Keywords
State of Health; Lithium-Ion Battery; PatchTST; Physics-Infromed Machine Learning
Abstract
Accurately forecasting the state of health of lithium-ion batteries is critical for improving performance, reliability and lifetime in energy storage applications. Battery capacity degrades into a nonlinear pattern over cycling due to electrochemical processes where neither purely data driven nor physics-based models can capture alone. This study looks at a hybrid framework combining that PatchTST patch-based transformer architecture with physics-based features derived from the solid electrolyte interphase and pseudo two-dimensional models. Physics inspired proxy features were computed from cycling data and concatenated with electrochemical measurements as added input channels before patch segmentation. There are five model configurations that were evaluated on the TJU NCA battery dataset: two standalone physics models, one data-driven transformer, and two hybrid variants. All transformer-based models were evaluated at the cycle level in real mAh units. The base PatchTST model got a MAE of 20.542 mAh and R2 of 0.9969, showing that the patch-based transformer learned battery degradation trajectories. The PatchTST+P2D outperformed the PatchTST+SEI in all metrics, supporting the hypothesis that multidimensional electrochemical variables provide more degradation context than scalar proxies. The results emphasize the promises and current limitations of physics-based transformer architectures for battery SOH prediction.
Thesis Completion Year
2026
Thesis Completion Semester
Spring
Thesis Chair
Rui Xie
College
College of Sciences
Department
Statistics
Thesis Discipline
Data Science
Language
English
Access Status
Open Access
Length of Campus Access
None
Campus Location
Orlando (Main) Campus
STARS Citation
Ravuri, Pavan, "Hybrid PatchTST and Physics-Based Framework for Predicting Lithium-Ion Battery State of Health" (2026). Honors Undergraduate Theses. 522.
https://stars.library.ucf.edu/hut2024/522
Included in
Data Science Commons, Longitudinal Data Analysis and Time Series Commons, Statistical Methodology Commons, Statistical Models Commons
Accessibility Statement
This item was created or digitized prior to April 24, 2027, or is a reproduction of legacy media created before that date. It is preserved in its original, unmodified state specifically for research, reference, or historical recordkeeping. In accordance with the ADA Title II Final Rule, the University Libraries provides accessible versions of archival materials upon request. To request an accommodation for this item, please submit an accessibility request form.