作　　者： (李荣雨）; (戚桂洪）;

机构地区： 南京工业大学计算机科学与技术学院,江苏南京211816

出　　处： 《计量学报》 2017年第5期650-655,共6页

摘　　要： 针对软测量建模样本的特性,提出一种基于布谷鸟选择性集成学习的在线贯序极限学习机(CSSEOSELML)软测量建模方法。首先,以多个OSELM组合成集成学习的框架,并给每个OSELM赋予权重并设定阈值,借助于布谷鸟算法(CS)从中选择出满足阈值条件的OSELM个体,重新组合成集成学习的子集。最终以该子集建立软测量模型,进行集成学习并做加权处理。以UCI标准数据集进行测试,同时对加氢裂化反应分馏塔航煤干点进行验证,仿真结果表明,该算法优于传统的方法,具有更高的预测精度和稳定性能。 According to the characteristics of soft sensor modeling samples, a soft sensor modeling method called cuckoo search selective ensemble of online sequential extreme learning machine （ CSSE-OSELM ） is proposed. First of all, OSELM individuals were assembled into the framework of integrated learning,and each OSELM individual is endowed with weight and threshold. Then,with the aid of the cuckoo search algorithm, choose out OSELM individuals,which satisfy the threshold condition ,to a subset of the combined into integrated learning. Finally soft sensor model is established with the subset, with integrating learning and weighted processing. Using UCI standard data sets to test, at the same time, the jet fuel for hydroeracking reaction distillation tower is verified, the simulation resuhs show that the algorithm is superior to the traditional method, with higher prediction precision and stable performance.

关 键 词： 计量学 软测量 选择性集成学习 布谷鸟算法