Recently, double-skin systems are highly considered due to their aesthetics and energy performance. To perform efficient control of the double-skin facade it is important to apply dynamic optimal control based on the predicted behaviour of the system. This study presents optimal control strategies of the double-skin facade using Generalized Pattern Search (GPS) and Genetic Algorithm (GA). The GPS and GA can be effectively used for nonlinear and nondifferentiable optimization problems. As a follow-up of the previous study (Yoon et al, 2009), the existing optimal controls of the double-skin facade are classified according to control strategies (rule-based approach, exhaustive search, and gradient-based search). This study aims to compare the optimal controls using GPS and GA with the aforementioned three control strategies. To compare control strategies, the lumped simulation model developed in the previous study is integrated in MATLAB optimization routines (patternsearch, ga) which determines optimal control variables (blind slat angle, airflow regime, opening ratio of the ventilation dampers). The optimal control minimizes the objective function in according to heating and cooling modes over a given time horizon. It was found that the optimal control based on GA performs better than the others, but takes longer time to search for converging solution sets.