About the Project

Geomagnetic index magic box 🪄📦

Our user interface displays graphs representing predictions for a day with 3-hour intervals of Kp, which are disturbances in Earth's geomagnetic field.

How It Works

Data Extraction: We initiate the data extraction process directly from the satellite in its original CDF (Common Data Format) with daily intervals. Although this stage posed a significant challenge due to the complexity of the CDF format, we have worked on optimizing and speeding up the deserialization, correlation, and encoding process. We have implemented substantial improvements in this phase, significantly reducing the time required to process raw data. This allowed us to streamline data preparation for subsequent analysis and neural network training.

Data Cleaning: We start by ensuring clean input data. We have two data cleaning processes: one to consolidate and review the data, extracting the relevant Kp-dependent information, and another to handle NaN values, which can adversely affect the neural network's training.

Neural Network: Next, we employ a neural network. We process the clean data, splitting it into training, validation, and testing sets. Our neural network is designed using the TensorFlow library, utilizing an LSTM architecture to predict future Kp values in 3-hour intervals.

Benefits

Satellite Lifespan Extension: Our project can help extend the lifespan of satellites by providing valuable data.

Goals

Our primary goal is to accurately predict the necessary data to prevent geomagnetic storms, which can be detrimental to electronic and telecommunication devices.

Technologies Used

Text Editors: Visual Studio Code for coding.
Programming Language: Python for data processing, using the TensorFlow library for the neural network.
Data Sources: Data was obtained from DSCOVR.
Frontend: Node.js and Three.js were utilized for data visualization and presentation.

Installation

gh repo clone Esleiter/GMF

cd frontend

npm install

npm start

2023 NASA Space Apps Challenge

GMF WIZARDS