The invention discloses a Gaussian elimination method for analyzing and optimizing power grid topology structures. According to the method, the aims of improving the power grid topology analysis efficiency and optimizing power grid structures are taken as starting points; resource advantages of reconfigurable optical processors are utilized to construct a large-digit generalized adding unit and alarge-digit generalized multiplying unit; by means of data splicing and edition technologies, the operations of multiple rows and multiple columns in one iteration of the Gaussian elimination method are processed in parallel; and triple loops such as decomposition, forward substitution and back substitution in the traditional algorithm are improved into one loop, so as to reduce the delay of network topology analysis of the Gaussian elimination method, and improve the complicated network data analysis and processing ability. The method aims at enhancing the ability, for dynamically processingcomplicated power grid topology structures, of the Gaussian elimination method, and is capable of effectively answering massive data problem of power grid topology analysis and improving the power grid topology structure analysis efficiency.