Method, system and product for predicting Kp index of future three days for coronal mass ejection event

Abstract

The invention belongs to the technical field of space physics and artificial intelligence interdiscipline and space environment forecasting, and particularly relates to a method, a system and a product for predicting Kp indexes of future three days for coronal mass ejection events. The method comprises the following steps: acquiring CME characteristic parameters, background solar wind characteristics and solar wind activity levels of a historical coronal mass ejection event, and Kp indexes within three days after the historical coronal mass ejection event reaches the earth, and constructing a data set; a prediction model is constructed based on an adjustment cosine similarity algorithm, and a feature weight combination of the prediction model is obtained through fitting by taking an average absolute error of time when coronal mass is projected to the earth as a loss function; and obtaining CME characteristic parameters and background solar wind characteristics of a to-be-predicted coronal mass ejection event, inputting the CME characteristic parameters and the background solar wind characteristics into the prediction model, and outputting a Kp index of the future three days corresponding to a historical event with the minimum distance from the to-be-predicted coronal mass ejection event so as to realize prediction.

Description

Technical Field

[0001] This invention belongs to the interdisciplinary field of space physics and artificial intelligence, and the field of space environment forecasting technology. In particular, it relates to a method, system and product for predicting the Kp index for the next three days in response to a coronal mass ejection event. Background Technology

[0002] Coronal mass ejections (CMEs), as the core solar activity triggering strong geomagnetic storms, can precisely quantify the intensity of their disturbances to the Earth's magnetic field using the Kp index. Therefore, research on predicting the Kp index for CMEs has crucial scientific value and practical significance. In practical applications, accurate Kp index prediction can provide early warnings for critical infrastructure such as power grids, satellites, aerospace, and communication navigation, effectively mitigating risks such as power grid paralysis, satellite radiation damage, and excessive radiation during high-altitude flight caused by geomagnetic induced currents (GIC), thus reducing the economic losses and social impacts of space weather disasters. From a space safety perspective, this research provides a scientific basis for radiation protection and spacecraft orbit adjustments for long-duration space missions (such as Mars exploration), ensuring the safety of astronauts and the successful implementation of missions.

[0003] Currently, Kp forecasts for periods longer than 3 days still rely on solar wind velocity extrapolation, with an accuracy rate of less than 50% for CME events. Most existing models assume a steady-state magnetosphere, neglecting the impact of initial conditions on the CME response—whether or not the CME's influence is considered in the model can significantly affect the predicted outcome. Current Kp forecasts have extremely low accuracy for CME-driven strong geomagnetic storms (Kp≥7). The triggering mechanism of strong geomagnetic storms is extremely complex, including the combined effects of CMEs and high-velocity coronal hole flows. Current Kp exponential forecasting models largely eliminate the CME effect, thus significantly increasing errors in practical forecasting applications.

[0004] Predicting the Kp index based on coronal mass ejections (CMEs) has always been a challenge in operational space environment forecasting. Current models only consider coronal holes or the 27-day recurrence relation of the Kp index, rarely taking CME factors into account. Currently, operational forecasts of the Kp index for CMEs largely rely on human experience, with very few models available for reference. Furthermore, these human-based forecasts require extensive reference to similar historical events, combined with the characteristics of the current event, to arrive at the final forecast result. Therefore, constructing a model for predicting the Kp index based on distance similarity can provide significant reference value for forecasters in operational forecasting. Compared to general forecasting models, this model not only predicts the Kp index for CME events but also outputs historically similar CME events, offering crucial reference value for operational forecasting. Summary of the Invention

[0005] There are few studies on predicting the Kp index by considering CME factors in existing technologies. The purpose of this invention is to overcome the shortcomings of existing technologies and propose a method, system and product for predicting the Kp index for the next 3 days based on coronal mass ejection events. It comprehensively considers the two key sources of information that cause geomagnetic storms—coronal mass ejection and coronal holes (solar wind information)—to construct a Kp index model for the next 3 days, providing a reference for current space weather forecasting services.

[0006] In view of this, the present invention proposes a method for predicting the Kp index for the next three days in response to a coronal mass ejection event, comprising: We acquire CME characteristic parameters, background solar wind characteristics, solar wind activity levels, and Kp index within three days after the historical coronal mass ejection event reaches Earth to construct a dataset. Based on the adjusted cosine similarity algorithm, a prediction model is constructed, and the average absolute error of the time when the coronal mass ejection reaches the Earth is used as the loss function. The feature weight combination of the prediction model is obtained by fitting. The system obtains the CME characteristic parameters and background solar wind characteristics of the coronal mass ejection (CME) event to be predicted, inputs them into the prediction model, and outputs the Kp index of the next three days corresponding to the historical event that is closest to the CME event to be predicted, thus achieving prediction.

[0007] As an improvement to the above method, the CME characteristic parameters include: center position angle, center mass maximum position angle, angular width, and linear velocity; the background solar wind characteristics include: solar wind average velocity, temperature, proton density, southward component of interplanetary magnetic field, and total magnetic field; the solar wind activity level is an index characterizing the solar activity level: the F10.7 index.

[0008] As an improvement to the above method, the dataset is constructed with all parameters of CME feature parameters, background solar wind features, and solar wind activity level, and labeled with the Kp index within three days after the historical coronal mass ejection event reached Earth.

[0009] As an improvement to the above method, the adjusted cosine similarity algorithm is used to eliminate observation bias interference and improve the matching accuracy of similar events. The similarity is obtained according to the following method. : Let the feature vector of the event to be predicted be... The feature vector of historical events is ,in, Let be the observed value of the i-th feature. The historical observation value of the i-th feature; Let be the mean of the feature vector of the event to be predicted. The mean of the feature vectors of historical events is denoted by n, which represents the total number of features (10). The formula for calculating the adjusted cosine similarity is: .

[0010] As an improvement to the above method, the feature weight combination of the prediction model includes: the weight of each dimension corresponding to the 10-dimensional feature vector.

[0011] As an improvement to the above method, the method further includes: outputting one or more historical event information that is most similar to the coronal mass ejection event to be predicted.

[0012] On the other hand, the present invention provides a system for predicting the Kp index for the next three days based on a coronal mass ejection event, implemented using the above method, the system comprising: The dataset construction module is used to obtain the CME characteristic parameters, background solar wind characteristics, solar wind activity level, and Kp index of historical coronal mass ejection events within three days after the event reaches Earth, and to construct the dataset. The model training module is used to build a prediction model based on the adjusted cosine similarity algorithm. The average absolute error of the time when the coronal mass ejection reaches Earth is used as the loss function, and the feature weight combination of the prediction model is obtained by fitting. The prediction output module is used to obtain the CME characteristic parameters and background solar wind characteristics of the coronal mass ejection event to be predicted, input them into the prediction model, and output the Kp index of the next three days corresponding to the historical event with the smallest distance from the coronal mass ejection event to be predicted, thus realizing the prediction.

[0013] Thirdly, the present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0014] Compared with the prior art, the advantages of the present invention are: 1. This invention overcomes the deficiency of current Kp value prediction not incorporating CME features, enabling the model to simultaneously include solar eruption and background solar wind information during construction.

[0015] 2. Significantly improved the predictive ability of geomagnetic storm events (especially large geomagnetic storms with Kp≥7). In historical cases, this method successfully captured the severe geomagnetic storm with Kp=9 on May 12, 2024, confirming the feasibility of the model in predicting the Kp index for CME.

[0016] 3. Since the average velocity of CME during solar maximum is generally higher than that during solar minimum, traditional cosine similarity is prone to misinterpreting this periodic amplitude difference as a difference in characteristic direction. The adjusted cosine similarity adopted in this invention can remove the periodic baseline bias through decentralization and focus on the relative distribution of CME's own characteristics (such as the difference between the patterns of "high speed - narrow angle width" and "low speed - wide angle width").

[0017] 4. To address the characteristics such as angular width and center position angle, which are easily affected by observation conditions, adjusting the cosine similarity can more accurately match historical CME events with the same physical evolution laws, thereby improving the prediction accuracy of Earth effects (such as the geomagnetic storm intensity Kp index). Attached Figure Description

[0018] Figure 1 This is a flowchart of a method for predicting the Kp index for the next 3 days based on a coronal mass ejection event; Figure 2 This is a technical roadmap of the present invention. Detailed Implementation

[0019] The key point of this invention is to rationally correlate and fuse CME features with background solar wind parameters, and apply distance similarity to predict the Kp index. For example... Figure 1 The following is a flowchart of the method. Figure 2 This is a technology roadmap.

[0020] 1. Data The data used in this application are derived from all events causing geomagnetic disturbances from 1996 to May 2024. The features employed include three categories: CME features, including the central position angle (CPA), central mass maximum position angle (MPA), angular width, and linear velocity; background solar wind features, including the solar wind mean velocity, temperature, proton density, southward component of the interplanetary magnetic field, and total magnetic field; and the F10.7 index, which characterizes the level of solar activity (the flux of optical radiation emitted by the solar atmosphere at a wavelength of 10.7 cm). These features are correlated with the Kp index within three days of the event's arrival on Earth. Compared to traditional Kp index prediction models, this model's input not only includes background solar wind conditions but also more fully considers the influence of CME.

[0021] 2. Model Building a. For each CME event, establish a corresponding feature bar. Each feature bar includes the dynamic characteristics of the CME, as well as the background solar wind parameters on the day of the CME outbreak and the F10.7 index on that day, for a total of 10-dimensional feature vectors; b. Using the mean absolute error of CME time to Earth as the loss function, find the most suitable combination of feature weights; Based on the principle of similarity, the distance between sample events is calculated, and the Kp value of the historical event with the smallest distance in the next 3 days is used as the output.

[0022] Let the feature vector of the CME event to be predicted be... The feature vector of historical CME events is ,in These are the observed values ​​of the i-th feature (e.g., the first dimension is angular width, the second dimension is velocity, etc.). For the i-th feature, the historical observation value is; Let be the mean of the feature vector of the event to be predicted. This represents the mean of the feature vectors of historical events. Compared to traditional cosine similarity, which only measures the directional consistency of feature vectors but suffers from systematic bias in actual observations, this adjusted cosine similarity eliminates baseline bias through decentralization, thus more accurately capturing the directional correlation of feature vectors.

[0023] The formula for calculating the adjusted cosine similarity of CME feature vectors is:

[0024] This formula can: Eliminating observational bias interference: For example, through numerous experiments, we have found that the average velocity of the CME during solar maxima is generally higher than that during solar minima. Traditional cosine similarity tends to misjudge this periodic amplitude difference as a difference in characteristic direction. Adjusting cosine similarity through decentralization can remove periodic baseline bias and focus on the relative distribution of CME's own characteristics (such as the pattern difference between "high speed - narrow angle width" and "low speed - wide angle width").

[0025] Improving the accuracy of matching similar events: Adjusting the cosine similarity can more accurately match historical CME events with the same physical evolution patterns, targeting features such as angular width and center position angle that are easily affected by observation conditions, thereby improving the prediction accuracy of Earth effects (such as the Kp index of geomagnetic storm intensity).

[0026] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and embodiments.

[0027] Example 1 To build a model for predicting the Kp index for coronal mass ejection (CME) events, the following steps were taken: First, a feature bar was constructed based on historical CME events, including CME characteristic parameters, solar wind characteristic parameters, and F10.7, comprising a total of 10 parameters. Second, a model was built based on a similarity algorithm to perform event matching. In the experiment, the weight of each parameter was fitted based on a loss function, and finally, the Kp index for the next 3 days was output.

[0028] This model addresses the limitation of current Kp index predictions that do not incorporate CME features. It optimizes the model by simultaneously incorporating two core types of information—solar eruptions and background solar wind—during its construction, providing crucial support for improving the accuracy of Kp forecasts under complex space weather events. The results of this model are compared with other models, as shown in Table 1. The model using this invention outperforms the vast majority of models.

[0029] Table 1 Comparison of Model Mean Absolute Error

[0030] Therefore, this model largely meets the requirements for richness of inputs while ensuring prediction accuracy, and it can output historical similar events, which has special reference significance for current operational space weather forecasting.

[0031] Example 2 Embodiment 2 of the present invention provides a system for predicting the Kp index for the next three days based on a coronal mass ejection event. The system is implemented using the method described above and includes: The dataset construction module is used to obtain the CME characteristic parameters, background solar wind characteristics, solar wind activity level, and Kp index of historical coronal mass ejection events within three days after the event reaches Earth, and to construct the dataset. The model training module is used to build a prediction model based on the adjusted cosine similarity algorithm. The average absolute error of the time when the coronal mass ejection reaches Earth is used as the loss function, and the feature weight combination of the prediction model is obtained by fitting. The prediction output module is used to obtain the CME characteristic parameters and background solar wind characteristics of the coronal mass ejection event to be predicted, input them into the prediction model, and output the Kp index of the next three days corresponding to the historical event with the smallest distance from the coronal mass ejection event to be predicted, thus realizing the prediction.

[0032] It is worth noting that in the embodiments of the above system, the modules included are divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional module are only for easy differentiation and are not used to limit the scope of protection of the present invention.

[0033] Example 3 Embodiments of the present invention also provide a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0034] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to the embodiments, those skilled in the art should understand that modifications or equivalent substitutions to the technical solutions of the present invention do not depart from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for predicting the Kp index for the next three days based on a coronal mass ejection event, comprising: We acquire CME characteristic parameters, background solar wind characteristics, solar wind activity levels, and Kp index within three days after the historical coronal mass ejection event reaches Earth to construct a dataset. Based on the adjusted cosine similarity algorithm, a prediction model is constructed, and the average absolute error of the time when the coronal mass ejection reaches the Earth is used as the loss function. The feature weight combination of the prediction model is obtained by fitting. The system obtains the CME characteristic parameters and background solar wind characteristics of the coronal mass ejection (CME) event to be predicted, inputs them into the prediction model, and outputs the Kp index of the next three days corresponding to the historical event that is closest to the CME event to be predicted, thus achieving prediction.

2. The method for predicting the Kp index for the next three days based on a coronal mass ejection event according to claim 1, characterized in that, The CME characteristic parameters include: center position angle, center mass maximum position angle, angular width, and linear velocity; the background solar wind characteristics include: solar wind mean velocity, temperature, proton density, southward component of interplanetary magnetic field, and total magnetic field; the solar wind activity level is an index characterizing the solar activity level: the F10.7 index.

3. The method for predicting the Kp index for the next three days based on a coronal mass ejection event according to claim 2, characterized in that, The dataset is constructed using all parameters of CME feature parameters, background solar wind features, and solar wind activity levels, and labeled with the Kp index within three days after the historical coronal mass ejection event reached Earth.

4. The method for predicting the Kp index for the next three days based on a coronal mass ejection event according to claim 2, characterized in that, The adjusted cosine similarity algorithm is used to eliminate observation bias interference and improve the matching accuracy of similar events. The similarity is obtained according to the following method. : Let the feature vector of the event to be predicted be... The feature vector of historical events is ,in, Let be the observed value of the i-th feature. The historical observation value of the i-th feature; Let be the mean of the feature vector of the event to be predicted. The mean of the feature vectors of historical events is denoted by n, which represents the total number of features (10). The formula for calculating the adjusted cosine similarity is: 。 5. The method for predicting the Kp index for the next three days based on a coronal mass ejection event according to claim 2, characterized in that, The feature weight combination of the prediction model includes the weight of each dimension corresponding to the 10-dimensional feature vector.

6. The method for predicting the Kp index for the next three days based on a coronal mass ejection event according to claim 1, characterized in that, The method also includes: outputting one or more historical event information that is most similar to the coronal mass ejection event to be predicted.

7. A system for predicting the Kp index for the next three days based on a coronal mass ejection event, implemented based on the method described in any one of claims 1-6, characterized in that, The system includes: The dataset construction module is used to obtain the CME characteristic parameters, background solar wind characteristics, solar wind activity level, and Kp index of historical coronal mass ejection events within three days after the event reaches Earth, and to construct the dataset. The model training module is used to build a prediction model based on the adjusted cosine similarity algorithm. It uses the mean absolute error of the arrival time of coronal mass ejections (CMEs) on Earth as the loss function and obtains the feature weight combination of the prediction model through fitting. The prediction output module is used to obtain the CME characteristic parameters and background solar wind characteristics of the coronal mass ejection event to be predicted, input them into the prediction model, and output the Kp index of the next three days corresponding to the historical event with the smallest distance from the coronal mass ejection event to be predicted, thus realizing the prediction.

8. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1-6.

Images