An empirical model was established based on measurements of devolatilization characteristics of coal particles (18.7%, 32.7%, and 43.0% VM; 0.14, 0.20, and 0.30 mm, respectively) heated at a heating rate of 10 3 -10 4 K/s by a well-controlled CO 2 laser with double shutters and a high-accuracy two-color pyrometer. The conclusions obtained are as follows: (1) particle temperature (T p ), volatile yield (V), heating rate (HR), and in situ energy density flux (F ed ) during devolatilization can be well-predicted by solving an energy conservation equation and a devolatilization rate equation with two competing reaction rates; (2) the preexponential factor and activation energy of the two reactions are A 1 = (1.7249-1.8936) Â 10 2 s -1 , E 1 = (2.6248-3.5447)Â10 4 J/mol, A 2 =2.6Â10 6 s -1 , and E 2 =1.6740Â10 5 J/mol, respectively, which are obtained by fitting our experiment data; (3) the final volatile yield (V f ) is dependent on the laser intensity (Q L ), the particle size (D n ), and the proximate volatile matter content (V 0 ), and it can be well-predicted by a regressive equation of V f with Q L , D n , and V 0 ; (4) a modified Merrick's heat capacity correlation can be used to predict the temperature history of coal particles heated at a rate of 10 3 -10 4 K/s with excellent agreement; (5) the heats of devolatilization, which are given as -0.5244, -0.6629, and -0.7348 MJ/kg for the three coals used in this study, are suitable values for predictions; (6) the sum of the heat of devolatilization (ΔH d ) and energy loss per unit mass volatile yield (ΔH vl ), which can be represented as ΔH d þ ΔH vl =-0.01046 MJ/kg, is a suitable value for predicting particle temperatures and volatile yields for all three coals; (7) the measured histories of weight loss are imperative for predicting the temperature and the devolatilization kinetics of coal particles heated at a high heating rate; and (8) the swellingshrinking ratios, absorptivity, emissivity, thermal capacity, and heat of devolatilization of coal particles at high heating rates should be measured for better predictions.