To overcome the slower convergence speed of neural network and easily falling into local minimum,this paper proposes a hybrid learning algorithm to optimize neural network. The mixed learning algorithm of genetic (GA) algorithm and conjugate gradient algorithm ( CGA) is used for neural network weight changes.According to the strong coupling problem between the total pressure of headbox and the level, a decoupling method based on the mixed algorithm of improved GA - CCA, neural network PID is proposed. The decoupling control between the two variable is realized successfully. The results of simulation show that the system have better control effect and anti - interference than neural network PID depcoupling.%为了克服神经网络存在的收敛速度慢、易于陷入局部极值等不足,采用一种混合学习算法优化神经网络,即将改进遗传算法(GA)和共轭梯度算法(CGA)的混合学习算法用于对神经网络权值的修改,针对流浆箱的总压和浆位之间存在耦合问题,提出一种基于GA-CGA混合优化算法的神经网络PID解耦方法,成功地实现了总压、浆位之间的解耦.仿真结果表明该系统比神经网络PID解耦具有更好的控制效果和抗干扰能力.
展开▼