In order to solve power quality problems more effectively such as harmonic pollution, low power factor as well as the sensitive loads not to work properly because of voltage sags, flickers and so on. The optimal configuration of custom power devices is studied. By considering power quality improving goals and total investment cost, a mathematical model is proposed for optimization. Besides, a mixed optimization algorithm that combines genetic algorithm and interior-point method is proposed for the optimal configuration of multi-type custom power devices. This mixed strategy uses genetic algorithm to find the optimal location and approximate optimal capacity of each device. Then the approximate capacity will be set as the initial value of the interior-point method, and more accurate capacity can be calculated through interior-point method. Besides, some improvements are made in the process of genetic algorithm and interior-point method. These improvements can speed up the convergence. Theoretical analysis and simulation results indicate that this mixed algorithm has better stability, and more accurate results can be obtained than through traditional genetic algorithm. This work is supported by National Natural Science Foundation of China (No. 51477184 and No. 61271001) and Fundamental Research Funds for the Central Universities (No. 14CX02085A).