In order to increase the operating efficiency and economic performance, this paper uses the group search optimizer to optimize the control variable of power systems so as to decrease the reactive loss. Group search optimizer is a new kind of swarm intelligence optimization algorithm, which divides group members into three types: producer, scrounger and ranger. The location of rangers is random, which can avoid entrapment in local minima effectively. The input of reactive power and the voltage transformer ratios, and the terminal voltage of generator are designated as control variables, after the iterative computations and power flow calculation by using the swarm search optimization algorithm, the minimum of the reactive loss and the corresponding control variable can be figured out. At last, the IEEE14-bus system and IEEE30-bus system are simulated with Matlab7.6 and contrasted with the others optimizers, the results show that the convergence rate of the group search optimizer is fast and stable, and superior in reactive power optimization.