With the integration of large clusters of variable renewable energy, plus the fact that modern power system has appeared great varieties of complicated characteristics. Improvement of safe and stable operation of the grid system has become a research focus in the new situation. Such being the case, a critical node identification method based on the theil entropy theory is proposed. By studying the characteristics of power flow and voltage variation, new methods for system safe operation improvement are explored. Firstly, proceeding from the mechanism of load variation on voltage and power flow, the theil entropy index of voltage growth rate is established to describe the equilibrium of voltage variety and the theil entropy index of weighted flow impact rate to describe the uniformity of power flow variation. Secondly, considering both subjective preferences and objective data, binary feature analysis and entropy weight method are combined to get the comprehensive index which can accurately identify the critical nodes for power grid, At last, the simulation results of IEEE30 node system and a southwest certain area power net-work has verified the validity and efficiency of the model. Further more, evaluation results has provided a reference for the construction of security defense system in modern power system.