This paper presents the application of a self-developed laser-induced breakdown spectroscopy (LIBS) online detection system for the rapid and quantitative analysis of coal combustion parameters, specifically sulfur content and calorific value, which are critical indicators affecting the efficiency of power plant boilers. A total of 57 coal samples from actual production were analyzed, with 43 dedicated to model development and 14 for validation. The system captured characteristic spectra, from which multiple spectral lines corresponding to sulfur, carbon, magnesium, silicon, iron, calcium, hydrogen, oxygen, and nitrogen were selected. Analytical models were initially developed using multiple linear regression (MLR) and subsequently refined with partial least squares regression (PLSR). The combined MLR and PLSR approach yielded superior analytical outcomes, enhancing the precision and accuracy of sulfur and calorific value determination in coal. The models achieved coefficients of determination (R²) of 0.925, 0.951, and 0.951, with average relative errors of 3.16%, 0.67%, and 0.52%, respectively.