The lithium iron phosphate power battery is an important component of mine-used lifesaving cabin, which state of charge (SOC) estimation accuracy directly influences the sheltering people"s safety. In view of common method in SOC estimation of battery, the paper proposed a SOC estimation of power battery in mine-used lifesaving cabin based on adaptive kalman filtering method. According to the analysis of the characteristics of the battery,The establishment of a more accord with the actual improvement equivalent second-order RC battery model and state space model of the battery. By pulse discharge experiments and improved RLS algorithm with forgetting factor, online identification model parameters, and the adaptive kalman filtering algorithm (AKF) used in this model, the online estimate the SOC of the battery. The experiment results shows that AKF can real-timely correct model error and estimate the dynamic changes of SOC with high accuracy and meet battery management system of mine-used lifesaving cabin requirements.