Advancing Disturbance-Storm Time Index Forecasting: A Comparative Analysis of Time Distributed CNN and Bi-LSTM Models in Geomagnetic Storm Prediction

Solar wind, consisting of charged particles such as electrons and protons emitted from the Sun's corona, plays a pivotal role in space weather and geomagnetic phenomena when it interacts with Earth's magnetosphere. Geomagnetic storms, quantified by the Disturbance-Storm Time (Dst) index, have significant implications for various systems, including power grids and communication networks. This study explores the use of machine learning models, specifically the Bidirectional Long-Short Term Memory (Bi-LSTM) and the innovative Time Distributed Convolutional Neural Network (TD-CNN), for predicting the Dst index. A comprehensive dataset spanning from 1998 to 2020, incorporating solar wind observations and Dst index measurements, serves as the foundation for model construction and training. The models' predictive performance is evaluated using Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) metrics. Notably, TD-CNN demonstrates superior accuracy and faster training times compared to Bi-LSTM, making it a promising choice for real-time Dst Index prediction. This study contributes valuable insights into TD-CNN's potential in space weather forecasting and solar activity analysis.