Battery SOC estimation accuracy is one of the important factors influencing the performance of electric vehicles. Considering that the traditional Kalman filter calls for the undersdanding of the statistical properties of system noise, a fuzzy adaptive Kalman filtering algorithm was presented, which was based on RC equivalent circuit model and identified by applying multiple linear regression method. Urban Dynamometer Driving Schedule simulation comparative results have shown that the proposed algorithm has higher SOC estimation accuracy than the conventional Kalman algorithm and can keep error within 2%.