The invention discloses a power failure risk computing method for an electric power system containing double-fed wind power plant. The power failure risk computing method comprises the steps of pre-establishing multiple models, wherein a simulation process comprises the steps of determining the initial working condition of a system, and setting an initial fault; determining whether a short circuit fault or line broken fault occurs or not, if the short circuit fault occurs, processing according to an off-network model of a wind turbine generator, and then performing island-division searching; determining a frequency fall-off and recovery condition of an electrical island according to a frequency stabilization model, and enabling the electrical island to restore to a power balancing state; carrying out alternating current power flow calculation on the system and determining whether the system is converged or not, if not, solving a power flow convergence boundary, and analyzing voltage weak points; switching loads for the voltage weak points; next, determining whether the system is restored to be stable or not, if not, determining that the current wind power plant system is subjected to global voltage collapse, cutting all loads; and if the system is restored to be stable, determining whether a tangent line exists or not, if so, returning to perform the island-division searching step, and or otherwise, finishing the simulation process. By adoption of the power failure risk computing method, the load reducing amount required by the system for system recovery can be reduced to the minimum, and cascading failure analysis error is lowered.