Abstract For early detection of the threat of a cascading failure we need to use the immune information system, the main subsystem of which is a virtual stability analyzer. We propose to implement the analyzer by usage of discrete linearized models of energy systems and apparatus of Lyapunov direct method, based on the study of discrete Lyapunov equation solutions, called Gramians. The functional of stability loss risk is equal to the square of H2-norm of the matrix transfer function, which is proportional to energy of the system. We propose to use a functional risk assessment through spectral decomposition of controllability and observability Gramians for evaluating of the energy accumulated in combinational dominant eigenvalues of the system. In comparison with the method of modal analysis the proposed approach allows to evaluate the impact of synergies between weak-stable eigenvalues and their combinations to the system stability loss risk and to detect the potential source of stability loss.