In this paper, a Prandtl-Ishlinskii hysteresis model (PI) is used to build a rate-independent hysteresis model for a class of conducting polymer actuators typified by tri-layer conjugated polymer actuators. Firstly, an off-line method is proposed to identify a discretization density function for the hysteresis model, and then a linear transfer function for the actuator is identified using the PI inverse model. Secondly, a neural network approach is proposed to realize an adaptive on-line identification method for the density function of the PI hysteresis model. In the back propagation (BP) algorithm for the neural network, the discretization PI operator is considered as an operational function of the neural network and the density function is considered as the power value. Finally, the simulation and experimental results are presented to demonstrate the validity of the model identification method and the actuator model.