A Sensor-Aware Physics-Based Framework for Continuous Smartphone Battery State and Lifetime Prediction

This paper proposes a continuous-time physics-based framework for smartphone battery state-of-charge (SOC) and time-to-empty (TTE) prediction using multi-source sensing and device operating information. A first-order RC equivalent circuit representation is introduced and extended by incorporating six major factors, including CPU workload, GPS activity, screen brightness, network connectivity, temperature, and battery aging, to characterize realistic battery depletion behavior. Based on the GreenHub dataset, a multi-factor coupled SOC model is developed, and K-means clustering combined with the fourth-order Runge-Kutta algorithm is employed to predict TTE under representative smartphone usage scenarios. Furthermore, an interpretable factor analysis framework integrating sensitivity coefficients, partial contribution rates, and entropy weights is introduced to quantify the contributions of different factors to battery consumption. Experimental results demonstrate that the proposed framework achieves an MAE of 4.2% and an RMSE of 6.8% for SOC prediction. The analysis reveals that CPU workload is the dominant factor affecting battery depletion, while GPS activity and screen usage also exhibit significant impacts. The proposed framework provides an interpretable solution for battery state perception and energy management in intelligent mobile sensing systems.