The generalized cyber net system is a kind of high-level Petri net, in which the weights are controlled by the places, and it can model the PID control law simply and effectively. Based on the single neuron PID control principle, the generalized cyber net system incorporates the learning rules of the neural network and is used to design the PID controller, which can decouple the nonlinear multivariable systems. The proposed method makes full use of the characteristics of the cyber net system, and the designed controller model realizes the unity of system control and parameter learning. The simulation analysis for the case of the double water tanks control system is carried out, and the results show the effectiveness of the proposed method.