入炉煤质特性在线检测对电厂锅炉优化运行、推动智能电站发展具有重要意义。近年来激光诱导击穿光谱（Laser-Induced Breakdown Spectroscopy，LIBS）技术因其强大的在线分析能力，广泛应用于煤质特性检测领域。依据电厂输煤系统的运行情况提出了一种在输送带上LIBS直接测量原煤煤质的在线检测方案，以满足入炉煤质在线检测的基本需求。为验证检测方案的可行性，搭建了入炉煤输送带模拟平台和LIBS测量系统，34个实际原煤样品作为测量对象放置在输送带上开展LIBS连续测量并获取光谱数据。根据直接测量原煤获取的光谱特征，确定CN388.176 nm作为光谱筛选和内标方法所用的特征峰。采用有效光谱筛选、光谱平均和合理内标选取的光谱预处理方法，减小无效光谱和光谱信号波动的不利影响。结合偏最小二乘（Partial Least Square，PLS）方法建立了相应的煤质定量分析模型，实现了灰分、挥发分和发热量的良好测量。灰分、挥发分和发热量模型的回归系数R2分别为0.98、0.97和0.97，预测集均方根误差分别为1.33%、1.03%和1.11 MJ/kg，平均绝对误差分别为1.24%、0.97%和0.88 MJ/kg。研究结果验证了LIBS在输送带上直接测量原煤煤质的可行性，为电厂提供了可行的煤质特性在线分析方案。

The on-line detection of incoming-coal property is of great significance to optimize the boiler operation and promote the development of intelligent power plant. In recent years，laser induced breakdown spectroscopy has been widely applied in coal property detection，due to its powerful on-line detection. Based on the operation of coal transporting system in the power plant，an on-line detection scheme for LIBS direct measurement of raw coal property on conveyor belt is proposed to meet the basic needs of online detection of incoming-coal property. To validate the feasibility of the detection scheme，the simulation platform for incoming-coal conveyor belt and the LIBS system were set up. 34 actual raw coal samples are placed on the conveyor belt as measurement objects to carry out continuous LIBS measurements and obtain spectral data. According to the spectral characteristics obtained from direct measurement of raw coal，CN388.176 nm is determined as the characteristic peak used in spectra selection and internal standard. Effective spectra selection，spectra averaging and internal standard method are used to reduce the adverse effects of invalid spectra and spectral signal fluctuations. Combined with the partial least squares(PLS）modeling method，the quantitative analysis models of coal property were established，and the good measurements of coal ash，volatiles and calorific value were achieved. The regression coefficient R2 of the ash，volatile，and calorific value are 0.98，0.97，and 0.97，respectively. The root mean square error of the prediction set are 1.33%，1.03%，and 1.11 MJ/kg，the mean absolute errors are 1.24%，0.97% and 0.88 MJ/kg，respectively. The results validate the feasibility of LIBS directly measuring raw coal quality on the conveyor belt and provide power plants a feasible scheme for online analysis of coal property.

0 引言

1 试验系统

2 试验方法与结果

   2.1 光谱筛选

   2.2 有效光谱平均

   2.3 内标方法

   2.4 PLS模型建立

   2.5 PLS模型结果

3 结论