Abstract: With the increasing penetration of renewable energy resources and flexible loads, the damping characteristics of critical electromechanical eigenmodes change with the time invariant operating conditions of power systems. When it comes to the dynamic stability evaluation, the QR method is accurate, but fails to satisfy the needs for online tracing. In order to calculate the time-varying eigenmodes, this paper proposes a method to reduce the state matrix order of the system via inertia equivalence. All state variables of an eigenmode are mapped to a new sub-space upon the modal clusters. However, the pattern matching and modal varying of eigenvalues in the calculation may lead to discontinuous results. To solve the problems, the relevant factor index of the dominant clusters is proposed and a framework is built to achieve fast calculation of critical eigenvalues in different conditions. The critical eigenmode calculation method is tested in the IEEE 145-bus system, and the simulation results demonstrate its accuracy and rapidity.