摘要
利用人工神经网络并行处理、容错以及任意逼近非线性函数的能力 ,针对感应电机这样一个各项参数随时间和运行状况的不同而变化的非线性系统 ,结合感应电机的外输入非线性自回归滑动平均模型(NARMAX) ,分析了不同人工神经网络 (ANN)结构和算法对系统辨识的影响。仿真实验结果中还看到前向BP网络存在着逼近“饱和”现象 ,即网络只能在一定程度上逼近辨识对象 ,性能指标会趋近于一极限 。
Utilizing the abilities of artifical neural networks in parallel disposal,fault-tolerart and discretional nonlinear function,according to such an inductive motor that is a nonlinear system with its changes by different time and operation situation,it discusses the influences by different ANN structures and algorithms combining with exagenous inputs NARMAX average model of inducation motor. The simulation results show that there exists 'saturation' in forward BP networks,that means networks can approach identification objects in a certain extent and capability target will tend to a high-point,which needs to be solved.
出处
《控制工程》
CSCD
2002年第4期71-72,90,共3页
Control Engineering of China
