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
- G01S13/88—Radar or analogous systems specially adapted for specific applications
- G01S13/95—Radar or analogous systems specially adapted for specific applications for meteorological use
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
- G01S7/41—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
- G01S7/411—Identification of targets based on measurements of radar reflectivity
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
- G01S7/41—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
- G01S7/417—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section involving the use of neural networks
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01W—METEOROLOGY
- G01W1/00—Meteorology
- G01W1/10—Devices for predicting weather conditions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Remote Sensing (AREA)
- Radar, Positioning & Navigation (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- General Health & Medical Sciences (AREA)
- Computer Networks & Wireless Communication (AREA)
- Health & Medical Sciences (AREA)
- Environmental & Geological Engineering (AREA)
- Environmental Sciences (AREA)
- Electromagnetism (AREA)
- Ecology (AREA)
- Biodiversity & Conservation Biology (AREA)
- Atmospheric Sciences (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention provides a short-term rainfall prediction method which is realized based on a ConvLSTM network and a SENet attention mechanism module, and comprises the steps of preprocessing meteorological satellite data; inverting precipitation with radar; constructing a ConvLSTM network fused with a SEnet attention module; leading the preprocessed radar echo image into a network for training to obtain a network training model; and generating a predicted precipitation image by using the trained model. The input data and the output data of the invention are radar echo images, the result expression is more visual, and the accuracy of the final result is higher. In addition, the invention fully utilizes the global characteristics, so that the learned global information is more reasonable. Moreover, the SEnet attention module is added, so that the effect of the network model is obviously improved. Finally, the invention does not change the size of the image in the training process, and perfectly keeps the detail and edge information of the image.
Description
Technical Field
The invention belongs to the technical field of image processing, and particularly relates to a short-critical precipitation prediction method based on ConvLSTM and SENEt attention mechanisms through radar echo images.
Background
The intelligent short-term rainfall forecast generally refers to the rainfall weather forecast within six hours, and the forecast precision can reach kilometer level and minute level. Compared with medium-long term and short term rainfall forecast, the method has higher requirements on the short-term forecast in terms of factors such as demand areas, forecast timeliness and the like. In recent years, the frequency and intensity of changes of marine climate and extreme weather events are increasing, and the influence of extreme weather on port operation is increasingly prominent. The short weather has great influence on the safety and normal operation of enterprises such as logistics and the like, and the extreme weather poses great threat to the property and even the life of people. The weather forecast is obtained in time, and the correct decision can be made by relevant organizations of governments and industries. At present, an end-to-end scheme can be provided by utilizing a deep learning method aiming at short-term rainfall forecast so as to solve the nonlinear complex problem. The method extracts the characteristics of a lower layer through a multi-layer network structure and nonlinear transformation to form an abstract high-layer representation so as to find the probability distribution characteristics of data and further predict the future precipitation condition. Because the artificial neural network can model a nonlinear system, the use of the deep neural network for short-term rainfall prediction has great potential, and the research of the subject has practical significance.
The precipitation prediction problem can be realized by predicting radar echo images after a certain moment through a series of radar echo images before the moment, wherein in a series of inputs before the moment, a certain relation exists between a previous input and a next input, and the prediction results are influenced to different degrees. The output of each time of the feedforward neural network only depends on the current input, and mutual influence of the inputs at different moments is not considered, so the feedforward neural network is not suitable for processing the space-time sequence problem, and therefore, the RNN (recurrent neural network) is provided, which is a network specially processing time sequence data, but is difficult to process overlong sequences, namely, only short-term memory and no long-term memory can be realized. To solve this problem, LSTM, a long-term memory network, a variant of RNN, has come into play that combines short-term memory and long-term memory through sophisticated gating. ConvLSTM is an optimization of LSTM, and by adding a convolution structure in the conversion from input to state and state to state, the space-time correlation can be better captured.
Disclosure of Invention
The invention aims to provide a short-term rainfall forecasting method based on ConvLSTM and SENEt attention mechanisms, the result expression is more visual by adopting the method, the accuracy of the final result is higher, and the global characteristics are fully utilized, so that the learned global information is more reasonable; meanwhile, the SENet attention module is added, so that the effect of the network model is remarkably improved under the condition of adding a small number of parameters. Finally, the invention does not change the size of the image in the training process, and perfectly keeps the detail and edge information of the image.
In order to achieve the above purpose, the invention provides a short-term rainfall prediction method, which is realized based on a ConvLSTM network and a SEnet attention mechanism module, and comprises the following steps:
step 2, radar inversion precipitation, namely calculating the distribution of the rain intensity and the accumulated rainfall of the corresponding area through radar echoes;
step 3, constructing a ConvLSTM network fused with the SEnet attention module for subsequent model training;
step 4, importing the preprocessed radar echo image into a network for training to obtain a network training model;
and 5, generating a predicted precipitation image by using the trained model.
The invention further improves the method in that the method also comprises a step 6 of evaluating the method by using the commonly used precipitation forecast indexes.
The invention has the further improvement that the data of the satellite is the data of the Fengyun No. 4 satellite, and the radar data is the radar echo diagram of the southeast region of China.
A further development of the invention consists in that in step 2 the so-called radar inversion of the precipitation into a Quantitative Precipitation Estimate (QPE), which isCalculating the distribution of rainfall intensity and the accumulated rainfall of the corresponding area through radar echoes; rainfall information is obtained according to an empirical formula Z-R relation existing between radar echo and rainfall, wherein the Z-R relation reflects the correlation between the radar echo and rainfall intensity, namely a radar reflection factor Z (unit: mm)6/m3) And rainfall intensity R (in mm/h) satisfy the following formula:
Z=aRb
the method comprises the following steps that a and b are empirical constants, a is approximately equal to 200, b is 1.5-2, and the a and b are determined according to various factors such as different times, different places, precipitation types and properties and the like; the different rainfall intensities are represented by different colors on the radar echo diagram, wherein the darker color represents the greater rainfall intensity, and the smaller or no rainfall is in turn represented.
A further development of the invention is that in step 3 the precipitation forecasts can be described as a spatio-temporal sequence prediction, where the input and prediction targets are spatio-temporal sequences; by expanding the fully-connected LSTM, a ConvLSTM network with convolution structures in both input-to-state and state-to-state transitions is provided, and the ConvLSTM network can capture spatio-temporal correlations; ConvLSTM has the following formula:
wherein, represents the convolution operation,representing a hadamard operation (corresponding multiplication).
A further refinement of the invention is that the sentet attention module block can be inserted after the non-linear activation function behind each convolutional layer in the ConvLSTM to be conveniently integrated into the ConvLSTM network.
A further improvement of the invention is to convert the predicted and true values into 0/1 matrices using a precipitation threshold of 0.5mm/h (indicating whether it is raining) and calculate hits (successful prediction), misses (not predicted) and false predictions, using the following as indicators for the performance of the evaluation model:
in order to achieve the object of the invention, the invention further provides an apparatus for implementing the aforementioned method for predicting short-term precipitation, the apparatus comprising at least one computing device including a memory, a processor and a computer program stored on the memory and executable on the processor.
The invention has the following beneficial effects: the method is based on ConvLSTM and SENET attention mechanisms, input data and output data of the method are radar echo images, and result expression is more visual. Meanwhile, a module capable of reasonably distributing weight is added behind the activation function of each convolution layer of ConvLSTM, so that the accuracy of the final result is higher. In addition, compared with the traditional LSTM network, the network provided by the invention makes full use of global characteristics, and suitable weight is memorized for a long time, so that the learned global information is more reasonable. Furthermore, the invention obviously improves the effect of the network model by adding the SEnet attention module under the condition of adding a small number of parameters. Finally, the invention does not change the size of the image in the training process, and perfectly keeps the detail and edge information of the image.
Drawings
FIG. 1 is a block flow diagram of the method of the present invention.
FIG. 2 is a diagram of the ConvLSTM unit structure in the present invention.
FIG. 3 shows the internal structure of ConvLSTM in the present invention.
Fig. 4 shows the coding network and prediction network of ConvLSTM in the present invention.
Fig. 5 is a basic structure of the SENet network in the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be described in detail with reference to the accompanying drawings and specific embodiments.
It should be emphasized that in describing the present invention, various formulas and constraints are identified with consistent labels, but the use of different labels to identify the same formula and/or constraint is not precluded and is provided for the purpose of more clearly illustrating the features of the present invention.
The invention uses a Pythroch library in a Python environment on a GeForce GTX 1080Ti processor. For network optimization, the present invention uses Adam optimizer and nonlinear activation function ReLU in experiments. During network training, the number of histories is set to 100, and the learning rate is set to 0.001. To achieve accurate convergence, the step size is set to 8.
As shown in FIG. 1, the method for predicting the short-critical precipitation based on the ConvLSTM network and the SENet attention mechanism module mainly comprises the following steps:
step 2, radar inversion precipitation, namely calculating the distribution of the rain intensity and the accumulated rainfall of the corresponding area through radar echoes;
step 3, constructing a ConvLSTM network fused with the SEnet attention module for subsequent model training;
step 4, importing the preprocessed radar echo image into a network for training to obtain a network training model;
step 5, generating a prediction precipitation image by using the trained model;
and 6, evaluating the method by using a common rainfall forecast index.
In particular, the principle and formula of the method used in step 2 are illustrated below.
The radar inversion precipitation is that the distribution of rain intensity and the accumulated rainfall in a corresponding area are actually calculated through radar echoes, and the radar inversion precipitation can be called quantitative precipitation estimation in meteorology, and is called QPE for short. And acquiring precipitation information according to an empirical formula Z-R relation existing between the radar echo and precipitation. The Z-R relationship reflects the correlation between radar echo and rainfall intensity, i.e. the radar reflection factor Z (unit: mm)6/m3) And rainfall intensity R (in mm/h) satisfy the following formula:
Z=aRb。
the method is characterized in that a and b are empirical constants, a is approximately equal to 200, b is 1.5-2, and the method is determined according to various factors such as different times, different places, precipitation types and properties and the like. The different rainfall intensities are represented by different colors on the radar echo diagram, wherein the darker color represents the greater rainfall intensity, and the smaller or no rainfall is in turn represented.
In particular, in step 3, the network structure and principle used thereof and the method of constructing the network are explained as follows. The precipitation forecasts problem can be described as a spatio-temporal sequence prediction problem, where the input and prediction targets are spatio-temporal sequences. By extending the fully-connected LSTM, it is proposed that a ConvLSTM network can better capture spatio-temporal correlations with ConvLSTM having convolution structures in both input-to-state and state-to-state transitions.
The structure of the ConvLSTM unit is shown in FIG. 1. ConvLSTM has the following formula:
wherein, represents a convolution operation;representing a hadamard operation (corresponding multiplication).
The ConvLSTM internal structure is shown in FIG. 2.
The coding network and the prediction network as shown in fig. 3 are formed by superimposing a plurality of ConvLSTM layers. The initial state and output of the prediction network are replicated from the final state of the coding network. Because the prediction target has the same dimensions as the input, connecting all states in the prediction network and inputting into a 1 x 1 convolutional layer can generate the final prediction result.
The attention mechanism is simple to say that attention is focused on important factors, and unimportant factors are omitted. The SENET network used by the invention is an embodiment of an attention mechanism, allows the network to perform feature recalibration, learns and uses global information, selectively emphasizes information features and suppresses less useful features, and the central idea is to predict a constant weight for each output channel, perform weighting operation on each channel, enhance effective information and suppress ineffective information. The basic structure is shown in fig. 4.
Ftr denotes a convolutional layer, implementing feature mapping, the features U are passed through a squeeze operation that generates channel descriptors by aggregating feature mappings across spatial dimensions (H x W), thus producing a globally distributed embedded channel feature response that allows information from the global acceptance domain of the network to be used by all of its layers. This is followed by an actuation operation in the form of a simple self-gating mechanism, taking the embedding as input, and generating a set of per-channel modulation weights. These weights are applied to the feature map U, generating the output of the SE block, which can be directly input into subsequent layers of the network.
Squeeze (fsq) represents feature compression by spatial dimension, with each two-dimensional feature channel one-dimensional vector having somewhat a global receptive field, and the output dimension matching the input number of feature channels. It characterizes the global distribution of responses over the feature channels and allows layers close to the input to get a global perception.
The excitation (fex) represents a mechanism similar to the gates in the recurrent neural network. Weights are produced for each eigen-channel by a parameter w that is learned to model the correlation between eigen-channels explicitly.
Scale (Fscale) represents that the weight of the output of the Excitation is regarded as the importance of each feature channel after feature selection, and then the original feature map is re-calibrated in the channel dimension by multiplying the weight to the previous feature channel by channel.
As shown in fig. 5, the SENet can be conveniently integrated into the ConvLSTM network by inserting the SENet module behind each convolutional layer in the ConvLSTM, which can trade off the performance improvement of the network by sacrificing very little computation cost.
In the short-term rainfall prediction method of the invention, a rainfall threshold (indicating whether it is raining) of 0.5mm/h is used to convert the predicted and true values into 0/1 matrix, and hits (successful prediction), misses (not predicted) and false predictions (wrong prediction) are calculated, using the following indexes for evaluating the performance of the model:
the apparatus for carrying out the method of the present invention comprises at least one computing device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor. The computer program, when loaded into the processor, implements the ConvLSTM network and SEnet attention module based precipitation prediction method of the present invention.
The method is based on ConvLSTM and SENET attention mechanisms, input data and output data of the method are radar echo images, and result expression is more visual. Meanwhile, a module capable of reasonably distributing weight is added behind the activation function of each convolution layer of ConvLSTM, so that the accuracy of the final result is higher. In addition, compared with the traditional LSTM network, the network provided by the invention makes full use of global characteristics, and suitable weight is memorized for a long time, so that the learned global information is more reasonable. Furthermore, the invention obviously improves the effect of the network model by adding the SEnet attention module under the condition of adding a small number of parameters. Finally, the invention does not change the size of the image in the training process, and perfectly keeps the detail and edge information of the image.
Although the present invention has been described in detail with reference to the preferred embodiments, it will be understood by those skilled in the art that various changes may be made and equivalents may be substituted for elements thereof without departing from the spirit and scope of the present invention.
Claims (8)
1. A short-rainfall prediction method is realized based on a ConvLSTM network and a SEnet attention mechanism module, and is characterized by comprising the following steps of:
step 1, preprocessing meteorological satellite data;
step 2, radar inversion precipitation, namely calculating the distribution of the rain intensity and the accumulated rainfall of the corresponding area through radar echoes;
step 3, constructing a ConvLSTM network fused with the SEnet attention module for subsequent model training;
step 4, importing the preprocessed radar echo image into a network for training to obtain a network training model;
and 5, generating a predicted precipitation image by using the trained model.
2. The method of short-term precipitation prediction according to claim 1, characterized in that: and 6, evaluating the method by using a commonly used rainfall forecast index.
3. The method of short-term precipitation prediction according to claim 2, characterized in that: the data of the satellite is selected from data of a wind cloud No. 4 satellite, and the radar data is selected from a radar echo map of the southeast region of China.
4. The method of short-term precipitation prediction according to claim 2, characterized in that: in step 2, the radar inverses precipitation into Quantitative Precipitation Estimation (QPE), which calculates the distribution of rain intensity and accumulated rainfall in the corresponding area through radar echo; rainfall information is obtained according to an empirical formula Z-R relation existing between radar echo and rainfall, wherein the Z-R relation reflects the correlation between the radar echo and rainfall intensity, namely a radar reflection factor Z (unit: mm)6/m3) And descendThe rain intensity R (unit mm/h) satisfies the following formula:
Z=aRb
the method comprises the following steps that a and b are empirical constants, a is approximately equal to 200, b is 1.5-2, and the a and b are determined according to various factors such as different times, different places, precipitation types and properties and the like; the different rainfall intensities are represented by different colors on the radar echo diagram, wherein the darker color represents the greater rainfall intensity, and the smaller or no rainfall is in turn represented.
5. The method of short-term precipitation prediction according to claim 3, characterized in that: in step 3, the precipitation forecasts can be described as a spatio-temporal sequence prediction, wherein the input and prediction targets are spatio-temporal sequences; by expanding the fully-connected LSTM, a ConvLSTM network with convolution structures in both input-to-state and state-to-state transitions is provided, and the ConvLSTM network can capture spatio-temporal correlations; ConvLSTM has the following formula:
6. The method of short-term precipitation prediction according to claim 4, characterized in that: the SENET attention module block can be inserted behind each convolutional layer in the ConvLSTM to be conveniently integrated into the ConvLSTM network.
7. The method of short-term precipitation prediction according to claim 5, characterized in that: the predicted and true values were converted to 0/1 matrices using a precipitation threshold of 0.5mm/h (indicating whether it was raining) and hits (successful prediction), misses (not predicted) and false predictions (wrong prediction) were calculated using the following indices to evaluate the performance of the model:
8. an apparatus, characterized by: the apparatus for implementing the method of any one of claims 1-7, the apparatus comprising at least one computing device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor.
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111498466.5A CN114139690B (en) | 2021-12-09 | Short-term precipitation prediction method and device | |
PCT/CN2022/125568 WO2023103587A1 (en) | 2021-12-09 | 2022-10-17 | Imminent precipitation forecast method and apparatus |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111498466.5A CN114139690B (en) | 2021-12-09 | Short-term precipitation prediction method and device |
Publications (2)
Publication Number | Publication Date |
---|---|
CN114139690A true CN114139690A (en) | 2022-03-04 |
CN114139690B CN114139690B (en) | 2024-07-26 |
Family
ID=
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114740549A (en) * | 2022-04-12 | 2022-07-12 | 北京墨迹风云科技股份有限公司 | Method and device for predicting short-term rainfall weather, electronic equipment and readable medium |
CN115220133A (en) * | 2022-07-19 | 2022-10-21 | 海南大学 | Multi-meteorological-element rainfall prediction method, device, equipment and storage medium |
WO2023103587A1 (en) * | 2021-12-09 | 2023-06-15 | 南京邮电大学 | Imminent precipitation forecast method and apparatus |
CN116451881B (en) * | 2023-06-16 | 2023-08-22 | 南京信息工程大学 | Short-time precipitation prediction method based on MSF-Net network model |
CN117313823A (en) * | 2023-11-28 | 2023-12-29 | 哈尔滨工业大学(深圳)(哈尔滨工业大学深圳科技创新研究院) | Mixed distributed parallel training method and system for convolutional neural network |
CN117368881A (en) * | 2023-12-08 | 2024-01-09 | 哈尔滨工业大学(深圳)(哈尔滨工业大学深圳科技创新研究院) | Multi-source data fusion long-sequence radar image prediction method and system |
CN118070846A (en) * | 2024-04-18 | 2024-05-24 | 南京信息工程大学 | Short-term precipitation prediction method, device and storage medium |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110363327A (en) * | 2019-06-04 | 2019-10-22 | 东南大学 | Short based on ConvLSTM and 3D-CNN faces Prediction of Precipitation method |
WO2020037965A1 (en) * | 2018-08-21 | 2020-02-27 | 北京大学深圳研究生院 | Method for multi-motion flow deep convolutional network model for video prediction |
CN111815037A (en) * | 2020-06-24 | 2020-10-23 | 河海大学 | Interpretable short-critical extreme rainfall prediction method based on attention mechanism |
CN112446419A (en) * | 2020-10-29 | 2021-03-05 | 中山大学 | Time-space neural network radar echo extrapolation forecasting method based on attention mechanism |
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2020037965A1 (en) * | 2018-08-21 | 2020-02-27 | 北京大学深圳研究生院 | Method for multi-motion flow deep convolutional network model for video prediction |
CN110363327A (en) * | 2019-06-04 | 2019-10-22 | 东南大学 | Short based on ConvLSTM and 3D-CNN faces Prediction of Precipitation method |
CN111815037A (en) * | 2020-06-24 | 2020-10-23 | 河海大学 | Interpretable short-critical extreme rainfall prediction method based on attention mechanism |
CN112446419A (en) * | 2020-10-29 | 2021-03-05 | 中山大学 | Time-space neural network radar echo extrapolation forecasting method based on attention mechanism |
Non-Patent Citations (1)
Title |
---|
杨仲江;马俊彦;王昊;: "序列结构的RNN模型在闪电预警中的应用", 灾害学, no. 02, 20 April 2020 (2020-04-20), pages 92 - 98 * |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2023103587A1 (en) * | 2021-12-09 | 2023-06-15 | 南京邮电大学 | Imminent precipitation forecast method and apparatus |
CN114740549A (en) * | 2022-04-12 | 2022-07-12 | 北京墨迹风云科技股份有限公司 | Method and device for predicting short-term rainfall weather, electronic equipment and readable medium |
CN114740549B (en) * | 2022-04-12 | 2024-02-06 | 北京墨迹风云科技股份有限公司 | Short-term precipitation weather prediction method and device, electronic equipment and readable medium |
CN115220133A (en) * | 2022-07-19 | 2022-10-21 | 海南大学 | Multi-meteorological-element rainfall prediction method, device, equipment and storage medium |
CN115220133B (en) * | 2022-07-19 | 2023-05-26 | 海南大学 | Rainfall prediction method, device and equipment for multiple meteorological elements and storage medium |
CN116451881B (en) * | 2023-06-16 | 2023-08-22 | 南京信息工程大学 | Short-time precipitation prediction method based on MSF-Net network model |
CN117313823A (en) * | 2023-11-28 | 2023-12-29 | 哈尔滨工业大学(深圳)(哈尔滨工业大学深圳科技创新研究院) | Mixed distributed parallel training method and system for convolutional neural network |
CN117313823B (en) * | 2023-11-28 | 2024-04-12 | 哈尔滨工业大学(深圳)(哈尔滨工业大学深圳科技创新研究院) | Mixed distributed parallel training method and system for convolutional neural network |
CN117368881A (en) * | 2023-12-08 | 2024-01-09 | 哈尔滨工业大学(深圳)(哈尔滨工业大学深圳科技创新研究院) | Multi-source data fusion long-sequence radar image prediction method and system |
CN117368881B (en) * | 2023-12-08 | 2024-03-26 | 哈尔滨工业大学(深圳)(哈尔滨工业大学深圳科技创新研究院) | Multi-source data fusion long-sequence radar image prediction method and system |
CN118070846A (en) * | 2024-04-18 | 2024-05-24 | 南京信息工程大学 | Short-term precipitation prediction method, device and storage medium |
Also Published As
Publication number | Publication date |
---|---|
WO2023103587A1 (en) | 2023-06-15 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2023103587A1 (en) | Imminent precipitation forecast method and apparatus | |
CN111639748B (en) | Watershed pollutant flux prediction method based on LSTM-BP space-time combination model | |
CN110363327B (en) | ConvLSTM and 3D-CNN-based short rainfall prediction method | |
Ma et al. | Transfer learning for long-interval consecutive missing values imputation without external features in air pollution time series | |
Todini | A model conditional processor to assess predictive uncertainty in flood forecasting | |
CN113496104B (en) | Precipitation prediction correction method and system based on deep learning | |
Burian et al. | Training artificial neural networks to perform rainfall disaggregation | |
CN113255972B (en) | Short-term rainfall prediction method based on Attention mechanism | |
CN111462324B (en) | Online spatiotemporal semantic fusion method and system | |
CN110648014A (en) | Regional wind power prediction method and system based on space-time quantile regression | |
CN112415521A (en) | CGRU (China-swarm optimization and RU-based radar echo nowcasting) method with strong space-time characteristics | |
CN115220133B (en) | Rainfall prediction method, device and equipment for multiple meteorological elements and storage medium | |
CN114943365A (en) | Rainfall estimation model establishing method fusing multi-source data and rainfall estimation method | |
CN114742206B (en) | Rainfall intensity estimation method for comprehensive multi-time space-scale Doppler radar data | |
CN112418149A (en) | Abnormal behavior detection method based on deep convolutional neural network | |
CN113610286A (en) | PM accounting for spatio-temporal correlations and meteorological factors2.5Concentration prediction method and device | |
CN115792853A (en) | Radar echo extrapolation method based on dynamic weight loss | |
CN115859620A (en) | Runoff reconstruction method based on multi-head attention mechanism and graph neural network | |
CN117494034A (en) | Air quality prediction method based on traffic congestion index and multi-source data fusion | |
CN116861347A (en) | Magnetic force abnormal data calculation method based on deep learning model | |
CN117131991A (en) | Urban rainfall prediction method and platform based on hybrid neural network | |
Xu et al. | Federated learning for interpretable short-term residential load forecasting in edge computing network | |
CN117436653A (en) | Prediction model construction method and prediction method for travel demands of network about vehicles | |
CN117195976A (en) | Traffic flow prediction method and system based on layered attention | |
CN114139690B (en) | Short-term precipitation prediction method and device |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant |