Prediction of Ionospheric Scintillation using Neural Network

Abstract

Global Navigation Satellite System (GNSS) signals are radio waves that travel through the ionosphere before reaching ground-based receivers. Irregularities in the Earth's ionosphere can make the amplitude and phase of radio signals change rapidly. An understanding of ionospheric scintillations is critical for mitigating positioning errors in GNSS-based applications. The aim of the research is to analyze the ionospheric scintillations over Parit Raja (1°52' N, 103°06’ E) from 2017 to 2021, and then to predict the ionospheric scintillations over Parit Raja using neural network. The phase amplitude,  data were collected from a Global Positioning System Ionospheric Scintillation and Total Electron Content Monitor (GISTM) receiver at UTHM. This study used the method of feedforward back propagation neural network to predict of. In this work, data from GISTM receiver from 2017-2021 were analyzed. Results show insignificant phase scintillation between 0.05 rad and 0.1 rad during this period. Various parameters may be utilized to evaluate the precision of the trained model produced by the NN model. Results show that in most set-up of number of neurons in the hidden layer(s), the configuration provides the same RMSE for the training and testing processes. Testing results show predicted values from the neural network are almost the same as the actual values. The error between the actual and predicted values is 4.13% for the phase scintillations. For the future, to facilitate more accurate predictions, the training data set needs to include a greater number of data sets. It is also recommended to combine with other methods such as Genetic Algorithm (GA) and Machine Learning (ML) to get a more accurate prediction.