对称NARMA-U模型及其神经网络自校正控制器

Symmetric NARMA-U Model and Its Neural Network Self-Tuning Controller

带预测误差补偿的改进NARMA-L2模型是由NARMA模型在自适应滤波动态工作点处一阶泰勒展开逼近得出的,在自适应滤波动态工作点处二阶泰勒展开逼近可得到对称NARMA-U模型,采用BP神经网络辨识对称NARMA-U模型参数,提出一广义目标函数,基于对称NARMA-U模型的非线性系统的神经网络自校正控制器,应用直接极小化指标函数自适应优化算法对BP神经网络连接权重值进行在线学习。仿真研究表明算法的响应优良。

NARMA-L2 model with prediction error compensation was approached by NARMA model's first-order Taylor expansion at adaptive filtering dynamic working point.By second order Taylor expansion approaching at the adaptive filtering dynamic working point,a symmetric NARMA-U model was obtained.By using BP neural net work identifying symmetrical NARMA-U model parameters,a generalized object function was developed.Based on the nonlinear system neural net work self-tuning controller of symmetrical NARMA-U model developed,on-line learning of BP neural net connection weight value was conducted by using adaptive optimization algorithm for direct minimization of index function.Simulation results indicate that the model shows excellent response。