基于微粒群优化算法的间歇过程迭代学习  

Particle Swarm Optimization Algorithm Based Iterative Learning Algorithm for Batch Processes

作　　者：贾立[1] 程大帅[1] 施继平[1] 邱铭森[2] 

机构地区：[1]上海大学机电工程与自动化学院自动化系电站自动化技术重点实验室,上海200072 [2]新加坡国立大学工程学院,新加坡119260

出　　处：《控制工程》2011年第3期341-344,共4页Control Engineering of China

基　　金：上海市基础研究重点项目(09CJ1406300);上海市教育委员会科研创新项目(09YZ08);上海大学"十一五"211建设项目资助;上海大学研究生创新基金资助

摘　　要：在实际的间歇过程优化控制中面临着模型参数不确定性的问题,并且基于梯度的优化控制技术在处理具有多个局部极值的目标函数时,往往会陷入局部极值点。针对该问题,提出一种基于微粒群优化算法的间歇过程产品质量迭代学习策略。结合间歇过程产品质量控制问题的特点和各种约束条件,利用微粒群优化算法在非线性问题求解及大尺度空间搜索上的优势,来改变间歇过程鲁棒优化收缩控制变量的搜索域,并用迭代法以迭代方式消除误差,从而使优化性能指标渐次达到最优。在理论研究的基础上,提出的算法用于一类典型的间歇过程重点产品质量优化控制中,仿真结果表明产品终点质量能够充分逼近期望值,并且控制轨迹收敛,从而验证了算法的有效性和实用价值。In the batch processes. The problem of model parameters uncertainty is confronted, however the traditional gradient-based optimization control technology easily falls into local extreme points when solving the objective function with multiple local extreme points. To overcome this problem, a PSO-based iterative learning strategy is proposed. Taking with the characteristics of product quality control issues and a variety of constraints and advantage of particle swarm optimization algorithm in solution for nonlinear problem and large-scale searching problem, the strategy of eliminating error using iterative algorithm was used to drive the solution to the optimal. Lastly, to verify the efficiency of the proposed algorithm, the algorithm was applied to a classical batch processes, the simulation results showed that the quality at the end of each batch can approximate the desire value sufficiently, and the input trajectory convergent. Thus the efficiency and practicability of the algorithm was verified.

关 键 词：间歇过程 微粒群优化 迭代学习 产品质量控制 

分 类 号：TP27[自动化与计算机技术—检测技术与自动化装置]