基于工业分析的煤质发热量预测

Prediction of coal heating value based on proximate analysis

为提升煤质工业分析数据对煤质发热量的预测精度,以167种中国煤和4 540种美国煤的信息为基础建立了基于SVR的煤质发热量预测模型。重构Majumder等和Parikh等提出的线性模型,用于对比非线性SVR和已有的基于线性的煤质发热量预测模型的预测能力。对比结果为:SVR对中国煤和美国煤的预测相对误差为2.16%和2.42%;重构后的Majumder模型为3.04%和4.61%;重构后的Parikh模型为3.39%和12.99%。工业分析对发热量的散点图也表明各工业分析组分与发热量之间没有明显的线性关系。研究结果表明,基于非线性SVR的煤质发热量预测模型具有更高的预测精度。

To predict the heating value of coals based on proximate analysis, a nonlinear model termed as support vector regression (SVR) was introduced in this study. A total of 167 Chinese coal samples and 4 540 U. S. coal samples were selected to develop and verify the SVR-based correlations. Some published linear correlations were also employed and redeveloped with the Chinese and U. S. coals to obtain a comparison with the SVR-based correlations developed in the present study. The comparison results indicate that the SVR-based correlations can be more accurate than the published linear correlations. The scatter plot of proximate analysis versus heating value also indicates that there is no obvious linear relationship between the proximate analysis and the heating value of coals. © 2015, China Coal Society. All right reserved.