CN-121997017-A - Method, system and product for predicting Kp index of three days in future aiming at coronal mass ejection event

CN121997017ACN 121997017 ACN121997017 ACN 121997017ACN-121997017-A

Abstract

The invention belongs to the technical field of space physics and artificial intelligence intersection disciplines and space environment prediction, and particularly relates to a method, a system and a product for predicting Kp indexes of three days in the future aiming at coronal mass projection events. The method comprises the steps of obtaining CME characteristic parameters, background solar wind characteristics and solar wind activity levels of historical coronal material projection events, constructing a data set by the aid of Kp indexes of the historical coronal material projection events within three days after the earth, constructing a prediction model based on an adjustment cosine similarity algorithm, obtaining characteristic weight combination of the prediction model by taking average absolute errors of arrival time of coronal material projection as a loss function through fitting, obtaining the CME characteristic parameters and the background solar wind characteristics of the coronal material projection events to be predicted, inputting the prediction model, and outputting the Kp indexes of three days in the future corresponding to the historical events with the smallest coronal material projection event distance to be predicted.

Inventors

SHI YURONG
LUO BINGXIAN
WANG JINGJING
CHEN YANHONG
LIU SIQING

Assignees

中国科学院国家空间科学中心

Dates

Publication Date: 20260508
Application Date: 20251224

Claims (8)

1. A method of predicting a three day future Kp index for coronal mass ejection events, comprising: acquiring CME characteristic parameters, background solar wind characteristics and solar wind activity levels of a historical coronal mass casting event and Kp indexes of the historical coronal mass casting event within three days after the earth, and constructing a data set; based on an adjustment cosine similarity algorithm, constructing a prediction model, taking the average absolute error of the time when the corona material is projected to reach the earth as a loss function, and fitting to obtain a characteristic weight combination of the prediction model; And acquiring CME characteristic parameters and background solar wind characteristics of the coronal mass casting event to be predicted, inputting a prediction model, and outputting a Kp index of three days in the future corresponding to a historical event with the minimum coronal mass casting event distance to be predicted, so as to realize the prediction.
2. The method of claim 1, wherein the CME characteristic parameters include a central position angle, a central mass maximum position angle, an angular width, and a linear velocity, the background solar wind characteristics include a solar wind average velocity, a temperature, a proton density, a interplanetary magnetic field southbound component, and a total magnetic field, and the solar wind activity level is an index representing a solar activity level, F10.7 index.
3. The method of claim 2, wherein the dataset constructs a 10-dimensional feature vector with all parameters of CME feature parameters, background solar wind features, and solar wind activity level, labeled with the Kp index of the historical coronal mass cast event for the three days after earth.
4. The method for predicting a three-day-in-future Kp index for coronal mass ejection events according to claim 2, wherein the cosine similarity adjustment algorithm is used for eliminating observation deviation interference, improving similarity event matching accuracy, and obtaining similarity according to the following method : Let the feature vector of the event to be predicted be The feature vector of the historical event is , wherein, As an observation of the i-th feature, Historical observations of the ith feature; As the mean value of the feature vectors of the event to be predicted, N represents the total number of features 10, which is the average value of the feature vectors of the historical events; the adjustment cosine similarity calculation formula is: 。
5. The method of claim 2, wherein the combination of feature weights of the predictive model includes weights for each dimension of a 10-dimensional feature vector.
6. The method of claim 1, further comprising outputting one or more pieces of historical event information most similar to the coronal mass casting event to be predicted.
7. A system for predicting the Kp index for three days in the future for coronal mass ejection events, implemented on the basis of the method of any one of claims 1-6, the system comprising: the data set construction module is used for acquiring CME characteristic parameters, background solar wind characteristics and solar wind activity levels of the historical coronal mass projection event and Kp indexes of the historical coronal mass projection event within three days after the earth to construct a data set; The model training module is used for constructing a prediction model based on an adjustment cosine similarity algorithm, taking the average absolute error of the arrival earth time of the coronal mass projection as a loss function, obtaining a characteristic weight combination of the prediction model through fitting, and The prediction output module is used for acquiring CME characteristic parameters and background solar wind characteristics of the coronal mass casting event to be predicted, inputting a prediction model, and outputting a future three-day Kp index corresponding to a historical event with the smallest coronal mass casting event distance to be predicted, so as to realize prediction.
8. A computer program product comprising a computer program which, when executed by a processor, implements the method according to any one of claims 1-6.

Description

Method, system and product for predicting Kp index of three days in future aiming at coronal mass ejection event Technical Field The invention belongs to the technical field of space physics and artificial intelligence intersection disciplines and space environment prediction, and particularly relates to a method, a system and a product for predicting Kp indexes of three days in the future aiming at coronal mass projection events. Background Coronal mass projection (Coronal mass ejections, CMEs) is used as a core solar activity for inducing a strong geomagnetic storm, and the disturbance intensity of the coronal mass projection on the earth magnetic field can be precisely quantified through Kp indexes, so that the research on predicting the Kp indexes by CME is carried out, and the method has key scientific value and practical significance. From practical application, accurate Kp index prediction can provide early warning for key infrastructures such as power grid, satellite, aerospace, communication navigation and the like, effectively avoid risks such as power grid paralysis, satellite radiation damage, high-altitude flight radiation superscale and the like caused by Geomagnetic Induction Current (GIC), and reduce economic loss and social influence caused by space weather disasters. From the space safety dimension, the research provides scientific basis for radiation protection of long-term aerospace tasks (such as Mars detection) and spacecraft orbit adjustment, and ensures life safety and smooth implementation of tasks of astronauts. The Kp forecast over 3 days still depends on the extrapolation of the solar wind speed, and the prediction accuracy of CME events is less than 50%. Most of the existing models assume that the magnetic layer is in a steady state, neglecting the effect of the initial state on the CME response-whether there may be a significant difference in the model in terms of the effect of the CME. The current Kp prediction has extremely low accuracy in predicting the CME-driven strong magnetic storm (Kp is more than or equal to 7). The triggering mechanism of strong magnetic storm is extremely complex, comprising the influence of superposition of CME and crown high-speed flow, and the effect of CME is basically eliminated by the current Kp index prediction model, so that the error is obviously increased in practical prediction application. In the actual space environment forecasting business, predicting Kp index based on coronal mass projection is always one of difficulties. The current correlation model only considers crown or Kp index 27 days reproducibility factors, and rarely considers CME factors. At present, most of Kp indexes for coronal mass projection prediction in service prediction depend on manual experience, a referenceable model is very little, and the manual experience prediction needs to fully refer to historical similar events, and then a final prediction result is given by combining the characteristics of the current events. Therefore, the coronal mass projection prediction Kp index model is constructed based on the distance similarity, and important reference value can be provided for a predictor in the actual prediction business. Compared with a general prediction model, the model not only predicts Kp indexes aiming at CME events, but also can output historically similar coronal mass ejection events, and has important reference value for real business prediction. Disclosure of Invention Aiming at the defect of the prior art, the invention provides a method, a system and a product for predicting a Kp index of 3 days in the future aiming at coronal mass casting events, which take CME factors into consideration in the prior art, comprehensively consider two key source information of magnetic storm initiation, namely coronal mass casting and crown (solar wind information), and construct a Kp index model of 3 days in the future, so as to provide reference for the current space weather forecast business. In view of this, the present invention proposes a method for predicting the Kp index for three days in the future for coronal mass ejection events, comprising: acquiring CME characteristic parameters, background solar wind characteristics and solar wind activity levels of a historical coronal mass casting event and Kp indexes of the historical coronal mass casting event within three days after the earth, and constructing a data set; based on an adjustment cosine similarity algorithm, constructing a prediction model, taking the average absolute error of the time when the corona material is projected to reach the earth as a loss function, and fitting to obtain a characteristic weight combination of the prediction model; And acquiring CME characteristic parameters and background solar wind characteristics of the coronal mass casting event to be predicted, inputting a prediction model, and outputting a Kp index of three days in the future corresponding to a historical event with