This work describes a measurement-based load modeling methodology, using voltage and current data registered during power system disturbances. These load models are used on computational simulations for predicting power system stability after disturbances of system operational points. Practical results are presented of the proposed methodology, defining load as a system whose output is power deviation from its operational state and input is voltage sags, both measured at 69 kV bus bars of São Francisco Hydroelectric Company (CHESF), points of common coupling (PCC) between CHESF and local distribution utilities. Therefore, distribution utilities are seen as loads supplied by CHESF’s transmission system. All devices consuming power from the PCC are aggregated into an equivalent model, parameterized to simulate most typical contingencies measured by these 69kV load bars. Optimization technique used for load model parameter estimation is Genetic Algorithm (GA), showing his flexibility on implementation and good coverage and accuracy in the final results. At the end, it will be presented a set of active and reactive power curves during and after voltages sags, measured on CHESF’s substations. These curves were used as estimation data to parameterize load models for each substation chosen.