The invention discloses a two-layer planning-based electric power system N-K fault analysis method, and relates to fault analysis of an electric power system. The fault analysis method comprises the following steps: for an upper layer problem: establishing a mathematical model in a direct current power flow condition, and determining an upper layer problem target function; in a condition of a given fault multiple number, obtaining a fault combination which can cause the maximum of power grid load shedding quantity; for the lower layer problem: performing simulation on an operating condition after power grid suffers from an accident specific to the fault combination in the upper layer problem decision making in the step 1); when various kinds of operating constraint conditions of the power grid are satisfied, determining the power grid load shedding quantity by regulating the output power of a generator unit to be the minimum; and when the upper layer problem target function value is equal to the maximum value of the minimum power grid load shedding quantity caused by each fault combination, obtaining the fault combination which causes the maximum of the power grid load shedding quantity. By adoption of the method, the power grid operation constraint and a re-correction process after the power grid accident are taken into full consideration, and the actual operation of the power grid can be satisfied.