The current under frequency load shedding (UFLS) scheme usually cuts the load of offline setting according to the frequency drop delay, and it ignores the load characteristics, which leads to the overcut and undercut problems. Grey correlation analysis is proposed to calculate correlation degree of shedding load with different importance level based on the indexes of load frequency regulation factor, unit power outage loss, and load rate. The load shedding order depends on the correlation degree. Considering system spinning reserve capacity and load regulation characteristics, the total shedding capacity is decided after using initial frequency change rate measured by WAMS technology to adaptively evaluate the system power gap. The total shedding capacity is assigned to corresponding load automatically in the order of load shedding. Four typical fault scenarios validate that the proposed scheme can decide the intelligent UFLS adaptively and obtain the optimum system dynamic frequency recovery characteristic curve with the minimum load shedding.