BattVAE-GP: Generative Modeling of Long-Horizon Battery Degradation with Uncertainty Quantification

Short summary

BattVAE-GP combines a Variational Autoencoder with a sparse multitask Gaussian Process to create a surrogate model for lithium-ion battery degradation. Trained on DFN/P2D simulation data from PyBaMM, the VAE latent space organizes degradation trajectories by cycle progression and charging rate, while the GP interpolates unseen C-rates with uncertainty estimates. The framework enables computationally efficient, uncertainty-aware State of Health predictions across operating conditions.

• •Hybrid VAE + Gaussian Process surrogate for battery degradation modeling
• •Latent space organizes trajectories by cycle and charging rate; GP interpolates unseen C-rates with uncertainty
• •Enables efficient uncertainty-aware SOH prediction vs expensive physics simulations

Generated with AI, which can make mistakes.