Magnetic field disturbance analysis methods, equipment, and storage media based on ground observation data

Abstract

This application discloses a method, device, and storage medium for analyzing magnetic field disturbances based on ground observation data, relating to the field of magnetic field prediction. The method includes: acquiring ground observation data; determining a target model; predicting first magnetic disturbance prediction data for each station using the target model; and determining a first prediction error of the target model based on the first magnetic disturbance prediction data and station observation data corresponding to each station; for each station, designating it as a rejection station, removing its station observation data from the ground observation data to obtain rejection data, and determining a verification model; for each verification model, predicting second magnetic disturbance prediction data using the verification model, and determining a second prediction error for the rejection station based on the second magnetic disturbance prediction data and the station observation data of the rejection station; and determining the prediction reliability of the target model based on the first prediction error and each of the second prediction errors. This application solves the problem of how to determine the reliability of geomagnetic disturbance predictions.

Description

Technical Field

[0001] This application relates to the field of magnetic field prediction technology, and in particular to a method, device and storage medium for analyzing magnetic field disturbances based on ground observation data. Background Technology

[0002] Geomagnetic disturbances are increasingly impacting space weather forecasting, power grid security, spacecraft navigation, and polar communications. Accurate forecasting of geomagnetic disturbances is crucial for implementing necessary protective measures and ensuring the stable operation of infrastructure. While it is currently possible to predict magnetic field disturbances at specific points, in practical applications, users, although able to obtain predicted values, often lack knowledge of the reliability of these predictions or rely solely on personal experience to estimate their reliability. Consequently, it is difficult to formulate scientifically sound decisions and response strategies based on the prediction results. For example, over-reliance on a highly uncertain prediction may lead to unnecessary warnings and wasted resources; conversely, underestimating the risks indicated by a highly reliable prediction could result in serious consequences. Therefore, determining the reliability of geomagnetic disturbance predictions is a pressing technical problem that needs to be solved.

[0003] The above content is only used to help understand the technical solutions of the embodiments of this application, and does not represent an admission that the above content is prior art. Summary of the Invention

[0004] The main objective of this application is to provide a method, device, and storage medium for analyzing magnetic field disturbances based on ground observation data, aiming to solve the problem of how to determine the reliability of geomagnetic disturbance predictions.

[0005] To achieve the above objectives, embodiments of this application provide a method for analyzing magnetic field disturbances based on ground observation data. The method includes:

[0006] Acquire ground observation data, wherein the ground observation data includes station observation data corresponding to multiple stations respectively;

[0007] A target model for analyzing magnetic field disturbances is determined based on the fitting of the ground observation data. The first magnetic disturbance prediction data of each station is predicted by the target model. The first prediction error of the target model is determined by combining the first magnetic disturbance prediction data and the station observation data corresponding to each station.

[0008] For each of the stations, the station is designated as a rejection station, and the station observation data of the rejection station is removed from the ground observation data to obtain the rejection observation data. A verification model for the rejection station, which is fitted based on the rejection observation data, is then determined for analyzing magnetic field disturbances.

[0009] For each of the aforementioned verification models, the second magnetic disturbance prediction data of the excluded station is predicted through the verification model, and the second prediction error of the excluded station is determined based on the second magnetic disturbance prediction data and the station observation data of the excluded station.

[0010] The prediction reliability of the target model is determined based on the first prediction error and each of the second prediction errors.

[0011] In one embodiment, a list of observation stations and observation periods are obtained;

[0012] For each station in the observation station list, based on the station type, the data acquisition component corresponding to the station type is invoked to acquire the initial observation data of the station during the observation period. The initial observation data includes sub-observation data corresponding to multiple observation times within the observation period. For each initial observation data, the initial observation data is divided into multiple window observation data corresponding to multiple time windows according to a preset duration unit. Target sub-observation data satisfying a preset median condition is determined within each window observation data, and these target sub-observation data are sequentially concatenated according to a preset time order to obtain standard observation data. For each station, the station observation data is determined based on the station's standard observation data. The preset median condition includes the target sub-observation data being the median within the window observation data.

[0013] In one embodiment, for each station, the step of determining the station observation data based on the station's standard observation data includes: for each station's standard observation data, removing abnormal target sub-observation data from the standard observation data, and if a first target time window that does not belong to a preset time alignment range is detected in the standard observation data, removing the target sub-observation data of the first target time window from the standard observation data;

[0014] If there is no target sub-observation data for the second target time window in the standard observation data, obtain the first adjacent window and the second adjacent window that are adjacent to the second target time window; wherein, the second target time window is any preset observation window within a preset time alignment range;

[0015] Linear interpolation is performed on the target sub-observation data of the first adjacent window and the target sub-observation data of the second adjacent window to obtain the target sub-observation data of the second target time window.

[0016] The standard observation data that does not include abnormal target sub-observation data, does not include target sub-observation data of the first target time window, and has target sub-observation data of each preset observation window within the preset time alignment range is used as the station observation data.

[0017] In one embodiment, the target model includes at least one target sub-model corresponding to a preset observation window, the target sub-model being used to analyze the magnetic field disturbances generated within the preset observation window;

[0018] The step of determining the target model for analyzing magnetic field disturbances based on the fitted ground observation data includes:

[0019] Target sub-observation data for each station corresponding to the same preset observation window are obtained from the ground observation data; based on each target sub-observation data and its corresponding observation coordinates, an observation matrix for the preset observation window is constructed; based on the preset current singularity grid and the observation matrix, the equivalent current coefficient of each singularity in the current singularity grid is deduced, and the target sub-model of the preset observation window is obtained by constructing the equivalent current coefficients in the singularity grid.

[0020] In one embodiment, the step of deriving the equivalent current coefficient of each singular point in the current singular point grid based on a preset current singular point grid and the observation matrix includes: correcting the observation matrix based on an error matrix constructed from preset observation error weights of each station to obtain a weighted matrix; calculating the initial magnetic disturbance prediction data of each station under the current singular point grid based on the preset initial equivalent coefficient of each singular point in the current singular point grid; constructing an initial prediction matrix based on the initial magnetic disturbance prediction data and observation coordinates corresponding to each station; constructing a final optimization problem corresponding to the equivalent current coefficient of each singular point in the current singular point grid based on the weighted matrix, the initial prediction matrix, and preset regularization parameters; and deriving the equivalent current coefficient of each singular point by performing a preset matrix decomposition on the final optimization problem.

[0021] In one embodiment, the first magnetic disturbance prediction data includes: first window prediction data of the station obtained by the target sub-model; the station observation data includes target sub-observation data with a preset observation window; the step of jointly determining the first prediction error of the target model based on the first magnetic disturbance prediction data and the station observation data corresponding to each of the stations includes:

[0022] For each station, the difference between the first window prediction data and the target sub-observation data within the same preset observation window is calculated to obtain the first window error; the first window errors are collectively used as the first prediction error.

[0023] In one embodiment, the verification model includes at least one verification sub-model with a preset observation window, the verification sub-model being used to analyze magnetic field disturbances generated within the preset observation window; the second magnetic disturbance prediction data includes: second window prediction data of the eliminated stations predicted by the verification sub-model; the station observation data includes target sub-observation data of the preset observation window;

[0024] The step of determining the second prediction error of the eliminated station based on the second magnetic disturbance prediction data and the station observation data of the eliminated station includes: for each preset observation window, calculating the difference between the second window prediction data and the target sub-observation data of the eliminated station within the same preset observation window to obtain the second window error of the preset observation window; and taking all the second window errors as the second prediction error.

[0025] In one embodiment, the prediction confidence includes station confidence; the step of determining the prediction confidence of the target model based on the first prediction error and each of the second prediction errors includes:

[0026] For each of the eliminated stations, the second prediction error is determined by comparing the second window error within the second prediction error with the first window error within the first prediction error within the same preset observation window, thus obtaining the verification error. For each eliminated station, the station reliability is determined based on the verification error of the eliminated station, wherein the station reliability is negatively correlated with the verification error.

[0027] Furthermore, to achieve the above objectives, this application also provides a magnetic field disturbance analysis device based on ground observation data. The magnetic field disturbance analysis device based on ground observation data includes: a memory, a processor, and a program for the magnetic field disturbance analysis method based on ground observation data stored in the memory and executable on the processor. When the program for the magnetic field disturbance analysis method based on ground observation data is executed by the processor, it can implement the steps of the magnetic field disturbance analysis method based on ground observation data as described above.

[0028] Furthermore, to achieve the above objectives, embodiments of this application also provide a computer-readable storage medium storing a program for implementing a magnetic field disturbance analysis method based on ground observation data. When the program for the magnetic field disturbance analysis method based on ground observation data is executed by a processor, it implements the steps of the magnetic field disturbance analysis method based on ground observation data as described above.

[0029] In addition, to achieve the above objectives, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the magnetic field disturbance analysis method based on ground observation data as described above.

[0030] The one or more technical solutions proposed in this application have at least the following technical effects: This embodiment can acquire ground observation data, which includes observation data from multiple stations. This embodiment can obtain a target model for magnetic field disturbance analysis by fitting the observation data from multiple stations. Then, it can predict the first magnetic disturbance prediction data for each station based on the target model. Based on the first magnetic disturbance prediction data and the station observation data, the first prediction error of the target model is determined. That is, this embodiment can determine the first prediction error of the target model by comparing the predicted value and the actual value of the magnetic field disturbance. Since the target model is obtained by fitting the observation data from each station, the first magnetic disturbance prediction data predicted by the target model for each station will naturally be close to the station observation data. Therefore, even if the first prediction error of the target model is determined, the reliability of the target model's prediction cannot be accurately explained. Therefore, in this embodiment, for each station, the station is designated as a rejection station, and the station observation data of the rejection station is removed from the ground observation data to obtain the observation rejection data. Then, a verification model can be fitted based on the observation rejection data. The verification model is also used for magnetic field disturbance analysis. Each station can have its own corresponding verification model. Then, for each verification model, the second magnetic disturbance prediction data of the rejection station is predicted through the verification model, thereby determining the second prediction error of the rejection station.

[0031] Since the validation model is based on fitting observational data after data removal, the fitting process lacks observational data from the removed stations. Therefore, the second geomagnetic disturbance prediction data for the removed stations predicted by the validation model will not be affected by the station observational data. This makes it easier to objectively evaluate the accuracy of the target model's geomagnetic disturbance prediction at unknown geographical locations through the second prediction error. However, since the second prediction error only reflects the geomagnetic disturbance prediction performance after the lack of observational data for a single station, the reliability of the target model's prediction cannot be accurately reflected solely by the second prediction errors. This is because if the prediction error of a removed station differs significantly from the prediction error of the validation model, it indicates that the target model may rely too heavily on known data. The stability of the target model's prediction of magnetic disturbance data may be poor, leading to poor prediction reliability. Therefore, it is necessary to determine the prediction stability of the target model based on the first prediction error and each second prediction error. Since the first prediction error is jointly determined by the first magnetic disturbance prediction data and the station's observation data respectively, and each station is considered as a rejected station with a second prediction error, it is convenient to compare the first prediction error and each second prediction error to determine the prediction stability and obtain the prediction reliability. This quantifies the reliability of the target model, allowing users to make corresponding decisions based on the prediction reliability and the magnetic disturbance prediction data obtained from the target model. Attached Figure Description

[0032] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with those described herein and, together with the specification, serve to explain the principles of those embodiments.

[0033] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0034] Figure 1 This is a flowchart illustrating one embodiment of the magnetic field disturbance analysis method based on ground observation data according to this application.

[0035] Figure 2 This is a schematic diagram comparing the predicted and observed values ​​of magnetic field disturbances in the magnetic field disturbance analysis method based on ground observation data in an embodiment of this application.

[0036] Figure 3 This is a schematic diagram of the functional modules in the magnetic field disturbance analysis method based on ground observation data in an embodiment of this application;

[0037] Figure 4 This is a schematic diagram of the process of removing station observation data station by station in the magnetic field disturbance analysis method based on ground observation data in the embodiments of this application;

[0038] Figure 5 This is a schematic diagram of the equipment structure of the hardware operating environment involved in the magnetic field disturbance analysis method based on ground observation data in the embodiments of this application.

[0039] The objectives, features, and advantages of the embodiments described in this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0040] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of the embodiments of this application and are not intended to limit the embodiments of this application.

[0041] To better understand the technical solutions of the embodiments of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0042] With the increasing significance of geomagnetic disturbances in fields such as space weather forecasting, power grid security, spacecraft navigation, and polar communications, accurate forecasting of geomagnetic disturbances is crucial for taking necessary protective measures and ensuring the stable operation of infrastructure. While it is currently possible to predict magnetic field disturbances at specific points, in practical applications, users, although able to obtain predicted values, often lack knowledge of the reliability of these predictions or rely solely on personal experience to estimate their reliability. Consequently, it is difficult to formulate scientifically sound decisions and response strategies based on the prediction results. For example, over-reliance on a highly uncertain prediction may lead to unnecessary warnings and wasted resources; conversely, underestimating the risks indicated by a highly reliable prediction could result in serious consequences. Therefore, determining the reliability of geomagnetic disturbance predictions is a pressing technical problem that needs to be solved.

[0043] In addition, in the current field of geomagnetic disturbance analysis, the available data is limited, or the various observation data are incompatible, requiring manual adjustment of the observation data, which is prone to errors and thus leads to low accuracy of geomagnetic disturbance analysis.

[0044] Therefore, this application provides a method for magnetic field disturbance analysis based on ground observation data. This method acquires ground observation data, including data from multiple stations. A target model for magnetic field disturbance analysis is obtained by fitting the observation data from multiple stations. Then, the first predicted magnetic disturbance data for each station is predicted based on the target model. The first prediction error of the target model is determined based on the first predicted magnetic disturbance data and the station observation data. That is, this method determines the first prediction error of the target model by comparing the predicted and actual values ​​of the magnetic field disturbance. Since the target model is obtained by fitting the observation data from each station, the first predicted magnetic disturbance data for each station will naturally be close to the actual station observation data. Therefore, even if the first prediction error of the target model is determined, the reliability of the target model's prediction cannot be accurately explained. Therefore, in this embodiment, for each station, the station is designated as a rejection station, and the station observation data of the rejection station is removed from the ground observation data to obtain the observation rejection data. Then, a verification model can be fitted based on the observation rejection data. The verification model is also used for magnetic field disturbance analysis. Each station can have its own corresponding verification model. Then, for each verification model, the second magnetic disturbance prediction data of the rejection station is predicted through the verification model, thereby determining the second prediction error of the rejection station.

[0045] Since the validation model is based on fitting observational data after data removal, the fitting process lacks observational data from the removed stations. Therefore, the second geomagnetic disturbance prediction data for the removed stations predicted by the validation model will not be affected by the station observational data. This makes it easier to objectively evaluate the accuracy of the target model's geomagnetic disturbance prediction at unknown geographical locations through the second prediction error. However, since the second prediction error only reflects the geomagnetic disturbance prediction performance after the lack of observational data for a single station, the reliability of the target model's prediction cannot be accurately reflected solely by the second prediction errors. This is because if the prediction error of a removed station differs significantly from the prediction error of the validation model, it indicates that the target model may rely too heavily on known data. The stability of the target model's prediction of magnetic disturbance data may be poor, leading to poor prediction reliability. Therefore, it is necessary to determine the prediction stability of the target model based on the first prediction error and each second prediction error. Since the first prediction error is jointly determined by the first magnetic disturbance prediction data and the station's observation data respectively, and each station is considered as a rejected station with a second prediction error, it is convenient to compare the first prediction error and each second prediction error to determine the prediction stability and obtain the prediction reliability. This quantifies the reliability of the target model, allowing users to make corresponding decisions based on the prediction reliability and the magnetic disturbance prediction data obtained from the target model.

[0046] Based on this, embodiments of this application provide a method for analyzing magnetic field disturbances based on ground observation data, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the magnetic field disturbance analysis method based on ground observation data according to this application. The magnetic field disturbance analysis method based on ground observation data includes steps S10 to S50:

[0047] Step S10: Obtain ground observation data, wherein the ground observation data includes observation data from multiple stations respectively;

[0048] It should be noted that ground-based observation data characterizes magnetic field disturbance data observed at the ground. This data can include observations from multiple stations. These stations can be USGS (United States Geological Survey Station), NRCAN (Natural Resources Canada Station), etc. This embodiment does not specifically limit the stations; each station can be used to observe magnetic field disturbance data. Station observation data reflects the magnetic field disturbance data observed at the station. This magnetic field disturbance data includes the three magnetic field components of the Earth's magnetic field: the northward magnetic field component, the eastward magnetic field component, and the vertical magnetic field component.

[0049] Step S20: Determine the target model for analyzing magnetic field disturbances obtained by fitting ground observation data, predict the first magnetic disturbance prediction data of each station through the target model, and jointly determine the first prediction error of the target model based on the first magnetic disturbance prediction data and station observation data corresponding to each station.

[0050] It should be noted that the target model is a model fitted from ground observation data. The target model is used to analyze magnetic field disturbances and can analyze magnetic field disturbance data from multiple geographical locations. These geographical locations can be any location; this embodiment does not impose specific limitations on this. The first magnetic disturbance prediction data is the geomagnetic disturbance data predicted by the target model, and each station has its own first magnetic disturbance prediction data. The first prediction error includes the station's prediction error, which is the difference between the station's first magnetic disturbance prediction data and the station's observed data. The larger the station's prediction error, the greater the difference between the first magnetic disturbance prediction data and the station's observed data.

[0051] For example, the station observation data corresponding to multiple stations are obtained, the target model is obtained using the observation data of multiple stations, the first magnetic disturbance prediction data of each station is predicted using the target model, and for each station, the station prediction error is calculated based on the first magnetic disturbance prediction data and the station observation data.

[0052] Step S30: For each station, the station is designated as a removal station, and the station observation data of the removal station is removed from the ground observation data to obtain the observation removal data. The verification model of the removal station for analyzing magnetic field disturbances is determined based on the observation removal data.

[0053] It should be noted that the "removed stations" refer to the ground-based observation data from which observations need to be removed. The "removed observation data" are the ground-based observation data from the stations whose data have been removed. Each station has corresponding removed observation data, and the validation model is obtained by fitting this data. The fitting process for the validation model can be the same as that for the target model, only the data used differs. The validation model can also be used to analyze magnetic field disturbances. Furthermore, it can be used to verify the accuracy of predicted geomagnetic disturbance data.

[0054] For example, for each station, the station is designated as a rejection station, and its observation data is removed from the ground observation data to obtain rejection data. A validation model is then fitted using this rejection data to perform magnetic field disturbance analysis, thereby facilitating the subsequent determination of the prediction reliability of the target model. In other words, in this embodiment, station prediction data can be rejected station by station, facilitating the fitting of validation sub-models one by one.

[0055] Step S40: For each verification model, predict the second magnetic disturbance prediction data of the stations to be removed by the verification model, and determine the second prediction error of the stations to be removed based on the second magnetic disturbance prediction data and the station observation data of the stations to be removed.

[0056] Step S50: Determine the prediction reliability of the target model based on the first prediction error and each of the second prediction errors.

[0057] It should be noted that each removed station has a validation model. The validation model for each removed station is obtained by fitting the observed data from that station. The second magnetic disturbance prediction data obtained from the validation model is the geomagnetic disturbance data from the removed station. The second prediction error is the difference between the second magnetic disturbance data from the removed station and the station's observed data. The larger the second prediction error, the greater the difference between the second magnetic disturbance data from the removed station and the station's observed data.

[0058] Prediction confidence reflects the reliability of the target model's predictions of geomagnetic disturbance data. Higher prediction confidence indicates higher reliability of the target model, while lower prediction confidence indicates lower reliability. Prediction confidence includes multiple station confidence levels, each with its own corresponding station confidence level. Station confidence reflects the reliability of the target model's predictions for removed stations and their surrounding areas.

[0059] For example, for each eliminated station, the second prediction error is compared with the first prediction error to determine the station's credibility. The station's credibility is the same as the station's credibility when eliminated. For instance, if station A exists and is eliminated, the station's credibility is the same as the station's credibility when eliminated. If the second prediction error of the eliminated station is closer to the station's prediction error in the first prediction error, it indicates that the target model's prediction for the eliminated station is more stable, and consequently, that the target model's prediction for the surrounding area of ​​the eliminated station is also relatively stable, resulting in higher station credibility. Conversely, if the second prediction error of the eliminated station is farther from the station's prediction error in the first prediction error, it indicates that the target model's prediction for the magnetic field of the eliminated station and its surrounding area is less stable, resulting in lower station credibility.

[0060] This embodiment can acquire ground observation data, including observation data from multiple stations. This embodiment can obtain a target model for magnetic field disturbance analysis by fitting the observation data from multiple stations. Then, based on the target model, the first predicted magnetic disturbance data for each station can be predicted. Based on the first predicted magnetic disturbance data and the station observation data, the first prediction error of the target model is determined. That is, this embodiment can determine the first prediction error of the target model by comparing the predicted value and the actual value of the magnetic field disturbance. Since the target model is obtained by fitting the observation data from each station, the first predicted magnetic disturbance data for each station will naturally be close to the station observation data. Therefore, even if the first prediction error of the target model is determined, the reliability of the target model's prediction cannot be accurately explained. Therefore, in this embodiment, for each station, the station is designated as a rejection station, and the station observation data of the rejection station is removed from the ground observation data to obtain the observation rejection data. Then, a verification model can be fitted based on the observation rejection data. The verification model is also used for magnetic field disturbance analysis. Each station can have its own corresponding verification model. Then, for each verification model, the second magnetic disturbance prediction data of the rejection station is predicted through the verification model, thereby determining the second prediction error of the rejection station.

[0061] Since the validation model is based on fitting observational data after data removal, the fitting process lacks observational data from the removed stations. Therefore, the second geomagnetic disturbance prediction data for the removed stations predicted by the validation model will not be affected by the station observational data. This makes it easier to objectively evaluate the accuracy of the target model's geomagnetic disturbance prediction at unknown geographical locations through the second prediction error. However, since the second prediction error only reflects the geomagnetic disturbance prediction performance after the lack of observational data for a single station, it is not possible to accurately reflect the reliability of the target model's prediction solely through the second prediction errors. This is because if the prediction error of a removed station differs significantly from the prediction error of the validation model and the target model, it indicates that the target model may be overly reliant on... Based on the known station prediction data, the stability of the target model's prediction of magnetic disturbance data may be poor, thus affecting the reliability of the prediction. Therefore, it is necessary to determine the prediction stability of the target model based on the first prediction error and each second prediction error. Since the first prediction error is determined by the first magnetic disturbance prediction data and the station's observation data respectively, and each station is considered as a excluded station, each excluded station has a second prediction error, it is convenient to compare the first prediction error and each second prediction error to determine the prediction stability and obtain the prediction reliability. This quantifies the reliability of the target model, allowing users to make corresponding decisions based on the prediction reliability and the magnetic disturbance prediction data obtained from the target model.

[0062] In a feasible embodiment, step S10 further includes steps S11 to S14:

[0063] Step S11: Obtain the list of observation stations and observation periods;

[0064] Step S12: For each station in the observation station list, according to the station type, call the data acquisition component corresponding to the station type to obtain the initial observation data of the station during the observation period. The initial observation data includes sub-observation data corresponding to multiple observation times within the observation period.

[0065] It should be noted that the station list includes stations from which observation data needs to be acquired. The station list can be preset based on actual conditions, and this embodiment does not impose specific limitations on it. The station list includes multiple stations. The observation period can also be preset based on actual conditions, and this embodiment does not impose specific limitations on it; the observation period refers to the time period during which geomagnetic disturbance data of the observed stations are collected.

[0066] Different station types require different data acquisition components. The station types in the station list can include USGS, NRCAN, and domestic types, etc., and this embodiment does not specifically limit them. For example, when the station type is USGS, the data acquisition component is the USGS type station API (Application Programming Interface) or FTP server (File Transfer Protocol Server). For instance, files with the .min extension can be downloaded in batches from the USGS type station through the station interface or FTP server, and all the .min files downloaded from that station can be used as the initial observation data for that station. The .min files include geomagnetic disturbance data observed by that station during the observation period. The initial observation data is geomagnetic disturbance data observed at the station without preprocessing, while the station observation data is geomagnetic disturbance data observed at the station after preprocessing. It can be considered that the station observation data is obtained by adjusting the format of the initial observation data.

[0067] For example, when the station type is NRCAN, the data acquisition component can be an access parsing component. This component accesses the webpage containing the NRCAN station and parses the webpage filename rules to automatically download the geomagnetic disturbance data for the observation period, thus obtaining the initial observation data. When the station type is domestic, the data acquisition component can be the data acquisition interface for domestic stations, allowing direct acquisition of the station's initial observation data for the observation period.

[0068] In this embodiment, during the download of geomagnetic disturbance data from the webpage corresponding to the station, interrupted downloads and error retry mechanisms are supported to ensure download integrity. This, in turn, facilitates the integrity of the initial observation data acquisition.

[0069] Step S13: For each initial observation data, the initial observation data is divided into multiple time windows according to the preset time division unit. In each window observation data, the target sub-observation data that meets the preset median condition is determined, and the target sub-observation data are spliced ​​together in the preset time order to obtain the standard observation data.

[0070] Step S14: For each station, determine the station observation data based on the station's standard observation data;

[0071] Among them, the preset median condition includes the target sub-observation data being the median of the window observation data.

[0072] It should be noted that the preset duration division unit can be set based on actual conditions. For example, the preset duration division unit can be 1 minute, or in other embodiments, it can be 30 seconds, etc. This embodiment does not specifically limit this. The duration of each time window is the same as the preset duration division unit. Multiple time windows are consecutive, and multiple time windows can be spliced ​​together to obtain the observation period. The window observation data is the observation data of the initial observation data within the time window. The window observation data includes the observation data corresponding to each of the multiple times, and the target sub-observation data is the median of the observation data within the window observation data.

[0073] Standard observation data is obtained by splicing together target sub-observation data according to a preset time order. For example, in the standard observation data, the observation time of the target sub-observation data that is ranked earlier is earlier than the observation time of the target sub-observation data that is ranked later. In this embodiment, the preprocessing step for the initial observation data includes steps S13 to S14. After preprocessing the initial observation data, station observation data can be obtained. After obtaining the standard observation data, abnormal data can be removed from each standard observation data, and each standard observation data can be time-aligned to ensure the time consistency of each standard observation data, thereby obtaining the station observation data for each station.

[0074] For example, for each initial observation data point, the initial observation data is divided into multiple time windows according to a preset time period. For each window observation data point, the median of the observation data within the window is used as the target sub-observation data. These target sub-observation data points are then concatenated sequentially according to a preset time order to obtain standard observation data. For example, the preset time order is such that earlier observation times precede later observation times, facilitating the sequential concatenation of target sub-observation data to obtain standard observation data. For each station, the standard observation data for that station undergoes outlier removal and time alignment processing to obtain station observation data. For example, the standard observation data can be time-aligned according to a preset time alignment range, ensuring that the standard observation data contains target sub-observation data for each preset observation window within the preset time alignment range. The preset time alignment range can be determined in advance based on actual conditions; this embodiment does not impose specific limitations on it. Each preset observation window within the preset time alignment range is within an observation period.

[0075] This embodiment reduces the amount and complexity of data processing by dividing the initial observation data and determining the median within each window of observation data. At the same time, it does not lead to inaccurate fitting due to the reduction in data, because the target sub-observation data included in the standard observation data is the median within the window of observation data, so that the observation situation of the window of observation data can be represented by the target sub-observation data.

[0076] In a feasible embodiment, step S14 further includes steps S141 to S144:

[0077] Step S141: For the standard observation data of each station, remove the abnormal target sub-observation data in the standard observation data, and if a first target time window that does not belong to the preset time alignment range is detected in the standard observation data, remove the target sub-observation data of the first target time window in the standard observation data.

[0078] It should be noted that since the standard observation data is actually obtained from station observations, environmental factors and other influences may cause anomalies in the data during the observation process. To avoid the impact of abnormal observation data on the subsequent fitting of the target model, it is necessary to remove abnormal target sub-observation data from the standard observation data. For example, when the magnetic field component of the target sub-observation data deviates from the historical average of the station by more than 3 times the root mean square, it is considered abnormal; or, when the change in the magnetic field component of the target sub-observation data exceeds a preset anomaly amplitude threshold, it is considered abnormal. The preset anomaly amplitude threshold can be set based on actual conditions, and this embodiment does not impose specific limitations on it.

[0079] The preset time alignment range includes multiple preset observation windows within the observation period. For example, the preset time alignment range can be determined based on the time window existing in any standard observation data so that other standard observation data are aligned with the preset time alignment range.

[0080] Because the time window in this embodiment is relatively short—for example, it is divided according to a preset time unit, typically one minute or less—it can be approximated as the observation moment due to its sufficiently short duration. Therefore, the first target time window is a moment present in the standard observation data but not within the preset time alignment range. Thus, the target sub-observation data for the first target time window can be removed from the standard observation data. When all standard observation data are aligned with the preset time alignment range, it helps ensure consistency between the standard observation data; this consistency is the consistency of the time window. This, in turn, facilitates ensuring the accuracy of subsequent predictions.

[0081] Step S142: If there is no target sub-observation data for the second target time window in the standard observation data, obtain the first adjacent window and the second adjacent window that are adjacent to the second target time window; wherein, the second target time window is any preset observation window within the preset time alignment range.

[0082] It should be noted that when dividing the initial observation data according to the preset time interval, there may be situations where some time windows do not contain observation data. For example, the second target time window may be a time window without target sub-observation data, so the target sub-observation data in the second target time window is empty. In this embodiment, when stitching together the target sub-observation data, each target sub-observation data will carry its corresponding time window. The second target time window is also within the preset time alignment range. The second target time window is a preset observation window within the preset time alignment range, but it is a time window that does not exist in the standard observation data. The first adjacent window and the second adjacent window are both adjacent to the second target time window, but the first adjacent window and the second adjacent window are different. For example, the time of the first adjacent window is earlier than that of the second adjacent window.

[0083] Step S143: Perform linear interpolation on the target sub-observation data of the first adjacent window and the target sub-observation data of the second adjacent window to obtain the target sub-observation data of the second target time window;

[0084] Step S144: Standard observation data that does not include abnormal target sub-observation data, does not include target sub-observation data of the first target time window, and has target sub-observation data of each preset observation window within the preset time alignment range is used as station observation data.

[0085] It should be noted that, based on the target sub-observation data of the first and second adjacent windows, linear interpolation is performed on the second target time window, thereby completing the target sub-observation data of the second target time window. The preset observation window is a pre-defined time window.

[0086] For example, the difference between the target sub-observation data of the first adjacent window and the target sub-observation data of the second adjacent window can be calculated to obtain the observation difference. The difference between the second target time window and the first adjacent window can be calculated to obtain the first window difference. The difference between the first adjacent window and the second adjacent window can be calculated to obtain the second window difference. The ratio of the first window difference to the second window difference can be calculated to obtain the difference ratio. The ratio of the observation difference to the difference ratio can be determined to obtain the observation ratio. The sum of the target sub-observation data of the first adjacent window and the observation ratio can be calculated to obtain the target sub-observation data of the second target time window.

[0087] For each standard observation data set, after executing steps S141 to S144, the station observation data of the standard observation data can be obtained, thus ensuring that the observation data of each station are aligned in time. That is, for any target sub-observation data time window in the station observation data, each station observation data set has a corresponding target sub-observation data set within the same time window. Furthermore, in this embodiment, the time window within each station observation data set is relative to the same time zone. This facilitates ensuring the time alignment of the station observation data.

[0088] In this embodiment, each target sub-observation data within each station's observation data carries a corresponding time window and observation coordinates. Each station's observation data can be represented as a matrix including a time series, observation coordinates, and three-component magnetic field disturbance data. The time series refers to the time window of each target sub-observation data in the station's observation data, the observation coordinates are the coordinates of the station where the target sub-observation data is located, and the three-component magnetic field disturbance data is the target sub-observation data.

[0089] This embodiment can eliminate abnormal observation data and ensure that the observation data of each station are aligned in time, thereby facilitating the accuracy of subsequent fitting of the target model.

[0090] In a feasible embodiment, the target model includes at least one target sub-model corresponding to a preset observation window, the target sub-model being used to analyze the magnetic field disturbances generated within the preset observation window; step S20 further includes steps S21 to S23:

[0091] Step S21: Obtain target sub-observation data for each station corresponding to the same preset observation window from the ground observation data;

[0092] It should be noted that the target model can include multiple target sub-models, which are also models used for magnetic field disturbance analysis. Each preset observation window has a corresponding target sub-model. Since geomagnetic disturbances generated at the same location at different times are different, when it is necessary to predict geomagnetic disturbances at the same location at different times, it is necessary to use the target sub-models corresponding to different times for prediction. Ground observation data includes observation data from multiple stations, and the time window of any target sub-observation data in the station observation data belongs to the preset time alignment range. Therefore, when it is necessary to fit the target sub-model of the preset observation window, multiple target sub-observation data corresponding to the same preset observation window can be obtained from the ground observation data. The multiple target sub-observation data within the same preset observation window come from different stations.

[0093] Step S22: Based on the observation data of each target sub-observation and their respective corresponding observation coordinates, construct the observation matrix of the preset observation window;

[0094] It should be noted that the observation matrix is ​​constructed within the same preset observation window from the target sub-observation data and observation coordinates. For example, Xij in the observation matrix represents the target sub-observation data in the i-th row and j-th column, where i can refer to the longitude coordinate of the target sub-observation data and j can refer to the latitude coordinate of the target sub-observation data. It can be assumed that there is an observation matrix within each preset observation window.

[0095] Step S23: Based on the pre-constructed current singularity grid and observation matrix, the equivalent current coefficient of each singularity in the current singularity grid is deduced, and the target sub-model of the pre-constructed observation window is obtained from the equivalent current coefficients in the singularity grid.

[0096] It should be noted that the current singularity grid is a SECS node grid. The current singularity grid includes multiple singularities, each of which is equivalent to a current source. In the SECS node grid, each grid node is considered a hypothetical current source. These current sources are not actual currents, but rather artificially defined in the mathematical model. They are cells used to construct the perturbation current field, enabling subsequent prediction of magnetic field perturbations at arbitrary geographical locations using the SECS node grid. Each singularity generates a magnetic field perturbation at each observation point.

[0097] For example, based on the preset number of nodes and the preset grid radius, a current singularity grid is constructed in the preset observation area, and the initial current coefficient of each singularity in the current singularity grid is configured as a preset initial equivalent coefficient; for each singularity in the current singularity grid, based on the preset initial equivalent coefficient of the singularity and the relative distance between the singularity and each observation point, the magnetic field disturbance generated by the singularity at each station is calculated. The preset initial equivalent coefficient can be a preset unit current coefficient, for example, it can be 1.

[0098] The preset number of nodes represents the number of singularities constructed, and the preset grid radius is the sum of the Earth's radius and the height of the current singularity grid. For example, the current singularity grid can be set at an altitude of Earth's radius + 110 km. Both the preset number of nodes and the preset grid radius can be determined based on actual conditions; this embodiment does not impose specific limitations on them. Each station is located within a preset observation area. The preset initial equivalent coefficients can also be determined based on actual conditions; this embodiment does not impose specific limitations on them. Each grid node in the current singularity grid can be used to place a current source (singularity), and each current source will generate magnetic field disturbances at the Earth's surface or the altitude of the satellite.

[0099] Since magnetic field disturbances are related to activities such as the solar wind, it is assumed that when the solar wind enters the Earth's magnetosphere, it causes changes in the current in the Earth's ionosphere. These changes in the ionospheric current, in turn, lead to part of the magnetic field disturbance. Therefore, magnetic field disturbances can be predicted using a grid of current singularities.

[0100] The relative distance between the singular point and the observation point varies, resulting in different magnetic field disturbances and current magnitudes at the singular point. Consequently, the magnetic field response data generated at the same observation point may also differ. Therefore, in this embodiment, the magnetic field disturbance generated by the singular point at each observation point is calculated based on the initial equivalent coefficient of the singular point and the relative distance between the singular point and each observation point.

[0101] The equivalent current coefficient can characterize the current magnitude at singularities. The observation matrix is ​​the actual observed magnetic field disturbance. By using the current singularity grid and the observation matrix, the equivalent current coefficient of each singularity in the current singularity grid is obtained through inversion, which is equivalent to reconstructing the ionospheric current corresponding to the current singularity grid.

[0102] Since the equivalent current coefficient is derived from the current singularity grid and the observation matrix, and the observation data is the actual observed magnetic field disturbance data, the inverted equivalent current coefficient is also accurate, so that the magnetic field disturbance at any geographical location can be predicted based on the equivalent current coefficient.

[0103] In a feasible embodiment, step S23 further includes steps S231 to S234:

[0104] Step S231: Based on the error matrix constructed from the preset observation error weights of each station, the observation matrix is ​​corrected to obtain the weighted matrix;

[0105] It should be noted that the preset misobservation error weights can be determined based on the user's trust in the observation points. This embodiment does not impose specific limitations on this. Each station has its own corresponding preset observation error weights. The more the user trusts the station, the smaller the corresponding preset observation error weight; conversely, the less the user trusts the station, the larger the corresponding preset observation error weight. The observation matrix can be corrected based on the observation error weights of each station to obtain a weighted matrix, thereby improving the accuracy of the equivalent current coefficient and thus improving the accuracy of the target sub-model.

[0106] For example, an error matrix is ​​constructed based on the preset observation error weights of each station. The observation matrix is ​​then corrected using the error matrix to obtain a weighted matrix. For instance, if the same observation coordinate has the same position in both the error matrix and the observation matrix, the observation matrix can be corrected by multiplying each target sub-observation data point in the observation matrix by the preset observation error weight of the station containing that sub-observation data. In this embodiment, the error matrix can be reused without repeated construction. For example, when fitting target sub-models with different preset observation windows, the error matrix can be directly reused without reconstruction. This improves the efficiency of target sub-model construction.

[0107] Step S232: Based on the preset initial equivalent coefficient of each singular point in the current singular point grid, calculate the initial magnetic disturbance prediction data of each station under the current singular point grid, and construct the initial prediction matrix based on the initial magnetic disturbance prediction data and observation coordinates corresponding to each station.

[0108] It should be noted that the preset initial equivalent coefficients can be predetermined, and the preset initial equivalent coefficients for each singular point can be the same, for example, they can all be 1. This embodiment does not impose specific limitations on this. The initial magnetic disturbance prediction data refers to the magnetic field disturbance generated by each singular point in the current singular point grid at the station when the current coefficient at each singular point is the preset initial equivalent coefficient. Each station has initial magnetic disturbance data. The initial prediction matrix is ​​constructed from the initial magnetic disturbance prediction data and observation coordinates of each station. For example, Pij in the initial prediction matrix is ​​the initial magnetic disturbance prediction data in the i-th row and j-th column, where the i-th row can refer to the longitude coordinates of the station, and the j-th column can refer to the latitude coordinates of the station.

[0109] Step S233: Based on the weighting matrix, the initial prediction matrix, and the preset regularization parameters, construct the final optimization problem corresponding to the equivalent current coefficient of each singular point in the current singularity grid.

[0110] It should be noted that the preset regularization parameter can be set based on actual conditions. For example, the preset regularization parameter can be set based on the distance between adjacent singular points in the current singular point grid. How to set the preset regularization parameter based on the distance between adjacent singular points is based on experience, and this embodiment does not impose specific limitations on this. Introducing the preset regularization parameter in this embodiment can avoid model overfitting and improve the accuracy of model construction. The final optimization problem is a least squares problem.

[0111] For example, the final optimization problem can be referred to as Formula 1:

[0112] (Formula 1);

[0113] Where x* is a vector composed of all equivalent current coefficients in the current singularity grid, A is the initial prediction matrix, b is the weighted observation matrix, λ is the preset regularization parameter, and x is the equivalent current coefficient.

[0114] Step S234: By performing a preset matrix decomposition on the final optimization problem, the equivalent current coefficient of each singular point is obtained by reverse deduction.

[0115] It should be noted that the preset matrix decomposition can be LU (triangular) decomposition, which can be performed on the matrix multiplied by the equivalent current coefficients in the final optimization problem to obtain the equivalent current coefficients at each singular point. In other embodiments, the preset matrix decomposition can also be SVD (singular value decomposition), and this embodiment does not specifically limit it. In this embodiment, LU triangular decomposition can improve the accuracy of the equivalent current coefficients and avoid redundant calculations, significantly improving the computational efficiency of large-scale time-series magnetic field inversion problems.

[0116] In this embodiment, the final optimization problem is a least squares problem. To better understand the final optimization problem in this embodiment, a brief example is given of the process for determining the final optimization problem: First, a matrix equation is constructed, which is: observation matrix d = initial prediction matrix G * equivalent current coefficient x. The problem to be solved in the matrix equation is x. Since there are only a few observation points on the ground, this matrix is ​​an ultra-sparse matrix. Assuming that there is a station at every location on the ground, then d is a completely known matrix. Then, G and d are also completely known, and x can be solved by directly calculating x = the inverse of G * d. However, the number of stations is finite, and d must be an ultra-sparse matrix. In this case, x = the inverse of G * d is mathematically invalid, which is an ill-conditioned problem. To solve this ill-conditioned problem, we need to construct a final optimization problem, namely the least squares problem, and solve the least squares problem to obtain the equivalent current coefficients. Solving the least squares problem involves finding the best match of the data by minimizing the sum of squares of the errors. Finding the best match of the data means finding the optimal equivalent current coefficients such that the initial prediction matrix G * equivalent current coefficient x is closest to the observation matrix d. Minimizing the sum of squares of the errors is to minimize the function of the equivalent current coefficient x. The function of the equivalent current coefficient can be referred to as the matrix equation, which can also be considered as a function of the equivalent current coefficient x. Minimizing the function of the equivalent current coefficient is to make the derivative of the equivalent current coefficient zero, which facilitates the least squares problem. Before differentiating the function of the equivalent current coefficient, the observation matrix needs to be corrected to obtain a weighted matrix so that the calculation of the equivalent current coefficient can be more accurate. Furthermore, a preset regularization parameter needs to be added to the matrix equation to improve the smoothness of the equivalent current coefficient and prevent abrupt changes in the equivalent current coefficient at adjacent singular points. Finally, the final squared problem can be obtained by referring to Formula 1.

[0117] In a feasible embodiment, the first magnetic disturbance prediction data includes: first window prediction data of the station obtained by the target sub-model; the station observation data includes target sub-observation data of a preset observation window; step S20 further includes steps A10 to A20:

[0118] Step A10: For each station, calculate the difference between the first window prediction data and the first target sub-observation data within the same preset observation window to obtain the first window error;

[0119] Step A20: The errors of each first window are combined as the first prediction error.

[0120] It should be noted that the station's observation coordinates can be input into the target sub-model, which then predicts the station's first window prediction data. This first window prediction data represents the predicted magnetic field disturbance data for the station within a preset observation window. The first magnetic disturbance prediction data can include the first window prediction data for each station within each preset observation window. Specifically, the first magnetic disturbance prediction data can include multiple first window prediction data sets. Within the same preset observation window, each station has its own first window prediction data set for that window, and the same station has first window prediction data for each preset observation window. For the same station, if it is necessary to predict magnetic field disturbance data for different preset observation windows, it is necessary to obtain the target sub-models corresponding to each preset observation window to predict the corresponding window prediction data.

[0121] Each station has a corresponding first window error within each preset observation window. Understandably, a single station may have multiple first window errors, each corresponding to a different preset observation window. For example, the first window error of station A within preset observation window t is the difference between the predicted data Y1 of station A within preset observation window t and the target sub-observation data G1 within preset observation window t. Both G1 and Y1 belong to preset observation window t. The first prediction error includes the first window error of each station within each preset observation window.

[0122] For example, after fitting the target sub-model for each preset observation window, the observation coordinates of each station can be input into the target sub-model for each preset observation window. The target sub-model then predicts the first window prediction data for each station within that preset observation window. For each station, the difference between the first window prediction data and the target sub-observation data within each preset observation window is calculated to obtain the first window error within each preset observation window. The first window errors of each station within each preset observation window are collectively used as the first prediction error. Therefore, this embodiment can calculate the corresponding first window error for each target sub-model, which helps to measure the prediction accuracy of the target sub-model.

[0123] In a feasible embodiment, the verification model includes at least one verification sub-model with a preset observation window, which is used to analyze the magnetic field disturbances generated within the preset observation window; the second magnetic disturbance prediction data includes: second window prediction data of the excluded stations predicted by the verification sub-model; the station observation data includes target sub-observation data of the preset observation window; step S40 further includes steps S41 to S42:

[0124] Step S41: For each preset observation window, calculate the difference between the second window prediction data and the target sub-observation data of the station within the same preset observation window, and obtain the second window error of the preset observation window.

[0125] Step S42: Combine the errors of each second window as the second prediction error.

[0126] It should be noted that the fitting process of the validation sub-model is the same as that of the target sub-model, and will not be repeated in this embodiment. Each preset observation window has a corresponding validation sub-model, and the same preset observation window can correspond to multiple validation sub-models. For example, if there are 3 stations a1~a3, in preset observation window B, there are validation sub-models c1~c3. c1 is the validation sub-model corresponding to a1, and c1 is obtained by fitting the observation data of station a1 that has been removed; c2 is the validation sub-model corresponding to a2, and c2 is obtained by fitting the observation data of station a2 that has been removed; c3 is the validation sub-model corresponding to a3, and c3 is obtained by fitting the observation data of station a3 that has been removed. For example, after obtaining the validation sub-model c1, the validation sub-model c1 predicts the second window prediction data of station a1.

[0127] The second magnetic disturbance prediction data includes the second window prediction data of the eliminated stations obtained from the verification sub-model. The second window prediction data is the magnetic field disturbance data predicted in the verification sub-model. For each verification sub-model, the station coordinates of the eliminated stations can be input into the verification sub-model, and the second window prediction data of the eliminated stations can be output.

[0128] Each station has a second prediction error, which includes the second prediction error corresponding to different preset observation windows. The second window error is the difference between the second window observation data and the target sub-observation data within the same preset observation window. The smaller the second window error, the closer the predicted value of the verification sub-model is to the true value of the magnetic field disturbance, and the more accurate the prediction of the verification sub-model is.

[0129] For example, for each preset observation window, the difference between the second window prediction data and the target sub-observation data of the station to be removed within each preset observation window is calculated to obtain the second window error of the preset observation window; the second window errors of the removed stations in each preset observation window are collectively used as the second prediction error. Each removed station has a corresponding second prediction error. The second prediction error of the removed station is the second prediction error of the station corresponding to that removed station. For example, if the station whose observation data is currently being removed is station A, then the current removed station is station A. If the second prediction error of the current removed station is obtained, then the second prediction error of station A is also obtained.

[0130] This embodiment determines the second prediction error for each station, which facilitates the subsequent determination of the reliability and stability of the target model based on the second and first prediction errors. Furthermore, the second prediction error reflects the prediction accuracy of the validation sub-model; for example, the smaller the mean of the errors in each second window within the second prediction error, the lower the prediction accuracy of the validation sub-model.

[0131] In a feasible embodiment, the predicted confidence level includes the station confidence level; step S50 further includes steps S51 to S52:

[0132] Step S51: For the second prediction error of each eliminated station, determine the difference between the second window error within the second prediction error and the first window error within the first prediction error within the same preset observation window, and obtain the verification error of the eliminated station.

[0133] Step S52: For each rejected station, determine the station's credibility based on the verification error of the rejected station, wherein the station's credibility is negatively correlated with the verification error.

[0134] It should be noted that for each excluded station, there is a verification error within each preset observation window. The verification error reflects the prediction difference between the verification sub-model and the target sub-model within the preset observation window. A smaller verification error indicates a smaller prediction difference, and a larger verification error indicates a larger prediction difference. The prediction difference refers to the discrepancy between the predicted magnetic field disturbance data. A smaller verification error indicates a more stable target sub-model, because a smaller verification error means that the observation data from stations without excluded stations are similar in both the verification sub-model and the target sub-model, indicating a more stable, reliable, and trustworthy target sub-model. A smaller verification error results in higher station trustworthiness, and a larger verification error results in lower station trustworthiness. For example, the station trustworthiness corresponding to the verification error can be found in the preset error trustworthiness mapping relationship; the preset error trustworthiness mapping relationship is a mapping relationship between preset errors and preset trustworthiness, and there is a negative correlation between preset errors and preset trustworthiness.

[0135] In this embodiment, each target sub-model has multiple corresponding station confidence levels. For example, each target sub-model has a station confidence level at each station. The station confidence level reflects the reliability of the target sub-model in predicting the magnetic field disturbances of that station and its surrounding area. The station confidence level can be a percentage; the higher the station confidence level, the higher the reliability of the target sub-model in predicting the magnetic field disturbances of that station and its surrounding area. The surrounding area can be a region with the station as the center and a radius of a preset radius. The preset radius can be set based on actual conditions, and this embodiment does not impose specific limitations on it. In this embodiment, the station confidence level can carry a preset observation window and a label for the stations to be removed. This label can be represented in the form of a number. When the label carried by the verification error is 0n0m, it indicates that the station confidence level is the station confidence level of the preset observation window numbered 0n and the station to be removed numbered 0m. It also indicates that the station confidence level is the station confidence level of the target sub-model under the preset observation window of 0n in predicting the station to be removed (0m) and its surrounding area.

[0136] For example, for each eliminated station's second prediction error, the difference between the second window error of the second prediction error within each preset observation window and the first window observation error within the first prediction error is calculated to obtain the verification error. For each eliminated station, based on the verification error of the eliminated station, the station confidence level of the target sub-model in predicting the eliminated station within the preset observation window where the verification error is located is determined. For any target sub-model, when predicting magnetic field disturbance data of any station or the surrounding area of ​​the station, the target sub-model can simultaneously output the station confidence level of the target sub-model in predicting the station, so that users can make corresponding decisions based on the prediction results.

[0137] Additionally, this embodiment can also display a visual comparison chart of predicted and observed values. For example, it can refer to... Figure 2 , Figure 2 This diagram illustrates the comparison between observed and predicted magnetic field disturbance values ​​at a single station. Figure 2 The predicted magnetic field disturbance values ​​represent the first predicted magnetic disturbance data obtained from the target model, while the observed magnetic field disturbance values ​​are the station observation data obtained from the stations. Figure 2 The diagram illustrates a comparison between predicted and observed magnetic field disturbances for the three components of the magnetic field. For example, in... Figure 2 The three magnetic field components are B X B Y B Z ;exist Figure 2 In the diagram, the horizontal axis represents time, and the vertical axis represents the intensity of the magnetic field disturbance, with units of nT; for example, in... Figure 2The document shows a comparison between the predicted and observed values ​​of magnetic field disturbances from 00:00 on April 15th to 00:00 on April 18th. This embodiment can automatically generate time-series comparison charts of the observed and predicted three components of the magnetic field disturbance and save them as high-quality image files. For example, this embodiment may include a visualization module. The visualization module can be input with station observation data, the first magnetic field prediction data, the time axis of the observation period, and the station's observation coordinates. The visualization module can call Matplotlib to plot the comparison charts of the observed and predicted values ​​of the three magnetic field components (B_X, B_Y, and B_Z) for each station. The comparison charts can be displayed in real time or directly saved as image files. Parameters such as font, resolution, and size can be set in the comparison charts, and this embodiment does not specifically limit these settings. For a better understanding of this embodiment, please refer to... Figure 3 , Figure 3 The diagram illustrates the functional modules of this embodiment. X10 represents the data source, which can originate from multiple stations, such as stations 1 to N, including both USGS and NRCAN stations. X20 represents the data acquisition module, used to acquire initial observation data from multiple stations. X30 represents the preprocessing module, used to preprocess the initial observation data to obtain station observation data. X40 represents the inversion fitting module, used to fit a target model based on the observation data from each station. The specific target model can include multiple target sub-models. X50 represents the target model, which is obtained from the observation data from each station through the inversion fitting module. The target model is used for magnetic field disturbance analysis. X60 refers to the station-by-station elimination analysis module. For example, this module is used to remove the observation data of each station from the ground observation data, obtaining observation elimination data, and determining a verification model for analyzing magnetic field disturbances based on the observation elimination data. Each eliminated station has a verification model. X70 refers to the verification model, which can be obtained through the station-by-station elimination analysis module. Each eliminated station has a verification model, which can also be used for magnetic field disturbance analysis. X80 refers to the visualization module, which can visually compare the magnetic field disturbance prediction data obtained from the target model with the station observation data, and also visually compare the magnetic field disturbance prediction data obtained from the verification model with the station observation data. In this embodiment, a first prediction error can be calculated using the first magnetic disturbance prediction data predicted by the target model and the station observation data, and a second prediction error can be calculated using the second predicted magnetic disturbance data from the verification model and the station observation data, so as to determine the prediction reliability through the first and second prediction errors.

[0138] To better understand this embodiment, you can also refer to Figure 4 , Figure 4 This document illustrates the flowchart of the process for obtaining a validation model by systematically removing station observation data. A brief explanation of the process for obtaining the validation model is provided, including steps Y10 to Y100. Step Y10: Start; Step Y20: Acquire ground observation data; Step Y30: Determine the stations to be removed from the ground observation data; any station can be selected as the removal station; Step Y40: Remove the observation data of the removal station from the ground observation data to obtain the observation removal data; Step Y50: Use the observation removal data to fit the validation model. Specifically, multiple validation sub-models can exist under the same observation removal data; for example, each preset observation window in the observation removal data can correspond to a validation sub-model. Step Y60: Predict the second magnetic disturbance prediction data of the stations to be removed using the verification model; specifically, the second magnetic disturbance prediction data may include the second window prediction data of the stations to be removed in each preset observation window; Step Y70: Compare the second magnetic disturbance prediction data with the station observation data of the stations to be removed to obtain the second prediction error; specifically, the second prediction error may include the second window error of the stations to be removed in each preset observation window; Determination step Y80: Determine whether there are any stations among the stations that have not been removed from the station observation data; if all stations have been removed from the station observation data, it can be considered that there are no stations among the stations that have not been removed from the station observation data, and step Y100 can be executed directly; if there are stations among the stations that have not been removed from the station observation data, step Y90 can be executed: determine new stations to be removed from the stations; specifically, it may be to determine new stations to be removed from the remaining stations that have not been removed from the station observation data, and return to step Y40.

[0139] This application provides a magnetic field disturbance analysis device based on ground observation data. The device includes: at least one processor; and a memory communicatively connected to the at least one processor. The memory stores instructions executable by the at least one processor, which, when executed, enable the at least one processor to perform the magnetic field disturbance analysis method based on ground observation data described in Embodiment 1 above. (Refer to the following...) Figure 5The diagram illustrates a structural schematic of a magnetic field disturbance analysis device based on ground observation data suitable for implementing embodiments of this application. The magnetic field disturbance analysis device based on ground observation data in the embodiments of this application may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 5 The magnetic field disturbance analysis device based on ground observation data shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0140] like Figure 5 As shown, the magnetic field disturbance analysis device based on ground observation data may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory 1002 or a program loaded from a storage device 1003 into a random access memory 1004. The random access memory 1004 also stores various programs and data required for the operation of the magnetic field disturbance analysis device based on ground observation data. The processing unit 1001, the read-only memory 1002, and the random access memory 1004 are interconnected via a bus 1005. An input / output interface 1006 is also connected to the bus. Typically, the following systems can be connected to the input / output interface 1006: input devices 1007 including, for example, a touch screen, touchpad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD), speaker, vibrator, etc.; storage devices 1003 including, for example, magnetic tape, hard disk, etc.; and communication devices 1009. Communication device 1009 allows the magnetic field disturbance analysis equipment based on ground observation data to communicate wirelessly or wiredly with other equipment to exchange data. Although the figure shows a magnetic field disturbance analysis equipment based on ground observation data with various systems, it should be understood that it is not required to implement or possess all of the systems shown. More or fewer systems may be implemented alternatively.

[0141] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, the embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from read-only memory 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application. The magnetic field disturbance analysis device based on ground observation data provided in this application, employing the magnetic field disturbance analysis method based on ground observation data in the above embodiments, can solve the technical problem of how to determine the reliability of geomagnetic disturbance prediction. Compared with the prior art, the beneficial effects of the magnetic field disturbance analysis device based on ground observation data provided in this application are the same as the beneficial effects of the magnetic field disturbance analysis method based on ground observation data provided in the above embodiments, and other technical features in the magnetic field disturbance analysis device based on ground observation data are the same as the features disclosed in the method of the previous embodiment, and will not be repeated here.

[0142] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples. The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0143] This embodiment provides a computer-readable storage medium having computer-readable program instructions stored thereon, which are used to execute the magnetic field disturbance analysis method based on ground observation data in Embodiment 1 above.

[0144] The computer-readable storage medium provided in this application embodiment may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor devices, apparatuses, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections with one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable EPROM (Electrical Programmable Read Only Memory) or flash memory, optical fiber, portable compact disk CD-ROM (compact discread-only memory), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution device, apparatus, or apparatus. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof. The aforementioned computer-readable storage medium may be included in a magnetic field disturbance analysis device based on ground observation data; or it may exist independently and not be assembled into a magnetic field disturbance analysis device based on ground observation data. The aforementioned computer-readable storage medium carries one or more programs. When these programs are executed by a magnetic field disturbance analysis device based on ground observation data, the device performs the following actions: acquires ground observation data, which includes station observation data corresponding to multiple stations; determines a target model for analyzing magnetic field disturbances, fitted based on the ground observation data; predicts first magnetic disturbance prediction data for each station using the target model; and jointly determines a first prediction error of the target model based on the first magnetic disturbance prediction data and the station observation data corresponding to each station; for each station, it is designated as a rejection station, and its station observation data is removed from the ground observation data to obtain rejection observation data; and determines a verification model for the rejected station, fitted based on the rejection observation data, for analyzing magnetic field disturbances; for each verification model, it predicts second magnetic disturbance prediction data for the rejected station using the verification model; and determines a second prediction error for the rejected station based on the second magnetic disturbance prediction data and the station observation data; and determines the prediction reliability of the target model based on the first prediction error and each of the second prediction errors.Computer program code for performing the operations of this disclosure can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a LAN (local area network) or WAN (wide area network)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0145] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of devices, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based device that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0146] The modules described in the embodiments of this disclosure can be implemented in software or hardware. The names of the modules do not necessarily limit the specific unit. The computer-readable storage medium provided in this application stores computer-readable program instructions for executing the above-described magnetic field disturbance analysis method based on ground observation data, aiming to solve the technical problem of determining the reliability of geomagnetic disturbance predictions. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as the beneficial effects of the magnetic field disturbance analysis method based on ground observation data provided in the above embodiments, and will not be repeated here.

[0147] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the magnetic field disturbance analysis method based on ground observation data as described above. The computer program product provided in this application aims to solve the technical problem of how to determine the reliability of geomagnetic disturbance predictions. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as those of the magnetic field disturbance analysis method based on ground observation data provided in the above embodiments, and will not be repeated here.

[0148] The above are merely preferred embodiments of the present application and do not limit the patent scope of the present application. Any equivalent structural or procedural transformations made using the description and drawings of the present application, or direct or indirect applications in other related technical fields, are similarly included within the patent processing scope of the present application.

Claims

1. A method of magnetic field disturbance analysis based on ground observation data, characterized by, The method comprises: acquiring ground observation data, wherein the ground observation data comprises station observation data corresponding to each station; determining a target model for analyzing magnetic field disturbance based on the ground observation data, predicting first magnetic disturbance prediction data of each station through the target model, and determining first prediction errors of the target model according to the first magnetic disturbance prediction data and the station observation data of each station respectively; for each station, taking the station as a removed station, removing the station observation data of the removed station from the ground observation data to obtain observation removal data, and determining a verification model of the removed station for analyzing magnetic field disturbance based on the observation removal data; for each verification model, predicting second magnetic disturbance prediction data of the removed station through the verification model, and determining second prediction errors of the removed station according to the second magnetic disturbance prediction data and the station observation data of the removed station; determining the prediction credibility of the target model according to the first prediction errors and the second prediction errors.

2. The method of claim 1, wherein the magnetic field disturbance analysis based on ground observation data is characterized by, The step of acquiring ground observation data comprises: acquiring an observation station list and an observation period; for each station in the observation station list, calling a data acquisition component corresponding to the station type of the station according to the station type to acquire initial observation data of the station in the observation period, wherein the initial observation data comprises sub-observation data corresponding to each observation time in the observation period; for each initial observation data, dividing the initial observation data into window observation data corresponding to a plurality of time windows according to a preset time length division unit, determining target sub-observation data satisfying a preset median condition in each window observation data, and sequentially splicing each target sub-observation data according to a preset time order to obtain standard observation data; for each station, determining the station observation data of the station according to the standard observation data of the station; wherein the preset median condition comprises that the target sub-observation data is the median in the window observation data.

3. The method of claim 1, wherein the magnetic field disturbance analysis based on ground observation data is characterized by, For each station, the step of determining the station observation data of the station according to the standard observation data of the station comprises: for each station, removing abnormal target sub-observation data in the standard observation data, and removing target sub-observation data in a first target time window in the standard observation data when it is detected that the first target time window does not belong to a preset time alignment range; when there is no target sub-observation data in a second target time window in the standard observation data, acquiring a first adjacent window and a second adjacent window adjacent to the second target time window; wherein the second target time window is any preset observation window in the preset time alignment range; linearly interpolating the target sub-observation data of the first adjacent window and the target sub-observation data of the second adjacent window to obtain target sub-observation data of the second target time window; the standard observation data excluding the target sub-observation data of the abnormality, excluding the target sub-observation data of the first target time window, and including the target sub-observation data of each preset observation window within the preset time alignment range, as the station observation data.

4. The method of claim 1, wherein the magnetic field disturbance analysis based on ground observation data is characterized by, the target model includes at least one target sub-model corresponding to a preset observation window, and the target sub-model is used for analyzing magnetic field disturbance generated in the preset observation window; the step of determining the target model for analyzing magnetic field disturbance based on the ground observation data fitting includes: obtaining target sub-observation data of each station corresponding to a same preset observation window from the ground observation data; constructing an observation matrix of the preset observation window according to each target sub-observation data and a corresponding observation coordinate; back-propagating equivalent current coefficients of each singular point in the current singular point grid to obtain a target sub-model of the preset observation window constructed by the equivalent current coefficients in the singular point grid.

5. The method of claim 4, wherein the magnetic field disturbance analysis based on ground observation data is characterized by, the step of back-propagating the equivalent current coefficients of each singular point in the current singular point grid according to the preset constructed current singular point grid and the observation matrix includes: modifying the observation matrix to obtain a weighted matrix according to an error matrix constructed by a preset observation error weight of each station; calculating initial magnetic disturbance prediction data of each station under the current singular point grid according to a preset initial equivalent coefficient of each singular point in the current singular point grid, and constructing an initial prediction matrix according to the initial magnetic disturbance prediction data of each station and a corresponding observation coordinate of each station; constructing a final optimization problem corresponding to the equivalent current coefficients of each singular point in the current singular point grid according to the weighted matrix, the initial prediction matrix, and a preset regularization parameter; back-propagating the equivalent current coefficients of each singular point by performing a preset matrix decomposition on the final optimization problem.

6. The method of claim 4, wherein the magnetic field disturbance analysis based on ground observation data is characterized by, the first magnetic disturbance prediction data includes first window prediction data of the station predicted by the target sub-model, and the station observation data includes target sub-observation data of a preset observation window; the step of collectively determining the first prediction error of the target model according to the first magnetic disturbance prediction data and the station observation data corresponding to each station respectively includes: for each station, calculating a difference between the first window prediction data and the target sub-observation data in a same preset observation window to obtain a first window error; collectively using each first window error as the first prediction error.

7. The method of claim 1, wherein the magnetic field disturbance analysis based on ground observation data is characterized by, The check model comprises at least one check sub-model of a preset observation window, and the check sub-model is used for analyzing the magnetic field disturbance generated in the preset observation window; the second magnetic disturbance prediction data comprises second window prediction data of the excluded station predicted by the check sub-model; and the station observation data comprises target sub-observation data of a preset observation window. The step of determining the second prediction error of the excluded station according to the second magnetic disturbance prediction data and the station observation data of the excluded station comprises: For each preset observation window, a difference between the second window prediction data and the target sub-observation data of the excluded station in the same preset observation window is calculated to obtain a second window error of the preset observation window; The second window errors are collectively used as the second prediction error.

8. The method of claim 1, wherein the magnetic field disturbance analysis based on ground observation data is characterized by, The prediction credibility comprises station credibility; The step of determining the prediction credibility of the target model according to the first prediction error and the second prediction error comprises: For each second prediction error of the excluded station, a difference between the second window error in the same preset observation window and the first window error in the first prediction error is determined to obtain a check error; For each excluded station, the station credibility is determined according to the check error of the excluded station, and the station credibility is negatively correlated with the check error.

9. A magnetic field disturbance analysis device based on ground observation data, characterized by, The magnetic field disturbance analysis device comprises: at least one processor; and a memory connected in communication with the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the steps of the magnetic field disturbance analysis method based on ground observation data according to any one of claims 1 to 8.

10. A storage medium, characterized by The storage medium is a computer readable storage medium, and the computer readable storage medium stores a program for implementing the magnetic field disturbance analysis method based on ground observation data. The program for implementing the magnetic field disturbance analysis method based on ground observation data is executed by the processor to implement the steps of the magnetic field disturbance analysis method based on ground observation data according to any one of claims 1 to 8.

Images