Abstract: Due to the different sampling frequencies of SCADA data and PMU data in power system,there is a problem of data delay. The SCADA data and PMU data are unified to the same time scale based on the method of PMU optimal buffer length,and then combines the unscented transformation with the exponential weight function robust algorithm to propose a two-stage unscented Kalman filter dynamic state estimation method. In each snapshot,firstly,the method combines the predicted value after unscented transformation and two-parameter exponential smoothing prediction with SCADA data for the first stage of filtering,and then combines the estimated value obtained by first stage with PMU data for the second stage of filtering. This method can obviously increase the measurement redundancy in the filtering process through two-stage filtering,and effectively reduce the impact of SCADA measurement with lower measurement accuracy on PMU measurement with higher measurement accuracy in the process of mixed data filtering. Based on the IEEE-39 buses standard system,the proposed method is simulated and analyzed. The simulation results show that the proposed method can effectively combine the PMU and SCADA data to calculate the dynamic state of the power system with high estimation accuracy and good robustness.