This paper proposed a auto-tuning parameter of fuzzy PID control system based on the structure of multi -layer neutral net-work. In this system, the determination of membership functions and rules of fuzzy inference translated into the optimization problem of con-nection weight coefficient and network structure in neutral network. Taking the minimum deviation of DO as the objective function, this pa-per used the improved Genetic Algorithm as the learning algorithm of fuzzy neutral network to optimize the parameter and structure of net-work. This method realized the online auto-tuning of PID parameter. Finally, by simulations, the results showed that this method greatly enhanced the self - learning capability and robustness of the system, and improved the dynamic and static performance of the control system.%构建一种基于多层神经网络结构的模糊PID参数自整定系统,将模糊规则和隶属甬数的选取转化为神经网络中连接权系数和网络结构的优化问题；以氧化沟内溶解氧偏差最小为目标函数,采用改进的遗传算法作为模糊神经网络的学习算法对网络的参数和结构进行优化,实现PID参数的在线自整定；仿真实验表明此方法较好地提高了氧化沟溶解氧系统的自学习能力和鲁棒性,使控制系统的动、静态性能都有较大的改善.
展开▼
