Abstract: In practical engineering, the distribution network always suffers from insufficient measurements, poor observability, weak communication and low quality of network model. Aiming at the problems, an adaptive self-optimizing state estimation method for complex and changeable distribution network was proposed. This method combined static state estimation with dynamic state estimation and consisted of several modules which were respectively suitable for redundant measurements, only high-frequency measurements refreshing and unobservable environment caused by data loss. In this method, the spatial-temporal analysis of the data environment was firstly implemented to clarify the primary task of current state estimation, so as to adaptively decide the start of the corresponding estimation module. The simulation results based on IEEE system verify that the high-precision state estimation is realized and the errors of network parameters are identified and corrected in measurement redundancy. The state of part of distribution network is tracked in real time with the help of the high-frequency measurements. In the case of missing data, the algorithm convergence is restored and the estimation accuracy is maintained at the same time.