As trees growth simulation algorithm(TGSA) has the disadvantage of bad convergence stability and it is difficult to find the global optimal solution in solving large scale reactive power optimization problem, a new reactive power optimization algorithm called chaotic trees growth simulation algorithm(CTGSA) is presented by using the characters that chaos optimization algorithm is sensitive to initial value and has ergodicity. An operator called chaos immigrant is introduced to the process of TGSA to improve the quality of feasible solutions in the growing point collection and keep the diversity of feasible solutions, by which the new algorithm has better convergence stability and optimization precision. When the algorithm is used for IEEE 30-bus system, the results show that the algorithm has strongly global optimization ability and convergence stability.