基于SVR和随机森林模型的动力煤高位发热量预测研究  

Prediction of higher heating value of fuel coal based on SVR and random forest models

作　　者：郭龙 郭文文 GUO Long;GUO Wenwen(Zhejiang Zheneng Fuxing Fuel Co.Ltd.,Hangzhou 310023,China;College of Mechanical Engineering,Department of Energy and Environment System Engineering,Zhejiang University of Science and Technology,Hangzhou 310023,China)

机构地区：[1]浙江浙能富兴燃料有限公司,浙江杭州310023 [2]浙江科技学院机械与能源工程学院,浙江杭州310023

出　　处：《能源工程》2024年第1期35-42,共8页Energy Engineering

摘　　要：采用两种统计学习算法(支持向量回归和随机森林)构建了5种模型(LSVR、PSVR、RSVR、SSVR和RF),评估了它们在预测典型动力煤高位发热量与工业分析数据关系方面的表现。结果表明,RSVR和RF模型能够准确预测高位发热量,特别是在HHV_(d)为26.00~28.00 MJ/kg和V_(d)为28%~34%内。RSVR和RF模型的平均百分比误差(MAPE)分别为0.97%和0.96%。选择了不同类型的动力煤验证模型的可用性和应用范围,通过与各类煤的绝对百分比误差比较,可以发现随机森林模型普遍表现出较好的适应性和稳定性。Five models(LSVR,PSVR,RSVR,SSVR,and RF)were constructed by using two statistical learning algorithms(Support Vector Regression and Random Forest),and their performance in predicting the relationship between the high level heat generation of typical power coals and the industrial analysis data was evaluated.The results showed that the RSVR and RF models were able to accurately predict the high-level heat generation,especially within HHV_(d) of 26.00~28.00 MJ/kg and V_(d) of 28~34%.The mean percentage errors(MAPE)of the RSVR and RF models were 0.97%and 0.96%,respectively.Different types of power coals were selected to verify the usability and application range of the models,and by comparing the absolute percentage errors with various types of coals,it can be found that the Random Forest model generally shows better adaptability and stability.

关 键 词：高位发热量预测 机器学习 动力煤 支持向量回归(SVR) 随机森林 

分 类 号：TM621[电气工程—电力系统及自动化]