Currently， the power system dynamic state estimation mainly adopts extended Kalman filter（EKF） method. Nevertheless， the EKF method normally suffers from slow convergence and low robustness. This paper introduces a new nonlinear method， namely unscented Kalman filter（UKF） method， to estimate the state of a power system. By applying the unscented transformation（UT）， the UKF method can avoid the error caused by the linearization process and the complicated calculation of Jacobian matrix in the EKF method， and therefore show better estimation accuracy and robustness than the EKF method. On the other hand， since wide area measurement system（WAMS） with its phasor measurements has advantages such as higher accuracy and network synchronization， the mixed measurement vector combining WAMS measurement and supervisory control and data acquisition system（SCADA） measurement is also introduced in the dynamic state estimation. The simulation shows that UKF method can estimate the state of power system more accurately than EKF method， and the introduction of WAMS can further improve the performance of the dynamic state estimation.