The questions of construction and application of polynomial models (PM) for classification of power system operational conditions in depending on steady-state stability margins are considered. The measurements of voltage phasors are used as input data to PM. A comparison of PM and artificial neural networks by use a criterion of classification errors was realized.