S-AT-KiBaM-Age: A Dual-Time-Scale Integrated Model for Smartphone Battery

Abstract

Predicting smartphone battery life remains difficult because hardware load, environmental conditions, and battery aging interact in complex ways. Most current methods rely on curve- fitting or black-box models that lack physical grounding and fail to capture real-world usage patterns. We propose the S-AT-KiBaM-Age model, a continuous-time framework that couples macro-scale aging with micro-scale discharge dynamics through thermo-electro-kinetic interactions. The model embeds the Kinetic Battery Model for charge transfer, applies the Arrhenius equation for temperature-dependent resistance, and tracks State of Health degradation through an empirical aging formula. By quantifying energy dissipation pathways—Joule heating, trapped charge, and voltage collapse—it identifies what drives rapid battery drain. Validation on the AndroWatts dataset shows an RMSE of 0.183 hours (2.86% relative error) across 24 usage-environment-aging scenarios. Sobol sensitivity analysis reveals that screen brightness (first-order index 0.55), ambient temperature, and battery aging dominate battery life, while background app management improves runtime by less than 1%. Users can extend battery life by 23-35% by keeping screen brightness at 30-50%, avoiding temperatures below 0°C or above 40°C, and using Wi-Fi instead of cellular in weak signal areas. The model generalizes to other portable devices via parameter scaling and enables physics-aware Model Predictive Control for OS-level power management. This work offers a physically grounded tool to reduce battery anxiety and improve energy efficiency.