Neural Network Emulator for Atmospheric Chemical ODE

• DOI: 10.48550/arxiv.2408.01829
• Publication Date: 2024-08-03
• Authors: Zhi-Song Liu; Petri Clusius; Michael Boy
• Keywords: Ode

Modeling atmospheric chemistry is complex and computationally intense. Given the recent success of Deep neural networks in digital signal processing, we propose a Neural Network Emulator for fast chemical concentration modeling. We consider as time-dependent Ordinary Differential Equation. To extract hidden correlations between initial states future time evolution, ChemNNE, an Attention based (NNE) that can model ODE process. efficiently simulate changes, sinusoidal embedding to estimate oscillating tendency over time. More importantly, use Fourier operator process efficient computation. also three physical-informed losses supervise training optimization. evaluate our model, large-scale dataset be used network evaluation. The extensive experiments show approach achieves state-of-the-art performance modeling accuracy computational speed.