Due to the slow calculation speed of state estimation in complex active distribution network, a distributed state estimation of complex active distribution network based on Lagrange relaxation technique is proposed. A partition model of complex active distribution network is established, in which the measurement configuration of complex active distribution network and distributed parallel computing efficiency are analyzed. The calculation speed and accuracy of each sub-region after partition are also considered. By using the Lagrange relaxation technique, the constraint conditions are absorbed into the objective function, which can reduce the difficulty of solving the partition model. The state estimation problem of the network can be solved in a distributed parallel environment by using the network decoupling method. The state estimation of each sub-region is carried out by using the state estimation method with exponential objective function which can automatically exclude bad data effects, which effectively reduces the system size and Jacobi matrix order of the problem, and improves the efficiency of the state estimation algorithm while ensuring the accuracy of the results. Simulation results show that the proposed method can achieve reasonable partition in a complex active distribution network, and the calculation speed of the state estimation is largely improved. This work is supported by Science and Technology Project of State Grid Corporation of China (No. PD71-14-032).