Cooperative Co-evolutionary Approach (CCA) is a new architectureof evolutionary computation. Based on CCA, the paper proposes a new methodfor reactive power optimization problem in power system, which is non-convex,non-linear, discrete, and usually with a large number of control variables. Accordingto the decomposition-coordination principle, the reactive power optimizationproblem is decomposed into a number of sub-problems, which is optimizedby a single evolutionary algorithm population. The populations interactwith each other through a common system model and co-evolve and result inthe continuous evolution of the whole system. The reactive power optimizationproblem is solved when the co-evolutionary process ends. Simulation resultsshow that compared with conventional Genetic Algorithm (GA), CCA not onlycan obtain better optimal results, but also has better convergence property. CCAreduce the over-long computational time of GA and is more suitable for solvinglarge-scale optimization problems