A novel subregion-based optimization algorithm is developed for multi-dimensional optimization of electromagnetic devices. In the algorithm, the whole design space is divided into a number of subregions. And then in each subregion optimal solution is found by using particle swarm optimization (PSO) based on the Kriging approximation of the objective and constraint functions. Among the optimal solutions obtained from the subregions, a global optimal solution is obtained. The algorithm efficiently reduces the expensive computational cost especially for a design of the large scale multi-dimensional problems.