Blade loss seriously affects the stage efficiency of steam turbine. The Genetic Algorithm is used to optimize the NURBS curve of blade profile in governing stage. With the maximum value of the function composed of total pressure loss coefficient, steam flow Angle and static pressure ratio, the optimized blade profile in peak regulating condition is obtained. The flow pattern of internal flow field, the load distribution of blade and the development law of cascade losses are studied before and after blade profile optimization. The results show that after blade profile optimization, larger rotor blade leading edge diameter can effectively reduce the influence of steam attack angle on the flow field when the volume flow rate is small. The smoother suction surface and thinner trailing edge can reduce the end wall loss in the cascade passage, which leads to the reduction of the influence area of secondary flow, the effective restraint of the boundary layer thickness on the blade surface, the reduction of blade loss, and the improvement of flow efficiency of the governing stage after optimization. The optimized blade not only improves the performance in the peak regulating condition, but also has good performance in the design condition.