A multi-factor coupled continuous-time model for smartphone battery discharge prediction

Abstract

Accurate prediction of smartphone battery life under complex usage scenarios is crucial for both user experience and system power management. This paper proposes a multi-scale dynamic model that integrates electrochemical mechanisms with user behavior patterns. The model employs a second-order equivalent circuit to capture battery dynamics, incorporating the real-time power consumption of key components (screen, CPU, network) along with temperature and aging effects. Validated on public datasets, the model demonstrates a high linear correlation (0.993) between predicted and expected discharge times. Sensitivity analysis identifies CPU load, screen brightness, and network type as the most critical factors affecting battery life. These findings provide prioritized optimization strategies for users and vendors, with CPU task management and Wi-Fi switching emerging as the most effective measures.