The work aims to improve the quality and efficiency of beer filling, so as to accurately control the liquid level of liquid storage tank. On the basis of traditional PID control, the method to adaptively control the liquid level in the liquid storage tank of beer filling machine (its RBF neural network was combined with PID) was proposed. The fast self-learning ability of RBF neural network was used to adjust the proportion, integral and differential coefficient of PID control. Compared with the traditional PID control, the response speed of RBF-PID control was fast, the overshoot was small and the time for the system to be stabilized had been greatly shortened. The proposed method has improved the control accuracy of the liquid level in the liquid storage tank of beer filling machine, which effectively enhances the quality and yield of beer.