一种并行协同粒子群优化的支持向量机预测模型  

Prediction model of support vector machine based on parallel cooperative particle swarm optimization

作　　者：王华秋[1] 廖晓峰[1] 

机构地区：[1]重庆大学 计算机学院

出　　处：《控制理论与应用》2006年第6期934-940,共7页Control Theory & Applications

基　　金：国家自然科学基金(60271019);重庆市教委基础研究项目(KJ060614);重庆市科委攻关项目(20020828).

摘　　要：转炉提钒过程是一个非常复杂的多元非线性反应过程,从统计学和反应机理等角度出发,难以建立终点控制静态模型．针对这样的问题,提出了并行协同粒子群优化的支持向量机预测模型,不仅克服了支持向量机偏差ε和折中参数C选择的随机性,而且较好地解决了大数据集的快速并行计算,缩短了计算时间,从而有利于连续生产操作．试验表明,用该模型预测转炉提钒的冷却剂加入量和吹氧时间,结果的误差减小,满足了终点命中率在90%以上的指标,具有工程实用性．Converter re-vanadium is a very complexdiverse and nonlinear reaction. From the point of view of statistics and reaction mechanism, it is difficult to build an endpoint control static modei. Considering this problem, we presented a prediction model using support vector machine （SVM） based on parallel cooperative particle swarm. This model not only perfectly solves the problem of random selection of SVM regression parameter such as e and C, but also provides rapid calculation for the problem with large data sets and reduces the computing time. Accordingly the model is beneficial to continuous production. The model was used to predict the quantity of refrigerant and time consumption of oxygen in converter re-vanadium, the results of experiments showed that the errors were reduced and the endpoint hitting ratio reached the target for over ninety percent.

关 键 词：并行协同粒子群 支持向量机 参数优化 转炉提钒 预测模型 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]