Taking the minimum total year energy loss as an objective function, a modified teaching-learning-based optimization (MTLBO) algorithm is proposed for reactive optimization of power systems. The proposed teaching-learning optimization algorithm is a new population-based optimization method which includes two phases: teacher phase and learner phase. To overcome local convergence, a new phase called “modified phase” based on a self-adaptive wavelet mutation strategy is added to the algorithm to improve the performance. The proposed method is applied to IEEE-30 bus test system and the simulation results are compared with other algorithms which verify the superiority of the proposed method. It shows that this method is an effective method for large-scale power system application.