摘要
提出了一种基于小波神经网络辨识的PID神经网络模型参考自适应控制方法。该方法采用小波神经网络作为辨识器,PID神经网络作为控制器在线调节。由于小波变换具有良好的时频局部特性,神经网络具有强大的非线性映射能力,自学习、自适应等优势,采用规范正交的小波函数作为神经网络的基函数构成小波神经网络,该网络兼有小波函数的紧支性、波动性以及神经网络的非线性映射能力,自学习、自适应能力等优点,仿真结果表明用该方法构成的控制系统收敛速度快,逼近精度高,鲁棒性好,优于一般的BP网络控制。
A new PID neural model reference adaptive control method based on wavelet neural network identification is presented in this paper. It adopts wavelet neural network as identifier, and PID neural network as controller. It introduces normalized wavelet basis functions as the basis of neural network. Benefited from wavelet transform' being constrictive and fluctuant, it shows excellent temporal-frequency localization property, whilst it possesses such merits as powerful ability of mapping nonlinear systems, self-learning, self-adaptation and so on, thus, this network converges quickly with high precision and good robustness. Results of numerical simulation test to be satisfactory and this proves the method to be feasible, effective and superior to general BP neural network control.
出处
《航空计算技术》
2004年第1期17-20,共4页
Aeronautical Computing Technique
