The reactive power optimization problem is studied and a kind of reactive power optimization scheme of power system based on adaptive refraction learning and elite search social spider optimization（ARLESSO）algorithm is proposed. As for such defect as not high local optimization and convergence precision by social spider optimization（SSO）algorithm

the multi-function sub ethnic partition strategy is introduced. The spider population is classified dynamically into elite group

disturbance group and maintain group in accordance with the size of individual fitness of the spider. For elite group and disturbance group

the elite search and adaptive refraction learning evolution mechanism are adopted respectively to improve the algorithm’s global depth search ability and population sample diversity. Based on this

the minimum network loss model for reactive power optimization is constructed and the ARLESSO is used to solve the problem. It is shown by the IEEE node test system simulation results that the proposed algorithm

compared with other reactive power optimization schemes

has stronger global searching ability and higher accuracy

and can give effectively the reactive power optimization result of power system.