基于改进GRU的白酒蒸汽运行预测模型研究  

Research on Fermented Grains Steam Running Condition Prediction Model of Make Wine Robot Based on Modified GRU

作　　者：张杰[1] 张子蓬[1] ZHANG Jie;ZHANG Zipeng(School of Computer Science,Hubei Univ.of Tech.,Wuhan 430068,China)

机构地区：[1]湖北工业大学计算机学院,湖北武汉430068

出　　处：《湖北工业大学学报》2022年第4期43-48,61,共7页Journal of Hubei University of Technology

基　　金：湖北省科技厅重大专项基金(2018ABA084)。

摘　　要：中国白酒酿造的关键步骤是上甑蒸馏,在探气上甑蒸馏过程中,白酒蒸汽的运行状态的准确预测起到重要作用,其变化受多种因素影响,为预测带来了很大的难度。针对该问题,提出基于改进的门控循环循环单元的白酒蒸汽运行状态预测模型。基于梯度提升树分析白酒蒸汽影响因素的重要性,根据重要性进行有序组合,然后输入门控循环单元,得到最优预测模型并实现白酒蒸汽状态预测。利用酒醅和蒸汽的历史时间序列信息搭建蒸汽状态预测模型,实现蒸汽状态的准确预测。本文使用劲牌酒厂上甑过程数据,最优均方根误差为2.276,平均绝对误差为1.623。实验结果表明本文提出的改进的门控循环单元模型可以较精确地预测白酒蒸汽运行状态的变化趋势。The key step of Chinese liquor brewing is steamer distillation.In the process of exploring steam,the accurate prediction of liquor steam plays an important role.Its changes are affected by various factors,which brings about great difficulty for prediction.Less research on this problem proposes liquor steam operation state prediction model based on improved gating cycle unit.The importance of liquor steam based on the gradient lifting tree is analyzed,orderly combined according to the importance,and then the gating cycle unit is input to obtain the optimal prediction model and realize liquor steam state prediction.The historical time series information of fermented grains and steam is used to build a steam state prediction model and realize the accurate prediction of steam state.Using the steamer process data,the optimal root mean square error is 2.276,and the average absolute error is 1.623.The results show that the improved gated cycle unit model can predict the trend of liquor steam.

关 键 词：白酒蒸汽运行状态 门控循环单元 时间序列回归 上甑机器人 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]