神经网络应用于非线性建模具有很多优点,但对迟滞这类多值映射非线性无能为力。一个新的基于神经网络的迟滞建模方法——拓展空间法被提出。通过坐标变换建立基本迟滞算子,将基本迟滞算子的输出与迟滞输入同时作为神经网络的输入,使神经网络的输入空间由一维上升为二维,从而使输入与输出之间形成一对一映射关系。最后的实验结果表明,通过拓展空间法神经网络能够对多值映射非线性进行建模。 Neural networks have many advantages for nonlinear modeling, but they can not do non-linearity for hysteresis. A new hysteresis modeling method based on neural network - extended space method was proposed. The basic hysteresis operator is established through the coordinate transformation, and the output of the basic hysteresis operator and the hysteretic input are taken as the inputs of the neural network at the same time. The input space of the neural network is increased from one dimension to two dimensions so that a pair of input and output is formed A mapping relationship. The final experimental results show that the multi-valued mapping nonlinearity can be modeled by expanding the space-law neural network.