CN112558187A - Method for forecasting solar flare outbreak based on 3D convolutional neural network - Google Patents
Method for forecasting solar flare outbreak based on 3D convolutional neural network Download PDFInfo
- Publication number
- CN112558187A CN112558187A CN202011494983.0A CN202011494983A CN112558187A CN 112558187 A CN112558187 A CN 112558187A CN 202011494983 A CN202011494983 A CN 202011494983A CN 112558187 A CN112558187 A CN 112558187A
- Authority
- CN
- China
- Prior art keywords
- layer
- neural network
- solar
- observation data
- convolutional neural
- 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
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 42
- 238000000034 method Methods 0.000 title claims abstract description 30
- 230000000694 effects Effects 0.000 claims abstract description 49
- 238000012549 training Methods 0.000 claims abstract description 25
- 238000012360 testing method Methods 0.000 claims abstract description 12
- 210000002569 neuron Anatomy 0.000 claims description 10
- 241000282414 Homo sapiens Species 0.000 claims description 5
- 238000004422 calculation algorithm Methods 0.000 claims description 5
- 238000007781 pre-processing Methods 0.000 claims description 5
- 238000005070 sampling Methods 0.000 claims description 4
- 239000000126 substance Substances 0.000 claims description 3
- 238000013480 data collection Methods 0.000 abstract description 2
- 238000005516 engineering process Methods 0.000 description 8
- 230000006870 function Effects 0.000 description 8
- 238000013528 artificial neural network Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 5
- 238000004364 calculation method Methods 0.000 description 4
- 230000007547 defect Effects 0.000 description 3
- 238000011161 development Methods 0.000 description 3
- 206010000496 acne Diseases 0.000 description 2
- 238000013277 forecasting method Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000003062 neural network model Methods 0.000 description 2
- 238000011056 performance test Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000007477 logistic regression Methods 0.000 description 1
- 230000000877 morphologic effect Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 239000002243 precursor Substances 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
Images
Classifications
-
- 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
- G06N3/045—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
-
- 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)
- Theoretical Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Environmental & Geological Engineering (AREA)
- Atmospheric Sciences (AREA)
- Biodiversity & Conservation Biology (AREA)
- Ecology (AREA)
- Environmental Sciences (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a method for forecasting solar flare outbreak based on a 3D convolutional neural network, which comprises the following steps: 1. adopting continuous observation data to construct an observation data cube; 2. training the 3D convolutional neural network model by adopting a training set; 3. and inputting the test set into the trained 3D convolutional neural network model for prediction. According to the method, a series of continuous solar activity area images in time are positioned, tracked, intercepted and corrected, so that the problem of picture distortion subsequently input into a model can be avoided, the result is more practical, and the processed data is matched with the specific 3D convolutional neural network, so that the time sequence information contained in continuous solar observation data can be fully extracted, the evolution process information of an activity area is captured, and the accuracy of the model for solar flare forecasting is effectively improved; meanwhile, due to the fact that continuous sun observation data are used, the complexity of observation data collection and arrangement work is effectively reduced.
Description
Technical Field
The invention relates to a method for forecasting solar flare outbreak based on a 3D convolutional neural network, and belongs to the field of astronomy technology and image processing.
Background
With the development of the science and technology level of human beings, in particular, the exploration and development activities of outer space are deepened continuously, and the influence of space weather on the human activities is deepened continuously. Solar flare outbreaks are an important embodiment of severe weather conditions in space, and high intensity solar flares pose a threat to radio communications, satellite navigation, and even space flight, as well as to the safety of astronauts. Therefore, the method accurately forecasts the outbreak of the solar flare and has important significance for avoiding or reducing the harm of the solar flare to human beings.
Solar flare is a severe solar outbreak, and research on the forecast of solar flare outbreaks has been an important subject in the industry. However, the specific physical mechanism of solar flare outbreaks is not well defined at present, and therefore almost all forecasting models are based on the statistical relationship between solar flare outbreaks and solar observation data.
As early as 1990, McIntosh proposed a morphological classification of sun blackheads, i.e., "McIntosh typing", and based thereon explored the association of blackheads with flare outbreaks. Later, with the continuous development of computer technology, machine learning related technology was widely applied to research work related to solar flare prediction. In 2007, Li et al put forward a Solar Flare Forecasting model combining a Support Vector Machine and a K-neighbor algorithm in a document 'Support Vector Machine combined with K-near satellites for Solar Flare Forecasting', so that the accuracy of Solar Flare Forecasting is further improved; in 2008, Wang et al proposed a Solar flare prediction model based on an artificial neural network in the document "Solar flare estimating model with associated Solar network technologies"; in 2009, Song et al proposed a method of predicting Solar Flare outbreaks using sequential logistic regression in the literature "statistical assessment of photonic Magnetic Features in immune Solar Flare precursors"; yu et al, 2010, used a Bayesian network to make Short Term forecasts of Solar Flare outbreaks in "Short-Term Solar Flare Level Prediction Using a Bayesian network. However, the early solar flare outbreak forecasting method based on the machine learning method depends on the manually extracted solar activity area parameters, and the effect is not very ideal.
In 2018, Huangxin et al put forward a Solar Flare Forecasting method Based on a convolutional neural network in a document 'Deep Learning Based Solar Flare Forecasting model.I. results for Line-of-sight magnetic'. In the same year, Park et al proposed a Convolutional Neural Network based on a Full-time magnetic map in the document "Application of the Deep probabilistic Neural Network to the Forecast of Solar Flare Using Full-disk Solar magnetic maps", which omits the step of preprocessing the input data and further simplifies the use of the Flare prediction model. In 2019, Zheng Yanhuang et al put forward a Hybrid Convolutional Neural Network that can be used for Flare Prediction in the document "Solar Flare Prediction with the Hybrid Deep conditional Neural Network", which realizes further accurate classification of Flare outbreak intensity. However, the solar flare prediction model based on the convolutional neural network uses the conventional 2D convolutional neural network, and the convolution kernel thereof can only perform convolution on a planar image, i.e. two-dimensional data, and cannot fully utilize time dimension information contained in continuous solar observation data.
Whether the solar flare happens or not is closely related to the evolution process of the activity area where the solar flare happens, whether the time dimension information contained in continuous solar observation data can be fully utilized or not can enable the forecasting model to learn the evolution process of the physical parameters of the activity area, and the bottleneck for limiting the performance of the solar flare forecasting model is formed.
Disclosure of Invention
The invention provides a method for forecasting solar flare outbreaks based on a 3D convolutional neural network, which can realize solar flare outbreak forecasting by fully utilizing time dimension information contained in continuous solar observation data.
The technical scheme of the invention is as follows: a method of forecasting solar flare outbreaks based on a 3D convolutional neural network, the method steps being as follows:
step 1, adopting continuous observation data to construct an observation data cube, and dividing the observation data cube into a training set and a test set;
and 3, inputting the test set into the trained 3D convolutional neural network model to obtain a prediction result.
The step 1 comprises the following steps:
acquiring original observation data and solar activity record data of a solar activity area;
preprocessing observation data;
and constructing and classifying the data cube.
The step 1 specifically comprises the following steps:
s1.1, acquiring original observation data of a solar activity area, namely a full-sun-face longitudinal magnetic map provided by SDO/HMI (software development architecture/human machine interface) and solar activity record data provided by NOAA (Nomina);
s1.2, positioning an active area on each full-day-surface longitudinal magnetic map based on solar activity record data provided by NOAA; tracking the position change of the solar activity area and intercepting the activity area image by combining the solar rotation speed; correcting the intercepted image by utilizing a spherical coordinate positioning algorithm; finally, continuous observation data of a plurality of groups of solar activity areas within a certain time range are obtained;
s1.3, sampling each group of continuous observation data obtained in S1.2 into m groups of observation data containing one frame at intervals of n × m minutes
A data cube of the frame; wherein n represents the time interval of shooting of the longitudinal magnetic map of the whole day surface, w is more than or equal to 210 and represents the total frame number of a group of continuous observation data, and m takes the value of 5, 6 or 7;
and S1.4, according to the solar activity record data provided by NOAA, dividing all data cubes into a outbreak type and a non-outbreak type, and further dividing the two types into a test set and a training set respectively.
The 3D convolutional neural network of step 2 consists of 12 layers:
the 1 st layer is an input layer with input size of 256 × s; wherein the content of the first and second substances,
w is more than or equal to 210 and represents the total frame number of a group of continuous observation data, and m is 5, 6 or 7;
the 3 rd layer is a batch standardization layer;
the 4 th layer is a nonlinear layer;
layer 6 is the second 3D convolutional layer, containing a total of 64 size 3 x 3D convolutional kernels;
the 7 th, 8 th and 9 th layers have the same structure and function as the 3 rd, 4 th and 5 th layers;
the 10 th layer and the 11 th layer are full connection layers, the number of neurons contained in the two layers is respectively 256 and 32, and the neurons between the two layers are connected with each other pairwise;
the 12 th layer is an output layer, and two types of output results are shared, namely the prediction result is 'burst' or 'no burst'.
The invention has the beneficial effects that: according to the method, a series of continuous solar activity area images in time are positioned, tracked, intercepted and corrected, so that the problem of picture distortion subsequently input into a model can be avoided, the result is more practical, and the processed data is matched with the specific 3D convolutional neural network, so that the time sequence information contained in continuous solar observation data can be fully extracted, the evolution process information of an activity area is captured, and the accuracy of the model for solar flare forecasting is effectively improved; meanwhile, due to the fact that continuous sun observation data are used, the complexity of observation data collection and arrangement work is effectively reduced.
Drawings
FIG. 1 is a schematic diagram of the principle of the 3D convolution technique on which the present invention is based;
FIG. 2 is a schematic diagram of a conventional convolution technique;
FIG. 3 is a schematic diagram of a data cube to which the present invention relates;
FIG. 4 is a block diagram of a 3D convolutional neural network model constructed in an embodiment of the present invention;
FIG. 5 is a schematic diagram of the prediction performance test result of the 3D convolutional neural network model according to the present invention.
Detailed Description
Example 1: a method of forecasting solar flare outbreaks based on a 3D convolutional neural network, comprising: step 1, constructing an observation data cube, and dividing the observation data cube into a training set and a test set; step 2, training the 3D convolutional neural network model by adopting a training set to obtain a trained 3D convolutional neural network model; and 3, inputting the test set into the trained 3D convolutional neural network model to obtain a prediction result and evaluating the result.
Further, step 1 may be configured to include: acquiring original observation data and solar activity record data of a solar activity area; preprocessing observation data; and constructing and classifying the data cube.
Further, the step 1 may specifically be:
s1.1, acquiring original observation data of a solar activity area, namely a full-sun-surface longitudinal magnetic map provided by an SDO/HMI (solar dynamics observation station) and solar activity record data provided by NOAA;
s1.2, positioning an active area on each full-day-surface longitudinal magnetic map based on solar activity record data (including position coordinate data) provided by NOAA; tracking the position change of the solar active area and intercepting the image of the active area by combining the autorotation speed of the sun at about 0.6 degree per hour; correcting the captured image by utilizing a spherical coordinate positioning algorithm so as to reduce the influence of projection effect on the imaging of the sun surface; finally, continuous observation data of a plurality of groups of solar activity areas within a certain time range are obtained;
s1.3, sampling each group of continuous observation data obtained in S1.2 into m groups of observation data containing one frame at intervals of n × m minutes
A data cube of the frame; wherein n represents the time interval of shooting of the longitudinal magnetic map of the whole day surface, w is more than or equal to 210 and represents the total frame number of a group of continuous observation data, and m takes the value of 5, 6 or 7;
the value of m is 5, 6 or 7, so that the defects that the data size for training is too small and the training effect is poor due to too small number of data cubes can be avoided; meanwhile, the defects that each subcube contains too few frames and cannot effectively contain time dimension information due to the fact that the number of the data cubes is too large can be avoided.
S1.4, according to solar activity record data provided by NOAA, dividing all data cubes into two types of 'outbreak' and 'no outbreak', and further respectively dividing the two types of data cubes into two types of data cubes according to the ratio of 1: the ratio of 4 is divided into a test set and a training set.
Further, the 3D convolutional neural network of step 2 may be set to be composed of 12 layers:
the 1 st layer is an input layer with input size of 256 × s; wherein the content of the first and second substances,
w is more than or equal to 210 and represents the total frame number of a group of continuous observation data, and m is 5, 6 or 7;
the 3 rd layer is a batch standardization layer, and the function of the batch standardization layer is that the input of each layer of the network model in the training process keeps the same distribution;
the 4 th layer is a nonlinear layer, and the function of the layer is to introduce nonlinear characteristics into the model;
the 5 th layer is a Dropout layer which randomly shields a certain proportion of neurons in the training process so as to improve the generalization capability of the model;
layer 6 is a second 3D convolution layer that functions similarly to layer 2, containing a total of 64 3 x 3 size 3D convolution kernels;
the 7 th, 8 th and 9 th layers have the same structure and function as the 3 rd, 4 th and 5 th layers;
the 10 th layer and the 11 th layer are full connection layers, the number of neurons contained in the two layers is respectively 256 and 32, and the neurons between the two layers are connected with each other pairwise;
the 12 th layer is an output layer, and two types of output results are shared, namely the prediction result is 'burst' or 'no burst'.
Set up the 3D convolution neural network of specific 12 layers of constructions through this application, on satisfying the basis that training speed is fast, realized stronger learning ability, also can avoid the number of piles too much to bring the not enough of overfitting, show through final simulation experiment simultaneously, the model input of this application is through specific preliminary treatment after-data, can have outstanding prediction performance.
Example 2: aiming at a method for forecasting solar flare outbreaks based on a 3D convolutional neural network, the following experimental steps are given:
the 3D convolution neural network model disclosed by the invention uses a 3D convolution technology as shown in figure 1, and is mainly characterized in that convolution calculation can be carried out on a three-dimensional space containing a time dimension. Compared with the conventional convolution technology shown in fig. 2, the 3D convolution technology increases a time dimension, so that information including the time dimension can be effectively extracted from continuous solar observation data, and the amount of information learned by the 3D convolution neural network model is effectively increased, so as to achieve the purpose of improving the accuracy of solar flare prediction.
The invention uses the data cube constructed by continuous observation data in 48 hours of each solar activity area as input, and then uses the 3D convolution neural network to automatically extract the characteristics from the data cube, thereby predicting whether the solar activity area will generate solar flare outbreak in the future.
The method comprises the following specific steps:
step 1: and acquiring the original observation data and the solar activity record data of the solar activity area.
In the embodiment of the invention, the original observation data of the solar activity area come from a full-sun longitudinal magnetic map shot by the SDO/HMI, the shooting time interval is 12 minutes, and the data can be obtained by http:// jsoc.stanford.edu/downloading; solar activity recording data is provided by NOAA, and information such as date, flare outbreak situation, coordinate position and the like of each activity area can be obtained by http:// www.ngdc.noaa.gov/downloading.
Step 2: and (5) preprocessing the observation data.
Since the observation data acquired in step 1 is a longitudinal full-solar magnetic map and the basic unit for forecasting solar flare is an activity area magnetic map, the following processing is required to be performed on the observation data:
A. and positioning the active area on each full-day magnetic map according to the coordinate position information of the active area contained in the active area data provided by the NOAA.
B. And tracking and intercepting the position change of the solar activity area by combining the rotation speed of the sun with the average 0.6 degree per hour, so as to avoid the defect that the position of the activity area on the solar surface changes along with the time.
C. And correcting the distortion of the spherical coordinate positioning algorithm.
And step 3: and constructing and classifying the data cube.
Fig. 3 shows a series of successive observations that were pre-processed for each activity zone. In order to expand the total amount of the data cubes as much as possible and ensure that the number of frames of each cube is not too small, each group of continuous observation data (wherein the frame interval time is 12 minutes, and the total number is 210 frames) obtained in the previous step is sampled into 7 data cubes containing 30 frames according to the sampling frequency of taking one frame every 84 minutes; and dividing all data cubes into two types of 'outbreak' and 'no outbreak' according to the solar activity record provided by NOAA, and further respectively dividing the two types into 1: the ratio of 4 is divided into a test set and a training set.
And 4, step 4: and establishing a 3D convolutional neural network model.
The structure of the 3D convolutional neural network model constructed by the embodiment of the present invention is shown in fig. 4:
level 1 is the input level with input size 256 x 30 (i.e., the size of the data cube described in step 3).
Layer 3 is a batch normalization layer whose main function is to keep the same distribution of inputs to each layer of the network model during the training process.
Layer 4 is a non-linear layer whose function is to introduce non-linear features into the model.
The 5 th layer is a Dropout layer which randomly shields a certain proportion of neurons in the training process so as to improve the generalization capability of the model.
Layer 6 is the second convolution layer, which contains 64 3D convolution kernels of size 3 x 3 in total, and is responsible for the second 3D convolution calculation.
The layers 7, 8 and 9 have the same structure and function as the layers 3, 4 and 5.
The 10 th layer and the 11 th layer are full connection layers, the number of neurons contained in the two layers is 256 and 32 respectively, and the neurons between the two layers are connected with each other pairwise.
The 12 th layer is an output layer, and two types of output results are shared, namely the prediction result is 'burst' or 'no burst'.
And 5: and training a 3D convolutional neural network model.
And 3, taking the training set in the data cube constructed and classified in the step 3 as input data of the 3D convolutional neural network, and performing iterative training for a plurality of times until convergence.
Step 6: and testing and evaluating the performance of the forecasting model.
As a typical binary model, there are usually three indicators for evaluating the performance:
the accuracy is as follows:
accuracy (burst):
precision (no burst):
recall (outbreak):
recall (no outbreak):
wherein, TP is the number of samples correctly predicted as outbreak, TN is the number of samples correctly predicted as not outbreak, FN is the number of samples incorrectly predicted as not outbreak, and FP is the number of samples incorrectly predicted as outbreak.
In the above, the accuracy rate indicates the proportion of all correctly predicted samples to all samples, the accuracy rate indicates the proportion of correctly predicted samples in samples predicted to burst (or not burst), and the recall rate indicates the proportion of correctly predicted samples in actual bursts (or not bursts).
The performance test result of the 3D convolutional neural network in the present invention is shown in fig. 5 by combining the above 5 indexes. As can be seen from fig. 5, the solar flare prediction method based on the 3D convolutional neural network has excellent prediction performance, which shows that the 3D convolutional neural network built by the invention can effectively extract and utilize the timing sequence information hidden in the continuous observation data, and utilize the extracted information to perform solar flare prediction.
The above description is only for the preferred embodiment of the present invention, and is not intended to limit the scope of the present invention.
While the present invention has been described in detail with reference to the embodiments shown in the drawings, the present invention is not limited to the embodiments, and various changes can be made without departing from the spirit of the present invention within the knowledge of those skilled in the art.
Claims (4)
1. A method for forecasting solar flare outbreaks based on a 3D convolutional neural network is characterized in that: the method comprises the following steps:
step 1, adopting continuous observation data to construct an observation data cube, and dividing the observation data cube into a training set and a test set;
step 2, training the 3D convolutional neural network model by adopting a training set to obtain a trained 3D convolutional neural network model;
and 3, inputting the test set into the trained 3D convolutional neural network model to obtain a prediction result.
2. A method of forecasting solar flare outbreaks based on 3D convolutional neural network as claimed in claim 1, characterized in that: the step 1 comprises the following steps:
acquiring original observation data and solar activity record data of a solar activity area;
preprocessing observation data;
and constructing and classifying the data cube.
3. A method of forecasting solar flare outbreaks based on 3D convolutional neural network as claimed in claim 2, characterized in that: the step 1 specifically comprises the following steps:
s1.1, acquiring original observation data of a solar activity area, namely a full-sun-face longitudinal magnetic map provided by SDO/HMI (software development architecture/human machine interface) and solar activity record data provided by NOAA (Nomina);
s1.2, positioning an active area on each full-day-surface longitudinal magnetic map based on solar activity record data provided by NOAA; tracking the position change of the solar activity area and intercepting the activity area image by combining the solar rotation speed; correcting the intercepted image by utilizing a spherical coordinate positioning algorithm; finally, continuous observation data of a plurality of groups of solar activity areas within a certain time range are obtained;
s1.3, sampling each group of continuous observation data obtained in S1.2 into m groups of observation data containing one frame at intervals of n × m minutes
Number of framesAccording to the cube; wherein n represents the time interval of shooting of the longitudinal magnetic map of the whole day surface, w is more than or equal to 210 and represents the total frame number of a group of continuous observation data, and m takes the value of 5, 6 or 7;
and S1.4, according to the solar activity record data provided by NOAA, dividing all data cubes into a outbreak type and a non-outbreak type, and further dividing the two types into a test set and a training set respectively.
4. A method of forecasting solar flare outbreaks based on 3D convolutional neural network as claimed in claim 1, characterized in that: the 3D convolutional neural network of step 2 consists of 12 layers:
the 1 st layer is an input layer with input size of 256 × s; wherein the content of the first and second substances,
w is more than or equal to 210 and represents the total frame number of a group of continuous observation data, and m is 5, 6 or 7;
layer 2 is a first 3D convolution layer containing 32 3D convolution kernels of size 5 x 5;
the 3 rd layer is a batch standardization layer;
the 4 th layer is a nonlinear layer;
layer 5 is a DropOut layer;
layer 6 is the second 3D convolutional layer, containing a total of 64 size 3 x 3D convolutional kernels;
the 7 th, 8 th and 9 th layers have the same structure and function as the 3 rd, 4 th and 5 th layers;
the 10 th layer and the 11 th layer are full connection layers, the number of neurons contained in the two layers is respectively 256 and 32, and the neurons between the two layers are connected with each other pairwise;
the 12 th layer is an output layer, and two types of output results are shared, namely the prediction result is 'burst' or 'no burst'.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011494983.0A CN112558187B (en) | 2020-12-17 | 2020-12-17 | Method for forecasting solar flare outbreak based on 3D convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011494983.0A CN112558187B (en) | 2020-12-17 | 2020-12-17 | Method for forecasting solar flare outbreak based on 3D convolutional neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112558187A true CN112558187A (en) | 2021-03-26 |
| CN112558187B CN112558187B (en) | 2022-08-12 |
Family
ID=75062919
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011494983.0A Active CN112558187B (en) | 2020-12-17 | 2020-12-17 | Method for forecasting solar flare outbreak based on 3D convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112558187B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113537460A (en) * | 2021-06-29 | 2021-10-22 | 哈尔滨工业大学 | Method for constructing multithreading neural network model suitable for flare prediction problem |
| CN113610762A (en) * | 2021-07-07 | 2021-11-05 | 中国科学院国家空间科学中心 | Early warning method and early warning system for solar flare |
| CN114912674A (en) * | 2022-05-05 | 2022-08-16 | 山东大学 | Solar radio burst forecasting method and system based on full-sun photosphere magnetic map |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105372723A (en) * | 2015-10-30 | 2016-03-02 | 中国科学院国家天文台 | Solar flare forecasting method based on convolutional neural network model |
| CN108717568A (en) * | 2018-05-16 | 2018-10-30 | 陕西师范大学 | A kind of image characteristics extraction and training method based on Three dimensional convolution neural network |
| KR102009460B1 (en) * | 2018-07-25 | 2019-08-09 | 경희대학교 산학협력단 | Device and method for forecasting solar flare based on sdo/hmi vector magnetic fields |
| CN111563420A (en) * | 2020-04-16 | 2020-08-21 | 自然资源部第一海洋研究所 | Sea surface solar flare area oil spilling multispectral detection method based on convolutional neural network |
-
2020
- 2020-12-17 CN CN202011494983.0A patent/CN112558187B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105372723A (en) * | 2015-10-30 | 2016-03-02 | 中国科学院国家天文台 | Solar flare forecasting method based on convolutional neural network model |
| CN108717568A (en) * | 2018-05-16 | 2018-10-30 | 陕西师范大学 | A kind of image characteristics extraction and training method based on Three dimensional convolution neural network |
| KR102009460B1 (en) * | 2018-07-25 | 2019-08-09 | 경희대학교 산학협력단 | Device and method for forecasting solar flare based on sdo/hmi vector magnetic fields |
| CN111563420A (en) * | 2020-04-16 | 2020-08-21 | 自然资源部第一海洋研究所 | Sea surface solar flare area oil spilling multispectral detection method based on convolutional neural network |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113537460A (en) * | 2021-06-29 | 2021-10-22 | 哈尔滨工业大学 | Method for constructing multithreading neural network model suitable for flare prediction problem |
| CN113610762A (en) * | 2021-07-07 | 2021-11-05 | 中国科学院国家空间科学中心 | Early warning method and early warning system for solar flare |
| CN113610762B (en) * | 2021-07-07 | 2024-02-23 | 中国科学院国家空间科学中心 | Early warning method and early warning system for solar flare |
| CN114912674A (en) * | 2022-05-05 | 2022-08-16 | 山东大学 | Solar radio burst forecasting method and system based on full-sun photosphere magnetic map |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112558187B (en) | 2022-08-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112558187B (en) | Method for forecasting solar flare outbreak based on 3D convolutional neural network | |
| Xu et al. | On-board real-time ship detection in HISEA-1 SAR images based on CFAR and lightweight deep learning | |
| CN107945204B (en) | Pixel-level image matting method based on generation countermeasure network | |
| CN110232350B (en) | Real-time water surface multi-moving-object detection and tracking method based on online learning | |
| CN111507982A (en) | Point cloud semantic segmentation method based on deep learning | |
| CN109523013B (en) | Air particulate matter pollution degree estimation method based on shallow convolutional neural network | |
| Wang et al. | On deep learning-based bias correction and downscaling of multiple climate models simulations | |
| Woźniak et al. | Deep neural network with transfer learning in remote object detection from drone | |
| CN110136162B (en) | Unmanned aerial vehicle visual angle remote sensing target tracking method and device | |
| CN112580407B (en) | Space target part identification method based on lightweight neural network model | |
| CN111461006B (en) | Optical remote sensing image tower position detection method based on deep migration learning | |
| Liu et al. | RB-Net: Training highly accurate and efficient binary neural networks with reshaped point-wise convolution and balanced activation | |
| CN112836820B (en) | Deep convolution network training method, device and system for image classification task | |
| CN110647977B (en) | Method for optimizing Tiny-YOLO network for detecting ship target on satellite | |
| Gong et al. | Enhancing spatial variability representation of radar nowcasting with generative adversarial networks | |
| Ghosh et al. | On-board ship detection for medium resolution optical sensors | |
| Park et al. | Rgb image prioritization using convolutional neural network on a microprocessor for nanosatellites | |
| CN113222209B (en) | Regional tail gas migration prediction method and system based on domain adaptation and storage medium | |
| Mastrofini et al. | Design and validation of a U-Net-based algorithm for star sensor image segmentation | |
| CN113221997A (en) | High-resolution image rape extraction method based on deep learning algorithm | |
| CN113989612A (en) | Remote sensing image target detection method based on attention and generation countermeasure network | |
| Leong et al. | Image classification unit: A U-Net convolutional neural network for on-orbit cloud detection aboard CubeSats | |
| Lu et al. | A method of ground-based cloud motion predict: CCLSTM+ SR-Net | |
| Tian et al. | Estimation model of global ionospheric irregularities: an artificial intelligence approach | |
| Hu et al. | CNN-based vehicle target recognition with residual compensation for circular SAR imaging |
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 | ||
| GR01 | Patent grant |