Different types of loads and distributed generations’ output show significant differences in different seasons and periods, namely timing characteristics. Based on timing characteristics of loads and distributed generations, this article takes the loads of peak valley division theory to divide a year for four seasons five periods. Combined with the time of use price, the cost of network loss is taken as objective function to establish optimal site selection of distributed generations. Finally, IEEE33node distribution network is used as example, and genetic algorithm is used to optimize ,verifying the rationality and feasibility of the proposed method.