CN114818860A - Typhoon track prediction method based on multivariate features - Google Patents
Typhoon track prediction method based on multivariate features Download PDFInfo
- Publication number
- CN114818860A CN114818860A CN202210284772.7A CN202210284772A CN114818860A CN 114818860 A CN114818860 A CN 114818860A CN 202210284772 A CN202210284772 A CN 202210284772A CN 114818860 A CN114818860 A CN 114818860A
- Authority
- CN
- China
- Prior art keywords
- typhoon
- dimensional
- seawater
- network
- input
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 26
- 239000013598 vector Substances 0.000 claims abstract description 26
- 238000013528 artificial neural network Methods 0.000 claims abstract description 10
- 238000012549 training Methods 0.000 claims abstract description 8
- 239000013535 sea water Substances 0.000 claims description 34
- 230000004927 fusion Effects 0.000 claims description 7
- 238000012545 processing Methods 0.000 claims description 6
- 230000007774 longterm Effects 0.000 claims description 5
- 230000000306 recurrent effect Effects 0.000 claims description 3
- 238000013135 deep learning Methods 0.000 description 3
- 238000013277 forecasting method Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01W—METEOROLOGY
- G01W1/00—Meteorology
- G01W1/10—Devices for predicting weather conditions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Environmental & Geological Engineering (AREA)
- Atmospheric Sciences (AREA)
- Biodiversity & Conservation Biology (AREA)
- Ecology (AREA)
- Environmental Sciences (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
A typhoon track prediction method based on multivariate features belongs to the technical field of typhoon prediction. The method comprises the steps of inputting various characteristics influencing typhoon tracks, wherein the characteristics comprise meteorological characteristics, ocean characteristics, geographic characteristics and physical characteristics, extracting the characteristics of different characteristics by using different networks based on the input data types, fusing the characteristics by adopting a characteristic weighting mode after extracting the different characteristics to obtain weighted multidimensional vectors, then using the weighted multidimensional vectors as the input of an LSTM neural network, enabling the output result to be the position of the typhoon at the next time point, and training a model by using a distributed training framework DistributedDataParallel of PyTorch. The method predicts the typhoon track based on the multivariate characteristics, can improve the accuracy of typhoon track prediction to a great extent, and reduces the harm of typhoon to the normal life of people.
Description
Technical Field
The invention relates to a typhoon track prediction method, in particular to a typhoon track prediction method based on multivariate characteristics, and belongs to the technical field of typhoon prediction.
Background
The traditional typhoon path forecasting method comprises numerical forecasting, statistical forecasting, regression forecasting and comprehensive forecasting. The traditional typhoon track prediction method has the following defects that 1. long-term space-time dependence in historical data is difficult to analyze, and long-term prediction performance is reduced. 2. The modeling method is complex and requires a large amount of computational resources and time cost, and the modeling method often needs to be updated. In recent years, deep learning and parameter optimization are rapidly developed, scientists propose a typhoon path forecasting method based on RNN and LSTM, and research shows that the typhoon path forecasting model can forecast typhoon paths 6 ー 24 hours in advance and is high in forecasting accuracy. Although the existing research applies a deep learning method and combines the characteristics of potential height, atmospheric pressure, wind field and the like to predict the typhoon track, the factors which influence the typhoon path and are considered by the methods are too simple and incomplete, and the accuracy of predicting the typhoon track is not high.
Therefore, a typhoon track prediction method based on multiple features is designed, the meteorological features including the potential height, the relative humidity, the east-west wind speed, the north-south wind speed and the air temperature, the physical quantities including the sea surface temperature and the seawater isothermal line of 26 ℃, the physical quantities related to the ocean mixing layer, the physical quantities related to the thermocline, the three-dimensional temperature of the seawater, the three-dimensional salinity of the seawater and the three-dimensional density of the seawater, the geographic features including the terrain and the ground transfer force, and the physical features including the position, the direction, the speed and the strength of the typhoon are comprehensively considered, the multiple features are used as the input of an LSTM network, the accuracy of typhoon track prediction is improved, and the method has practical significance and good application background.
Disclosure of Invention
Aiming at the problem that the factors influencing the typhoon track are simply considered in the prior art, the invention provides a typhoon track prediction method based on multivariate features.
The invention is realized by the following technical scheme:
s1, inputting various characteristics influencing a typhoon track, wherein the characteristics comprise meteorological characteristics, ocean characteristics, geographic characteristics and physical characteristics, and extracting the characteristics of different characteristics by using different networks based on the input data types of the characteristics, wherein the data is mainly collected from an area between 0-60-degree N latitude and 100-degree E-180-degree E longitude;
s2, after extracting different features, performing feature fusion in a feature weighting mode, fusing the feature fusion into 70-dimensional feature vectors with weights, and then taking the 70-dimensional feature vectors as input of a predictor;
s3, because the recurrent neural network RNN has great advantages in processing time sequence data and the neural network LSTM improves the long-term dependence problem in the RNN, the LSTM neural network is selected as a predictor, and a multi-dimensional vector of a specified time sequence is input into the LSTM, so that the output result is the position of typhoon at the next time point;
and S4, reducing the time required by model convergence by using a distributed training frame DistributedDataParalll of PyTorch, accelerating the training speed and better processing a large amount of data.
Wherein, the image characteristics in the step S1 are as follows: the potential height, the relative humidity, the east-west wind speed, the north-south wind speed and the air temperature are respectively acquired at 200hPa, 500hPa, 700hPa, 850hPa and 1000hPa for each characteristic, data are acquired at intervals and are made into two-dimensional three-channel pictures with the format of 3 x 25 x 33, the images are processed by a latest RegNet network, filling is set to be 7 x 7 according to the input size, the 200hPa, 500hPa, 700hPa, 850hPa and 1000hPa are taken as five factors influencing the typhoon track, so the output dimension of the last full-connection layer of the network is set to be five, namely the output result is a five-dimensional vector.
Wherein, the ocean characteristics in step S1 are as follows: sea surface temperature, physical quantities related to seawater 26 ℃ isotherms, physical quantities related to an ocean mixing layer, physical quantities related to a thermocline, three-dimensional temperature of seawater, three-dimensional salinity of seawater, and three-dimensional density of seawater, wherein the input of the sea surface temperature is a two-dimensional three-channel picture in a format of 3 x 242 x 326, which is processed with a latest RegNet network and is filled to 7 x 7, wherein the physical quantities related to the seawater 26 ℃ isotherms comprise the depth of the seawater 26 ℃ isotherm and the heat content above the seawater 26 ℃ isotherm, the physical quantities related to the ocean mixing layer comprise the thickness of the mixing layer and the heat content of the mixing layer, the physical quantities related to the thermocline comprise the depth of the thermocline, the heat content above the thermocline, and the intensity of the thermocline, the inputs of which are two-dimensional three-channel pictures, and are in a format of 3 x 63 x 84, which are all processed with the latest RegNet network and are filled to 3 x 4, and the output dimension of the last full connection layer of the network is set to be five, namely the output result is a five-dimensional vector, so that each feature can output one five-dimensional vector, the input dimensions of the three features of the three-dimensional temperature of the seawater, the three-dimensional salinity of the seawater and the three-dimensional density of the seawater are three-dimensional, ResNet 3D is adopted to process the features, the input formats of the three features are all 3 multiplied by 17 multiplied by 63 multiplied by 84, so that filling is set to be 1 multiplied by 3 multiplied by 4, and the output dimension of the last full connection layer of the network is set to be five, so that each feature can correspond to one five-dimensional vector.
Wherein, the geographic features in step S1 are as follows: terrain, ground rotation force, physical characteristics such as: the position and direction, speed, intensity of the typhoon are generally considered to be unchanged, i.e. the terrain is the same at all times, and since there is almost no terrain influence on the sea, it is set to 0, the input of terrain data is a two-dimensional tristimulus channel of 3 × 121 × 161, processed with the latest RegNet network and filled to 7 × 7, which is estimated by the formula of the ground rotation force F2 mv ω sin θ, where m is the mass of the object, v is the speed of the object, ω is the angular velocity of the earth's rotation, θ is the latitude before the object starts to move, considering that the mass of the typhoon is difficult to estimate, we use the gradient of the ground rotation force to represent the influence of the ground rotation force on the typhoon, whose magnitude is equal to 2v ω sin θ, the position and direction of the typhoon are represented by latitude and longitude, the speed of the typhoon is represented by the average speed of two consecutive moments, the intensity of the typhoon depends on its maximum wind speed, therefore, the strength of the typhoon is represented by the maximum wind speed of the typhoon, and the ground rotation force, the position, the direction, the speed and the strength of the typhoon are processed by a fully-connected network, so that the wind power is output as a five-dimensional vector.
The invention has the following beneficial effects:
compared with a common method for predicting the typhoon track by utilizing deep learning, the method disclosed by the invention predicts the typhoon track by utilizing different network fusion according to different input data types, considers various factors influencing the typhoon track, synthesizes multivariate characteristics to predict the typhoon track, and can more accurately predict the typhoon track.
Drawings
FIG. 1 is a flow chart of the method of the present invention
FIG. 2 is a network architecture diagram of the present invention
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described below by way of figures and examples.
Example 1: referring to fig. 1, a typhoon trajectory prediction method based on multivariate features includes the following steps:
s1, inputting various characteristics influencing a typhoon track, wherein the characteristics comprise meteorological characteristics, ocean characteristics, geographic characteristics and physical characteristics, and extracting the characteristics of different characteristics by using different networks based on the input data types of the characteristics, wherein the data is mainly collected from an area between 0-60-degree N latitude and 100-degree E-180-degree E longitude;
s2, after extracting different features, performing feature fusion in a feature weighting mode, fusing the feature fusion into 70-dimensional feature vectors with weights, and then taking the 70-dimensional feature vectors as input of a predictor;
s3, because the recurrent neural network RNN has great advantages in processing time sequence data and the neural network LSTM improves the long-term dependence problem in the RNN, the LSTM neural network is selected as a predictor, and a multi-dimensional vector of a specified time sequence is input into the LSTM, so that the output result is the position of typhoon at the next time point;
and S4, reducing the time required by model convergence by using a distributed training frame DistributedDataParalll of PyTorch, accelerating the training speed and better processing a large amount of data.
Wherein, the image characteristics in the step S1 are as follows: the potential height, the relative humidity, the east-west wind speed, the north-south wind speed and the air temperature are respectively acquired at 200hPa, 500hPa, 700hPa, 850hPa and 1000hPa for each characteristic, data are acquired at intervals and are made into two-dimensional three-channel pictures with the format of 3 x 25 x 33, the images are processed by a latest RegNet network, filling is set to be 7 x 7 according to the input size, the 200hPa, 500hPa, 700hPa, 850hPa and 1000hPa are taken as five factors influencing the typhoon track, so the output dimension of the last full-connection layer of the network is set to be five, namely the output result is a five-dimensional vector.
Wherein, the ocean characteristics in step S1 are as follows: sea surface temperature, physical quantities related to seawater 26 ℃ isotherms, physical quantities related to an ocean mixing layer, physical quantities related to a thermocline, three-dimensional temperature of seawater, three-dimensional salinity of seawater, and three-dimensional density of seawater, wherein the input of the sea surface temperature is a two-dimensional three-channel picture in a format of 3 x 242 x 326, which is processed with a latest RegNet network and is filled to 7 x 7, wherein the physical quantities related to the seawater 26 ℃ isotherms comprise the depth of the seawater 26 ℃ isotherm and the heat content above the seawater 26 ℃ isotherm, the physical quantities related to the ocean mixing layer comprise the thickness of the mixing layer and the heat content of the mixing layer, the physical quantities related to the thermocline comprise the depth of the thermocline, the heat content above the thermocline, and the intensity of the thermocline, the inputs of which are two-dimensional three-channel pictures, and are in a format of 3 x 63 x 84, which are all processed with the latest RegNet network and are filled to 3 x 4, and the output dimension of the last full connection layer of the network is set to be five, namely the output result is a five-dimensional vector, so that each feature can output one five-dimensional vector, the input dimensions of the three features of the three-dimensional temperature of the seawater, the three-dimensional salinity of the seawater and the three-dimensional density of the seawater are three-dimensional, ResNet 3D is adopted to process the features, the input formats of the three features are all 3 multiplied by 17 multiplied by 63 multiplied by 84, so that filling is set to be 1 multiplied by 3 multiplied by 4, and the output dimension of the last full connection layer of the network is set to be five, so that each feature can correspond to one five-dimensional vector.
The geographic features in step S1 are as follows: terrain, ground rotation force, physical characteristics such as: the position and direction, speed, intensity of the typhoon are generally considered to be unchanged, i.e. the terrain is the same at all times, and since there is almost no terrain influence on the sea, it is set to 0, the input of the terrain data is a two-dimensional tristimulus channel of 3 × 121 × 161, processed with the latest RegNet network and filled to 7 × 7, which is processed with the formula of the ground force F2 mv ω sin θ, where m is the mass of the object, v is the speed of the object, ω is the angular velocity of the earth's rotation, θ is the latitude before the object starts to move, considering that the mass of the typhoon is difficult to estimate, we use the gradient of the ground force to represent the influence of the ground force on the typhoon, whose magnitude is equal to 2v ω sin θ, the position and direction of the typhoon are represented by latitude and longitude, the speed of the typhoon is represented by the average speed of two consecutive moments, the intensity of the typhoon is dependent on its maximum wind speed, therefore, the strength of the typhoon is expressed by the maximum wind speed of the typhoon, and the ground rotation force, the position, the direction, the speed and the strength of the typhoon are processed by a fully-connected network, so that the strength is output as a five-dimensional vector.
The above-described embodiments are merely illustrative of the present patent and do not limit the scope of the patent, and those skilled in the art may make modifications to the parts thereof without departing from the spirit and scope of the patent.
Claims (4)
1. A typhoon track prediction method based on multivariate features is characterized by comprising the following steps:
s1, inputting various characteristics influencing a typhoon track, wherein the characteristics comprise meteorological characteristics, ocean characteristics, geographic characteristics and physical characteristics, and extracting the characteristics of different characteristics by using different networks based on the input data types of the characteristics, wherein the data is mainly collected from an area between 0-60-degree N latitude and 100-degree E-180-degree E longitude;
s2, after extracting different features, performing feature fusion in a feature weighting mode, fusing the feature fusion into 70-dimensional feature vectors with weights, and then taking the 70-dimensional feature vectors as input of a predictor;
s3, because the recurrent neural network RNN has great advantages in processing time sequence data and the neural network LSTM improves the long-term dependence problem in the RNN, the LSTM neural network is selected as a predictor, and a multi-dimensional vector of a specified time sequence is input into the LSTM, so that the output result is the position of typhoon at the next time point;
and S4, reducing the time required by model convergence by using a distributed training frame DistributedDataParalll of PyTorch, accelerating the training speed and better processing a large amount of data.
2. The method for predicting a typhoon trajectory with multiple characteristics according to claim 1, wherein:
the image characteristics in step S1 are as follows: the potential height, the relative humidity, the east-west wind speed, the north-south wind speed and the air temperature are respectively collected at 200hPa, 500hPa, 700hPa, 850hPa and 1000hPa for each characteristic, data are collected at intervals and two-dimensional three-channel pictures are made, the format of the pictures is 3 multiplied by 25 multiplied by 33, the images are processed by a latest RegNet network, filling is set to be 7 multiplied by 7 according to the input size, the 200hPa, 500hPa, 700hPa, 850hPa and 1000hPa are regarded as five factors influencing typhoon trajectories, and therefore the output dimension of the last full-connection layer of the network is set to be five, namely the output result is a five-dimensional vector.
3. The method for predicting a typhoon trajectory with multiple characteristics according to claim 1, wherein:
the sea features in step S1 are as follows: sea surface temperature, physical quantities related to seawater 26 ℃ isotherms, physical quantities related to an ocean mixing layer, physical quantities related to a thermocline, three-dimensional temperature of seawater, three-dimensional salinity of seawater, and three-dimensional density of seawater, wherein the input of the sea surface temperature is a two-dimensional three-channel picture in a format of 3 x 242 x 326, which is processed with a latest RegNet network and is filled to 7 x 7, wherein the physical quantities related to the seawater 26 ℃ isotherms comprise the depth of the seawater 26 ° isotherm and the heat content above the seawater 26 ° isotherm, the physical quantities related to the ocean mixing layer comprise the thickness of the mixing layer and the heat content of the mixing layer, the physical quantities related to the thermocline comprise the depth of the thermocline, the heat content above the thermocline, and the intensity of the thermocline, the inputs of which are two-dimensional three-channel pictures, and are in a format of 3 x 63 x 84, which are all processed with the latest RegNet network and are filled to 3 x 4, and the output dimension of the last full connection layer of the network is set to be five, namely the output result is a five-dimensional vector, so that each feature can output one five-dimensional vector, the input dimensions of the three features of the three-dimensional temperature of the seawater, the three-dimensional salinity of the seawater and the three-dimensional density of the seawater are three-dimensional, ResNet 3D is adopted to process the features, the input formats of the three features are all 3 multiplied by 17 multiplied by 63 multiplied by 84, so that filling is set to be 1 multiplied by 3 multiplied by 4, and the output dimension of the last full connection layer of the network is set to be five, so that each feature can correspond to one five-dimensional vector.
4. The method for predicting a typhoon trajectory with multiple characteristics according to claim 1, wherein:
the geographic features in step S1 are as follows: terrain, ground rotation force, physical characteristics such as: the position and direction, speed, intensity of the typhoon are generally considered to be unchanged, i.e. the terrain is the same at all times, and since there is almost no terrain influence on the sea, it is set to 0, the input of terrain data is a two-dimensional tristimulus channel of 3 × 121 × 161, processed with the latest RegNet network and filled to 7 × 7, which is estimated by the formula of the ground rotation force F2 mv ω sin θ, where m is the mass of the object, v is the speed of the object, ω is the angular velocity of the earth's rotation, θ is the latitude before the object starts to move, considering that the mass of the typhoon is difficult to estimate, we use the gradient of the ground rotation force to represent the influence of the ground rotation force on the typhoon, whose magnitude is equal to 2v ω sin θ, the position and direction of the typhoon are represented by latitude and longitude, the speed of the typhoon is represented by the average speed of two consecutive moments, the intensity of the typhoon depends on its maximum wind speed, therefore, the strength of the typhoon is expressed by the maximum wind speed of the typhoon, and the ground rotation force, the position, the direction, the speed and the strength of the typhoon are processed by a fully-connected network, so that the strength is output as a five-dimensional vector.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210284772.7A CN114818860A (en) | 2022-03-24 | 2022-03-24 | Typhoon track prediction method based on multivariate features |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210284772.7A CN114818860A (en) | 2022-03-24 | 2022-03-24 | Typhoon track prediction method based on multivariate features |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114818860A true CN114818860A (en) | 2022-07-29 |
Family
ID=82530314
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210284772.7A Pending CN114818860A (en) | 2022-03-24 | 2022-03-24 | Typhoon track prediction method based on multivariate features |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114818860A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115016041A (en) * | 2022-08-09 | 2022-09-06 | 中国科学院自动化研究所 | Typhoon intensity determination method and device, electronic equipment and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2012046959A1 (en) * | 2010-10-07 | 2012-04-12 | 서울대학교산학협력단 | Prediction model for summer typhoon number and track for each group |
| CN109902885A (en) * | 2019-04-09 | 2019-06-18 | 中国人民解放军国防科技大学 | Typhoon prediction method based on deep learning mixed CNN-LSTM model |
| CN112507633A (en) * | 2020-12-03 | 2021-03-16 | 广东电网有限责任公司电力科学研究院 | Method and system for predicting and early warning wind speed of transmission tower |
| CN113222210A (en) * | 2021-03-30 | 2021-08-06 | 浙江工业大学 | Typhoon short-term prediction method and device based on multi-mode generation confrontation |
| US20220027761A1 (en) * | 2020-07-21 | 2022-01-27 | Cosco Shipping Technology (beijing) Co., Ltd. | Typhoon trajectory prediction method, device, electronic equipment and computer readable medium |
-
2022
- 2022-03-24 CN CN202210284772.7A patent/CN114818860A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2012046959A1 (en) * | 2010-10-07 | 2012-04-12 | 서울대학교산학협력단 | Prediction model for summer typhoon number and track for each group |
| CN109902885A (en) * | 2019-04-09 | 2019-06-18 | 中国人民解放军国防科技大学 | Typhoon prediction method based on deep learning mixed CNN-LSTM model |
| US20220027761A1 (en) * | 2020-07-21 | 2022-01-27 | Cosco Shipping Technology (beijing) Co., Ltd. | Typhoon trajectory prediction method, device, electronic equipment and computer readable medium |
| CN112507633A (en) * | 2020-12-03 | 2021-03-16 | 广东电网有限责任公司电力科学研究院 | Method and system for predicting and early warning wind speed of transmission tower |
| CN113222210A (en) * | 2021-03-30 | 2021-08-06 | 浙江工业大学 | Typhoon short-term prediction method and device based on multi-mode generation confrontation |
Non-Patent Citations (2)
| Title |
|---|
| CHEN RUI 等: "A hybrid CNN-LSTM model for typhoon formation forecasting", 《GEOINFORMATICA：AN INTERNATIONAL JOURNAL OF ADVANCES OF COMPUTER SCIENCE FOR GEOGRAPHIC》, 31 December 2019 (2019-12-31), pages 1 - 3 * |
| 徐高扬;刘姚;: "LSTM网络在台风路径预测中的应用", 计算机与现代化, no. 05, 15 May 2019 (2019-05-15), pages 68 - 72 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115016041A (en) * | 2022-08-09 | 2022-09-06 | 中国科学院自动化研究所 | Typhoon intensity determination method and device, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109410575B (en) | Road network state prediction method based on capsule network and nested long-time memory neural network | |
| CN113283095A (en) | Evolutionary digital twin watershed construction method | |
| CN112946784B (en) | SuperDARN radar convection diagram short-term forecasting method based on deep learning | |
| CN114612769B (en) | Integrated sensing infrared imaging ship detection method integrated with local structure information | |
| CN110781458B (en) | Method for predicting surface solar irradiance based on mixed regression model | |
| WO2021196528A1 (en) | Global ionospheric total electron content prediction method based on deep recurrent neural network | |
| CN114818860A (en) | Typhoon track prediction method based on multivariate features | |
| Sampath et al. | Estimation of rooftop solar energy generation using Satellite Image Segmentation | |
| CN112558185A (en) | Bidirectional GRU typhoon track intelligent prediction and forecast system based on attention mechanism, computer equipment and storage medium | |
| CN113496310A (en) | Atmospheric pollutant prediction method and system based on deep learning model | |
| Giffard-Roisin et al. | Deep learning for hurricane track forecasting from aligned spatio-temporal climate datasets | |
| Li et al. | Energy-efficient task offloading of edge-aided maritime UAV systems | |
| CN113743577A (en) | Fine grid data partition construction method and system for mesoscale vortex identification | |
| CN115984330A (en) | Boundary-aware target tracking model and target tracking method | |
| Lei et al. | Real-time Anomaly Target Detection and Recognition in Intelligent Surveillance Systems based on SLAM | |
| CN112132880B (en) | Real-time dense depth estimation method based on sparse measurement and monocular RGB image | |
| Zhao | Spatial-Temporal Correlation-Based LSTM Algorithm and Its Application in PM2. 5 Prediction. | |
| Luo et al. | Ship motion trajectory and prediction based on vector analysis | |
| Zhu et al. | Adversarial fine-grained adaptation network for cross-scene classification | |
| CN115840261A (en) | Typhoon precipitation short-term prediction model optimization and prediction method | |
| Jin et al. | A deep learning model for joint prediction of three-dimensional ocean temperature, salinity and flow fields | |
| Xie et al. | Stcanet: Spatiotemporal coupled attention network for ocean surface Current prediction | |
| CN115561834A (en) | Meteorological short-term and temporary forecasting all-in-one machine based on artificial intelligence | |
| CN115358419A (en) | Federal distillation-based indoor positioning method for Internet of things | |
| CN114972429A (en) | Target tracking method and system for cloud edge collaborative self-adaptive inference path planning |
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 | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20221220 Address after: 110004, Lane 3, Lane 11, Wenhua Road, Heping District, Liaoning, Shenyang Applicant after: Northeastern University Address before: No. 195, Chuangxin Road, Baita Town, Hunnan District, Shenyang, Liaoning 110015 Applicant before: Xia Xinyu Applicant before: Gao Yibo |