提高激光诱导击穿光谱的在线测定煤质方法研究  

Improvementof coal quality online detectionmethodof laser-induced breakdown spectroscopy

作　　者：张业才 郑见云 郝红亮 张永 赵上勇 侯宗余[3,4] 王哲 ZHANG Ye-cai;ZHENG Jian-yun;HAO Hong-liang;ZHANG Yong;ZHAO Shang-yong;HOU Zong-yu;WANG Zhe(Anhui Branch of National Energy Group,Hefei 230051,China;Guoneng Bengbu Power Generation Co.,Ltd.,Bengbu 233000,China;State Key Laboratory of Power System and Power Generation Equipment Control and Simulation,Tsinghua University,Tsinghua BP Clean Energy Research and Education Center,Beijing 100084,China;Shanxi Clean Energy Research Institute of Tsinghua University,Taiyuan 030032,China)

机构地区：[1]国家能源集团安徽公司,安徽合肥230051 [2]国能蚌埠发电有限公司,安徽蚌埠233000 [3]清华大学电力系统及发电设备控制和仿真国家重点实验室,清华—BP清洁能源研究与教育中心,北京100084 [4]清华大学山西清洁能源研究院,山西太原030032

出　　处：《煤炭科技》2022年第6期30-35,共6页Coal Science & Technology Magazine

基　　金：国家能源集团2030先导项目(GJNY2030XDXM-19-08.1);国家自然科学基金(61675110)。

摘　　要：灰分和热值是影响煤炭工业生产与利用的重要因素,如何快速、精确地进行在线煤质测定也是开采和使用中亟需解决的关键问题。利用激光诱导击穿光谱技术(LIBS)结合偏最小二乘回归(PLSR)和极限学习机模型(K-ELM)用于在线煤炭灰分和热值分析,通过对样品中Si、Al、Fe、Ca、Mg、Na、K、Ti、N、H等元素特征谱线进行主导因素选择(DF),进一步提高预测模型的分析精度。结果发现,利用DF方法,PLSR和K-ELM模型的预测精度均有所提高;相比于常用的PLSR预测模型,DF-K-ELM模型的煤炭灰分和热值的决定系数分别提高了3.4%和5.6%,均方根误差分别降低了0.214和0.297。所研究的煤炭灰分和热值预测模型对于提高LIBS煤质在线分析具有重要的参考价值。Ash content and calorific value are important factors affecting the production and utilization of coal industry.How to quickly and accurately determine the online coal quality is also a key problem to be solved in mining and use.Laser induced breakdown spectroscopy(LIBS) combined with Partial Least Squares Regression(PLSR) and Kernel Based Extreme Learning Machine(K ELM) was used for online coal ash and calorific value analysis.Characteristic spectral lines of H and other elements are selected as dominant factors(DF) to further improve the analysis accuracy of the prediction model.Results show that the prediction accuracy of PLSR and K ELM models is improved by DF method;Compared with the commonly used PLSR prediction model, the determination coefficients of ash content and calorific value of DF K ELM model are increased by 3.4% and 5.6% respectively, and the root mean square error is reduced by 0.214 and 0.297 respectively.Therefore, the coal ash and calorific value prediction model studied has important reference value for improving LIBS coal quality online analysis.

关 键 词：煤炭 灰分 热值 激光诱导击穿光谱 偏最小二乘回归 极限学习机 

分 类 号：TD353[矿业工程—矿井建设]