In view of multi-objective reactive power optimization issue of distribution network containing distributed power supply in both reducing active power network loss of system and simultaneously improving voltage level

a kind of non-dominated sorting differential evolution algorithm（I-NSDE） is proposed.Based on the non-dominated sorting differential evolution algorithm（NSDE）

opposition based learning and self-adaption of control parameters is introduced to enhance the diversity and the global convergence of algorithm.The IEEE 33 node system distribution system is taken as calculation example for simulation analysis.It is shown by the result that the proposed algorithm can maintain the diversity of Pareto-optimal solutions and

at the same time

has better global convergence

which can provide various reactive power optimization proposals for network operators.