In view of the gradual increase of penetration of distributed generators（DG）and separate planning of active power and reactive power，the concept of generalized power sources is proposed to name all power sources in the distribution network capable of supplying active power or reactive power.Considering the time-sequence characteristic and randomness of load and uncontrollable wind power and photovoltaic factor，a multistate model with time-sequence is developed.A bi-level programming model adopting the chance constrained programming is proposed，which considers the reactive regulating ability of DG and adjustment of generalized power sources with load varying during operation，so that the calculation of the objective function and indices of network loss and voltage is accurate.The model takes the minimum annual cost and minimum annual operation cost above the confidence level as the upper and lower level planning objective function，respectively.The elitist strategy genetic algorithm based on segmented chromosome coding is applied for solving the bi-level programming model.The simulation results of an example show that comprehensively optimal configuration of active and reactive power sources can improve resource availability of the distribution network，meanwhile adjustment of DG reactive output can make full use of its reactive potential.The method considering DG reactive power can play a role in DG reactive control and guarantee the economical and safe operation of distribution networks.