Solar Update Neural Network for Improved Event Prediction (SUNNIE)

Several NASA Heliophysics models have been developed to predict solar energetic particle (SEP) events. As presented in the CCMC model catalog, most of these are either physics-based or empirical models with very few models utilizing machine learning (ML) techniques. However, none of the existing models leverage the combined power of physics models and ML: Physics informed neural networks (PINNs). To the best of our knowledge, there currently exists no PINN implementation of 3D time-dependent Magnetohydrodynamics (MHD), and there currently does not exist a PINN surrogate model for physics-based CCMC models, specifically, WSA-ENLIL-cone model. P3P proposes to build a Solar Update Neural Network for Improved Event prediction (SUNNIE) that leverages the cutting edge in Physics Informed Neural Network (PINN) architecture to enhance NASA’s CCMC models for Solar Energetic Particle (SEP) event forecasting by integrating underlying numerical solvers directly into the training cycle. This will enhance SEP event predictability via higher spatial and temporal resolution, while requiring fewer data points and thereby also reducing computational footprint. The implementation of SUNNIE offers significant benefits for a wide range of applications, from improving satellite and spacecraft safety to aiding in the strategic planning of space missions. By providing a more nuanced and accurate view of space weather events, SUNNIE aims to be an indispensable tool for agencies like NASA and the US Space Force, ensuring that space operations can be conducted more safely and effectively in the face of unpredictable solar activity.