基于动态递归神经网络的玻璃窑炉温度双重控制  

Dual-Control of Glass Furnace Temperature Based on Dynamic Recurrent Neural Networks

作　　者：张岚清[1] 党选举[1] 成娟娟[1] 

机构地区：[1]桂林电子科技大学计算机与控制学院,广西桂林541004

出　　处：《兵工自动化》2007年第5期67-69,共3页Ordnance Industry Automation

摘　　要：将双重控制策略应用到玻璃窑炉温度控制系统,实现油流量和油压的解耦控制。其主控制器采用动态递归神经网络逆控制器,应用反馈误差学习方法在线学习,并通过改进的L-M算法训练神经网络。副控制器采用常规PI控制器。主、副控制器共同作用于被控对象,保证系统具有良好的控制品质。仿真结果表明该控制策略有效。Dual-control method is applied to the controlling of the glass furnace temperature system, and decoupling control between oil pressure and flux is implemented. The inverse controller, which based on dynamic recursion neural networks, is used as the main controller, and is trained online by feedback error learning strategy. An improved Levenberg-Marquardt algorithm is proposed and applied to adjust the weights of the neural networks. A normal PI controller is used as the ancillary controller. Both the main and ancillary controller act on the testing object, they work jointly to obtain high control quality. Simulation results show the validity of this control strategy.

关 键 词：窑炉温度控制系统 双重控制 动态递归神经网络 解耦控制 L-M(Levenberg-Marquardt)算法 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]