Abstract: This article proposed a new technique using Genetic Algorithms to realize Real-time fuzzy control, discussed how to form the rules, to train the parameters, the relation between the training scale and the generalization of the model comparing results of different training scales. The model has less input parameters, less rules, and smaller errors than most of other models. It has advantages in both time and space complexity. It is easy to control, and has no restriction of experience knowledge. This method is applicable to Real-time system applications.