Thermal–Electrochemical Interactions and Data-Driven Diagnostics for Lithium-Ion Battery Degradation Under Stress Conditions

Understanding how lithium-ion batteries degrade under coupled thermal and electrochemical stress is essential for ensuring their reliability in electric vehicles and stationary storage systems, particularly in hot regions. Our investigation integrates controlled cycling experiments, advanced data analytics, and machine-learning diagnostics to identify the mechanisms driving performance loss under elevated temperature, overcharge conditions, and standard operation.

A comprehensive test matrix was conducted on LG and Panasonic cells, cycled under multiple voltage and temperature settings. Standard charge–discharge at 25–45 °C exhibited smooth, monotonic capacity fade, while overcharge protocols triggered abrupt degradation events—manifested as sharp drops in discharge capacity, erratic energy throughput, and intermittent spikes in cumulative capacity loss. These anomalies indicate unstable electrochemical pathways and accelerated aging behavior when voltage or thermal stress is imposed.

Building on these experimental insights, machine-learning frameworks were applied to extract early-life degradation signatures and predict knee-onset behavior. Classical and ensemble models revealed that naïve data-splitting approaches inflate prediction accuracy, whereas group-wise validation provides more realistic generalization across cells. Complementary analysis using knee-point modeling demonstrated that early-cycle features can provide actionable indicators of forthcoming accelerated aging, enabling timely intervention and more reliable State-of-Health assessment.

During the presentation, we will discuss the interplay among thermal stress, overcharge behavior, and degradation kinetics; showcase the diagnostic value of capacity- and energy-based indicators; and demonstrate how robust data-driven models can strengthen battery safety, lifespan forecasting, and thermal management design.