摘要
提出在利用前馈神经网络对非线性系统建模的基础上,对系统输出实现递推多步预测,并且结合自整定PID方法,实现非线性系统控制,神经网络在线辨识时采用学习速度较快的扩展Kalman滤波方法,仿真实验表明了该方法的有效性.
In order to control a nonlinear system, a model of recurrent multi-step prediction of the future process outputs which is constructed based on multiplayer neural network is developed and combined with a self-tuning PID controller. In the control process, an extended Kalman filter algorithm is introduced to quicken neural training. Simulation studies show the effectiveness and good performance.
出处
《河北工业大学学报》
CAS
2003年第2期21-24,共4页
Journal of Hebei University of Technology
基金
国家自然科学基金资助项目(60174021)
国家科技攻关计划资助项目(2001BA204801-02)