The state estimation of active distribution network is an essential component of the distribution management system. However, the accuracy of the estimation results is affected by the measurement position significantly. In order to improve the accuracy of state estimation and optimize the real-time database, this paper established an optimal measurement location model which combined with a state estimation algorithm based on parallel belief propagation to minimize the state estimation error of an active distribution network. Furthermore, an improved immune particle swarm optimization algorithm which optimize the initial position of the particles is proposed to solve the model. Finally, an optimal configuration scheme of the measuring device is obtained by simulation. Under the scheme, the precision of the state estimation is improved obviously.