With the large-scale grid interconnection of the new energy power generation, stochastic load disturbance gives the new challenges about how to maintain the stability and optimization operation of power systems. For the problems of stochastic load disturbance and parameter uncertainties in automatic generation control process, a parameter optimization algorithm of the state feedback controller based on robust variance constraint is proposed. Based on the inequality constraints of robust variance control principles, the game relationship between stochastic disturbance rejection and dynamic performance improvement of step disturbance response of closed loop system is analyzed. The constrained optimization problems considering steady state variance and control energy output constraint is constructed and LMI is used to design the controller parameters. Using genetic algorithm, closed loop poles circle center and positive scalar are optimized to design the control strategy with the optimal performance index. Two-area power system model is provided to demonstrate that the presented method can effectively restrain stochastic load disturbance and have good control performance and robustness.