基于ARIMA-LSTM模型的卷烟制丝质量预测研究
Research on Quality Prediction of Tobacco Primary Processing Based on ARIMA-LSTM Model
李志敏 1叶春明 1闫文凯1
作者信息
摘要
为了应对卷烟制丝工艺中的不确定性所带来的质量风险,提出了一种将差分移动平均自回归(ARIMA)模型与深度学习中的长短期记忆单元(LSTM)模型相结合的制丝质量数据预测方法,并用该方法对某卷烟企业制丝工艺中的加热处理(HT)出口温度和叶丝增温增湿的入口水分进行预测,以均方根误差(RMSE)、平均绝对误差(MAE)和平均绝对百分比误差(MAPE)为指标来评估各模型的预测精度.结果表明,相较于单一 ARIMA和LSTM模型,所提出的ARIMA-LSTM组合模型的预测值更精准,在对HT出口温度进行预测时,组合模型的RMSE、MAE、MAPE值分别至少降低了 60.1％、63.1％和63％;在对叶丝增温增湿的入口水分进行预测时,组合模型的RMSE、MAE、MAPE值分别至少降低了 49.5％、49.4％和49.3％,有望为卷烟企业及时制定或调整生产方案提供合理的参考依据.
Abstract
In order to address quality risks associated with the uncertainty in the tobacco primary process,a tobacco forming quality data prediction method that combines the autoregressive integrated moving average(ARIMA)model with the long short-term memory(LSTM)model in deep learning is proposed,and the method is used to predict the outlet tempera-ture of heat treatment(HT)and the inlet moisture of leaf filament warming and humidification in the tobacco primary process of a tobacco enterprise.It evaluates the prediction accuracy of each model using metrics such as root mean square er-ror(RMSE),mean absolute error(MAE),and mean absolute percentage error(MAPE).The results show that the com-bined ARIMA-LSTM model provides more accurate predictions compared to the individual ARIMA and LSTM models.When predicting the HT export temperature,the combined model reduces the RMSE,MAE,and MAPE values by at least 60.1％,63.1％,and 63％respectively.When predicting the inlet moisture content of the leaf conditioning process,the combined model reduces the RMSE,MAE,and MAPE values by at least 49.5％,49.4％,and 49.3％respectively.This method has the potential to provide a reasonable reference for tobacco companies to timely formulate or adjust their produc-tion plans.
关键词
ARIMA模型/长短期记忆网络/残差/时序预测/卷烟制丝质量Key words
ARIMA model/long short-term memory networks/residuals/time series forecasting/tobacco primary quality引用本文复制引用
出版年
2024
国家科技期刊平台
NSTL
万方数据