摘要
针对时变的非线性系统,提出一种基于神经网络的迭代优化预测控制。它将传统的预测控制策略与神经网络逼近任意非线性函数的能力结合,预测系统未来输出,然后用迭代学习方法优化预测控制器,即通过一阶泰勒展开的方法,把非线性优化问题转化为线性优化问题。不仅简化计算,同时避免用神经网络优化控制器时,由于调节参数过多、调节速度慢而导致系统闭环稳定性和鲁棒性差的问题。仿真结果表明,该控制方案具有良好的控制品质,并适应对象参数的变化,具有较强的鲁棒性和自适应性。
Aiming at the time - varying nonlinear system, this paper proposes an iterative optimal predictive control based on Neural Network, which combines the traditional predictive control optimization strategy with neural network' s capability of approximating to nonlinear function to predict system output, then using iterative optimal methods to optimize the predictive controller, namely by using first - order Taylor expansion method to transform nonlinear optimal question into linear optimal question. Using neural network as controller can not only simplify computation, but also avoid the problems of unstability and non - robustness of closed loop system caused by adjusting too much parameters and low speed. Simulation results show that the proposed method has prefect control performance, strong robustness and self- adaptive ability.
出处
《计算机仿真》
CSCD
2006年第10期147-150,共4页
Computer Simulation
关键词
非线性系统
预测控制
神经网络
自适应控制
迭代控制
Nonlinear system
Predictive control
Neural network
Adaptive control
lterative control