biromiro / pignn-multivp

💨 This repository contains the experiments with Physics-Informed Graph Neural Networks (PiGNNs) for Solar Wind Modelling to improve prediction accuracy and computational efficiency.

Physics-Informed Graph Neural Networks for Solar Wind Forecasting

This repository contains the source code and experimental results for the PiGNNs (Physics-Informed Graph Neural Networks) approach as Solar Wind Forecasting Models.

Overview

The PiGNN approach explores the integration of physics-informed techniques with graph neural networks for improved solar wind forecasting. Unlike traditional data-driven methods, PiGNN models incorporate domain-specific knowledge to enhance prediction accuracy while maintaining computational efficiency.

Structure

• src/: Contains the source code for implementing the PiGNN approach.
• results/: Contains visualizations and summaries of experimental results obtained during the study.

Data

The model uses data derived from magnetogram observations and solar wind parameters processed through the MULTI-VP model. Note that due to data privacy and licensing issues, the dataset used in the study is not publicly available in this repository. For access to the data or further inquiries, contact the authors directly.

License

This project is licensed under the MIT License - see the LICENSE.md file for details.