With a large number of distributed generation and electric vehicles connected to the distribution network, the unpredictable output of DG and the complex monitoring of load are the difficult problems for the operation and management of distribution network. Considering the questions of large errors and being susceptible to bad data in traditional Unscented Kalman Filter (UKF) prediction, the adaptive factor is constructed by using the new vector and an Adaptive Unscented Kalman Filter (AUKF) algorithm is proposed to state estimation of distribution network. When the system load is abrupt and the measurement has bad data, the corresponding predictor covariance matrix is modified online by the adaptive factor to reduce the influence of prediction error on the estimation accuracy. The simulation analysis is carried out in three-phase unbalanced distribution network. The results show that the AUKF algorithm is more accurate and robust than the UKF estimation, and the validity of presented algorithm is verified. This work is supported by National Natural Science Foundation of China (No. 61540067).