Nonlinear Macrohysteretic System identification based on neural network is one of the hotspots in current study. In this paper, the principle of neural network identification is introduced. By studying the design and algorithms of BP and RBF neural network, analysis and comparison is done with the simulation of MATLAB. The simulation results show that: RBF neural network, with faster convergence speed and higher accuracy, is better than BP on conformance, while BP neural network is better than RBF on generalization performance. According to the characteristics of the two networks, this conclusion can be used as a reference of the neural network model identification.