Coal property analysis using laser-induced breakdown spectroscopy

Abstract

Fast or online coal property analysis would greatly improve coal pricing reliability and combustion optimization for power generation, especially in China. A non-linearized multivariate dominant factor based partial least square (PLS) model was applied to analyze coal properties through laser-induced breakdown spectroscopy (LIBS). The dominant factor explicitly modeled the direct correlation between coal property and spectral line intensities, whereas residual errors were corrected by PLS method with full spectral information. The results demonstrated an overall improvement over conventional PLS method for ash content, volatile content, and calorific value measurement. For example, the root mean square error of prediction of calorific value was decreased from 1.63 MJ kg−1 to 1.33 MJ kg−1; and the average relative error of all samples was reduced from 3.55% to 2.71%. Although current LIBS application results for coal proximate analysis have not met the national standard, LIBS is shown to be fully capable of providing useful information for coal combustion optimization and reliable references for coal pricing, showing LIBS has great potential for coal property analysis for the power generation industry.