Genetic algorithms (GAs) are global optimization methods that can be applied to almost all problems, requiring only proper fitness function to evaluate. However, one problem of general GA is slow convergence. An improved GA is presented to speed up the efficiency of searching for global optimum in this author. The concept of this proposed method uses a few cost to obtain better individuals in initial population, and the evolution of GA is divided into two-stage with the concept of the genetic evolution process, which uses to improve efficiency. An improved GA with finite-difference time-domain (FDTD) will be applied to optimize mushroom-type EBG structures, which can obtain a wide range stop-band by adjusting the position of via with different patch size cascaded without changing via size, then the simulation and measurement results are also compared. In addition, the novel steps will be presented to design broadband mushroom-type EBG structures with smaller size systematically.