Smartphone Battery Life Prediction Based on Multi-Module Power Decomposition and Second-Order Thevenin Thermo-Electro Coupled Model

Abstract

This paper proposes a continuous-time forecasting method that integrates power consumption decomposition with battery state evolution for predicting smartphone battery life. The method unifies the modeling of power-consuming modules—such as the screen, CPU, network, GPS, and background activities—and combines them with a second-order Thevenin thermoelectric coupling model to describe the dynamic changes in SOC, voltage, polarization, and temperature. In the parameter estimation phase, least squares and weighted least squares estimation are introduced to improve the stability of the model fit. In the prediction phase, the remaining discharge time is calculated using the SOC decay relationship, and prediction reliability is evaluated through uncertainty propagation and sensitivity analysis. The results indicate that this model can reliably characterize battery discharge trends under various load conditions and demonstrates good transferability and practical value.