Improving the laser-induced breakdown spectroscopy (LIBS) data for analyzing sulfur content and calorific value of coal

Abstract

If the sulfur concentration of coal is increased, there is a possibility that the amount of waste water discharged during desulfurization process is increased. In a coal-fired power plant, high sulfur concentrations accelerate equipment corrosion and can be expensive to replace it. In order to prevent these problems, a new powerful technique for measuring sulfur concentration of mixed coals in real time was studied. Laser-induced breakdown spectroscopy (LIBS) has many advantages compared to the conventional methods, no need for sample preparation and high sensitivity for light elements. Therefore, many researches have been done on properties such as calorific value and ash. Monitoring the calorific value of coal in-situ has the advantage of reducing the amount of unburned carbon by inducing complete combustion by injecting an accurate amount of combustion air immediately. In addition, since the emission of sulfur from coal-fired power plants depends entirely on the amount contained in the coal, accurate prediction is effective in preventing air pollution. However, there is a limit to analyze sulfur with LIBS since only a weak sulfur emission lines are detected at the main measurement wavelength range of LIBS spectrometer. Although in-situ measurements were not directly tested in this study, experiments were conducted in the laboratory condition to improve the data processing ability. This ability is the most important in field measurements to produce results quickly and accurately. Several blended coals were specially used for the experiment to replicate the actual coal-fired power plant condition. Although not well used in LIBS study, two well-known data processing methods have been used in near-infrared spectroscopy were employed to enhance the weak sulfur emission lines. The performances of partial least square regression (PLSR) model were compared by the parameters such as determination coefficient ( ), relative error, and root mean square error (RMSE). Moreover, the prediction ability for the unknown sample was also evaluated to verify whether this can be applied to real-time analysis system or not. The RMSE average were calculated and compared with the results of previous studies. As a result, the values from this study were 6.02%p smaller for calibration line and 4.5%p smaller for validation line than the previous study using near-infrared spectroscopy. The RMSE average values of calorific value were calculated to be less than 1%.