In order to enhance the ability of wind farm actively integrated into grid, the super short-term prediction model of wind speed and power is established. Then with the goal of the maximum wind power capture and the smooth output power and taking generator speed and pitch angle as control variables, the minimization control criterion is set out and the corresponding multi-objective optimization models are established to reduce mechanical and pitch angle stresses. The genetic algorithm is used to solve the optimal solution which is applied to wind power generation system to optimize performances. A 1.5 MW variable speed constant frequency wind power generation system is studied by the computer simulation. The results show that the output power is increased and the turbulence of power in low frequency is reduced compared with the traditional maximum power point tracking (MPPT) control strategy. Key words : wind power generation optimization control maximum wind power capture power prediction smooth output power