Computation requirement for forming the state matrix in small signal stability analysis of power systems is rapidly increased with the expansion of system scale. By analyzing the forming process of state matrix, reconstructing the related formulas, introducing reduced-order matrix inversion, and other approaches, improvement in computing speed is achieved in this paper. According to the Plug-in Modeling Technique, the first-order transmission block equation, correlation matrix equation and the first derivative equations are mixed; and the state matrix expression is obtained by further reconstruction and transformation. Partitioning and descending order is used in matrix inversion; triple form is used in matrix storage; parallel computation technique is used in processing cycle and independent program sections. Comparison and analysis on two different scale of power systems shows the effectiveness of proposed method.