采用遗传算法训练对角递归神经网络预测控制器
TRAINING OF DIAGONAL RECURRENT NEURAL NETWORK PREDICTIVE CONTROLLER USING GENETIC ALGORITHMS
-
摘要: 本文提出了一种基于广义预测控制的神经网络预测控制方案．预测控制器由对角递归神经网络预测控制器和前向神经网络静态补偿器组成．两种神经网络均采用遗传算法进行训练．仿真实验表明，对于带纯时延的非线性被控对象，采用遗传算法设计的对角递归神经网络预测控制器具有令人满意的控制性能．Abstract: This paper proposes a neural network predictive control scheme based on generalized predictive control(GPC).The predictive controller is made of diagonal recurrent neural network predictive controller(DRNPC) and feed forward neural network steady state compensator (FNC).Two kinds of neural networks are trained using genetic algorithms.The simulation results show satosfactory performance of the neural network predictive controller for nonlinear plants with dead time.
下载:
