The ball type pulverizer system is widely used in power plant in China because of its better coal adaptability, easy maintenanceand high reliability. However, its primary disadvantageis high power consumption. Itis not only the big consumer of power plant ,but also thebig latenteconomizer. Some existentresearches mostly focus on pulverizer, not on the ball type pulverizer system. An operation optimization method is presented in the disquisitionbased on the system identification. By selecting data upon orthogonal norm , acquiring the system model through Least Squares method, and according to the theory of optimization, the authors obtain the optimal values of controlledparameters.This method can avoid disadvantageboth of searching for the input-output specialty through mechanismmodeling and big data sample space for Manual Neural Network. It is valuable in practice for the ball type pulverizer system operation optimization and economizing energy in power plant.