文献类型：期刊文献

中文题名：基于极限学习机的在线参数更新方法及工业应用

英文题名：Online parametric update method based on extreme learning machine and its industrial application

作者：王再辰[1];程辉[1];赵亮[1]

机构：[1]华东理工大学能源化工过程智能制造教育部重点实验室,上海200237

年份：2023

卷号：46

期号：22

起止页码：126

中文期刊名：现代电子技术

外文期刊名：Modern Electronics Technique

收录：CSTPCD

基金：国家自然科学基金面上项目:化学过程生命周期评价与多目标鲁棒优化方法及其应用(22178103)。

语种：中文

中文关键词：在线序列简化核极限学习机(OS-RKELM);简化核极限学习机(RKELM);遗忘因子;在线序列;参数更新;乙烯裂解炉

外文关键词：online sequence simplified kernel extreme learning machine;reduced kernel extreme learning machine;forgetting factor;online sequence;parameter updates;ethylene cracking furnace

摘要：针对乙烯裂解炉结焦导致裂解炉机理改变,从而引起的模型预测不准确问题,提出一种带有遗忘因子的在线序列简化核极限学习机算法(FOS-RKELM)。该算法基于在线序列的简化核极限学习机,数据可以在线实时添加到网络中,从而提高模型的适应度;通过引入遗忘因子提高最近学习数据对模型的贡献,增强模型在线学习的能力;引入聚类算法优化、简化核极限学习机(RKELM),提高算法的稳定性。结果表明:所提算法在Mackey-Glass时滞混沌序列上取得了较好的预测效果;在乙烯产物收率预测问题上,与在线序列简化核极限学习机(OS-RKELM)、简化核极限学习机(RKELM)、BP神经网络和径向基学习机(RBF)算法相比,该算法平均绝对误差显著减小,证明了该算法的有效性。
In response to the problem of inaccurate model prediction caused by coking of ethylene cracking furnace and changes in cracking furnace mechanism,a forget operator online sequential reduced kernel extreme learning machine(FOS-RKELM)is proposed.In this method,based on the online sequential reduced kernel extreme learning machine,data can be added to the network in real time,so as to improve the adaptability of the model.By introducing a forget factor,the contribution of the most recent learning data to the model is increased to enhance the model's ability to learn online.The clustering algorithm optimization and the reduced kernel extreme learning machine(RKELM)are introduced to improve the stability of the algorithm.The results show that the proposed algorithm has achieved good prediction performance on Mackey Glass time-delay chaotic sequences.In the prediction of ethylene product yield,compared with online sequence simplified kernel limit learning machine(OS-RKELM),RKELM,BP neural network,and radial learning machine(RBF)algorithm,the average absolute error of this algorithm is reduced significantly,proving its effectiveness.