Enhanced S‐ARIMAX model performance and state‐of‐health prediction accuracy with battery pack degradation factor combinations

Abstract

This study addresses the critical need for accurate state‐of‐health (SOH) predictions in lithium‐ion batteries, crucial for maintaining efficient battery operation and extending lifespan. We propose a novel approach utilizing the Seasonal autoregressive integrated moving average with exogenous variables (S‐ARIMAX) model, integrating multiple battery degradation factors as exogenous variables for precise SOH estimation. Our methodology involves correlation analysis to identify optimal factors, which are then incorporated into distinct cases of the S‐ARIMAX model. We evaluate the model's performance using mean absolute percentage error (MAPE) and identify voltage deviation as a highly correlated exogenous variable significantly impacting predictive accuracy. Notably, our findings underscore the S‐ARIMAX model's superiority in longer cycles, demonstrating a MAPE as low as 0.1113 at 160 cycles, emphasizing the model's adeptness at capturing cyclic patterns for precise long‐term SOH forecasts. Comparing S‐ARIMAX and ARIMAX models highlights the pivotal role of seasonality, particularly in long‐term predictions. The study's quantitative findings emphasize the necessity of integrating highly correlated variables and accounting for seasonal patterns in optimizing lithium‐ion battery SOH predictions. Our approach presents practical implications for maintaining battery performance and lifespan in electric vehicles and energy storage systems, underlining the importance of selecting strongly correlated variables and integrating seasonal components for accurate and reliable SOH forecasts.