The Extended Kalman Filtering (EKF) algorithm is the mainstream algorithm for State of charge (SOC) estimation. The uncertainty noise covariance matrix can slow down the EKF algorithm's correction of state variables, and even make the filtering process unstable or divergent, thereby reducing the accuracy of SOC estimation. Here, an improved Extended Kalman Filter algorithm with improved Grey Wolf Optimization is developed to address the above issue. Firstly, a first-order Thevenin lithium battery model is established to identify the battery parameters. Secondly, the Grey Wolf Optimization algorithm is improved and it is used to optimize the random noise covariance matrix of EKF. Finally, under different working conditions, the Simulation and battery test platform are built under different working conditions to demonstrate the validity of proposed algorithm, and the proposed algorithm can decrease SOC estimation error by 1%.