A Statistical and Predictive Framework for Evaluating Temperature Effects on Lithium-Ion Battery Lifespan

Lithium-ion batteries are fundamentally applied in electric vehicles, consumer electronics, and renewable energy systems. Among operational factors, temperature strongly influences efficiency, safety, and longevity. This study presents a data-driven investigation of the thermal impact on battery performance using life-cycle data from NASA's Prognostics Center of Excellence and the Hawaii Natural Energy Institute (HNEI). After preprocessing and feature engineering from charge, discharge, and impedance cycles, we performed hypothesis testing with Pearson and Spearman correlations, t-tests, and ANOVA. Results confirm strong inverse relationships between battery temperature and key metrics, including discharge time, capacity, and remaining useful life (RUL). The main contribution is a two-stage predictive modeling pipeline. First, battery temperature is estimated from discharge time and capacity using a multivariate linear regression model (R² = 0.88, RMSE = 0.25). The predicted temperature is then used in a second model to estimate RUL, achieving high predictive performance (R² = 0.98, RMSE = 25.95 cycles). Close agreement with an oracle baseline using true temperature confirms that error propagation is minimal and the pipeline is robust. This interpretable framework enables sensorless thermal inference and reliable integration into thermally aware battery management systems. The proposed methodology offers a reproducible framework for predictive diagnostics.