Abstract: To solve transformer ratio optimization problem for large scale power grid monthly operational mode, a combinatorial optimization method based on cluster parallel computing is proposed in this paper.The searching range of optimal ratio is reduced through methods such as determining the upper and lower bound, ratio combination within electrical section, and step-by-step optimization, then the optimal tap position can be given out through linear programming in quasi-linear interval based on sensitivity. The computing efficiency is improved by enhancing the convergence of interior point algorithm. The method can effectively avoid the problem of ‘combination explosion’. The effectiveness is verified by a practical case of the China Southern Power Grid.