Techniques for the analysis and design of multivariable system controllers in the frequency domain are used to verify the possibility of achieving robust control using reduced-order decentralized controllers with the objective of robust stability of a power system. In order to achieve robustness, the selection of the most effective signals and inputs for application to the controllers is essential. Accordingly, two frequency analysis techniques were used: relative gain matrix (RGM) and singular value analysis. The proposed techniques were applied to a three-area load-frequency system. Finally, it was possible to realize a first-order efficient controller that remains robust, even in the presence of faults or disturbances. The controller used was an H¥ controller and the optimization method applied was based on genetic algorithms. © 2016, ICICI International .121321331