Most of existing node vulnerability assessment methods fail to present a full view of evaluation and the index weights are inaccurate. To solve these problems, this paper starts analysis from the topology of the system, the current safe operation of the power system as well as the consequence of a node failure, and constructs a hierarchical node vulnerability assessment index system taking the structural vulnerability of power grid, voltage margin after accident, overall system load rate, and line flow impact into account. On this basis, a comprehensive evaluation model of AHP-gray relevance is proposed, integrating indicator value coefficients into objective grey-relevance. And a dynamic resolution coefficient adjustment strategy is used to achieve the combination of both subjective preference and objective data. The simulation results show that this method can identify the vulnerable nodes in the power grid more effectively than the traditional methods, which means the proposed model has certain practical value. This work is supported by National Natural Science Foundation of China (No. 51277134).