多模型预测控制在磨矿分级过程中的应用.pdfVIP

第27 卷 第11 期 控 制 与 决 策 2012 年 11 月 Vol. 27 No. 11 Control and Decision Nov. 2012 文章编号: 1001-0920 (2012) 11-1715-05 多模型预测控制在磨矿分级过程中的应用 马天雨, 桂卫华, 阳春华, 王雅琳 (中南大学信息科学与工程学院，长沙410083) 摘 要: 磨矿分级过程的控制目标是将一、二级溢流浓度和细度稳定控制在质量指标区间内. 磨矿分级过程是胖系 统, 完成动态优化目标后控制器仍有剩余自由度, 因此需考虑局部稳态经济优化. 针对这一目标, 提出一种考虑局部 稳态经济目标的多模型预测控制方案. 首先, 建立了基于现场数据库的球磨机和分级机传函矩阵模型; 然后考虑局部 经济性能, 将稳态经济目标以罚函数形式嵌入动态优化目标函数; 为消除球磨机换球引起的模型失配的影响, 建立了 一种基于换球规律的多模型切换策略. 仿真结果表明了所提出控制方案的有效性. 关键词: 磨矿分级过程；罚函数；多模型预测控制 中图分类号: TP393 文献标志码: A Multiple model predictive control applied in grinding and classification process MA Tian-yu, GUI Wei-hua, YANG Chun-hua, WANG Ya-lin (School of Information Science and Engineering，Central South University，Changsha 410083，China. Correspondent: MA Tian-yu ，E-mail ：pymty@) Abstract: Stably controlling concentration and fineness of first and second overflow in their quality index range are the control objectives of grinding and classification process. Grinding and classification process is a fat system, and the controller still has free degree after its dynamic optimization objectives realized, so local steady-state economic optimization is considered. For this objective, a multiple model predictive control considering local steady-state economic objectives is proposed. Firstly, based on the field database, transfer function matrix models of ball-mill and classifications are built up. By considering local economic performance, steady-state economic objectives are embedded into dynamic objectives as a penalty function. To eliminate the effect of model mismatch, based on the law of ball changing, a multiple model switching strategy is built. The simula