KR-20260066347-A - Service providing apparatus and method for predicting occurrence of sporadic E layer

KR20260066347AKR 20260066347 AKR20260066347 AKR 20260066347AKR-20260066347-A

Abstract

The present invention relates to a device and method for providing a service for predicting the occurrence of a sporadic E layer, and more specifically, to a device and method for providing a service for predicting the occurrence of a sporadic E layer in the ionosphere by utilizing a plurality of prediction models and assigning weights to each prediction model to provide an optimal prediction result. The present invention utilizes a plurality of deep learning-based prediction models that consider both time-series and non-time-series patterns to accurately predict the occurrence of a sporadic E layer in the ionosphere E layer, determines weights for each prediction model based on the prediction accuracy calculated by comparing the prediction results of the probability of occurrence of the sporadic E layer calculated by each prediction model with actual observed sporadic E layer occurrence data, and then applies these weights to the prediction results of each prediction model to derive a final prediction result, thereby having the effect of improving the prediction accuracy for the occurrence of the sporadic E layer by optimally combining the strengths of each prediction model.

Inventors

최규철
신대규

Assignees

(주)에스이랩

Dates

Publication Date: 20260512
Application Date: 20241104

Claims (10)

A data collection unit that collects actual data including weather information, solar activity information, and sporadic E-layer occurrence information; A data preprocessing unit that preprocesses the above actual measurement data to generate training data; A prediction unit that performs training by applying the training data to a plurality of different deep learning-based prediction models, generates input information based on the prediction request information upon receiving prediction request information including current environmental data and a target time for prediction, applies the input information to each of the plurality of prediction models to calculate prediction information regarding the probability of occurrence of the sporadic E layer through each of the plurality of prediction models, and generates prediction result information including the plurality of prediction information; A model evaluation unit that collects sporadic E-layer occurrence information observed at the predicted target time through the data collection unit, calculates the prediction accuracy for each of the plurality of prediction models by comparing it with the prediction information for each of the plurality of prediction models, determines a weight for each of the plurality of prediction models based on the prediction accuracy, and generates weight setting information for the weights for each of the plurality of prediction models; and A prediction result providing unit that, upon receiving prediction result information including multiple prediction model-specific prediction information corresponding to new prediction request information for a different prediction target time from the prediction unit, calculates weighted prediction information by applying a weight corresponding to the prediction information to each of the multiple prediction information included in the prediction result information based on the weight setting information, and provides final prediction result information regarding the probability of occurrence of a sporadic E layer at the different prediction target time based on the multiple weighted prediction information corresponding to each of the multiple prediction models. A service providing device for predicting sporadic E-layer occurrence including
In claim 1, The above weather information includes information regarding weather, season, latitude, and longitude, and The above solar activity information includes the geomagnetic disturbance index and the number of sunspots, and A service providing device for predicting sporadic E-layer occurrence, characterized in that the sporadic E-layer occurrence information includes foEs values measured through an ionosphere observer.
In claim 1, The above model evaluation unit generates feedback information including actual data at a target prediction time and the corresponding prediction information according to the prediction information corresponding to the prediction model for each of the plurality of prediction models, and provides the corresponding feedback information to the prediction unit. A service providing device for predicting the occurrence of a sporadic E layer, characterized in that the prediction unit, upon receiving feedback information from the model evaluation unit, preprocesses the feedback information through the data preprocessing unit to generate training data for feedback, and trains the training data for feedback on a prediction model corresponding to the feedback information.
In claim 1, A service providing device for predicting the occurrence of a sporadic E layer, characterized in that the above environmental data includes current weather, season, latitude, longitude, geomagnetic disturbance index, and number of sunspots.
In claim 1, The above prediction result providing unit is, A service providing device for predicting the occurrence of a sporadic E layer, characterized by excluding prediction information of a prediction model whose weight according to the weight setting information is less than or equal to a preset threshold among a plurality of prediction information included in the prediction result information, obtaining weighted prediction information by applying a weight to the prediction information for each prediction model that exceeds the preset threshold, and generating final prediction result information based on one or more of the obtained weighted prediction information.
In claim 1, The above multiple prediction models are, A service providing device for predicting sporadic E-layer occurrence, characterized by including at least two of an MLP (Multi-Layer Perceptron) model, an RF (Random Forest) model, and an LSTM (Long Short Term Memory) model.
In claim 1, The above model evaluation unit is, After determining the environmental data of the period in which the occurrence of the sporadic E layer was predicted successfully and the environmental data of the period in which it was not predicted based on pre-set judgment criteria for each prediction model, the environmental data of the period in which the occurrence of the sporadic E layer was predicted successfully and the environmental data of the period in which it was not predicted are compared and analyzed to generate condition information regarding the environmental condition with the highest success probability for the prediction of the occurrence of the sporadic E layer for each prediction model, and provide this information to the prediction result providing unit. The above prediction result providing unit is, A service providing device for predicting the occurrence of a sporadic E layer, characterized by increasing the weight of the prediction information of a specific prediction model by a preset specified value when the current environmental data according to the new prediction request information satisfies the environmental conditions of the specific prediction model in conjunction with the above prediction unit.
In claim 1, The above prediction result providing unit is, A service providing device for predicting the occurrence of Sporadic E-layer, characterized by storing weight setting information received from the model evaluation unit in a history DB as weight history information including time information regarding the time of creation of said weight setting information, and upon receiving new weight setting information from the model evaluation unit, extracting one or more weight history information that is generated within a preset period based on the time of creation of said new weight setting information and stored in the history DB, and applying the highest reflection ratio to the new weight setting information with the most recent time of creation and the lowest reflection ratio to the weight history information with the oldest time of creation according to preset reflection criteria for said new weight setting information and said extracted weight history information, while applying mutually different reflection ratios, and then generating weight setting information to be applied to prediction information for each prediction model based on the new weight setting information with the applied reflection ratios and the one or more weight history information.
In claim 1, The above model evaluation unit is, After obtaining sporadic E-layer occurrence information at the target prediction time included in the final prediction result information, the prediction accuracy is calculated by comparing it with the final prediction result information, and if the prediction accuracy is below a preset threshold, update information is generated to reduce one or more model-specific weights used in calculating the final prediction result information by a preset amount and provided to the prediction result providing unit. A service providing device for predicting the occurrence of a sporadic E layer, characterized in that the prediction result providing unit updates the weight setting information based on the update information.
In a service provision method for predicting the occurrence of sporadic E-layer of a service providing device, A step of collecting actual data including meteorological information, solar activity information, and sporadic E-layer occurrence information; A step of generating training data by preprocessing the above actual measurement data; A step of performing training by applying the training data to a plurality of different deep learning-based prediction models, and upon receiving prediction request information including current environmental data and a target time for prediction, generating input information based on the prediction request information, applying the input information to each of the plurality of prediction models to calculate prediction information regarding the probability of occurrence of the sporadic E layer through each of the plurality of prediction models, and generating prediction result information including the plurality of prediction information; A step of collecting sporadic E-layer occurrence information observed at the predicted target time, calculating prediction accuracy for each of the plurality of prediction models by comparing it with prediction information for each of the plurality of prediction models, determining weights for each of the plurality of prediction models based on the prediction accuracy, and generating weight setting information for the weights for each of the plurality of prediction models; and Step of receiving prediction result information including multiple prediction model-specific prediction information corresponding to new prediction request information for another prediction target time, calculating weighted prediction information by applying a weight corresponding to the prediction information to each of the multiple prediction information included in the prediction result information based on the weight setting information, and providing final prediction result information regarding the probability of occurrence of the sporadic E layer at the other prediction target time based on the multiple weighted prediction information corresponding to each of the multiple prediction models. A method for providing a service for predicting the occurrence of a sporadic E layer, including

Description

Service providing apparatus and method for predicting occurrence of sporadic E layer The present invention relates to a service providing apparatus and method for predicting the occurrence of a sporadic E layer, and more specifically, to a service providing apparatus and method for predicting the occurrence of a sporadic E layer in the ionosphere that utilizes a plurality of prediction models to accurately predict the occurrence of a sporadic E layer and assigns weights to each prediction model to provide an optimal prediction result. The ionosphere is a region of plasma generated by the ionization of the neutral atmosphere by solar radiation; it is closely related to solar activity and serves as a medium for shortwave communication. Consequently, disturbances caused by changes in electron density within the ionosphere affect shortwave and satellite communication environments. In the ionosphere's E layer, a sporadic E layer—a phenomenon of temporary and localized increase in electron density—occurs. The sporadic E layer is generally observed at altitudes of 90 to 120 km, lasts for tens of minutes to several hours, and has a thickness ranging from several km to tens to hundreds of km. In addition, sporadic E layers are characterized by having a density that is about 2 to 3 times higher than that of surrounding E layers, and they not only have very different regional characteristics depending on topography and latitude, but sporadic E layers occurring in the mid-latitudes of the Northern Hemisphere also have seasonal tendencies, such as occurring frequently in the summer. When a sporadic E layer is generated, it can have a serious impact on radio communications in the HF and VHF bands. Generally, sporadic E layers reflect radio waves below 10 MHz or absorb some radio waves above 10 MHz, but sometimes they reflect up to 50 MHz, and even up to 450 MHz in the VHF range, causing serious disturbances to HF and VHF communications. Furthermore, the sporadic E layer affects satellite communication frequencies in the GHz band, causing amplitude flashes, and disrupts radio waves emitted from a transmitting station by failing to pass through the sporadic E layer and being reflected, preventing them from reaching the desired receiving station. The sporadic E layer is very difficult to predict accurately because its occurrence patterns vary depending on various factors such as weather, season, latitude, and longitude, and also differ by location. In particular, due to its random occurrence characteristics, methods to infer it indirectly are very limited. Although research on the sporadic E layer has been conducted by many researchers, studies presented in the form of quantified models are very rare, and there is no method to directly measure the sporadic E layer, making real-time verification difficult. Currently, there are cases where ionosphere observation instruments are operated to observe changes in the ionosphere in real time, and models are operated to predict changes in ionosphere electron density based on the observation results. However, these existing models have limitations in predicting the occurrence of sporadic E layers that appear temporarily and locally. Therefore, there is an urgent need for technology that can accurately predict the occurrence of sporadic E layers to prevent and prepare for communication failures. FIG. 1 is a configuration diagram of a service providing system for predicting sporadic E-layer occurrence according to an embodiment of the present invention. FIG. 2 is a detailed configuration diagram of a service providing device for predicting sporadic E-layer occurrence according to an embodiment of the present invention. FIG. 3 is a flowchart of a method for providing a service for predicting the occurrence of a sporadic E layer according to an embodiment of the present invention. FIG. 4 is an exemplary diagram of solar activity information according to an embodiment of the present invention. FIGS. 5 and 6 are exemplary diagrams of sporadic E-layer generation information according to an embodiment of the present invention. FIG. 7 is an exemplary diagram of the data preprocessing process of a service providing device according to an embodiment of the present invention. FIG. 8 is an example diagram of the calculation of the prediction accuracy of prediction information regarding the probability of occurrence of a sporadic E layer of a service providing device according to an embodiment of the present invention. Detailed embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a configuration diagram of a service providing system for predicting sporadic E-layer occurrence according to an embodiment of the present invention. As described above, a service providing system according to an embodiment of the present invention may include a user terminal, an ionospheric observer providing ionospheric observation data, a weather server providing weathe