基于模型预测的松散回潮水分控制研究

松散回潮系统具有大时滞、干扰变量多等特性,出口水分控制难度大。为解决松散回潮出口水分控制不稳定及反馈不及时等问题,采用相关性分析筛选出与松散回潮出口水分的相关变量,使用多元线性回归和神经网络建立松散回潮出口水分的预测模型,对2种模型的效果进行验证,在此基础上建立加水量实时在线优化控制模型。选取某烟厂生产数据对预测模型进行效果验证。结果表明,多元线性回归模型和神经网络模型预测的平均相对误差分别为0.3%、0.5%,多元线性回归模型表现出更好的变化趋势描述能力。实施控制后松散回潮出口水分均值偏差由0.151减小为0.098,松散回潮控制的准确性和稳定性显著提高,有效克服调控滞后问题,控制过程的智能化和精细化水平提升。

The loosening and conditioning system has the characteristics of large time delay and many disturbance variables,which makes it difficult to control the outlet moisture.In order to solve the problems of unstable loose return outlet moisture control and untimely feedback,correlation analysis was used to screen out the variables related to outlet moisture,and the prediction models of outlet moisture were established using multiple linear regression and neural network,and the effects of the two models were verified,based on which a real-time online optimal control model of water addition was established.The production data of a tobacco plant was selected to verify the effect of the prediction model.The results showed that the average relative errors of the prediction of the multiple linear regression model and the neural network model were 0.3%and 0.5%,respectively,and the multiple linear regression model showed better ability to describe the change trend.After the implementation of the control,the deviation of the average outlet moisture was reduced from 0.151 to 0.098,which significantly improved the accuracy and stability of the control process,effectively overcame the time-delay control problem,and enhanced the intelligence and accuracy of the control process.