Method, apparatus and system for predicting southward component of interplanetary magnetic field using bi-lstm
A Bi-LSTM-based prediction model for the southward component of the interplanetary magnetic field improves accuracy and real-time capability, addressing the limitations of existing methods by effectively handling extreme conditions and ensuring stable predictions for satellite and communication systems.
Patent Information
- Application Number
- KR1020250059892
- Authority / Receiving Office
- KR · KR
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2025-05-08
- Publication Date
- 2026-11-17
AI Technical Summary
Existing methods for predicting the southward component of the interplanetary magnetic field are inadequate in terms of accuracy and real-time capability, particularly during extreme or rapidly changing conditions, posing risks to satellite operations, power grids, and communication systems.
A machine learning model, specifically a Bi-LSTM-based prediction model, is trained using time series data of magnetic field components and related solar wind variables, with preprocessing to remove noise and outliers, and optimized using techniques like dropout and early stopping to enhance prediction accuracy.
The model achieves superior prediction accuracy and stability, reducing errors by up to 25-29% compared to conventional methods, enabling reliable long-term forecasting for space weather responses.
Smart Images

Figure PAT00002_ABST
Abstract
Description
Technology Field
[0001] The present disclosure relates to a method, apparatus, and system for predicting the southward component of an interplanetary magnetic field using a machine learning model. Background Technology
[0002] The southward component of the interplanetary magnetic field acts as a key factor in triggering various space weather phenomena, such as geomagnetic storms, through interactions with the Earth's magnetosphere. In particular, if this component extends strongly in a southward direction (negative direction), it activates magnetic reconnection between the solar wind and the Earth's magnetic field, which can lead to severe impacts on various technology-based infrastructures, including satellite malfunctions, power grid instability, and communication disruptions. Accordingly, the technology to precisely predict changes in the southward component of the interplanetary magnetic field is recognized as a critical technical challenge for ensuring the stability of satellite operations, the aerospace industry, power grid protection, and communication systems.
[0003] Traditionally, magnetohydrodynamics (MHD)-based physical models and statistical-based empirical models have been primarily utilized to predict the southward component of interplanetary magnetic fields. While physical models enable relatively reliable predictions by numerically simulating the complex interactions between the solar wind and the magnetosphere, they face limitations in real-time prediction due to the high computational costs and time-consuming computations required. Conversely, empirical models offer the advantage of rapid prediction based on statistical correlations between historical observational data; however, they are limited in prediction accuracy because they fail to adequately reflect the nonlinear and rapidly changing characteristics of the southward component of the interplanetary magnetic field.
[0004] Against this backdrop, research applying machine learning techniques to simultaneously improve prediction accuracy and processing speed has recently been actively underway. In particular, while some results have emerged with the introduction of deep learning-based approaches, most existing models focus on average magnetic field conditions, leaving their prediction performance regarding strong southward magnetic fields or situations of rapid change still inadequate. Despite the increasing demand for prediction models capable of effectively capturing extreme change situations, sufficient technology to meet this need has not yet been secured.
[0005] Therefore, the industry urgently requires the development of technology capable of predicting the southward component of the interplanetary magnetic field with greater precision and near real-time accuracy for responding to space weather and protecting related infrastructure. The problem to be solved
[0006] The present disclosure aims to accurately predict the strength of the southward component of the interplanetary magnetic field in real time.
[0007] The problems that this disclosure aims to solve are not limited to those described above, and other unmentioned problems will be clearly understood by a person skilled in the art from the description below. means of solving the problem
[0008] A method for predicting magnetic field components performed by one or more processors according to one embodiment of the present disclosure, wherein the method for predicting magnetic field components comprises: a step of generating training data for training a prediction model unit based on raw training data including a previously stored actual value of a magnetic field component; and a step of performing training on the prediction model unit based on the training data; wherein the step of generating training data may be characterized by comprising: a step of determining output training data including one or more magnetic field component values during a first time interval satisfying a target data criterion; and a step of determining input training data including one or more magnetic field-related variable data during a second time interval consecutive to the first time interval.
[0009] In addition, the above target data criteria may be characterized by being determined based on one or more magnetic field component statistical values related to previously stored actual magnetic field component values.
[0010] In addition, the above target data standard may be characterized as a magnetic field value that is less than or equal to a preset multiple of the standard deviation from the average of the previously stored actual magnetic field component values.
[0011] In addition, the above target data standard may be characterized as a magnetic field value that is less than or equal to one multiple of the standard deviation of the previously stored actual magnetic field component value.
[0012] In addition, the above target data standard may be characterized by having a magnetic field component value of -3nT or less.
[0013] In addition, the one or more magnetic field-related variable data included in the input learning data may be characterized by including at least one solar wind-related data or one or more different magnetic field component data.
[0014] In addition, the above solar wind-related data may be characterized by including at least one of solar wind velocity, solar wind density, or solar wind temperature.
[0015] In addition, the above training data may be characterized by being composed of the above raw training data divided by month.
[0016] In addition, the prediction model unit may be characterized as being a machine learning model configured to enable processing of time series data.
[0017] In addition, the above prediction model unit may be characterized by being implemented using Bi-LSTM (Bi-directional Long Short Term Memory).
[0018] A magnetic field component prediction device according to some embodiments of the present invention comprises: one or more processors; and a memory; wherein the one or more processors generate training data for training a prediction model unit based on raw training data including previously stored actual values of magnetic field components, and perform training on the prediction model unit based on the training data; wherein the one or more processors generating the training data may be characterized by determining output training data including one or more magnetic field component values during a first time interval that satisfies target data criteria, and determining input training data including one or more magnetic field-related variable data during a second time interval that is continuous with the first time interval. Effects of the invention
[0019] By automating parameter extraction and boundary condition setting for CAE analysis according to the present disclosure, the efficiency of CAE analysis can be increased.
[0020] The effects according to the present disclosure are not limited to those described above, and other unmentioned effects will be clearly understood by a person skilled in the art from the description below. Brief explanation of the drawing
[0021] FIG. 1 is a block diagram showing at least part of the configuration of a magnetic field southward component prediction device according to some embodiments of the present disclosure. FIG. 2 is a conceptual diagram showing the configuration of a prediction model unit according to some embodiments of the present disclosure. FIG. 3 is a conceptual diagram showing the characteristics of the output data of a prediction model unit according to some embodiments of the present disclosure. FIG. 4 is a conceptual diagram showing input data and output data of a prediction model unit according to some embodiments of the present disclosure. FIGS. 5a to 5c are drawings showing the prediction experimental results of a magnetic field southward component prediction device according to some embodiments of the present disclosure. FIG. 6 is a diagram comparing the error level of a magnetic field southward component prediction device according to some embodiments of the present disclosure with that of the prior art. FIG. 7 is a block diagram showing the configuration of the magnetic field southward component prediction device of the present disclosure. Specific details for implementing the invention
[0022] Hereinafter, exemplary embodiments according to the present invention will be described in detail with reference to the contents described in the attached drawings. However, the present invention is not limited or restricted by exemplary embodiments. Unless otherwise defined, all terms used in this specification (including technical and scientific terms) shall be used in a meaning that is commonly understood by those skilled in the art to which this disclosure belongs, but this may vary depending on the intent of those skilled in the art, case law, the emergence of new technology, etc.
[0023] Furthermore, terms defined in commonly used dictionaries are not to be interpreted ideally or excessively unless explicitly and specifically defined otherwise. In certain cases, terms have been selected at the applicant's discretion, and in such cases, their meanings will be described in detail in the relevant explanatory sections. Accordingly, terms used in this disclosure should be defined not merely by their names, but based on their meanings and the content throughout this disclosure.
[0024] Throughout this specification, when a part is described as "comprising" a certain component, this means that, unless specifically stated otherwise, it does not exclude other components but may include additional components. Furthermore, the singular form used in this specification includes the plural form unless specifically stated otherwise. Additionally, the expression "at least one of a, b, and / or c" as used throughout this specification may encompass 'a alone', 'b alone', 'c alone', 'a and b', 'a and c', 'b and c', or 'a, b, and c all'.
[0025] Meanwhile, terms such as "first and / or second" used in this specification may be used to describe various components, but they are used solely for the purpose of distinguishing one component from another and are not intended to limit the scope to the components referred to by such terms. For example, without departing from the scope of the present invention, the first component may be named the second component, and the second component may also be named the first component.
[0026] Additionally, terms such as “…part,” “…module,” etc., as described in this specification refer to a unit that processes at least one function or operation, which may be implemented in hardware or software, or a combination of hardware and software. Furthermore, embodiments of this disclosure may be represented in this specification by functional block configurations and various processing steps. These functional blocks may be implemented by various numbers of hardware and / or software configurations that execute specific functions. For example, embodiments of this disclosure may employ integrated circuit configurations such as memory, processing, logic, look-up tables, etc., which can execute various functions under the control of one or more microprocessors or other control devices.
[0027] Similar to how the components disclosed herein may be executed as software programs or software elements, embodiments of the present disclosure may be implemented in programming or scripting languages such as C, C++, Java, assembler, etc., including various algorithms implemented as combinations of data structures, processes, routines, or other programming configurations. Functional aspects may be implemented as algorithms executed on one or more processors. Additionally, the present embodiments may employ prior art for at least one of electronic configuration, signal processing, and data processing. Terms such as “mechanism,” “element,” “means,” and “configuration” may be used broadly and are not limited to mechanical and physical configurations. The above terms may include the meaning of a series of software processes (routines) in conjunction with a processor, etc.
[0028] Each block of the process flow diagrams attached to this specification and combinations of the flow diagrams may be executed by computer program instructions. Since these computer program instructions may be loaded into the processor of a general-purpose computer, a computer for special purposes, or other programmable data processing equipment, the instructions executed through the processor of the computer or other programmable data processing equipment create means for performing the functions described in the flow diagram block(s).
[0029] These computer program instructions may be stored in computer-available or computer-readable memory that can be directed toward a computer or other programmable data processing equipment to implement a function in a specific way, and the instructions stored in said computer-available or computer-readable memory may also produce a manufactured item containing instruction means that performs the function described in the flowchart block(s).
[0030] Since computer program instructions can be loaded onto a computer or other programmable data processing equipment, instructions that perform a series of operation steps on the computer or other programmable data processing equipment to create a process executed by the computer can also provide steps for executing the functions described in the flowchart block(s).
[0031] Additionally, each block may represent a module, segment, or part of code containing one or more executable instructions for executing a specified logical function(s). Furthermore, in some alternative execution examples, the functions mentioned in the blocks may occur out of order. For instance, two blocks described in succession may actually be executed substantially simultaneously, or the blocks may be executed in reverse order according to their corresponding functions.
[0032] The “electronic device” or “terminal” mentioned in this specification may be implemented as a computer or portable terminal capable of connecting to a server or other terminal via a network. Here, the computer includes, for example, a notebook, desktop, or laptop equipped with a web browser, and the portable terminal may include, for example, any type of handheld wireless communication device that ensures portability and mobility, such as a communication-based terminal like IMT (International Mobile Telecommunication), CDMA (Code Division Multiple Access), W-CDMA (W-Code Division Multiple Access), or LTE (Long Term Evolution), as well as a smartphone or tablet PC. Additionally, the “electronic device” or “terminal” mentioned in this specification may also include a processor, memory for storing and executing program data, permanent storage such as a disk drive, a communication port for communicating with an external device, and user interface devices such as a touch panel, a key, or a button.
[0033] In the present disclosure, methods implemented as software modules or algorithms may be stored on a computer-readable recording medium as computer-readable code or program instructions executable on a processor. The computer-readable recording medium may include magnetic storage media (e.g., ROM (read-only memory), RAM (random-access memory), floppy disks, hard disks, etc.) and optical reading media (e.g., CD-ROM, DVD (Digital Versatile Disc)). The computer-readable recording medium may be distributed and executed across networked computer systems.
[0034] Hereinafter, various embodiments of the present disclosure will be described in detail with reference to the accompanying drawings. In describing the embodiments, technical details that are well known in the art to which the present invention pertains and are not directly related to the present invention will be omitted. This is to ensure that the essence of the present invention is conveyed more clearly without obscuring it by omitting unnecessary explanations. For the same reason, some components in the accompanying drawings may be exaggerated, omitted, or schematically depicted. Furthermore, the size of each component does not entirely reflect its actual size. Throughout this specification, the same reference numerals may refer to the same or corresponding components.
[0036] FIG. 1 is a block diagram showing at least part of the configuration of a magnetic field southward component prediction device according to some embodiments of the present disclosure.
[0037] Referring to FIG. 1, a magnetic field southward component prediction device (100) according to some embodiments of the present disclosure may include a preprocessing unit (110).
[0038] In some embodiments, the preprocessing unit (110) may be a module implemented in hardware, software and / or a combination thereof for performing preprocessing on data input to the prediction model unit (130).
[0039] In some embodiments, the preprocessing unit (110) may perform preprocessing on raw time series learning data of the magnetic field southward component input to the prediction model unit (130). In one embodiment, the preprocessing performed by the preprocessing unit (110) may include the removal of noise components. In one embodiment, the removal of noise components performed by the preprocessing unit (110) may include the removal of outliers, etc.
[0040] In another embodiment, the preprocessing unit (110) may perform preprocessing on raw time series learning data of the magnetic field southward component for training the prediction model unit (130). At this time, the preprocessing of the learning data performed by the preprocessing unit (110) may include outlier removal and / or noise removal. In one embodiment, the noise removal of the learning data by the preprocessing unit (110) may include the removal of raw time series learning data that does not satisfy a preset target data criterion. At this time, the target data criterion may use a statistical value that has been stored for the raw time series learning data of the magnetic field southward component. For example, the target data criterion may be set by introducing various statistical values such as the mean, variance, standard deviation and / or median of the raw time series learning data of the magnetic field southward component.
[0041] In some other embodiments, the preprocessing unit (110) may divide raw data into multiple intervals and transmit them to the prediction model unit (130). For example, the preprocessing unit (110) may divide the raw data into pre-set time interval units to form input data and transmit it as input to the prediction model unit (130).
[0042] Referring again to FIG. 1, a magnetic field southward component prediction device (100) according to some embodiments may include a learning unit (120) for learning a prediction model unit (130).
[0043] The learning unit (120) may perform a series of processing steps to train a deep learning model included in the prediction model unit (130) based on time series data of the magnetic field southward component, and may include one or more processors and / or parallel computing resources for this purpose. For example, the learning unit may include a central processing unit (CPU), a graphics processing unit (GPU), a tensor processing unit (TPU), or other computing circuits, and may include an instruction set or computer-readable code for executing a learning algorithm.
[0044] The learning unit may include computational instructions for calculating a loss function, backpropagation, and parameter optimization. For example, the loss function may be the Mean Squared Error (MSE), Mean Absolute Error (MAE), or Cross Entropy, each of which may be included in the form of mathematical computational instructions executable by computational resources.
[0045] Additionally, the learning unit may execute an optimization routine to update the learning parameters of the prediction model, and for this purpose, it may include a set of optimization algorithm instructions related to Stochastic Gradient Descent (SGD), Adam, RMSProp, AdaGrad, etc. These instruction sets may consist of numerical optimization routines that compute gradient information for each parameter and calculate update values based thereon.
[0046] In addition, the learning unit may include a storage structure capable of storing or setting hyperparameters such as the learning rate, batch size, and number of epochs, and control commands that control model training using said parameters. In some embodiments, processing modules related to command codes for automatically searching for combinations of hyperparameters, such as grid search, random search, and Bayesian optimization, may be further included.
[0047] In addition, to prevent overfitting and improve generalization performance, the training unit may include control logic and operation instructions related to dropout, normalization (L1 or L2 norm), early stopping, or cross-validation, and each may be configured to be executable by one or more operation resources.
[0048] In some embodiments, the learning unit (120) may organize raw learning data or learning data into pre-set interval-unit folds to perform cross-validation. For example, in one embodiment, a single fold for cross-validation may be constructed using raw learning data and / or learning data in annual, semi-annual, quarterly, monthly, weekly, and / or daily units.
[0049] After the learning is completed, the learning unit can transmit the learned parameters to the prediction model unit (130) or record them in a storage module, such as a memory, accessible by the magnetic field southward prediction component prediction device (100) according to the present disclosure, so that the parameters can be used in the subsequent prediction process.
[0050] Referring again to FIG. 1, a prediction model unit (130) according to some embodiments is a component that generates a prediction result of a magnetic field southward component corresponding to a preset prediction period based on data transmitted from a preprocessing unit (110), and can be implemented in hardware, software and / or a combination thereof.
[0051] The prediction model unit (130) may include an instruction set or code for executing a machine learning algorithm and one or more processors and / or computational resources capable of processing it. For the prediction of time series data, the prediction model unit may be composed of a deep learning model based on a Recurrent Neural Network (RNN), a Long Short-Term Memory (LSTM), or a Transformer, and such a configuration may include a computational structure for analyzing the temporal correlation of input time series data and predicting future data.
[0052] The prediction model unit can learn internal parameters through the execution of learning commands and optimization algorithms provided by the learning unit (120), and then perform a prediction function. In the initialization phase, the model parameters can be set through random initialization, statistical initialization, and / or pre-training.
[0053] As an example of the prediction model section, a Bidirectional Long Short-Term Memory (Bi-LSTM) structure may be applied. By including forward and backward LSTM cells that can simultaneously consider information regarding the past and future directions for the input time series, the Bi-LSTM can enable predictions that reflect richer contextual information. This Bi-LSTM-based prediction model can effectively process data where patterns based on time order are important, such as the southward component of a magnetic field, and can provide high prediction accuracy when combined with multilayering and normalization techniques.
[0055] FIG. 2 is a conceptual diagram showing the configuration of a prediction model unit according to some embodiments of the present disclosure.
[0056] Referring to FIG. 2, a prediction model unit (130) according to some embodiments of the present disclosure may be implemented using a deep learning model of a structure for processing time series data.
[0057] In one embodiment, the prediction model unit (130) may be composed of a plurality of time series processing layers, a non-linear activation layer, and an output layer for processing input data. In one embodiment, the input data and output data may include data for one or more time intervals. In one embodiment, the input data may include one or more input variable data necessary to predict the strength of the magnetic field southward component.
[0058] Referring to FIG. 2, in one example of the present disclosure, the input data may be two-dimensional sequence data having dimensions of [24, 7], and may include data for 24 different or consecutive time intervals prior to the prediction reference time for each of the total 7 input variables. In one embodiment, when the output data is 12-dimensional vector data, it may include magnetic field southward component values for 12 different or consecutive time intervals. This configuration of input and output data allows the fluctuation patterns of magnetic field components and their variables to be expressed as high-resolution time series, which can be usefully utilized for real-time satellite operation or communication system alerts.
[0059] In one embodiment of the present disclosure illustrated in FIG. 2, the prediction model unit (130) may include a Bi-LSTM layer, and each layer may be configured to include a plurality (e.g., 128) of hidden units. The Bi-LSTM layer calculates the hidden state based on a bidirectional information flow (forward and backward) for an input sequence, which can contribute to improving the prediction accuracy of the southward component of the magnetic field by simultaneously considering the contextual information of the time sequence within the time series data. The output of each Bi-LSTM layer may be processed through a linear or non-linear transformation layer (e.g., ReLU (Rectified Linear Unit)) including an activation function.
[0060] However, the Bi-LSTM structure and ReLU function illustrated in FIG. 2 are merely exemplary embodiments and are not limited thereto. The time series processing layer can be replaced with various time series neural network structures other than Bi-LSTM, such as unidirectional LSTM, GRU (Gated Recurrent Unit), Transformer, or Temporal Convolutional Network (TCN). Additionally, the activation function can be implemented by replacing it with other non-linear functions other than ReLU, such as Sigmoid, Tanh, GELU (Gaussian Error Linear Unit), or Swish.
[0061] Meanwhile, the overall structure of the prediction model unit (130) can be generalized into a multilayer neural network architecture consisting of an input layer, a time series prediction layer, a nonlinear activation layer, and an output layer, and each layer can be implemented as an independent processing unit or configured as an integrated computational structure.
[0063] FIG. 3 is a conceptual diagram showing the characteristics of the output data of a prediction model unit according to some embodiments of the present disclosure.
[0064] FIG. 3 is a conceptual diagram illustrating input and output training data regions utilized by a prediction model unit according to some embodiments of the present disclosure for configuring training data, and target data criteria for determining them.
[0065] Referring to FIG. 3, the prediction model unit may include input training data (220) and output training data (230) determined based on a preset target data standard (210).
[0066] In some embodiments, one or more processors included in the magnetic field south direction component prediction device may operate in such a way that, in order to construct the training data of the prediction model, they first select output training data (230) according to the target data criteria (210) and derive corresponding input training data (220).
[0067] The target data criterion (210) can be set as a threshold condition to identify a section in which the magnetic field southward component to be predicted is strongly southward (i.e., numerically smaller negative value) above a certain level. For example, the target data criterion may be a time series section in which the value of the magnetic field southward component is below a preset threshold value, or at least one of the mean, median, or quantile of the component is below the threshold value.
[0068] A reference value can be set based on statistical characteristics of the magnetic field component values. As an example, the mean, variance, standard deviation, median, and / or interquartile range (IQR) may be utilized, and a value located a specific multiple of the standard deviation from the mean may be used as a reference. In one embodiment, the reference value may be a value corresponding to -1 standard deviation from the mean. Through this reference, the output training data (230) can be identified as a time series interval in which a statistically significant southward magnetic field is observed.
[0069] It is generally known that predicting the strength of the southward component of a magnetic field is technically very difficult. This is because, due to the time-series characteristics of the strength of the southward component of a magnetic field, its average value converges to zero, and since most data is concentrated at weak values, most predicted values converge to zero when machine learning methodologies are applied to prediction. Therefore, as described in the present disclosure, if prediction data is set based on a statistically significant threshold value and training and prediction are performed on this data, the problems arising from prediction using existing machine learning methodologies can be overcome. In this case, it may be necessary to select a threshold value that is not excessively sparse in order to secure an appropriate amount of training data.
[0070] When output training data (230) is selected, the processor may determine time series data collected during a certain period prior to the start of the corresponding section as input training data (220). At this time, the input training data (220) may not only include magnetic field component values but may also include additional variable data to improve prediction accuracy.
[0071] For example, the input training data (220) may include physical variables related to the solar wind, such as solar wind speed (SW speed), solar wind density (SW density), solar wind temperature (SW temperature), solar flux index, and geomagnetic activity index (Kp index), in addition to the southward component of the magnetic field, or data on other directional components (Bx, By, Bz) and total amount (Bt) of the magnetic field vector. These multivariate inputs help the model to capture the causes of complex magnetic field changes.
[0072] A more detailed description of the input learning data (220) and its composition is explained with reference to FIG. 4, which will be described later.
[0074] FIG. 4 is a conceptual diagram showing input data and output data of a prediction model unit according to some embodiments of the present disclosure.
[0075] FIG. 4 is a conceptual diagram illustrating the configuration of input data and output data of a prediction model unit according to some embodiments of the present disclosure.
[0076] Referring to FIG. 4, the prediction model unit may receive and process input data (131) to generate output data (132), and the input data may include one or more variable data necessary to predict the future time series of the southward component of the magnetic field.
[0077] The input data (131) may be constructed based on time series data corresponding to a continuous N-hour period prior to the prediction reference point. In this case, the input data may include multiple variable data for each of the multiple time points divided into time intervals W. The variable data may include information related to solar wind and information on the strength of the multi-directional components of the magnetic field to improve prediction performance.
[0078] More specifically, in this embodiment, the input data (131) may include solar wind speed, solar wind density, solar wind temperature, magnetic field components (Bx, By, Bz), and total magnetic field strength (Bt). This information may be defined based on a geomagnetic coordinate system (GSM coordinate system) and is input into the model in the form of a multidimensional time series tensor in which the above variables are arranged for each time interval.
[0079] The output data (132) may consist of time series data for a specific direction magnetic field component, for example, a south-facing magnetic field component, for M hours after the prediction reference point. The output data also corresponds to multiple time points divided into time intervals W, and may be calculated in the form of a vector containing continuous values of the magnetic field component to be predicted. The prediction interval M hours can be set in various ways according to system requirements; for example, if the prediction is 6 hours, M=6 and W=30 minutes, it consists of prediction values for a total of 12 time points.
[0080] In one embodiment, the total length (N) of the input time series data may be set to be longer than the total length (M) of the output time series data. When the time division unit (W) is the same, the input data reflects a longer observation interval, thereby supporting the model in learning a richer temporal context.
[0081] This configuration can be dynamically configured according to the selection criteria of the input training data (220) and output training data (230) described in FIG. 3, and can be automatically determined by the dataset configuration module in a real-time or offline environment.
[0083] FIGS. 5a to 5c are drawings showing the prediction experimental results of a magnetic field southward component prediction device according to some embodiments of the present disclosure.
[0084] Figures 5a to 5c respectively show cases where the strength of the southward component of the magnetic field is moderate (Fig. 5a, "Moderate level case"), strong (Fig. 5b, "Strong case"), and extremely strong (Fig. 5c, "Extreme case"). In each figure, the black solid line represents the input time series data prior to the prediction point, the green solid line represents the actual observed value (Target), the red solid line represents the average of the predicted value, and the red shading represents the confidence interval (predicted variance range) of the predicted value.
[0085] In the normal level case of Fig. 5a, it can be confirmed that the prediction accuracy is excellent, as the model's predicted values are very close to the actual observed values in an environment where the strength of the southward component of the magnetic field is relatively stable. The difference between the predicted mean and the actual values is minimal, and the predicted variance also converges within a narrow range, demonstrating high reliability.
[0086] The strong case in Fig. 5b and the extreme case in Fig. 5c involve situations where the strength of the southward component of the magnetic field decreases sharply to approximately -25nT and -50nT or more, respectively. Even in such high-variability environments, the prediction model stably tracks the overall trend of magnetic field changes, and the predicted average value maintains a directionality similar to the actual observed value. In particular, in Fig. 5c, even if there is a difference from the actual value in some parts of the prediction interval, the robust generalization performance of the model can be demonstrated by accurately reflecting the overall trend and including most of the actual value within the prediction interval.
[0088] FIG. 6 is a diagram comparing the error level of a magnetic field southward component prediction device according to some embodiments of the present disclosure with that of the prior art.
[0089] Table 1 below quantifies the contents of Figure 6 and shows the results of a quantitative comparison of the Root Mean Square Error (RMSE) for each prediction time point.
[0090] RMSE Comparison between the present invention and conventional models (Unit: nT) The present invention Conventional model Prediction interval Bi-LSTM-based prediction model section Multilayer perceptron Multiple regression model 30 minutes 1.75 1.63 1.51 1 hour 1.92 2.02 2.03 1 hour 30 minutes 2.07 2.22 2.31 2 hours 2.05 2.44 2.58 2 hours 30 minutes 2.10 2.63 2.78 3 hours 2.12 2.82 2.97 3 hours 30 minutes 2.18 2.96 3.08 4 hours 2.25 3.05 3.16 4 hours 30 minutes 2.29 3.14 3.22 5 hours 2.30 3.18 3.23 5 hours 30 minutes 2.35 3.24 3.26 6 hours 2.55 3.27 3.30
[0092] As shown in FIG. 6 and Table 1, it can be confirmed that the magnetic field southward component prediction device (Bi-LSTM) using the prediction model unit according to some embodiments of the present disclosure maintains a lower RMSE compared to the conventionally used Multi Layer Perceptron (MLP) and Multi Linear Regressor (MLR) models throughout the entire prediction interval, even though the prediction error gradually increases as the prediction time increases.
[0093] In particular, the MLR model shows the lowest error at the 30-minute prediction time, but from the time the prediction time exceeds 1 hour, the model (Bi-LSTM) associated with the magnetic field southward component prediction device according to the embodiment of the present invention continuously maintains the best prediction accuracy, and as the prediction time increases, the performance gap with existing models becomes more pronounced.
[0094] For example, at a prediction time of 3 hours, the RMSE of the model according to the present invention is 2.12 nT, showing an error reduction effect of approximately 25% and 29% compared to MLP (2.82 nT) and MLR (2.97 nT), respectively. Even at the longest prediction time of 6 hours, the prediction model according to the present disclosure records an RMSE of 2.55 nT, maintaining excellent performance compared to MLP (3.27 nT) and MLR (3.30 nT).
[0095] The prediction model according to the present disclosure is much less sensitive to the accumulation of errors due to time delay compared to existing multilayer perceptron or linear regression-based prediction models, and provides a structure capable of maintaining stable and consistent performance even in long-term prediction intervals. In particular, for variables in which nonlinearity and time dependence interact in a complex manner in a time series, such as the southward component of a magnetic field, the present invention simultaneously secures prediction accuracy and generalization performance by applying a Bi-LSTM structure capable of learning the time series context bidirectionally.
[0096] This performance offers practical applicability in various industrial sectors requiring reliable long-term forecasting, such as satellite orbit maintenance, communication failure prediction, and space weather response systems. Furthermore, the gradual rise in the error curve over time underpins its technical superiority, as it guarantees stable results even in real-time operating environments.
[0098] FIG. 7 is a block diagram showing the configuration of a magnetic field southward component prediction device of the present disclosure.
[0099] The computing device (800) may include a memory (810), a processor (820), a communication unit (830), and an input / output interface (840), and as shown in FIG. 7, the computing device (800) may be configured to communicate information and / or data through a network using the communication unit (830).
[0100] The memory (810) may include any non-transient computer-readable recording medium. According to one embodiment, the memory (810) may include a permanent mass storage device such as random access memory (RAM), read-only memory (ROM), disk drive, solid state drive (SSD), flash memory, etc. As another example, a permanent mass storage device such as ROM, SSD, flash memory, disk drive, etc. may be included in the computing device (800) as a separate permanent storage device distinct from the memory. Additionally, an operating system and at least one program code may be stored in the memory (810).
[0101] These software components may be loaded from a computer-readable recording medium separate from the memory (810). This separate computer-readable recording medium may include a recording medium that can be directly connected to the computing device (800), for example, a computer-readable recording medium such as a floppy drive, disk, tape, DVD / CD-ROM drive, or memory card. As another example, the software components may be loaded into the memory (810) via a communication unit (830) rather than a computer-readable recording medium. For example, at least one program may be loaded into the memory (810) based on a computer program installed by files provided through the communication unit (830) by developers or a file distribution system that distributes installation files for applications.
[0102] The program described herein may include program instructions, data files, and data structures, either individually or in combination. The program may be designed and created using machine code or high-level language code. The program may be specifically designed to implement the invention described above, or it may be implemented using various functions or definitions that are known and available to those skilled in the art in the field of computer software. A program for implementing the invention described above may be recorded on a recording medium readable by a processor.
[0103] Memory can store a program that performs the aforementioned operation and the operation described below, and the processor can execute the program stored in memory. When there are multiple processors and memory, it is possible for them to be integrated on a single chip or to be provided in physically separate locations. Memory may include volatile memory such as S-RAM (Static Random Access Memory, S-RAM) and D-RAM (Dynamic Random Access Memory) for temporarily storing data. Additionally, memory may include non-volatile memory such as ROM (Read Only Memory), EPROM (Erasable Programmable Read Only Memory), and EEPROM (Electrically Erasable Programmable Read Only Memory) for long-term storage of control programs and control data.
[0105] The processor (820) may be configured to process instructions of a computer program by performing basic arithmetic, logic, and input / output operations. Instructions may be provided to another user terminal (not shown) or another external system by memory (810) or communication unit (830).
[0106] The processor may include various logic circuits and arithmetic circuits, process data according to a program provided from memory, and generate control signals based on the processing results. In this case, the memory and the processor may each be implemented as separate chips. Alternatively, the memory and the processor may be implemented as a single chip.
[0107] The processor (820) may include one or more processors. In this case, the one or more processors may be homogeneous or heterogeneous processors. For example, the processor (820) may include heterogeneous processors such as a central processing unit (CPU), a graphics processing unit (GPU), a tensor processing unit (TPU) and / or a neural processing unit (NPU). For convenience of explanation, one or more homogeneous or heterogeneous processors may be referred to as 'processors' within this specification.
[0108] The communication unit (830) may provide a configuration or function for a user terminal (not shown) and a computing device (800) to communicate with each other via a network, and may provide a configuration or function for the computing device (800) to communicate with an external system (e.g., a separate cloud system). For example, control signals, commands, data, etc. provided under the control of the processor (820) of the computing device (800) may be transmitted to the user terminal and / or the external system through the communication unit (830) and the network, and through the communication unit of the user terminal and / or the external system.
[0109] The wired communication unit may include various wired communication units such as a Local Area Network (LAN) module, a Wide Area Network (WAN) module, or a Value Added Network (VAN) module, as well as various cable communication units such as USB (Universal Serial Bus), HDMI (High Definition Multimedia Interface), DVI (Digital Visual Interface), RS-232 (recommended standard 232), power line communication, or POTS (plain old telephone service).
[0110] In addition to Wi-Fi modules and WiBro (Wireless broadband) modules, the wireless communication unit may include a wireless communication unit that supports various wireless communication methods such as GSM (global System for Mobile Communication), CDMA (Code Division Multiple Access), WCDMA (Wideband Code Division Multiple Access), UMTS (universal mobile telecommunications system), TDMA (Time Division Multiple Access), LTE (Long Term Evolution), 4G, 5G, and 6G.
[0111] The wireless communication unit may include a wireless communication interface comprising an antenna and a transmitter that transmit a Wi-Fi signal. Additionally, the wireless communication unit may further include a Wi-Fi signal conversion module that modulates a digital control signal output from the control unit through the wireless communication interface into an analog wireless signal under the control of the control unit.
[0112] The wireless communication unit may include a wireless communication interface comprising an antenna and a receiver for receiving a Wi-Fi signal. Additionally, the wireless communication unit may further include a Wi-Fi signal conversion module for demodulating an analog wireless signal received through the wireless communication interface into a digital control signal.
[0113] The short-range communication unit is for short-range communication and can support short-range communication by using at least one of Bluetooth, RFID (Radio Frequency Identification), Infrared Data Association (IrDA), UWB (Ultra Wideband), ZigBee, NFC (Near Field Communication), Wi-Fi (Wireless-Fidelity), Wi-Fi Direct, and Wireless USB (Wireless Universal Serial Bus) technologies.
[0114] Additionally, the input / output interface (840) of the computing device (800) may be a means for interfacing with a device (not shown) for input or output that is connected to the computing device (800) or that the computing device (800) may include.
[0115] The input unit is for receiving information and data, including audio information (or signals), text, etc., from a network or a user, and may include at least one microphone and at least one of a user input unit. Data collected from the input unit may be analyzed and processed into a user control command.
[0116] The user input unit is intended to receive information from a user, and when information is input through the user input unit, the control unit can control the operation of the device to correspond to the input information. Such a user input unit may include a hardware physical key (e.g., a button, dome switch, jog wheel, jog switch, etc. located on at least one of the front, rear, and side of the device) and a software touch key. As an example, the touch key may consist of a virtual key, soft key, or visual key displayed on a touchscreen-type display unit through software processing, or may consist of a touch key placed on a part other than the touchscreen. Meanwhile, the virtual key or visual key may have various forms and may be displayed on the touchscreen, for example, as a graphic, text, icon, video, or a combination thereof.
[0117] In FIG. 7, the input / output interface (840) is shown as an element configured separately from the processor (820), but is not limited thereto, and the input / output interface (840) may be configured to be included in the processor (820). The computing device (800) may include more components than those in FIG. 7. However, there is no need to clearly illustrate most of the prior art components.
[0118] The processor (820) of the computing device (800) may be configured to manage, process, and / or store information and / or data received from a plurality of user terminals and / or a plurality of external systems.
[0119] The methods and / or various embodiments described above may be realized in digital electronic circuits, computer hardware, firmware, software, and / or combinations thereof. Various embodiments of the present disclosure may be executed by a data processing device, for example, one or more programmable processors and / or one or more computing devices, or may be implemented as a computer program stored on a computer-readable recording medium and / or a computer program stored on a computer-readable recording medium. The computer program described above may be written in any form of programming language, including a compiled language or an interpreted language, and may be distributed in any form, such as a standalone program, a module, a subroutine, etc. The computer program may be distributed through a single computing device, a plurality of computing devices connected through the same network, and / or a plurality of computing devices distributed to be connected through a plurality of different networks.
[0120] The methods and / or various embodiments described above may be performed by one or more processors configured to execute one or more computer programs that process, store, and / or manage any functions, functions, etc. by operating based on raw samples or generating output data. For example, the methods and / or various embodiments of the present disclosure may be performed by special-purpose logic circuits such as a Field Programmable Gate Array (FPGA) or an Application Specific Integrated Circuit (ASIC), and an apparatus and / or system for performing the methods and / or embodiments of the present disclosure may be implemented as a special-purpose logic circuit such as an FPGA or an ASIC.
[0121] One or more processors executing a computer program may include one or more processors of a general-purpose or special-purpose microprocessor and / or any type of digital computing device. The processor may receive instructions and / or data from each of read-only memory and random access memory, or receive instructions and / or data from read-only memory and random access memory. In the present invention, components of a computing device performing the methods and / or embodiments may include one or more processors for executing instructions and one or more memory devices for storing instructions and / or data.
[0122] According to one embodiment, a computing device may exchange data with one or more mass storage devices for storing data. For example, the computing device may receive and / or receive data from a magnetic disc or an optical disc, and may transfer data to a magnetic disc or an optical disc. A computer-readable storage medium suitable for storing instructions and / or data associated with a computer program may include, but is not limited to, any form of non-volatile memory including semiconductor memory devices such as EPROM (Erasable Programmable Read-Only Memory), EEPROM (Electrically Erasable PROM), and flash memory devices. For example, a computer-readable storage medium may include magnetic discs such as internal hard disks or removable disks, optical magnetic discs, CD-ROMs, and DVD-ROMs.
[0123] To provide interaction with a user, the computing device may include, but is not limited to, a display device for providing or displaying information to the user (e.g., CRT (Cathode Ray Tube), LCD (Liquid Crystal Display), etc.) and a pointing device (e.g., keyboard, mouse, trackball, etc.) on which the user can provide input and / or commands, etc. on the computing device. That is, the computing device may further include any other type of device for providing interaction with the user. For example, the computing device may provide any form of sensory feedback to the user for interaction with the user, including visual feedback, auditory feedback and / or tactile feedback. In this regard, the user may provide input to the computing device through various gestures such as visual, vocal, and motion.
[0124] In the present invention, various embodiments may be implemented in a computing system comprising backend components (e.g., data servers), middleware components (e.g., application servers), and / or frontend components. In this case, the components may be interconnected by any form or medium of digital data communication, such as a communication network. For example, the communication network may include a Local Area Network (LAN), a Wide Area Network (WAN), etc.
[0125] A computing device based on the exemplary embodiments described herein may be implemented using hardware and / or software configured to interact with a user, including a user device, a user interface (UI) device, a user terminal, or a client device. For example, the computing device may include a portable computing device such as a laptop computer. Additionally or alternatively, the computing device may include, but is not limited to, Personal Digital Assistants (PDAs), tablet PCs, game consoles, wearable devices, Internet of Things (IoT) devices, Virtual Reality (VR) devices, Augmented Reality (AR) devices, etc. The computing device may further include other types of devices configured to interact with a user. Additionally, the computing device may include a portable communication device suitable for wireless communication over a network such as a mobile communication network (e.g., a mobile phone, a smartphone, a wireless cellular phone, etc.). A computing device may be configured to communicate wirelessly with a network server using wireless communication technologies and / or protocols such as radio frequency (RF), microwave frequency (MWF) and / or infrared frequency (IRF).
[0126] The above descriptions are specific embodiments for carrying out the present disclosure. The present disclosure will include not only the embodiments described above, but also embodiments that can be simply modified or easily modified. Furthermore, the present disclosure will include technologies that can be easily modified and implemented using the embodiments described above. Accordingly, the scope of the present disclosure should not be limited to the embodiments described above, but should be defined by the claims set forth below as well as equivalents to the claims of the present disclosure.
Claims
Claim 1 A magnetic field component prediction method performed by one or more processors, wherein the magnetic field component prediction method comprises: a step of generating training data for training a prediction model unit based on raw training data including previously stored actual magnetic field component values; and a step of performing training on the prediction model unit based on the training data; wherein the step of generating training data comprises: a step of determining output training data including one or more magnetic field component values during a first time interval satisfying target data criteria; and a step of determining input training data including one or more magnetic field-related variable data during a second time interval consecutive to the first time interval. Claim 2 A magnetic field component prediction method according to claim 1, characterized in that the target data standard is determined based on one or more magnetic field component statistical values related to previously stored actual magnetic field component values. Claim 3 A method for predicting magnetic field components according to claim 2, wherein the target data standard is a magnetic field value less than or equal to a preset multiple of the standard deviation from the mean of the previously stored actual magnetic field component values. Claim 4 A method for predicting magnetic field components according to claim 3, wherein the target data standard is a magnetic field value less than or equal to one multiple of the standard deviation of a previously stored actual magnetic field component value. Claim 5 A method for predicting magnetic field components according to claim 1, characterized in that the target data reference is a magnetic field component value of -3nT or less. Claim 6 A method for predicting magnetic field components according to claim 1, wherein the one or more magnetic field-related variable data included in the input learning data comprises at least one solar wind-related data or one or more different magnetic field component data. Claim 7 A method for predicting magnetic field components according to claim 6, wherein the solar wind-related data comprises at least one of solar wind velocity, solar wind density, or solar wind temperature. Claim 8 A method for predicting magnetic field components according to claim 1, wherein the training data is configured by dividing the raw training data by month. Claim 9 A method for predicting magnetic field components according to claim 1, wherein the prediction model unit is a machine learning model configured to enable processing of time series data. Claim 10 A method for predicting magnetic field components according to claim 9, wherein the prediction model unit is implemented using Bi-LSTM (Bi-directional Long Short Term Memory). Claim 11 A magnetic field component prediction device comprising: one or more processors; and memory; wherein the one or more processors generate training data for training a prediction model unit based on raw training data including previously stored actual values of magnetic field components, and perform training on the prediction model unit based on the training data; wherein the one or more processors generating the training data determine output training data including one or more magnetic field component values during a first time interval satisfying target data criteria, and determine input training data including one or more magnetic field-related variable data during a second time interval consecutive to the first time interval.
