基于随机森林回归算法的水泥立式磨磨内压差预测  

Prediction of the Pressure Difference Inside the Cement Vertical Mill Based on the Random Forest Regression Algorithm

作　　者：孙胜难 袁铸钢[1] 刘钊 马洪浩 SUN Shengnan;YUAN Zhugang;LIU Zhao;MA Honghao(School of Electrical Engineering,University of Jinan,Jinan 250022,China)

机构地区：[1]济南大学自动化与电气工程学院,山东济南250022

出　　处：《洛阳理工学院学报（自然科学版）》2024年第2期44-50,共7页Journal of Luoyang Institute of Science and Technology：Natural Science Edition

基　　金：济南市“新高校20条”项目(2021GXRC096)。

摘　　要：水泥立式磨粉磨过程中,准确预测磨内压差对控制和稳定运行至关重要。为了使磨内压差预测值与实际值更接近,提出一种基于随机森林回归(Random Forest Regression, RFR)的预测模型,通过装袋算法(Bagging)和随机特征子空间(Random Subspace Method, RSM)强化原决策回归树(Classification And Regression Tree, CART)的泛化能力。利用灰色关联度分析法,验证了关键变量选取的合理性;利用RFR算法建立磨内压差预测模型,得到新的磨内压差值。对比反向传播神经网络(Back Propagation Neural Network, BPNN)预测模型、支持向量机(Support Vector Machines, SVM)预测模型、径向基函数(Radial Basis Function, RBF)神经网络预测模型、长短期记忆(Long Short Term Memory, LSTM)神经网络预测模型的预测结果,RFR模型的预测效果优于其他4种预测模型,在磨内压差预测方面表现良好。In the process of cement vertical milling,accurate prediction of the mill internal pressure difference is crucial to control and stable operation.To align the predicted values of the mill internal pressure difference more closely with the actual values,a prediction model based on Random Forest Regression(RFR)is proposed.This model enhances the generalization ability of the original Classification and Regression Tree(CART)through Bagging and Random Subspace Method(RSM).The rationality of key variable selection is verified using grey relational analysis.Compared to the prediction results of the Back Propagation Neural Network(BPNN),Support Vector Machines(SVM),Radial Basis Function(RBF)Neural Network,and Long Short Term Memory(LSTM)Neural Network prediction models,the RFR model demonstrates superior predictive performance and exhibits excellent performance in predicting mill internal pressure differences.

关 键 词：水泥立式磨 磨内压差 随机森林回归 灰色关联度 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]