Method and apparatus for forecasting time siries atmospheric radio refractivity index based on artificial intelligence

Abstract

A method for predicting atmospheric refractive index in a time series based on artificial intelligence is disclosed, the method comprising: a step of processing weather data measured over a predetermined period by at least one weather station into weather data at specific altitude intervals in a time series order; a step of calculating the atmospheric refractive index at specific altitude intervals using the processed weather data; and a step of inputting the atmospheric refractive index into a multi-layer perceptron neural network to generate a predicted atmospheric refractive index at a specific point in time.

Description

Technology Field

[0001] The present disclosure relates to a method and apparatus for predicting a time-series atmospheric refractive index based on artificial intelligence, and more specifically, to a method and apparatus for predicting a time-series atmospheric refractive index based on artificial intelligence that predicts the predicted atmospheric refractive index at a specific point in time from an artificial intelligence model learned based on the atmospheric refractive index of a specific altitude interval in a time-series order. Background Technology

[0003] Electromagnetic waves have the characteristic of undergoing distortion, such as refraction, reflection, and attenuation, at the boundary between two media with different refractive indices.

[0004] In communication between a satellite and a ground station, the Earth's atmosphere through which radio waves pass consists of multiple layers with different refractive indices, and the characteristics of radio waves reaching the ground from the satellite through the atmosphere are determined by the atmospheric refractive indices of the troposphere and the stratosphere.

[0005] Although the atmospheric refractive index can be accurately determined from direct measurements by rawinsondes—weather instruments that measure temperature, pressure, humidity, and wind in the upper atmosphere—there are physical, temporal, and economic limitations to obtaining these values ​​in real-time at all locations and time zones for the prediction of propagation characteristics.

[0006] Therefore, a method of indirect calculation using techniques such as interpolation, rather than direct measurement, is required; conventionally, a method was used to estimate the atmospheric refractive index at unobserved points through spatial interpolation (Inverse Distance Weighting, IDW) based on atmospheric refractive indices measured at some points.

[0007] However, estimating the atmospheric refractive index at unobserved points using spatial interpolation has lower accuracy compared to the method calculated based on direct measurement, and since accuracy is proportional to the number of measurement data used in spatial interpolation, values ​​measured at various points are required to increase the accuracy of atmospheric refractive index prediction by spatial interpolation.

[0008] On the other hand, the number of weather stations providing the meteorological data necessary for predicting atmospheric refractive index is limited. Consequently, there is a need for a method to calculate the atmospheric refractive index corresponding to a specific point in time in a given region by utilizing historical measurement data from non-operational weather stations.

[0009] In other words, it is required to secure an artificial intelligence model that increases the number of data available for spatial interpolation by using a deep learning model that learns the time series patterns of weather data from weather stations operated in the past to predict the atmospheric refractive index at a future point in time. The problem to be solved

[0011] The technical objective of the present disclosure is to provide an AI-based time-series atmospheric refractive index prediction method and apparatus that predicts the predicted atmospheric refractive index at a specific point in time from an AI model learned based on the atmospheric refractive index of specific altitude intervals in a time-series order.

[0012] The technical problems to be solved in this disclosure are not limited to those mentioned above, and other technical problems not mentioned will be clearly understood by those skilled in the art to which this disclosure belongs from the description below. means of solving the problem

[0014] According to one aspect of the present disclosure, a method for predicting atmospheric refractive index based on artificial intelligence is disclosed. The method comprises the steps of: processing weather data measured over a predetermined period by at least one weather station into weather data at specific altitude intervals in a time series order; calculating the atmospheric refractive index at specific altitude intervals using the processed weather data; and inputting the atmospheric refractive index into a multi-layer perceptron neural network to generate a predicted atmospheric refractive index at a specific point in time.

[0015] According to another aspect of the present disclosure, an artificial intelligence-based time-series atmospheric refractive index prediction device is disclosed. The device comprises a memory for storing at least one instruction and a processor for executing the at least one instruction stored in the memory based on data obtained from the memory. The processor processes weather data measured over a predetermined period by at least one weather station into weather data at specific altitude intervals in a time-series order, calculates the atmospheric refractive index at specific altitude intervals using the processed weather data, and inputs the atmospheric refractive index into a multilayer perceptron neural network to generate a predicted atmospheric refractive index at a specific point in time.

[0016] According to one aspect of the present disclosure, the step of processing into weather data of specific altitude intervals in the time series order comprises calculating the weather data of specific altitude intervals using a piecewise cubic Hermite interpolation polynomial, and processing the weather data of specific altitude intervals according to the time series order for each specific altitude.

[0017] According to one aspect of the present disclosure, the step of calculating the atmospheric refractive index of the specific altitude interval is to calculate the atmospheric refractive index of the specific altitude interval corresponding to the processed meteorological data using a millimeter-wave propagation model.

[0018] According to one aspect of the present disclosure, the step of calculating the atmospheric refractive index of the specific altitude interval calculates the atmospheric refractive index of the specific altitude interval corresponding to the processed meteorological data using an ITU-R model (International Telecommunication Union - Radiocommunication Model) that takes the atmospheric pressure, temperature, and partial pressure of water vapor of the meteorological data of the specific altitude interval as variables.

[0019] According to one aspect of the present disclosure, the meteorological data includes at least one of temperature, atmospheric pressure, partial pressure of water vapor, dew point, relative humidity, wind speed, wind direction, snow, rain, clouds, and aerosols.

[0020] According to one aspect of the present disclosure, prior to the step of generating the predicted atmospheric refractive index, the multilayer perceptron neural network is further provided with the step of training the network by processing the atmospheric refractive index of the specific altitude interval as training data to generate the predicted atmospheric refractive index at a specific time point.

[0021] According to one aspect of the present disclosure, the training data is configured by using a sliding window technique to input 28 consecutive atmospheric refractive indices in the time series order among the atmospheric refractive indices of the specific altitude interval, and to target 6 consecutive atmospheric refractive indices following the 28 atmospheric refractive indices.

[0022] The features briefly summarized above regarding the present disclosure are merely exemplary aspects of the detailed description of the present disclosure that follows and do not limit the scope of the present disclosure. Effects of the invention

[0024] According to the present disclosure, a method and apparatus for predicting a time-series atmospheric refractive index based on artificial intelligence can be provided, which predicts the predicted atmospheric refractive index at a specific point in time from an artificial intelligence model learned based on the atmospheric refractive index of a specific altitude interval in a time-series order.

[0025] In addition, according to the present disclosure, data available for spatial interpolation can be obtained to increase the accuracy of the prediction of atmospheric refractive index at unobserved points.

[0026] In addition, according to the present disclosure, not only can the vertical atmospheric refractive index according to non-homogeneous altitudes be calculated, but the atmospheric refractive index at a future point in time at a specific altitude can also be predicted.

[0027] The technical effects intended to be achieved in this disclosure are not limited to those mentioned above, and other technical effects not mentioned will be clearly understood by those skilled in the art to which this disclosure belongs from the description below. Brief explanation of the drawing

[0029] FIG. 1 is a diagram illustrating communication between an atmospheric refractive index prediction device and a weather station according to the present disclosure. FIG. 2 is a diagram showing a module constituting an atmospheric refractive index prediction device according to the present disclosure. Figure 3 is a flowchart illustrating the process of training a neural network that outputs a predicted atmospheric refractive index. Figure 4 is a diagram illustrating weather data corresponding to a specific altitude arranged in a time series. Figure 5 is a diagram illustrating the construction of training data from weather data corresponding to specific altitudes arranged in chronological order. FIG. 6 shows a multi-perceptron neural network for predicting atmospheric refractive index according to the present disclosure. Figure 7 is a flowchart illustrating the process of generating atmospheric refractive index at a specific time point using a trained multi-perceptron neural network. Specific details for implementing the invention

[0030] Hereinafter, embodiments of the present disclosure are described in detail with reference to the attached drawings so that those skilled in the art can easily implement them. However, the present disclosure may be embodied in various different forms and is not limited to the embodiments described herein.

[0031] In describing the embodiments of the present disclosure, if it is determined that a detailed description of known configurations or functions could obscure the essence of the present disclosure, such detailed description is omitted. Additionally, parts of the drawings unrelated to the description of the present disclosure have been omitted, and similar parts are denoted by similar reference numerals.

[0032] In the present disclosure, when a component is described as being "connected," "combined," or "joined" with another component, this may include not only a direct connection but also an indirect connection in which another component exists in between. Furthermore, when a component is described as "comprising" or "having" another component, this means that, unless specifically stated otherwise, it does not exclude the other component but may include additional components.

[0033] In the present disclosure, terms such as first, second, etc. are used solely for the purpose of distinguishing one component from another component and do not limit the order or importance of the components unless specifically stated otherwise. Accordingly, within the scope of the present disclosure, a first component in one embodiment may be referred to as a second component in another embodiment, and likewise, a second component in one embodiment may be referred to as a first component in another embodiment.

[0034] In this disclosure, distinct components are intended to clearly describe their respective features and do not imply that the components are separate. That is, multiple components may be integrated to form a single hardware or software unit, or a single component may be distributed to form multiple hardware or software units. Accordingly, such integrated or distributed embodiments are included within the scope of this disclosure, even if not otherwise mentioned.

[0035] In the present disclosure, each of the phrases such as “A or B”, “at least one of A and B”, “at least one of A or B”, “A, B or C”, “at least one of A, B and C”, and “at least one of A, B, C or combination thereof” may include any one of the items listed together in the corresponding phrase, or all possible combinations thereof.

[0036] In the present disclosure, the components described in various embodiments do not necessarily mean essential components, and some may be optional components. Accordingly, embodiments consisting of a subset of the components described in one embodiment are also included within the scope of the present disclosure. Furthermore, embodiments including other components in addition to the components described in various embodiments are also included within the scope of the present disclosure.

[0037] The advantages and features of the present invention and the methods for achieving them will become clear by referring to the embodiments described below in detail together with the accompanying drawings. However, the present invention is not limited to the embodiments presented below but can be implemented in various different forms, and these embodiments are provided merely to ensure that the disclosure of the present invention is complete and to fully inform those skilled in the art of the scope of the invention.

[0038] FIG. 1 is a diagram illustrating communication between an atmospheric refractive index prediction device and a weather station according to the present disclosure.

[0039] As illustrated in FIG. 1, the atmospheric refractive index prediction device (100) can collect weather data from at least one weather station through a network.

[0040] The method of communication between the atmospheric refractive index prediction device (100) and the weather station (200-1, 200-2, 200-3, 200-4, 200-5; hereinafter, drawing numbers omitted) is not limited and, for example, a mobile communication network, wired internet, wireless internet, broadcasting network, satellite network, etc., may be utilized, and is not limited thereto, may also include short-range wireless communication between devices. As an example, the network may include a PAN (Personal Area Network), LAN (Local Area Network), CAN (Campus Area Network), MAN (Metropolitan Area Network), WAN (Wide Area Network), BBN (Broadband Network), etc. Additionally, the network may include any one or more of network topologies including a bus network, star network, ring network, mesh network, star-bus network, tree or hierarchical network, etc., but is not limited thereto.

[0041] A weather station measures weather data and stores the measured weather data along with the date, time, location, etc. In addition, the weather station may process or transform the information described above into other forms. The weather station may collect weather data at various altitudes, such as the troposphere and stratosphere, by launching weather observation equipment, such as a radiosonde, from the surface of the earth twice a day. The weather station may collect weather data at predetermined time intervals using a radiosonde, and the weather data may include at least one of temperature, atmospheric pressure, partial pressure of water vapor, dew point, relative humidity, wind speed, wind direction, snow, rain, clouds, and aerosols according to altitude.

[0042] Weather data can be utilized as training data for a multi-perceptron neural network that predicts atmospheric refractive index according to the present disclosure. That is, the atmospheric refractive index prediction device can predict the atmospheric refractive index at a future point in time through a trained multi-perceptron neural network, and can increase the accuracy of atmospheric refractive index prediction for locations where observed weather data does not exist by using data on augmented atmospheric refractive index. This will be explained in detail through FIGS. 3 and 4.

[0043] The atmospheric refractive index prediction device (100) may include a portable user terminal such as a smartphone, laptop, or wearable device, and may include a fixed management device such as a PC or server.

[0044] An atmospheric refractive index prediction device (100) processes weather data measured over a certain period by at least one weather station into weather data at specific altitude intervals in a time series order to predict the atmospheric refractive index at a specific point in time, calculates the atmospheric refractive index at specific altitude intervals using the processed weather data, and inputs the atmospheric refractive index into a multi-perceptron neural network to generate the predicted atmospheric refractive index at a specific point in time. The operation of calculating the predicted atmospheric refractive index of the atmospheric refractive index prediction device is explained in detail through FIG. 7.

[0045] FIG. 2 is a diagram showing a module constituting an atmospheric refractive index prediction device according to the present disclosure.

[0046] An atmospheric refractive index prediction device (100) may include a processor (105), memory (110), a communication unit (115), and an interface unit (120), and the processor (105) may include a learning unit (105a) and an atmospheric refractive index prediction unit (105b). Each of the above components is not an essential component and may be omitted by including one or more components. Additionally, it may be configured to include other components.

[0047] The communication unit (115) can support mutual communication between the atmospheric refractive index prediction device (100) and the weather station using a communication means included in the network described in FIG. 1. The communication unit (115) according to the present disclosure can receive weather data from the weather station.

[0048] The interface section (120) can function as a user interface for controlling the atmospheric refractive index prediction device (100), and although not shown, it may be equipped with a display section composed of a touch screen capable of detecting user input.

[0049] The memory (110) may be a non-removable memory or a removable memory and may store applications and various data for controlling the atmospheric refractive index prediction device (100), and may load applications or read and write data upon the request of the processor (105). The memory (100) according to the present disclosure may store an application and at least one instruction for processing weather data measured over a certain period by at least one weather station into weather data at specific altitude intervals in a time series order, calculating the atmospheric refractive index at specific altitude intervals using the processed weather data, inputting the atmospheric refractive index into a multi-layer perceptron neural network to train the multi-layer perceptron neural network to generate a predicted atmospheric refractive index, and using the trained multi-layer perceptron neural network to generate a predicted atmospheric refractive index at a specific point in time.

[0050] The processor (105) can perform overall control of the atmospheric refractive index prediction device (100). The processor (105) can be configured to execute applications and instructions stored in memory (110). In connection with the present disclosure, the processor (105) can train a multi-perceptron neural network to generate a predicted atmospheric refractive index using applications, instructions, and data stored in memory (110), and can generate an atmospheric refractive index at a specific point in time through the trained multi-perceptron neural network.

[0051] In the present disclosure, the processor (105) may be implemented as a single processing module, for example. Alternatively, the processing according to the above-described matters may be performed in a plurality of processing modules, for instance, through a learning unit (105a) that trains a multi-perceptron neural network and an atmospheric refractive index prediction unit (105b) that outputs a predicted atmospheric refractive index through the trained multi-perceptron neural network. In the present disclosure, for convenience of description, the processor (120) is collectively referred to as a plurality of processing modules (105a, 105b), and the processing of the processor (120) will be described below.

[0052] Figure 3 is a flowchart illustrating the process of training a neural network that outputs a predicted atmospheric refractive index.

[0053] Referring to FIG. 3, the processor (105) collects weather data measured at weather stations (S310). Specifically, the processor (105) can collect weather data collected over a certain period from at least one weather station. Additionally, if the weather data from each weather station is already stored in the atmospheric refractive index prediction device (100), the weather data for a specific period can be collected by searching for the weather data for a specific period among the stored weather data.

[0054] Next, the processor (105) processes the collected weather data into weather data at specific altitude intervals in a time series order (S320). Each weather station is located at a different height, and the altitude of the weather data measured at the weather station may vary from station to station. Additionally, there may be missing weather data at different altitudes depending on the weather conditions. Accordingly, the processor (120) applies a formula to the collected weather data to standardize the weather data and reinforce missing data, thereby converting or processing it into weather data at specific altitude intervals, e.g., h0, h0+h, h0+h*2, h0+h*3, …, hn.

[0055] For example, the above formula may include a piecewise cubic Hermite interpolation polynomial (PCHIP). Specifically, the processor (120) uses the piecewise cubic Hermite interpolation polynomial to convert weather data into continuous data, extracts weather data for unmeasured altitudes, and processes the weather data into specific altitude intervals.

[0056] In addition, the processor (120) processes the weather data processed at specific altitude intervals according to the specific altitude in a chronological order. That is, it concatenates the data in chronological order according to the specific altitude. Accordingly, the weather data corresponding to the specific altitude is arranged in a chronological order. An example of this is illustrated in FIG. 5.

[0057] Figure 4 is a diagram illustrating weather data corresponding to a specific altitude arranged in a time series.

[0058] Figure 4 illustrates an example of arranging temperature, pressure, and relative humidity in a time series among weather data at a specific altitude, for example, an altitude of 50m above sea level. The intervals of the time series may vary depending on user specifications or system settings. In addition, just as weather data is processed at specific altitude intervals, if weather data for a specific time period is missing, the missing weather data for that time period can be augmented through interpolation.

[0059] Returning to FIG. 3, the processor (105) calculates the atmospheric refractive index for a specific altitude interval using weather data for a specific altitude interval (S330). For example, the processor (105) can calculate the atmospheric refractive index for a specific altitude interval using a Millimeter-Wave Propagation Model (MPM) to calculate the atmospheric refractive index for a specific altitude interval using weather data for a specific altitude interval. In this case, the Millimeter-Wave Propagation Model may include MPM93. As another example, the processor (105) can calculate the atmospheric refractive index for a specific altitude interval using an ITU-R model to calculate the atmospheric refractive index for a specific altitude interval using weather data for a specific altitude interval. Specifically, the processor (105) can calculate the atmospheric refractive index using an ITU-R model (Equation 1) according to the following equation.

[0060] [Mathematical Formula 1]

[0061]

[0062] Here, N represents the atmospheric refractive index [N-unit], P represents atmospheric pressure [mbar], T represents the temperature [K], and e represents the partial pressure of water vapor [mbar].

[0063] Next, the processor (105) inputs atmospheric refractive index to train a multi-perceptron neural network and outputs a predicted atmospheric refractive index at a specific time (S340).

[0064] The processor (105) can construct training data using atmospheric refractive index generated through the process described above, based on weather data from at least one weather station. Table 1 below shows the weather data collection period and the number of weather data collected for each domestic high-altitude weather station.

[0065] Weather station (number) Observation start date Collection end date Number of weather data available until the end of collection Baengnyeongdo (47102) May 1, 2018 November 2, 2022 2,760 Bukgangneung (47104) December 24, 2015 November 2, 2022 4,478 Osan (47122) March 31, 1957 November 2, 2022 47,832 Pohang (47138) November 12, 1966 November 2, 2022 40,356 Gwangju (47158) January 6, 1978 November 2, 2022 33,210 Heuksando (47169) May 7, 2003 November 2, 2022 13,706 National Typhoon Center (47186) May 23, 2016 November 2, 2022 4,176

[0066] As can be seen from Table 1, weather data measured at domestic weather stations available for learning is limited, and in order to calculate the predicted atmospheric refractive index at a specific point in the future, the weather data in time series is transformed into a supervised learning problem so that a multilayer perceptron neural network can learn the patterns contained therein.

[0067] The processor (105) labels the correct answer using the calculated atmospheric refractive index to train a multilayer perceptron neural network, and the labeling can be performed by auto-labeling or user input. Additionally, the processor (105) classifies the atmospheric refractive index so that the training data consists of input variables and targets.

[0068] For example, the processor (105) may use a sliding window technique to configure training data for training a multilayer perceptron neural network.

[0069] Specifically, the processor (105) can classify weather data at specific altitude intervals in chronological order using a sliding window method, calculate the atmospheric refractive index using the classified weather data, and input it as training data. Additionally, based on the atmospheric refractive index calculated using weather data at specific altitude intervals in chronological order, the processor can classify the data using a sliding window method and input the classified atmospheric refractive index as training data. This is explained through FIG. 6.

[0070] Figure 5 is a diagram illustrating the construction of training data from weather data corresponding to specific altitudes arranged in chronological order.

[0071] Specifically, FIG. 5 is a diagram illustrating, for example, data corresponding to temperature among weather data at a specific altitude. The processor (105) can construct a learning data set (Data 1, Data 2, …, Data n) by sliding the weather data in a time series order at a specific altitude, for example, with the size of the Input Window (N) set to 28 and the size of the Output Window (K) set to 6. Similarly, the processor (105) can construct a learning data set for atmospheric pressure, water vapor partial pressure, etc., based on the method described above, and finally, calculate the atmospheric refractive index based on this and use it as learning data for a multi-perceptron neural network.

[0072] Returning to FIG. 3, the processor (105) generates multiple training data using an Input Window and an Output Window of a specific size through the process described above. In the context of supervised learning problem solving, the Input Window is the input variable and the Output Window is the target. The process of training a multi-perceptron neural network using the training data is explained through FIG. 7.

[0073] FIG. 6 shows a multi-perceptron neural network for predicting atmospheric refractive index according to the present disclosure.

[0074] For example, the number of nodes in the input layer of a multilayer perceptron neural network may match the size of the input window, and the number of nodes in the output layer may match the size of the output window. The number of nodes in the input and output layers is a key variable determining the prediction performance of a multilayer perceptron neural network, and these variables may be determined differently depending on the model's performance, user specifications, or system settings.

[0075] A multilayer perceptron neural network learns the relationship between the Input Window and the Output Window from the training dataset as it repeats epochs, which are training cycles. In other words, the multilayer perceptron neural network learns to best predict the correct answer when specific data is input. Consequently, the multilayer perceptron neural network learns to minimize the error in the output.

[0076] Specifically, in a multi-perceptron neural network, the atmospheric refractive index calculated from weather data included in the Input Window is input to the input layer, and the atmospheric refractive index calculated from weather data included in the Output Window is input to the output layer. Accordingly, an error is calculated as the difference between the output value of the multi-perceptron neural network and the atmospheric refractive index calculated from the weather data included in the Output Window. Through an iterative learning process, the multi-perceptron neural network updates the weights and biases (α, β, γ) of each node by backpropagating from the output layer to the input layer in a direction that reduces the error.

[0077] In the multi-perceptron neural network according to the present disclosure, a sigmoid function (σ) may be used as the activation function. The following [Equation 2], [Equation 3], and [Equation 4] represent equations in which calculations are performed at nodes of the input layer, hidden layer, and output layer.

[0078] [Mathematical Formula 2]

[0079]

[0081] [Mathematical Formula 3]

[0082]

[0084] [Mathematical Formula 4]

[0085]

[0086] In the mathematical formula above, N represents the size of the Input Window, K represents the size of the Output Window, and M represents the number of nodes in the hidden layer. The total number of weights and biases updated in a single epoch is am.

[0087] Specifically, represents the value of the Mth node in the hidden layer, and represents the value of the Kth node of the output layer, and represents the Nth node vector of the input layer, and represents the Mth weight vector of the input layer, and represents the Mth bias value of the hidden layer, and represents the Kth bias value of the output layer, and represents the Mth weight vector of the hidden layer.

[0088] During the weight update process, the direction of learning for each node's weight can be determined by the learning rate. The computation of the artificial neural network on the training data and the backpropagation of errors can constitute a single learning cycle (epoch), and the learning rate can be applied differently depending on the number of iterations of the artificial neural network's learning cycle.

[0089] In the training of a multilayer perceptron neural network, the training data can generally be a subset of the actual data (i.e., the data to be processed using the trained multilayer perceptron neural network).

[0090] Accordingly, there may exist a learning cycle in which errors on training data decrease but errors on actual data increase. For example, overfitting is a phenomenon in which errors on actual data increase due to excessive training on training data, and it can act as a cause of increased errors in the algorithm. To prevent overfitting of the multi-perceptron neural network according to the present disclosure, methods such as regularization, dropout (a method of intentionally deactivating some nodes of the network during the training process), and the use of a batch normalization layer may be applied.

[0091] In addition, to avoid underfitting or overfitting models in the evaluation of the trained multi-perceptron neural network, the model is finally selected through cross-validation. For example, for cross-validation, the training set can be divided into 90% of the training data, and the remaining training data into the validation set and test set. To prevent overfitting or underfitting, the number of training iterations is determined by the intersection point of the loss graphs of the training set and validation set according to the change in epochs. The prediction performance of the multi-perceptron neural network for predicting the predicted refractive index according to the present disclosure showed the most optimal performance when the Input Window size (N) is 28 and the Output Window size (T) is 6. Accordingly, the optimized multi-layer perceptron neural network has a structure in which there are 28 nodes in the input layer, 6 nodes in the output layer, and 20 nodes in the hidden layer. That is, the multi-perceptron neural network according to the present disclosure may operate by using 28 atmospheric refractive indices as input values ​​in the input layer, and outputting 6 predicted refractive indices calculated through 20 hidden layers (drop out = 0.2) in the output layer. The optimizer used for training the multi-perceptron neural network may be Adam (Adaptive moment estimation), and the loss function may be MSE (Mean Squared Error), but is not limited thereto. The batch size of the multi-perceptron neural network according to the present disclosure is 32, and the learning rate is 0.001.

[0092] Returning to FIG. 3, the processor (105) performs learning through the process described above, then compares the predicted atmospheric refractive index output from the output layer with the atmospheric refractive index input as a target using a loss function to determine whether learning is required, and if learning is required, repeats the process described above (S350).

[0093] Hereinafter, with reference to Fig. 7, the process of generating the predicted atmospheric refractive index at a specific point in time through a multi-perceptron neural network that has been trained by the process described above is explained.

[0094] Figure 7 is a flowchart illustrating the process of generating atmospheric refractive index at a specific time point using a trained multi-perceptron neural network.

[0095] Referring to FIG. 7, the processor (105) collects weather data (710). For example, the processor (105) may collect weather data from at least one weather station to generate a predicted atmospheric refractive index, and may utilize weather data already stored in memory (110).

[0096] Next, the processor (105) processes the collected weather data into weather data at specific altitude intervals in a time-series order (S720). Similar to step S320 of the process of learning a multi-perceptron neural network described in FIG. 3, the data is processed into specific altitude intervals using a piecewise third-order Hermite interpolation polynomial. Next, the processor (105) connects the weather data processed into specific altitude intervals in chronological order to arrange the weather data in a time-series order by specific altitude. Subsequently, the processor (105) calculates the atmospheric refractive index at specific altitude intervals using the weather data at specific altitude intervals (S730). For example, a millimeter wave propagation model may be used as in step S330, and specifically, either the MPM93 model or the ITU-R model may be used, but is not limited thereto.

[0097] The processor (105) can input the atmospheric refractive index calculated by the process described above into a multi-perceptron neural network to generate a predicted atmospheric refractive index at a specific point in time (S740). Specifically, according to one embodiment of the present disclosure, if the size of the Input Window is 28 and the size of the Output Window is 6, inputting 28 atmospheric refractive indices in a time series generates 6 predicted atmospheric refractive indices in a time series for a future point in time. For example, if the atmospheric refractive index corresponding to a specific altitude for a total of 28 days from January 1, 2024 to January 28, 2024 is input, a predicted atmospheric refractive index corresponding to the specific altitude for a total of 4 days from January 29 to February 1 can be generated.

[0098] The exemplary methods of the present disclosure described above are expressed as a series of operations for clarity of explanation, but this is not intended to limit the order in which the steps are performed, and if necessary, each step may be performed simultaneously or in a different order. To implement the method according to the present disclosure, additional steps may be included in addition to the steps exemplified, steps excluding some steps and including the remaining steps, or steps excluding some steps and including additional steps.

[0099] The various embodiments of the present disclosure are not intended to list all possible combinations but to describe representative aspects of the present disclosure, and the matters described in the various embodiments may be applied independently or in combination of two or more.

[0100] In addition, various embodiments of the present disclosure may be implemented by hardware, firmware, software, or a combination thereof. In the case of implementation by hardware, it may be implemented by one or more ASICs (Application Specific Integrated Circuits), DSPs (Digital Signal Processors), DSPDs (Digital Signal Processing Devices), PLDs (Programmable Logic Devices), FPGAs (Field Programmable Gate Arrays), general processors, controllers, microcontrollers, microprocessors, etc.

[0101] The scope of the present disclosure includes software or machine-executable instructions (e.g., operating system, application, firmware, program, etc.) that enable an operation according to a method of various embodiments to be executed on a device or computer, and a non-transitory computer-readable medium on which such software or instructions, etc. are stored and executable on a device or computer. Explanation of the symbols

[0103] 100: Atmospheric Refractive Index Prediction Device 105 : Processor 105a : Learning section 105b : Atmospheric Refractive Index Prediction Section 110 : Memory 115: Communications Department 120 : Interface section

Claims

Claim 1 A method for predicting atmospheric refractive index based on artificial intelligence, comprising: a step of processing weather data measured over a predetermined period by at least one weather station into weather data of a specific altitude interval in a time series order; a step of calculating the atmospheric refractive index of the specific altitude interval using the processed weather data; and a step of inputting the atmospheric refractive index into a multi-layer perceptron neural network to generate a predicted atmospheric refractive index at a specific point in time. Claim 2 A method for predicting atmospheric refractive index according to claim 1, wherein the step of processing into meteorological data of specific altitude intervals in the time series order comprises calculating the meteorological data of the specific altitude intervals using a piecewise cubic Hermite interpolation polynomial, and processing the meteorological data of the specific altitude intervals according to the time series order for each specific altitude. Claim 3 A method for predicting atmospheric refractive index according to claim 1, wherein the step of calculating the atmospheric refractive index of the specific altitude interval is to calculate the atmospheric refractive index of the specific altitude interval corresponding to the processed weather data using a millimeter-wave propagation model. Claim 4 A method for predicting atmospheric refractive index according to claim 1, wherein the step of calculating the atmospheric refractive index of a specific altitude interval is to calculate the atmospheric refractive index of the specific altitude interval corresponding to the processed weather data using an ITU-R model (International Telecommunication Union - Radiocommunication Model) that takes the atmospheric pressure, temperature, and partial pressure of water vapor of the weather data of the specific altitude interval as variables. Claim 5 A method for predicting atmospheric refractive index according to claim 1, wherein the meteorological data includes at least one of temperature, atmospheric pressure, partial pressure of water vapor, dew point, relative humidity, wind speed, wind direction, snow, rain, clouds, and aerosols. Claim 6 A method for predicting atmospheric refractive index according to claim 1, further comprising, prior to the step of generating the predicted atmospheric refractive index, the step of training the multilayer perceptron neural network by processing the atmospheric refractive index of the specific altitude interval as training data to generate the predicted atmospheric refractive index at a specific point in time. Claim 7 In claim 6, the above-mentioned learning data is configured by using a sliding window technique to input 28 consecutive atmospheric refractive indices in the time series order among the atmospheric refractive indices of the above-mentioned specific altitude intervals, and targeting 6 consecutive atmospheric refractive indices following the 28 atmospheric refractive indices. Claim 8 An artificial intelligence-based time-series atmospheric refractive index prediction device comprises: a memory for storing at least one instruction; and a processor for executing the at least one instruction stored in the memory based on data obtained from the memory, wherein the processor processes weather data measured over a predetermined period by at least one weather station into weather data of a specific altitude interval in a time-series order, calculates the atmospheric refractive index of the specific altitude interval using the processed weather data, and inputs the atmospheric refractive index into a multilayer perceptron neural network to generate a predicted atmospheric refractive index at a specific point in time. Claim 9 An atmospheric refractive index prediction device according to claim 8, wherein the processor calculates the weather data of the specific altitude interval using a piecewise cubic Hermite interpolation polynomial, and processes the weather data of the specific altitude interval according to the specific altitude and time series order to process the weather data of the specific altitude interval in the time series order. Claim 10 In claim 8, the processor is an atmospheric refractive index prediction device that calculates the atmospheric refractive index of the specific altitude interval corresponding to the processed weather data using a millimeter-wave propagation model. Claim 11 In claim 8, the above processor calculates the atmospheric refractive index of the above specific altitude interval corresponding to the processed meteorological data using an ITU-R model (International Telecommunication Union - Radiocommunication Model) that takes the atmospheric pressure, temperature, and partial pressure of water vapor of the meteorological data of the above specific altitude interval as variables, an atmospheric refractive index prediction device. Claim 12 In claim 8, the atmospheric refractive index prediction device, wherein the meteorological data comprises at least one of temperature, atmospheric pressure, partial pressure of water vapor, dew point, relative humidity, wind speed, wind direction, snow, rain, clouds, and aerosols. Claim 13 In claim 8, the above processor is an atmospheric refractive index prediction device that trains a multilayer perceptron neural network to generate the predicted atmospheric refractive index at a specific point in time by processing the atmospheric refractive index of the above specific altitude interval as training data. Claim 14 An atmospheric refractive index prediction device according to claim 13, wherein the above-mentioned learning data is configured by using a sliding window technique to input 28 atmospheric refractive indices that are consecutive in the time series order among the atmospheric refractive indices of the above-mentioned specific altitude intervals, and targeting 6 atmospheric refractive indices that are consecutive after the above-mentioned 28 atmospheric refractive indices.

Images