基于高斯微粒群优化的动态神经网络延迟系统辨识

为提高神经网络对未知非线性大滞后动态系统的泛化能力,提出一种基于高斯微粒群优化的自适应动态前馈神经网络.在输入层与隐含层之间、隐含层与输出层之间分别加入动态延迟算子,可以高效地辨识出系统纯滞后时间,建立精确系统模型.此外,采用高斯函数和混沌映射方法平衡微粒群算法全局寻优能力,以克服提前收敛的缺陷,从而快速有效地自适应优化网络中的参数.仿真实验表明了该方法在非线性人滞后系统辨识中的有效性.

Dynamic neural network on Gaussian particle swarm optimization for delay system identification

In order to improve the generalization capacity of neural network for poorly known nonlinear dynamic system
with long time-delay, a dynamic feedforward neural network on Gaussian particle swarm optimization algorithm is proposed.
The dynamic delay operators are added between the input and hidden layer, output and last hidden layer, which can effectively
identify the precise pure delay time. The nonlinear dynamic system model is built exactly to predict the system change. On
the other hand, Gaussian function and chaos mapping are adopted to balance the global optimization ability to overcome
premature convergence, which can optimize the parameters in the neural network structure adaptively. Compared with other
methods in the experiments, the effectiveness of the proposed method on nonlinear long-delay system identification is shown.