CN116882594A - Geomagnetic ap index medium-term forecasting method based on limited recursion and deep learning - Google Patents
Geomagnetic ap index medium-term forecasting method based on limited recursion and deep learning Download PDFInfo
- Publication number
- CN116882594A CN116882594A CN202311140184.7A CN202311140184A CN116882594A CN 116882594 A CN116882594 A CN 116882594A CN 202311140184 A CN202311140184 A CN 202311140184A CN 116882594 A CN116882594 A CN 116882594A
- Authority
- CN
- China
- Prior art keywords
- index
- model
- prediction
- value
- day
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; 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
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; 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
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Tourism & Hospitality (AREA)
- Economics (AREA)
- General Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- General Business, Economics & Management (AREA)
- Software Systems (AREA)
- Marketing (AREA)
- Mathematical Physics (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Development Economics (AREA)
- Data Mining & Analysis (AREA)
- Primary Health Care (AREA)
- Quality & Reliability (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Educational Administration (AREA)
- Operations Research (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The application discloses a geomagnetic ap index medium-term forecasting method and device based on limited recursion and deep learning, comprising the following steps: acquiring an ap index prediction model; respectively obtaining the Ap index, F10.7 and predicted values of a plurality of step sizes of the solar active area characteristics; when the Ap index is predicted, the Ap index, F10.7, the solar active area characteristics and the history value of the Ap index are input into the Ap index prediction model to obtain the predicted value of the Ap index output by the model; taking the predicted value as an input value of an Ap index of the next step length, and inputting the predicted value of the Ap index, F10.7 and the next step length of the solar active area characteristic and the historical values of the Ap index, F10.7 and the solar active area characteristic into the Ap index prediction model to obtain a new predicted value of the Ap index output by the model; and obtaining ap index predicted values of a plurality of step sizes through multiple rounds of iterative prediction. The application can realize the medium-term prediction of geomagnetic ap indexes.
Description
Technical Field
The application relates to the technical field of geomagnetic explosion prediction, in particular to a geomagnetic ap index medium-term prediction method, a geomagnetic ap index medium-term prediction device, geomagnetic ap index medium-term prediction equipment, a geomagnetic explosion prediction readable storage medium and a computer program product based on limited recursion and deep learning.
Background
Geomagnetic explosions refer to a phenomenon in which the earth's magnetic field suddenly changes, and are usually accompanied by high-energy particle events such as solar storm and corona mass projection. Geomagnetic explosions may affect infrastructure such as electric facilities and satellite communications in modern society. Therefore, accuracy and reliability of geomagnetic explosion prediction become more and more important. At present, the existing geomagnetic explosion ap index prediction model in the field mainly comprises two main types: analytical models based on physical laws and timing models based on machine learning (or deep learning) methods.
For an analytic model based on a physical rule, the analytic model is often complex due to the complexity of a geomagnetic field system, and the uncertainty of the corona substance eruption phenomenon with the greatest influence on the geomagnetic field further reduces the accuracy of the analytic model in the magnetic explosion period prediction.
For a time sequence model based on a machine learning (or deep learning) method, the accuracy of the currently available machine learning (or deep learning) algorithm is often better when single step and test are performed, but various difficulties are faced when multi-step long (i.e. medium-long) prediction is performed. The following briefly illustrates some of the processing methods and drawbacks that may be employed for the multi-step or medium-long prediction method.
1) Recursive single-step prediction
The recursion method is to make single-step prediction through the historical value, and then take the predicted value as the historical value of the next stage to continue the single-step prediction after making the prediction. The main advantages of the recursive method are two: (1) the single-step prediction accuracy is relatively high; (2) only one model needs to be trained, and single-quantity multi-step prediction can be performed by a recursion method.
However, the disadvantages of the recursive method are also evident: the output value must be able to fill the input value that is missing at the next time, and for the ap index, which requires many parameters to predict together, multi-step long recursive prediction cannot be performed.
2) Direct prediction
Direct prediction is essentially a single-step prediction. The method of direct prediction is used for multi-step long prediction with the step length of n, in fact, n single-value prediction models are established, and single values after step length time of 1 to n are respectively and specially predicted.
In use, the disadvantage of direct prediction was found: (1) the number of models to be trained is excessive; (2) as n increases, the actual correlation between the quantity to be predicted at time t+n and all the input quantities at the previous n times becomes lower (information of input parameters between t+1 and t+n is missing), so that the accuracy of model prediction drops rapidly, and direct prediction is generally only effective in a short period (1-3 steps).
3) Single-magnitude multi-step prediction
The prediction values of the future steps of the same quantity are directly output through a model, the output values are arranged according to time sequence, the previous value is predicted preferentially, and meanwhile, the prediction of the next value is participated. This approach has the advantage of reducing the number of models to be trained, while the prediction at a later time takes advantage of the prediction information at a previous time. The disadvantages are: (1) the model outputs multiple values at the same time, meaning that the accuracy of each value is degraded; (2) in addition to the amount to be predicted, other input parameters between t and t+n information cannot be utilized.
4) Recursive multivariable single-step prediction
The recursive method of multivariable single-step prediction predicts a plurality of variables at a time, and has the defects that: (1) the model outputs multiple values at the same time, meaning that the accuracy of each value is degraded; (2) the multiple variable errors multiply together and the model performance drops rapidly.
Based on the existing prediction means, a large improvement space exists for the accuracy of geomagnetic explosion ap index prediction, and optimization and upgrading of a prediction tool are needed.
Disclosure of Invention
Aiming at the technical problems existing in the prior art, the application provides a geomagnetic ap index medium-term forecasting method, device, equipment, readable storage medium and computer program product based on limited recursion and deep learning, which can solve at least one technical problem.
The embodiment of the application provides a geomagnetic ap index medium-term forecasting method based on limited recursion and deep learning, which comprises the following steps:
acquiring an Ap index prediction model, wherein the Ap index prediction model is generated after training a first neural network model and is based on a planetary equivalent daily amplitude Ap index, solar radiation electric flux F10.7, solar active area characteristics and a geomagnetic disturbance equivalent planetary amplitude Ap index historical value;
respectively obtaining predicted values of multiple steps of the Ap index, the F10.7 and the solar active area characteristics, wherein the predicted values of the multiple steps of the Ap index are generated by a preset Ap index predicted model, the predicted values of the multiple steps of the F10.7 are generated by a preset F10.7 predicted model, and the predicted values of the multiple steps of the solar active area characteristics are generated by a preset solar active area predicted model;
when the Ap index is predicted, the Ap index, F10.7, the solar active area characteristics and the history value of the Ap index are input into the Ap index prediction model to obtain the predicted value of the Ap index output by the model;
taking the predicted value of the Ap index as the input value of the Ap index of the next step length, and inputting the predicted value of the Ap index, F10.7 and the next step length of the solar active area characteristic, and the historical values of the Ap index, F10.7 and the solar active area characteristic into the Ap index prediction model to obtain a new predicted value of the Ap index output by the model;
And obtaining ap index predicted values of a plurality of step sizes through multiple rounds of iterative prediction.
Optionally, according to a method of an embodiment of the present application, the first neural network model includes a transducer model; the history value comprises history measured values of K step sizes, K is more than or equal to 16, and the value of a single step size L is more than or equal to 1 and less than or equal to 6.
Optionally, according to the method of the embodiment of the present application, the preset Ap index prediction model includes a long-short-term memory network LSTM model;
the training data of the Ap index prediction model comprises an Ap index, F10.7, solar active area characteristics and a history value of solar X rays for a plurality of continuous days, and the output value of the model is a single-day prediction value of the Ap index;
the Ap index prediction model includes three sub-models for predicting Ap indexes of a future first day, a future second day, and a future third day, respectively.
Optionally, according to the method of the embodiment of the present application, the preset F10.7 prediction model includes a long-short-term memory network LSTM model;
the training data of the F10.7 prediction model comprises historical values of F10.7 for a plurality of continuous days, and the output value of the model is a single-day prediction value of F10.7;
and predicting F10.7 of a plurality of step sizes in an iterative mode, and adding the F10.7 predicted value output by the LSTM model of the previous round to the input value of the next round until a preset number of F10.7 predicted values are obtained.
Optionally, according to the method of the embodiment of the present application, the preset solar activity area prediction model includes a solar activity area interpolation model;
the processing procedure of the solar active area interpolation model comprises the following steps:
acquiring influence factors of solar activity areas on the n-4 th day, the n-3 th day, the n-2 th day and the n-1 th day on the Ap index on the n-th day;
acquiring influence factors of solar activity areas on the n-3 th day, the n-2 th day and the n-1 th day on the Ap index on the n+1 th day;
acquiring influence factors of solar activity areas on the n-2 th day and the n-1 th day on the Ap index on the n+2 th day;
taking the obtained influence factors as characteristic values of the solar active area, and adding the characteristic values into an interpolation model;
the characteristic value of a solar active area in the interpolation model is complemented by a linear interpolation method to obtain AP index influence factors of the nth day to the (n+1) th day and the (n+1) th day to the (n+2) th day, wherein the missing value is taken as a value of the previous day or an average value of the previous m days;
and obtaining a solar active area interpolation model for generating a single-day predicted value of the solar active area characteristic.
Optionally, according to the method of the embodiment of the present application, the multiple step ap exponent predicted values include R step ap exponent predicted values, r+.24.
Optionally, according to the method of the embodiment of the present application, when training the first neural network model, five halves are performed on training data, where four halves are used as a training data set and one half is used as a test data set, training is performed respectively, and after training is completed, five ap index prediction models are generated; in the prediction, prediction is performed through five ap index prediction models respectively, and the average value of the obtained five prediction values is used as the prediction value of the ap index.
The embodiment of the application provides a geomagnetic ap index medium-term forecasting device based on limited recursion and deep learning, which comprises the following steps:
the first acquisition module is used for acquiring an Ap index prediction model, wherein the Ap index prediction model is generated after training a first neural network model and is based on a planetary equivalent daily amplitude Ap index, solar radio flux F10.7, solar active area characteristics and geomagnetic disturbance equivalent planetary amplitude Ap index historical values;
the second acquisition module is used for respectively acquiring the Ap index, the F10.7 and the predicted values of the solar active area characteristic in a plurality of steps, wherein the predicted values of the Ap index in a plurality of steps are generated by a preset Ap index prediction model, the predicted values of the F10.7 in a plurality of steps are generated by a preset F10.7 prediction model, and the predicted values of the solar active area characteristic in a plurality of steps are generated by a preset solar active area prediction model; and
The prediction processing module is used for inputting the Ap index, F10.7, the solar active area characteristics and the historical values of the Ap index into the Ap index prediction model to obtain the predicted values of the Ap index output by the model when predicting the Ap index; the method comprises the steps of obtaining a predicted value of an Ap index, and obtaining a predicted value of the Ap index, wherein the predicted value of the Ap index is used as an input value of the Ap index of the next step length, and the predicted value of the Ap index, F10.7 and the next step length of the solar active area characteristic, and the historical values of the Ap index, F10.7 and the solar active area characteristic are all input into the Ap index prediction model to obtain a new predicted value of the Ap index output by the model; and the system is used for obtaining ap index predicted values of a plurality of step sizes through multiple rounds of iterative prediction, and outputting a task planning result.
The application also proposes an electronic device comprising: a processor and a memory storing computer program instructions; the electronic device, when executing the computer program instructions, implements the method as described above.
The application also proposes a computer-readable storage medium having stored thereon computer program instructions which, when executed by a processor, implement a method as described above.
The application also proposes a computer program product comprising computer program instructions which, when executed by a processor, implement a method as described above.
According to the embodiment of the application, the combined prediction is carried out by using a plurality of models, the used models comprise an F10.7 prediction model, an Ap index prediction model and a solar active area interpolation model, the middle-term prediction of geomagnetic Ap index 24 time step is realized in a recursion mode, the method can be used for predicting geomagnetic explosion three days in advance, and the accuracy of a prediction result is high.
Drawings
In order to more clearly describe the technical solution of the embodiments of the present application, the following description briefly describes the drawings in the embodiments of the present application.
Fig. 1 is a flowchart of a geomagnetic ap index mid-term forecasting method based on finite recursion and deep learning according to an embodiment of the present application.
FIG. 2 is a schematic illustration of the absence of features from the solar active area in accordance with an embodiment of the present application.
Fig. 3 is a schematic diagram of a characteristic sample of a solar active area according to an embodiment of the present application.
FIG. 4 is a schematic diagram of an ap index prediction model according to an embodiment of the present application.
FIG. 5 is a schematic diagram of predicting input parameter values for 3 days in the future using an ap index prediction model in accordance with an embodiment of the present application.
FIG. 6 is a schematic diagram of predicting future 3-day ap index based on finite recursion and deep learning in accordance with an embodiment of the present application.
Fig. 7 is a block diagram of a geomagnetic ap index mid-term forecasting apparatus based on finite recursion and deep learning according to an embodiment of the present application.
Fig. 8 is a schematic diagram of an electronic device for implementing a geomagnetic ap index mid-term forecasting method based on finite recursion and deep learning according to an embodiment of the present application.
Detailed Description
The principles and spirit of the present application will be described below with reference to several exemplary embodiments. It will be appreciated that such embodiments are provided to make the principles and spirit of the application clear and thorough, and enabling those skilled in the art to better understand and practice the principles and spirit of the application. The exemplary embodiments provided herein are merely some, but not all embodiments of the application. All other embodiments, which can be made by one of ordinary skill in the art without undue burden from the embodiments herein, are within the scope of the present application.
Embodiments of the present application relate to a terminal device and/or a server. Those skilled in the art will appreciate that embodiments of the application may be implemented as a system, apparatus, device, method, computer readable storage medium, or computer program product. Accordingly, the present disclosure may be embodied in at least one of the following forms: complete hardware, complete software, or a combination of hardware and software. According to the embodiment of the application, the application discloses a method and a device for calculating a large-scale satellite orbit based on a Newton interpolation formula and a Hohner law, electronic equipment and a computer readable storage medium.
In this document, terms such as first, second, third, etc. are used solely to distinguish one entity (or action) from another entity (or action) without necessarily requiring or implying any order or relationship between such entities (or actions).
The following briefly describes related concepts and technical terms, etc. that may be involved in the embodiments of the present application.
Geomagnetic explosions are typically caused by solar black sub-activity. The solar black is a dark spot naturally formed on the surface of the sun and is a product of a solar magnetic field. When the solar black appears on the surface of the sun, a charged particle stream with very high energy is emitted to the solar ring layer, namely, solar wind. When the charged particle flow in the solar wind reaches the vicinity of the earth, the charged particle flow affects the geomagnetic field, and severe geomagnetic disturbance in the global scope is caused, which is called geomagnetic explosion. The impact of geomagnetic explosions on human life is manifold. First, geomagnetic explosions destroy the ionosphere in the atmosphere, causing the attitude of satellites to change, thereby affecting radio communication and satellite communication. Secondly, navigation systems near the earth can be greatly interfered by geomagnetic explosion, and potential safety hazards are brought to traffic facilities such as airplanes, ships and the like. In addition, geomagnetic explosion can cause abnormal increase or sudden decrease of current and voltage in a power grid, so that the service life of the transformer is shortened, and even the transformer is burnt out to cause permanent damage.
The following briefly describes relevant parameters that may be involved in embodiments of the present application:
1) ap index
The ap index is not only the index to be predicted by the application, but also called the equivalent planetary amplitude of the geomagnetic disturbance, the unit is 2nT, and a new ap value is generated every 3 hours. According to the relevant criteria, an ap value greater than 39 corresponds to a period of magnetic storm. The maximum value of the ap index is 400.
2) Ap index
The Ap index is an index of global geomagnetic disturbance intensity, also called planetary equivalent daily amplitude, is an index for measuring geomagnetic activity level, generates a new Ap value every day, is an average value of 8 Ap values in each day, and therefore, the accurate prediction value of the Ap plays a very important index role in the prediction of the Ap.
3）F10.7
F10.7, i.e. 10.7cm solar radiation flux. The measurement of F10.7 is an average value of solar radio radiation intensities measured in one hour over a 100 mhz band centered at 2800 mhz on the ground, with a value updated daily. F10.7 is closely related to the activity of the sun surface, and there is also a long-short periodic variation law of 11 years and 27 days, which is an important index for representing the activity of the sun, and geomagnetic field disturbance has a strong correlation with the activity level of the sun.
The F10.7 history record comprises an observation value and an adjustment value (1 AU), and the observation value of F10.7 can be directly used as an input parameter in consideration of the relationship between the influence of the sun on the geomagnetism and the earth-day distance.
4) Solar active area features
The spatial weather forecast center sorts the observed solar active area picture data of the previous day into daily reports. We sort all the historical daily data into tabular form data. Solar active areas can be divided into two types in the data: the first type is a solar active area of a visible black sub-group, and the second type is a solar active area of an H-alpha wave band with light spots but without the visible black sub-group.
After feature engineering, 20 overall influence factors of the solar activity area on the Ap index can be obtained every day, and the 20 factors respectively represent the influence conditions of different types of solar activity areas on the Ap value 1-4 days ago.
5) Solar X-ray
Solar X-rays are also an important representation of the solar activity level, and typically during large flare bursts of the sun, the radiation in the X-ray band will be enhanced by several orders of magnitude relative to the electromagnetic radiation in other bands before flare occurs. The long wave flux (0.1-0.8 nm) and the short wave flux (0.05-0.4 nm) can be extracted from Gaussian satellite data and used as two independent parameters for calculation.
Fig. 1 is a flowchart of a geomagnetic ap index mid-term forecasting method based on finite recursion and deep learning according to an embodiment of the present application. The method comprises the following steps:
S101, acquiring an Ap index prediction model, wherein the Ap index prediction model is generated after training a first neural network model and is based on a planetary equivalent daily amplitude Ap index, solar radiation electric flux F10.7, solar active area characteristics and geomagnetic disturbance equivalent planetary amplitude Ap index historical values;
s102, respectively obtaining predicted values of multiple steps of the Ap index, F10.7 and the solar active area characteristics, wherein the predicted values of the multiple steps of the Ap index are generated by a preset Ap index prediction model, the predicted values of the multiple steps of F10.7 are generated by a preset F10.7 prediction model, and the predicted values of the multiple steps of the solar active area characteristics are generated by a preset solar active area prediction model;
s103, when the Ap index is predicted, inputting the Ap index, F10.7, the solar active area characteristics and the historical values of the Ap index into the Ap index prediction model to obtain the predicted values of the Ap index output by the model;
s104, taking the predicted value of the Ap index as the input value of the Ap index of the next step length, and inputting the predicted value of the Ap index, F10.7 and the next step length of the solar active area characteristic, and the historical values of the Ap index, F10.7 and the solar active area characteristic into the Ap index prediction model to obtain a new predicted value of the Ap index output by the model;
S105, carrying out multiple rounds of iterative prediction to obtain ap index predicted values of multiple step sizes.
By adopting the geomagnetic ap index medium-term prediction method based on limited recursion and deep learning, which is disclosed by the embodiment of the application, the medium-term prediction of the geomagnetic ap index 24 time step is realized in a recursion mode by using a plurality of models for combined prediction, the method can be used for predicting geomagnetic explosion three days in advance, and the accuracy of a prediction result is high.
In step S101, according to some embodiments of the present application, first, a history of the planetary equivalent daily amplitude Ap index, F10.7, solar activity area characteristics, and geomagnetic disturbance equivalent planetary amplitude Ap index may be obtained. As one example, the historical value may be a measured value corresponding to a plurality of steps. For example, historical measurements of each parameter having a historical value span of 4 days (step size of 3 hours/step, 4 days = 32 time steps) are obtained. Based on these historical values, a predetermined first neural network model is trained to generate an ap index prediction model.
As one example, the first neural network model may employ a transducer model. A transducer model is a neural network model that is generally considered to learn context and thus meaning by tracking relationships in sequence data. Alternatively, in training the transducer model, the training data may be a historical measurement of K steps, where K is greater than or equal to 16 and the individual steps L have a value in the range 1.ltoreq.L.ltoreq.6. For example, k=24, l=3 hours, which means that the transducer model is trained using historical measurements (as true values) of 24 steps (3 hours for a single step), and an ap index prediction model is obtained after training.
In order to predict the Ap index by the aforementioned Ap index prediction model, the Ap index, F10.7, and the predicted values of the multiple steps of the solar active area feature are also required.
According to the embodiment of the application, the predicted values of the multiple steps of the Ap index can be generated by a preset 'Ap index predicted model', the predicted values of the multiple steps of the F10.7 can be generated by a preset 'F10.7 predicted model', and the predicted values of the multiple steps of the solar active area characteristic can be generated by a preset 'solar active area predicted model'. The following describes a specific description of a predictive model for each parameter that may be used in the embodiments of the present application.
Ap index prediction model
The Ap index prediction model may be generated by long and short term memory network LSTM model training, the training data including Ap index, F10.7, solar active area characteristics, solar X-ray continuous multi-day history values. Multiple predictor models may be trained, as needed, to predict Ap indices for a given future date, respectively.
For example, the Ap index, the F10.7, the solar active area characteristics, and the solar X-ray history (for example, the history of 27 consecutive days) are input into the LSTM structure for training, and after training, an Ap index prediction model is generated, and the model adopts a direct prediction strategy, and the output of the model is a single-day prediction value of Ap.
In addition, proper training data can be selected according to the prediction requirement, a plurality of groups of models can be trained, for example, three groups of different Ap index prediction models are trained and used as three predictor models for respectively predicting the Ap values of one day, two days and three days in the future, and the prediction task of the Ap index is jointly completed.
(II) F10.7 predictive model
The prediction model of F10.7 can be generated by training a long-short-term memory network LSTM model, the training data is a history value of continuous multiple days (for example, 270 continuous days) of F10.7, and after training, the output value of the model is a single-day prediction value of F10.7.
The input parameters of the F10.7 prediction model can be regarded as the historical values of the output parameters (F10.7 prediction values), so that after a group of F10.7 prediction models are trained, F10.7 with multiple steps can be predicted in an iterative mode, that is, the prediction value of the previous round is taken as the newly added input value of the next round, and the F10.7 with multiple time steps is predicted by an iterative method to obtain multiple prediction values, so that the effect of combined prediction is achieved.
(III) solar active area prediction model
The magnetic field of the sun surface is a complex field, and it is generally considered that five basic magnetic fields (diffusion magnetic field, dynamic magnetic field, compression magnetic field, altitude magnetic field and wrapping magnetic field) exist on the sun surface, and interaction and complex coupling relation exist between the basic magnetic field types, so that the basic magnetic fields participate in the magnetic activity of the sun and the sun burst phenomenon on different scales, and the basic magnetic fields are represented as the characteristics of a solar activity area and can be described by the data of the solar activity area.
In the field, from solar activity area data, the influence factors of the activity areas of five basic magnetic field types on the solar surface on the current day on the Ap index of 1-4 days in the future can be extracted.
In an embodiment of the present application, assuming that it is desired to predict solar activity area characteristics for the next 3 days, the solar activity area characteristics for each day may be predicted based on factors in which the solar activity area characteristics for the first 4 days of each day affect the Ap index for that day.
For convenience of description, the first day of the future 3 days to be predicted is denoted as the nth day, the second day is denoted as the n+1th day, the third day is denoted as the n+2th day, and the solar active area characteristic loss due to the time is schematically shown in fig. 2.
The predicted values of a plurality of step sizes of the solar active area characteristics are generated by a preset solar active area prediction model. The preset solar active area prediction model comprises a solar active area interpolation model; the processing procedure of the solar active area interpolation model comprises the following steps: acquiring influence factors of solar activity areas on the n-4 th day, the n-3 th day, the n-2 th day and the n-1 th day on the Ap index on the n-th day; acquiring influence factors of solar activity areas on the n-3 th day, the n-2 th day and the n-1 th day on the Ap index on the n+1 th day; acquiring influence factors of solar activity areas on the n-2 th day and the n-1 th day on the Ap index on the n+2 th day; taking the obtained influence factors as characteristic values of the solar active area, and adding the characteristic values into an interpolation model; the characteristic value of a solar active area in the interpolation model is complemented by a linear interpolation method to obtain AP index influence factors of the nth day to the (n+1) th day and the (n+1) th day to the (n+2) th day, wherein the missing value is taken as a value of the previous day or an average value of the previous m days; and obtaining a solar active area interpolation model for generating a single-day predicted value of the solar active area characteristic.
Fig. 3 is a schematic diagram of a characteristic sample of a solar active area according to an embodiment of the present application. A sample of features of the solar active area is shown in FIG. 3, wherein the alpha1-alpha4 columns represent the factors of the effect of the alpha field type active area on the current day for the first 1-4 days, respectively. In the present embodiment, taking the curve in the figure as an example, when the active area data 1 month and 5 days ago is owned, the actually owned feature is above the curve. In this embodiment, the interpolation algorithm takes the value of the last day for the missing value, and the mathematical formula can be expressed as:
wherein,,is the characteristic value of the t day; />Is characteristic value of t+1st day.
In other embodiments, where the interpolation algorithm averages the missing values over the first 3 days, the mathematical formula may be expressed as:
wherein,,is the characteristic value of the t-2 th day; />Is the characteristic value of the t-1 day; />Is the characteristic value of the t day;is characteristic value of t+1st day.
Based on the embodiment of the application, when the Ap index of the next step is predicted, the Ap index, F10.7, the solar active area characteristic and the history value of the Ap index are input into the Ap index prediction model to obtain the predicted value of the Ap index output by the model. FIG. 4 is a schematic diagram of an ap index prediction model according to an embodiment of the present application. In this embodiment, the Ap index prediction model adopts a transform structure, as shown in fig. 4, and the dashed box represents input parameters of the model, that is, ap values, f10.7, solar active area characteristics, and historical values of the Ap values, where the time span of the input values is generally 32 steps (step=3 hours/step), that is, 4 days. The table in fig. 4 corresponds to the output of the model, that is, the ap index predicted value at the next moment.
When the first neural network model is trained, an integrated algorithm is adopted to carry out joint prediction, the specific method is that training data are divided into five parts, four parts are used as training data sets, one part is used as a test data set, training is respectively carried out, and five ap index prediction models are generated after the training is finished; in the prediction, prediction is performed through five ap index prediction models respectively, and the average value of the obtained five prediction values is used as the prediction value of the ap index. In this embodiment, in prediction, five models perform equal weight joint prediction, and the mathematical formula may be expressed as:
wherein,,predicted values for the ith model; />Is the final prediction result.
FIG. 5 is a schematic diagram of predicting input parameter values for 3 days in the future using an ap index prediction model in accordance with an embodiment of the present application. And respectively predicting the Ap index and the F10.7 of the future 3 days by using an Ap index prediction model and an F10.7 prediction model, and complementing the characteristics of the solar active area by using a solar active area interpolation model. After parameter prediction and complementation, the data situation is shown in fig. 5, and the data situation comprises a history interval and a prediction period, wherein the history interval spans 4 days, namely 32 steps, and comprises F10.7, an Ap index, a solar active area characteristic value and an Ap index which are respectively provided with 32 steps; the span of the prediction period is 3 days, namely 24 steps, and comprises the existing 24 steps of F10.7, ap index and predicted values of the characteristic values of the solar activity area. FIG. 6 is a schematic diagram of predicting future 3-day ap index based on finite recursion and deep learning in accordance with an embodiment of the present application. In this embodiment, the finite recursion method is used to predict the ap index, as shown in fig. 6, where the input data interval of the ap index prediction model corresponds to the historical interval value in the solid line frame 1 in the graph, and the time span of the historical interval is 32 time steps of 4 days. The ap predictable interval is the first step size of "to be predicted" in the interval to be predicted, indicated by solid line box 2, which corresponds to 3 hours 1 time step, i.e. an ap value.
According to the embodiment of the application, the predicted value of the Ap index is taken as the input value of the Ap index of the next step length, and the predicted value of the Ap index, F10.7 and the next step length of the solar active area characteristic, and the historical values of the Ap index, F10.7 and the solar active area characteristic are all input into the Ap index prediction model to obtain a new predicted value of the Ap index output by the Ap index prediction model. In this embodiment, after the first prediction is performed by the ap index prediction model, as shown in fig. 6, an ap prediction value in the solid line box is obtained. The next time step can be predicted by taking the same as an input value of the next time step.
Based on the embodiment of the application, the ap index predicted values of a plurality of step sizes are obtained through multiple rounds of iterative prediction. In this embodiment, as shown in fig. 6, it is the solid line frame 1 and the solid line frame 2 that are both moved by one step in the arrow direction. Similarly, three-day prediction of the ap index is completed after 24 steps of prediction, i.e., 24 steps later the solid line frame 2 moves out of the prediction interval. By the finite recursion method, recursive prediction of the ap index for three days in the future is realized by using a set of models, namely 24-step medium-term prediction of the ap index is realized.
Compared with a direct prediction method, the geomagnetic ap index medium-term prediction method based on limited recursion and deep learning of the invention makes full use of historical data and short-term prediction data, and is particularly characterized in that:
(1) ap's predictive value implies X-ray information;
(2) the solar active area features are fully utilized, the direct method can only utilize the leftmost associated features (influence factors of n-1 to n-4 days on n days) in the figure 2, and the method utilizes all the associated features (n+1, n+2 days-related influence factors) in the figure 2;
(3) for the prediction of the latter step, the prediction results of the former step (including the prediction values of Ap, ap and F10.7) are used.
Therefore, the invention has higher accuracy for the prediction of the ap index, and the accuracy advantage becomes more obvious with the increase of the prediction step length.
Compared with a single-quantity multi-step prediction method, the geomagnetic ap index medium-term prediction method based on limited recursion and deep learning has higher accuracy, and the specific reasons are as follows:
(1) the ap model focuses more on single-step predictions;
(2) the input parameters use short-term prediction results of two excellent models (F10.7 prediction model, ap prediction model).
Therefore, the prediction of the ap index according to the present application has higher accuracy than the single-magnitude multi-step prediction method.
By designing the whole model, recursive prediction in a limited interval is realized, and the whole complexity of the model (middle 24-step prediction model) is greatly reduced on the basis of ensuring the prediction accuracy. Specific comparative references are as follows:
(1) the application makes 24-step prediction, only 1 group of models are needed; the deep learning model can learn very complex modes and rules; features can be automatically extracted, and the requirement of manual feature engineering is reduced;
(2) the direct prediction method performs 24-step prediction, and 24 groups of models are needed;
(3) the single-quantity multi-step prediction method is used for 24-step prediction, 24 values are required to be output, which means that the algorithm needs to design a more complex loss function;
(4) the normal recursive method cannot be used for this example.
Correspondingly, the present application also provides a geomagnetic ap index mid-term forecasting device based on finite recursion and deep learning, as shown in fig. 7, the geomagnetic ap index mid-term forecasting device 100 based on finite recursion and deep learning includes:
a first obtaining module 110, configured to obtain an Ap index prediction model, where the Ap index prediction model is based on a history value of a planetary equivalent daily amplitude Ap index, solar radiation electric flux F10.7, a solar active area feature, and a geomagnetic disturbance equivalent planetary amplitude Ap index, and is generated after training a first neural network model;
The second obtaining module 120 is configured to obtain predicted values of multiple steps of the Ap index, the F10.7, and the solar active area feature, where the predicted values of multiple steps of the Ap index are generated by a preset Ap index prediction model, the predicted values of multiple steps of the F10.7 are generated by a preset F10.7 prediction model, and the predicted values of multiple steps of the solar active area feature are generated by a preset solar active area prediction model; and
the prediction processing module 130 is configured to input the Ap index, the F10.7, the solar active area feature, and the historical value of the Ap index into the Ap index prediction model to obtain a predicted value of the Ap index output by the model when predicting the Ap index of the next step; the method comprises the steps of obtaining a predicted value of an Ap index, and obtaining a predicted value of the Ap index, wherein the predicted value of the Ap index is used as an input value of the Ap index of the next step length, and the predicted value of the Ap index, F10.7 and the next step length of the solar active area characteristic, and the historical values of the Ap index, F10.7 and the solar active area characteristic are all input into the Ap index prediction model to obtain a new predicted value of the Ap index output by the model; and the system is used for obtaining ap index predicted values of a plurality of step sizes through multiple rounds of iterative prediction, and outputting a task planning result.
The electronic device in the embodiment of the application can be user terminal equipment, a server, other computing devices and a cloud server. Fig. 8 shows a schematic diagram of a hardware structure of an electronic device according to an embodiment of the present application, where the electronic device may include a processor 601 and a memory 602 storing computer program instructions, where the processor 601 implements the flow or functions of any of the methods of the embodiments described above when executing the computer program instructions.
In particular, the processor 601 may include a Central Processing Unit (CPU), or an application specific integrated circuit (Application Specific Integrated Circuit, ASIC), or may be configured as one or more integrated circuits that implement embodiments of the present application. Memory 602 may include mass storage for data or instructions. For example, the memory 602 may be at least one of: hard Disk Drive (HDD), read-only memory (ROM), random-access memory (RAM), floppy Disk Drive, flash memory, optical Disk, magneto-optical Disk, magnetic tape, universal serial bus (Universal Serial Bus, USB) Drive, or other physical/tangible memory storage device. As another example, the memory 602 may include removable or non-removable (or fixed) media. For another example, memory 602 may be internal or external to the integrated gateway disaster recovery device. The memory 602 may be a non-volatile solid state memory. In other words, generally the memory 602 includes a tangible (non-transitory) computer-readable storage medium (e.g., a memory device) encoded with computer-executable instructions and when the software is executed (e.g., by one or more processors) may perform the operations described by the methods of embodiments of the application. The processor 601 implements the flow or functions of any of the methods of the above embodiments by reading and executing computer program instructions stored in the memory 602.
In one example, the electronic device shown in fig. 8 may also include a communication interface 603 and a bus 610. The processor 601, the memory 602, and the communication interface 603 are connected to each other through a bus 610 and perform communication with each other. The communication interface 603 is mainly used to implement communications between modules, apparatuses, units, and/or devices in the embodiments of the present application. Bus 610 includes hardware, software, or both, and may couple components of the online data flow billing device to each other. For example, the bus may include at least one of: accelerated Graphics Port (AGP) or other graphics bus, enhanced Industry Standard Architecture (EISA) bus, front Side Bus (FSB), hyperTransport (HT) interconnect, industry Standard Architecture (ISA) bus, infiniBand interconnect, low Pin Count (LPC) bus, memory bus, micro channel architecture (MCa) bus, peripheral Component Interconnect (PCI) bus, PCI-Express (PCI-X) bus, serial Advanced Technology Attachment (SATA) bus, video electronics standards Association local (VLB) bus, or other suitable bus. Bus 610 may include one or more buses. Although embodiments of the application describe or illustrate a particular bus, embodiments of the application contemplate any suitable bus or interconnection.
In connection with the methods of the above embodiments, embodiments of the present application also provide a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the flow or function of any of the methods of the above embodiments.
In addition, the embodiment of the present application further provides a computer program product, where the computer program product stores computer program instructions, and the computer program instructions implement the flow or the function of any one of the methods in the above embodiments when the computer program instructions are executed by a processor.
The foregoing exemplarily describes the flow diagrams and/or block diagrams of methods, apparatus, systems, and computer program products according to embodiments of the present application, and describes various aspects related thereto. It will be understood that each block of the flowchart illustrations and/or block diagrams, or combinations of blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions, special purpose hardware which perform the specified functions or acts, and combinations of special purpose hardware and computer instructions. For example, these computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the present application, enable the implementation of the functions/acts specified in the flowchart and/or block diagram block or blocks. Such a processor may be a general purpose processor, a special purpose processor, an application specific processor, or a field programmable logic circuit.
Functional blocks shown in the block diagrams of the embodiments of the present application can be implemented in hardware, software, firmware, or a combination thereof. When implemented in hardware, it may be, for example, an electronic circuit, an Application Specific Integrated Circuit (ASIC), suitable firmware, a plug-in, a function card, or the like; when implemented in software, are the programs or code segments used to perform the required tasks. The program or code segments can be stored in a memory or transmitted over transmission media or communication links through data signals carried in carrier waves. The code segments may be downloaded via computer networks such as the internet, intranets, etc.
It should be noted that the present application is not limited to the specific configurations and processes described above or shown in the drawings. The foregoing is merely specific embodiments of the present application, and it will be clearly understood by those skilled in the art that, for convenience and brevity of description, specific working processes of the described system, apparatus, module or unit may refer to corresponding processes in the method embodiments, and need not be repeated. It should be understood that the scope of the present application is not limited thereto, and any person skilled in the art may conceive various equivalent modifications or substitutions within the technical scope of the present application, which are intended to be included in the scope of the present application.
Claims (10)
1. A geomagnetic ap exponential medium term forecasting method based on finite recursion and deep learning, characterized by comprising the following steps:
acquiring an Ap index prediction model, wherein the Ap index prediction model is generated after training a first neural network model and is based on a planetary equivalent daily amplitude Ap index, solar radiation electric flux F10.7, solar active area characteristics and a geomagnetic disturbance equivalent planetary amplitude Ap index historical value;
respectively obtaining predicted values of multiple steps of the Ap index, the F10.7 and the solar active area characteristics, wherein the predicted values of the multiple steps of the Ap index are generated by a preset Ap index predicted model, the predicted values of the multiple steps of the F10.7 are generated by a preset F10.7 predicted model, and the predicted values of the multiple steps of the solar active area characteristics are generated by a preset solar active area predicted model;
when the Ap index is predicted, the Ap index, F10.7, the solar active area characteristics and the history value of the Ap index are input into the Ap index prediction model to obtain the predicted value of the Ap index output by the model;
taking the predicted value of the Ap index as the input value of the Ap index of the next step length, and inputting the predicted value of the Ap index, F10.7 and the next step length of the solar active area characteristic, and the historical values of the Ap index, F10.7 and the solar active area characteristic into the Ap index prediction model to obtain a new predicted value of the Ap index output by the model;
And obtaining ap index predicted values of a plurality of step sizes through multiple rounds of iterative prediction.
2. The method of claim 1, wherein the first neural network model comprises a transducer model; the history value comprises history measured values of K step sizes, K is more than or equal to 16, and the value of a single step size L is more than or equal to 1 and less than or equal to 6.
3. The method of claim 1, wherein the pre-set Ap index prediction model comprises a long-short-term memory network LSTM model;
the training data of the Ap index prediction model comprises an Ap index, F10.7, solar active area characteristics and a history value of solar X rays for a plurality of continuous days, and the output value of the model is a single-day prediction value of the Ap index;
the Ap index prediction model includes three sub-models for predicting Ap indexes of a future first day, a future second day, and a future third day, respectively.
4. The method of claim 1, wherein the pre-set F10.7 predictive model comprises a long-short-term memory network LSTM model;
the training data of the F10.7 prediction model comprises historical values of F10.7 for a plurality of continuous days, and the output value of the model is a single-day prediction value of F10.7;
and predicting F10.7 of a plurality of step sizes in an iterative mode, and adding the F10.7 predicted value output by the LSTM model of the previous round to the input value of the next round until a preset number of F10.7 predicted values are obtained.
5. The method of claim 1, wherein the pre-set solar active area prediction model comprises a solar active area interpolation model;
the processing procedure of the solar active area interpolation model comprises the following steps:
acquiring influence factors of solar activity areas on the n-4 th day, the n-3 th day, the n-2 th day and the n-1 th day on the Ap index on the n-th day;
acquiring influence factors of solar activity areas on the n-3 th day, the n-2 th day and the n-1 th day on the Ap index on the n+1 th day;
acquiring influence factors of solar activity areas on the n-2 th day and the n-1 th day on the Ap index on the n+2 th day;
taking the obtained influence factors as characteristic values of the solar active area, and adding the characteristic values into an interpolation model;
the characteristic value of a solar active area in the interpolation model is complemented by a linear interpolation method to obtain AP index influence factors of the nth day to the (n+1) th day and the (n+1) th day to the (n+2) th day, wherein the missing value is taken as a value of the previous day or an average value of the previous m days;
and obtaining a solar active area interpolation model for generating a single-day predicted value of the solar active area characteristic.
6. The method of claim 1, wherein the plurality of step ap exponent predictors includes R step ap exponent predictors, R being ≡24.
7. The method according to claim 1, wherein when training the first neural network model, five halves are performed on training data, four of the halves are used as training data sets, one of the halves is used as a test data set, training is performed respectively, and five ap index prediction models are generated after training is completed; in the prediction, prediction is performed through five ap index prediction models respectively, and the average value of the obtained five prediction values is used as the prediction value of the ap index.
8. A geomagnetic ap exponential medium term forecasting device based on finite recursion and deep learning, characterized by comprising:
the first acquisition module is used for acquiring an Ap index prediction model, wherein the Ap index prediction model is generated after training a first neural network model and is based on a planetary equivalent daily amplitude Ap index, solar radio flux F10.7, solar active area characteristics and geomagnetic disturbance equivalent planetary amplitude Ap index historical values;
the second acquisition module is used for respectively acquiring the Ap index, the F10.7 and the predicted values of the solar active area characteristic in a plurality of steps, wherein the predicted values of the Ap index in a plurality of steps are generated by a preset Ap index prediction model, the predicted values of the F10.7 in a plurality of steps are generated by a preset F10.7 prediction model, and the predicted values of the solar active area characteristic in a plurality of steps are generated by a preset solar active area prediction model; and
The prediction processing module is used for inputting the Ap index, F10.7, the solar active area characteristics and the historical values of the Ap index into the Ap index prediction model to obtain the predicted values of the Ap index output by the model when predicting the Ap index; the method comprises the steps of obtaining a predicted value of an Ap index, and obtaining a predicted value of the Ap index, wherein the predicted value of the Ap index is used as an input value of the Ap index of the next step length, and the predicted value of the Ap index, F10.7 and the next step length of the solar active area characteristic, and the historical values of the Ap index, F10.7 and the solar active area characteristic are all input into the Ap index prediction model to obtain a new predicted value of the Ap index output by the model; and the system is used for obtaining ap index predicted values of a plurality of step sizes through multiple rounds of iterative prediction, and outputting a task planning result.
9. An electronic device, the electronic device comprising: a processor and a memory storing computer program instructions; the electronic device, when executing the computer program instructions, implements the method of any of claims 1-7.
10. A computer readable storage medium, characterized in that the computer storage medium has stored thereon computer program instructions which, when executed by a processor, implement the method according to any of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311140184.7A CN116882594B (en) | 2023-09-06 | 2023-09-06 | Geomagnetic ap index medium-term forecasting method based on limited recursion and deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311140184.7A CN116882594B (en) | 2023-09-06 | 2023-09-06 | Geomagnetic ap index medium-term forecasting method based on limited recursion and deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116882594A true CN116882594A (en) | 2023-10-13 |
| CN116882594B CN116882594B (en) | 2023-12-01 |
Family
ID=88260887
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311140184.7A Active CN116882594B (en) | 2023-09-06 | 2023-09-06 | Geomagnetic ap index medium-term forecasting method based on limited recursion and deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116882594B (en) |
Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5673366A (en) * | 1996-01-05 | 1997-09-30 | Maynard; Nelson C. | System and method for geomagnetic and ionospheric forecasting |
| US20060229813A1 (en) * | 2005-03-30 | 2006-10-12 | Tobiska William K | Ionospheric forecast system (IFS) |
| CN107563574A (en) * | 2017-09-29 | 2018-01-09 | 北京理工大学 | A kind of earth's magnetic field Forecasting Methodology based on long memory models Recognition with Recurrent Neural Network in short-term |
| KR102056719B1 (en) * | 2018-11-30 | 2019-12-17 | 한국 천문 연구원 | Prediction system and method for the disturbance storm time index using magnetic field from geosynchronous orbit |
| CN111797573A (en) * | 2020-07-13 | 2020-10-20 | 中国地震局地震预测研究所 | Ionized layer electron concentration total content time sequence prediction method based on deep learning technology |
| CN111967679A (en) * | 2020-08-20 | 2020-11-20 | 江苏师范大学 | Ionized layer total electron content forecasting method based on TCN model |
| CN113379107A (en) * | 2021-05-26 | 2021-09-10 | 江苏师范大学 | Regional ionized layer TEC forecasting method based on LSTM and GCN |
| CN114997474A (en) * | 2022-05-17 | 2022-09-02 | 桂林理工大学 | Short-term ionized layer TEC combined forecasting method |
| KR20220151246A (en) * | 2021-05-06 | 2022-11-15 | 대한민국(국방부 공군참모총장) | Learning method and learning device for predicting geomagnetic disturbance index by using extreme ultraviolet radiation video of sun, and testing method and testing device using the same |
| CN115951430A (en) * | 2023-03-15 | 2023-04-11 | 数字太空(北京)智能技术研究院有限公司 | Geomagnetic index forecasting method based on LSTM |
| CN116070784A (en) * | 2023-03-07 | 2023-05-05 | 数字太空(北京)智能技术研究院有限公司 | Method for extracting integral influence factor of solar active area on environmental index |
| CN116681153A (en) * | 2023-04-25 | 2023-09-01 | 哈尔滨工业大学（深圳） | TEC forecasting method based on LSTM neural network and GNSS historical observation data |
-
2023
- 2023-09-06 CN CN202311140184.7A patent/CN116882594B/en active Active
Patent Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5673366A (en) * | 1996-01-05 | 1997-09-30 | Maynard; Nelson C. | System and method for geomagnetic and ionospheric forecasting |
| US20060229813A1 (en) * | 2005-03-30 | 2006-10-12 | Tobiska William K | Ionospheric forecast system (IFS) |
| CN107563574A (en) * | 2017-09-29 | 2018-01-09 | 北京理工大学 | A kind of earth's magnetic field Forecasting Methodology based on long memory models Recognition with Recurrent Neural Network in short-term |
| KR102056719B1 (en) * | 2018-11-30 | 2019-12-17 | 한국 천문 연구원 | Prediction system and method for the disturbance storm time index using magnetic field from geosynchronous orbit |
| CN111797573A (en) * | 2020-07-13 | 2020-10-20 | 中国地震局地震预测研究所 | Ionized layer electron concentration total content time sequence prediction method based on deep learning technology |
| CN111967679A (en) * | 2020-08-20 | 2020-11-20 | 江苏师范大学 | Ionized layer total electron content forecasting method based on TCN model |
| KR20220151246A (en) * | 2021-05-06 | 2022-11-15 | 대한민국(국방부 공군참모총장) | Learning method and learning device for predicting geomagnetic disturbance index by using extreme ultraviolet radiation video of sun, and testing method and testing device using the same |
| CN113379107A (en) * | 2021-05-26 | 2021-09-10 | 江苏师范大学 | Regional ionized layer TEC forecasting method based on LSTM and GCN |
| CN114997474A (en) * | 2022-05-17 | 2022-09-02 | 桂林理工大学 | Short-term ionized layer TEC combined forecasting method |
| CN116070784A (en) * | 2023-03-07 | 2023-05-05 | 数字太空(北京)智能技术研究院有限公司 | Method for extracting integral influence factor of solar active area on environmental index |
| CN115951430A (en) * | 2023-03-15 | 2023-04-11 | 数字太空(北京)智能技术研究院有限公司 | Geomagnetic index forecasting method based on LSTM |
| CN116681153A (en) * | 2023-04-25 | 2023-09-01 | 哈尔滨工业大学（深圳） | TEC forecasting method based on LSTM neural network and GNSS historical observation data |
Non-Patent Citations (4)
| Title |
|---|
| 刘四清: "典型空间环境参数预报方法和模式研究典型空间环境参数预报方法和模式研究", 中国博士论文全文数据库 * |
| 常欣卓 等: "基于非线性自回归神经网络的局部大气密度预测方法", 中国科学技术大学学报, vol. 47, no. 12, pages 1015 - 1022 * |
| 张富彬 等: "基于深度学习的全球电离层TEC预测", 电波科学学报, vol. 36, no. 004, pages 553 - 561 * |
| 龚建村 等: "神经网络方法在太阳质子事件短期预报中的应用", 空间科学学报, no. 06, pages 443 - 451 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116882594B (en) | 2023-12-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN100524288C (en) | Space weather prediction system and method | |
| CN114547553B (en) | Inversion method, device and equipment for carbon dioxide emission and storage medium | |
| EP1611464B1 (en) | System and method for spectral corrected lightning detection | |
| CN112285807B (en) | Meteorological information prediction method and device | |
| CN115951430B (en) | Geomagnetic index forecasting method based on LSTM | |
| CN112616160A (en) | A kind of intelligent shortwave frequency cross-band real-time prediction method and system | |
| EP3441794B1 (en) | Device and method for obtaining local geomagnetic disturbances at mid-latitudes | |
| CN116882594B (en) | Geomagnetic ap index medium-term forecasting method based on limited recursion and deep learning | |
| CN117406242A (en) | Time transfer ionosphere error correction algorithm and system, electronic equipment and medium | |
| CN115062526B (en) | A three-dimensional ionospheric electron concentration distribution model training method based on deep learning | |
| Amariutei et al. | On the prediction of the auroral westward electrojet index | |
| CN113093225B (en) | Wide-area and local-area fused high-precision ionospheric scintillation model establishment method | |
| Sawadogo et al. | Koudougou (Burkina Faso, Africa), GPS‐TEC Response to Recurrent Geomagnetic Storms during Solar Cycle 24 Declining Phase | |
| US20230153699A1 (en) | Machine learning generated predictive model to forecast the dynamic flux distributions of ultra-relativistic electrons | |
| Paouris et al. | Statistical analysis on the current capability to predict the Ap Geomagnetic Index | |
| Fukuda et al. | Determining 70 percent probability-circle radii of tropical cyclone track forecasts with multiple ensembles | |
| Akir et al. | Preliminary vertical TEC prediction using neural network: Input data selection and preparation | |
| CN114089443A (en) | UHF frequency band ionosphere scintillation event forecasting method based on TEC integral quantity and seasonal change coefficient | |
| Ünal et al. | Performance of IRI-based ionospheric critical frequency calculations with reference to forecasting | |
| RU2630535C2 (en) | Method of short-term forecasting registration time for coronal mass ejection (cme) | |
| Ziangas et al. | Simulation Framework for Q/V-Band Excess Path Loss Forecasting in LEO Satellite Links Using Deep Learning | |
| CN119796532B (en) | Ultra-low orbit support method, system and equipment for near-Earth satellites | |
| CN119293455B (en) | Atmospheric density prediction method, device and equipment based on multi-model fusion | |
| CN120387682A (en) | Spacecraft space environment safety assurance system and method based on three-step calculation | |
| Hruska et al. | Solar-terrestrial predictions |
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 |