Power batteries are the power source of electric vehicles，so the estimation precision of the battery SOC directly affects the vehicle's handling and safety. In order to improve the accuracy of the online SOC estimation, a fuzzy rule is introduced in the filtering stage to adjust the filtering parameter β of the variable structure to suppress the buffeting of the convergence stage. This paper utilises double fuzzy-variable structure filtering to estimate the SOC. The first fuzzy-variable structure filter filters the terminal voltage of the power battery. Another fuzzy-variable structure filter uses the filtered terminal voltage data to estimate the SOC. During the simulation, 1%, 2%, and 3% Gaussian noise was added to the experimental data to simulate the sensor error characteristics in the actual industry and to verify the effectiveness of the double-fuzzy variable structure filtering algorithm in practice. The error of terminal voltage and the real voltage before and after fuzzy-variable structure filtering are compared thereby the effectiveness of filtering in reducing the terminal voltage error is verified. Under the three kinds of Gaussian noise conditions, the SOC is estimated by using terminal voltages before and after filtration. The simulation results verifies that the filtered terminal voltage can effectively improve the SOC estimation accuracy. The three kinds of different SOC initial values have proved that the algorithm has little dependence on the initial value. The experiment and simulation show that the double-fuzzy variable structure filtering is suitable for SOC estimation with low sensor accuracy and unknown initial value. It also has high estimation accuracy and high convergence speed. It provides a practical SOC estimation scheme for the battery management systems.