A weighted least squares method for determination of power system equivalents

The effect of several important factors, like the acceleration of convergence, selection of initial gradient, choice of parameter models, selection of output variables, etc., on the convergence of the parameters of a power system equivalent is investigated. It is observed that a fifth-order parameter model with two output variables provides little faster convergence near a minimum compared to the reduced-order models. The reduced-order models with lower outputs are computationally efficient and provide reasonable values for the more important parameters. A check on the stability of the model has been proposed to save further computation time.