A new robust adaptive fading (AF) unscented Kalman filter (UKF) for decentralized dynamic state estimation (DSE) has been presented. It utilizes the AF-UKF to minimize possible scale mismatches in the state and measurement noise covariance matrices of the KF and robust statistics to combat bad data or outliers. Simulation results on the NPCC 48-machine 140-bus system showed that the proposed algorithm yields more accurate and reliable performance than conventional methods under bad-data and noise covariance mismatches.