改进的Elman网络在非线性系统辨识中的应用

通过对Elman网络的研究,提出一种新型的基于输入层、隐层、输出层神经元递归的动态递归神经网络,给出Elman网络的标准BP学习算法,针对标准BP算法的收敛速度慢和容易收敛于局部极小点的缺点,利用非线性动量项自适应变步长的BP算法进行改进,从而提高算法的收敛速度,避免陷入局部极小点的问题.通过在系统辨识中的应用,表明该网络收敛速度快,模型精度高,并具有较强的自适应性和鲁棒性,适合于动态系统的实时辨识.

Application of Improved Elman Network in Nonlinear System Identification

Through studying on Elman network,this paper provides a new neural dynamic recurrent neural network based on input,output,hide layer and gives its algorithms.Normal BP algorithm is improved by non-linear adaptive step-varied back propagation,which increases the studying speed of algorithm and avoids trapping in the problem of local minimum.By using the algorithm in system identification, this network could be convergent quickly.High precision of models,stronger adaptability and robustness are provided.It is suitable for real-time identification of dynamic systems.