Abstract: Due to the similar characteristics of complex network between power grid and the Internet， this paper proposes an identifying method based on a modified PageRank algorithm for critical nodes in large power grid. Firstly， the original Google matrix of power grid is constructed on the basis of the grid topology and power flow direction. Secondly， the contribution of different node types to load supply and power transmission is considered to modify the Google matrix and the deprived matrix is obtained. Then， the influence of node failure on power fluctuation and voltage deviation is integrated as grid security factors. Finally， the PageRank algorithm is used to rank the nodes by their importance and identify the critical nodes which may affect load supply and system security. Taking the IEEE 39-node system as an example， the effectiveness of the proposed method is verified by the simulation of intentional attacks on critical nodes and the comparison with other methods.

Key words: PageRank algorithm, critical nodes, load supply, grid security