【系统仿真学报】溷凝剂投量的RBF网络预测控制系统.pdfVIP

第21 卷第17 期 系 统 仿 真 学 报© Vol. 21 No. 17 2009 年9 月 Journal of System Simulation Sep., 2009 混凝剂投量的RBF 网络预测控制系统 王军栋，刘思远,齐维贵 （哈尔滨工业大学电气工程及其自动化学院，哈尔滨 150001 ） 摘 要：净水厂水处理工艺中的混凝投药过程是一个大滞后、非线性的复杂系统。针对其特点并结 合国内目前投药控制的现状，提出了一种带有前馈补偿的 RBF 神经网络预测控制新方法。该方法 利用人工神经网络建立投药量的预测模型，然后用出水浊度与设定值之间的预测偏差构成闭环控制 系统。通过实时的在线滚动优化，实现了投药量的最优投加。经过仿真试验表明，出水浊度能够比 较准确地跟踪设定值的变化，同时该方法所需混凝剂量减少，控制效果明显。 关键词：混凝投药；浊度；RBF 网络；预测控制 中图分类号：TP391.9 文献标识码：A 文章编号：1004-731X (2009) 17-5540-04 RBF Network Predictive Control of Coagulant Dosage WANG Jun-dong, LIU Si-yuan, QI Wei-gui (Department of Electrical Engineering and Automation, Harbin Institute of Technology, Harbin 150001, China) Abstract: Coagulation in water purification plants is a complex system with nonlinear and large time-delay. By analyzing its characters and the status of domestic dosage control, a new method of predictive control based on RBF network with forward compensation was proposed. A predicting model was established by artificial neural network, and then a closed-loop control system was constructed by the predictive deviation between turbidity of outlet water and the set-value. Through real-time on-line rolling optimization, optimal chemical dose rate was realized. The simulation of coagulant dose process shows that the turbidity of outlet water can track the set-value precisely, and coagulant consumption is reduced, and the control effect is obvious. Key words: coagulant dosage; turbidity; RBF network; predictive control 引 言1 1 具有前馈补偿的神经网络预测控制方案 在地表水处理工艺中，混凝