The paper presents methods for studying the dynamics of nonlinear processes in relation to assessing the risks of stability loss on the basis of accumulated knowledge about the operation of power supply systems and methods for analyzing power supply modes. The methods are a complex of linear point discrete predictive identification models for a wide class of nonlinear objects, as well as spectral decompositions of Gramians for linear models.