基于深度集成学习的甘蔗压榨抽出率预测方法  

Prediction Method of Sugarcane Extraction Rate Based on Deep Integrated Learning

作　　者：蒙艳玫[1] 张月 段青山[2] MENG Yanmei;ZHANG Yue;DUAN Qingshan(College of Mechanical Engineering,Guangxi Univ.,Nanning,530004,China;College of Light Industry and Food Engineering,Guangxi Univ.,Nanning 530004,China)

机构地区：[1]广西大学机械工程学院,南宁530004 [2]广西大学轻工与食品工程学院,南宁530004

出　　处：《三峡大学学报（自然科学版）》2023年第4期101-107,共7页Journal of China Three Gorges University：Natural Sciences

基　　金：国家自然科学基金(61763001,51465003,12062001);广西自然科学基金(2021JJA110041)。

摘　　要：先进的甘蔗压榨建模方法能够给生产提供指导,有利于提高糖厂的经济效益并节约能源.本文选择深度极限学习机(DELM)和长短期记忆网络(LSTM)作为基学习器,极端梯度提升(XGBoost)作为元学习器,构建了Stacking深度集成学习模型,用于甘蔗压榨抽出率的在线预测;并通过计算和实验,验证该方法的可行性和有效性.与其他模型相比较,本文所提模型的预测精度高5%~12%,并且对数据的敏感性更低,泛化性更好,能够适应甘蔗压榨的不同工况.The advanced method of sugarcane press modeling can provide the guidance for the production,improve the economic benefit of sugar mills and save the energy.In this paper,deep extreme learning machine(DELM)and long and short term memory network(LSTM)are selected as the base learner and extreme Gradient Lift(XGBoost)is chosen as the meta-learner.The deep integrated learning model is constructed for online prediction of sugarcane extrusion rate.The feasibility and effectiveness of the proposed method are verified by the calculation and the experiments.Compared with other models,the prediction accuracy of the proposed model is 5%-12%higher,and the sensitivity to the data is lower,the generalization is better,and it can adapt to the different conditions of sugarcane pressing.

关 键 词：甘蔗压榨抽出率 集成学习 深度学习 数据驱动建模 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]