A formulation is proposed to obtain dynamic equivalents which preserve closely the set of electromechanical modes associated to generators of the studied system. The external subsystem is replaced with fictitious generators whose ulterior parameters are included in the solution. Such formulation is posed as an optimization problem without constraints that can be solved by a variety of methods. In this paper, genetic algorithms proved to be effective due to the starting point randomness. Through the diversity of operating conditions taken into account, robust dynamic equivalents result. The root mean square (rms) error helps to measure the proximity of time solutions between full and reduced models.