The greenhouse is a complex system. There is strong coupling relationship among its environmental factors in side greenhouse. So it's difficult to get satisfying effect by using conventional control methods. Based on the characteristics of the greenhouse, a new fuzzy neural network controller (FNNC) was proposed to create a proper condition for crop growth. The improved genetic algorithm was used to train the architecture of the fuzzy neural network controller, which was adopted in the greenhouse. Numerical experiment results showed that the greenhouse equipped with this fuzzy neural network controller had such features as responding quickly, smooth transition and programming simply.