This article discusses the control strategy of the steel casting system which possesses the characteristics of large inertia, time-varying and nonlinear. Aiming at getting the minimum deviation of liquid level, the control strategy uses the genetic algorithm to off-line optimize the parameters (cij,bj) of the Gaussian membership function and the network structure of fuzzy controller which affect the overall system firstly. Then, BP algorithm is used to online regulate and optimize the weight parameters of the control output which affect the system partly. Finally, the intelligent control system of the liquid level which is based on GA-FNC is simulated. The results show that the method can enhance the ability of self-learning and robustness of the system greatly and improve the stability of steel casting system significantly.