时延Elman递归神经网络及其在PMSM的混沌控制中的应用

针对Elman递归神经网络存在的高深度、低分辨率问题,提出了一个结构简单的时延Elman递归神经网络模型。通过在Elman递归神经网络中引入多步的时延结构和反馈结构增强网络的记忆深度和分辨率。针对永磁同步电动机(PMSM)中存在的混沌运动,设计了时延Elman递归神经网络控制器和辨识器,推导出时延Elman递归神经网络的动态反传算法。运用离散型Lyapunov稳定判据,推导出此神经网络控制器和辨识器的权值自适应学习速率的取值范围,确保了控制系统的稳定性和快速收敛性。仿真结果表明,作者提出的时延Elman递归神经网络在动态系统的辨识和控制等方面具有良好的性能。

Time delay Elman recurrent neural network and its application in PMSM chaos control

For the problem of high depth and low resolution ratio in the memory of Elman recurrent neural network,a time delay Elman recurrent neural network(TERNN) model with a simple configuration was proposed.Both the depth and the resolution ratio in the memory of this network were improved by introducing a multi-step time delay and recurrent mechanism.In consideration of chaos in the movement of permanent magnet synchronous motor(PMSM),a controller and an identifier based on TERNN were designed.Moreover,a dynamic recurrent back propagation algorithm for adjusting TERNN was derived.By discrete Lyapunov stability creterion,the regions of weight adaptive learning rates in the controller and the identifier were obtained so as to guarantee stability and fast convergence of the controlled system.Simulation results illustrate that the proposed TERNN has good performance in both identification and control for dynamic systems.