State estimation provides reliable status information supporting accurate dispatch of complicated active distribution systems. However, there are many buses in the distribution system, and state estimation can only be accomplished with an insufficient number of phase measurement units, with little confidence of accuracy. Moreover, the voltage instability by distributed generation uncertainties and power flow changes by soft open points bring new challenges to state estimation. Therefore, an optimal phasor measurement unit placement model considering economy, bus voltage out-of-limit weight and state estimation accuracy is investigated in the paper. First, Monte Carlo simulation calculates the bus voltage out-of-limit weights to approximate distribution generation uncertainties. Then, a virtual bus representing soft open point is introduced, displaying power flow changes with its own control signal in state estimation. Finally, placement locations are obtained from new optimization objective and constraints. Simulations and analyses on the modified IEEE 33-bus distribution network validate the proposed model.