基于极限学习机的玉米干燥系统出机水分含量预测模型  

A Model for Predicting the Outgoing Moisture Content of Corn Drying System Based on Extreme Learning Machine

作　　者：邢思敏 高香兰 林子木 王德华 曹英丽[2] 曹毅 刘国辉 XING Si-min;GAO Xiang-lan;LIN Zi-mu;WANG De-hua;CAO Ying-li;CAO Yi;LIU Guo-hui(Liaoning Grain Science Research Institute,Shenyang 110032,China;College of Information and Electrical Engineering,Shenyang Agricultural University,Shenyang 110161,China)

机构地区：[1]辽宁省粮食科学研究所,沈阳110032 [2]沈阳农业大学信息与电气工程学院,沈阳110161

出　　处：《沈阳农业大学学报》2023年第5期619-626,共8页Journal of Shenyang Agricultural University

基　　金：辽宁省自然科学基金项目(2022-MS-068)。

摘　　要：精准预测玉米出机水分含量是实现玉米干燥过程自动控制的关键。为提升玉米干燥系统自动化程度,实现玉米出机水分的快速、无损、精准预测,研究依托粮食干燥模拟试验系统,通过调节干燥过程的热风温度、干燥段温度和排粮频率,设计8组多批次玉米干燥平行试验,分时段获得玉米干燥过程6个点位的温湿度特征数据,以及相对应的出机玉米水分含量数据;通过相邻平均法(adjacent average smoothing,AAS)对12组温湿度特征数据进行平滑处理,平滑后的数据点集合毛刺减少、波动明显缓解,12组特征数据与玉米出机水分含量的Pearson相关性大幅提升,提高幅度在0.061~0.105之间,为后续构建玉米水分预测模型提供了数据基础;进而,研究以相邻平均法平滑后的12组特征数据作为模型输入量,构建了基于极限学习机(extreme learning ma⁃chine,ELM)的玉米干燥过程出机水分含量预测模型并进行评价分析。结果表明:基于ELM的玉米出机水分含量预测模型以sig⁃moid为激活函数、神经元个数设定为22时,模型运行速率较快、预测效果较好,其中训练集和验证集拟合优度R2分别为0.9886和0.9812,验证集R^(2)与训练集接近,表明该模型拟合效果较好、未出现过拟合情况,水分含量预测均方根误差(root mean square er⁃ror,RMSE)分别为0.2811%和0.3821%,预测误差总体较低。综上表明,模型在训练与验证过程中均具备较好的拟合效果和较高的预测准确性。该模型为玉米干燥过程的自动化控制和工艺优化提供了数据参考和技术支持。Accurately predicting the moisture content of corn outlet is the key to realize the automatic control of corn drying process.In order to improve the degree of automation of corn drying system and realize the rapid,non-destructive and accurate prediction of corn moisture content,this study relies on the grain drying simulation test system to design 8 groups multiple batches of corn drying parallel experiments by adjusting the hot air temperature,drying section temperature and grain discharge frequency,and obtained temperature and humidity characteristic data of six points and the corresponding moisture content data of corn during drying process at different time intervals.The smoothing of 12 sets temperature and humidity feature data were performed using the adjacent average smoothing(AAS)method,and the smoothed data point sets had fewer burrs and less fluctuation,resulting in a Pearson correlation significant improvement of 0.061 to 0.105 between the smoothed feature data and the moisture content of corn outlet,which provides a data basis for the subsequent construction of corn moisture prediction model;further,this study uses 12 sets of feature data smoothed by AAS as model inputs,and then a prediction model for the moisture content of corn outlet during the drying process based on the extreme learning machine(ELM)was constructed and evaluated.The results showed that the prediction model for the outgoing moisture content of corn based on ELM works at a faster rate and has a better prediction effect when the number of neurons is 22 with sigmoid as the activation function.The R2 of the training set and validation set are 0.9886 and 0.9812,respectively,and the R^(2) of the validation set is close to that of the training set,indicating that the model fits well and does not overfit;the root mean square error(RMSE)of moisture content prediction is 0.2811%and 0.3821%,respectively,and the prediction error is generally low.In summary,it shows that the model has a very good fitting effect and high prediction accuracy in b

关 键 词：玉米干燥 预测模型 极限学习机 相邻平均法 

分 类 号：S513[农业科学—作物学] S127