When choosing initial feasible solutions， the electrical characteristics of power grid may often be overlooked， lowering computational efficiency in power distribution network optimization algorithms. By proceeding from the power grid characteristics， network simplification and switches grouping are presented in this paper by analyzing the distribution network topology model based on common information model（CIM）. By considering the inductive network reconfiguration constraints， solution space satisfying the radial-running condition is formed. A case study using the genetic algorithms shows that the computational efficiency of optimization algorithms will increase exponential multiples by topology contraction and removal of the infeasible solutions.