This paper proposed and validated a new coal devolatilization model that can predict yields of respective gas and tar components, and be directly coupled with an elementary step-like reaction model. The new model was an extension of the cheinical percolation devolatilization (CPD) model. The CPD model is one of the existing primary pyrolysis models that consider the coal chemical structure. The mole fraction of labile bridges, cross links, peripheral groups, and average molecular weight of monomer in coals examined in this study were determined by a C-13 NMR analysis. The aromatic ring dusters size distribution was determined by a coal pyrolysis test using a Curie point pyrolyzer. The thermal decomposition process of the coal chemical structure was expressed by nine elementary reactions. From the above, the extended CPD model can predict gas and tar components as respective chemical species (H2O, CO2, CO, CH4, benzene, naphthalene and phenanthrene) and consequently capture the secondary decomposition and polymerization in gas phase by coupling the detailed gas-phase chemistry. The proposed model was validated by comparing with experiments using a pressurized drop tube furnace (PDTF). The results showed that the trend of light gases and soot yield in the experiments could be successfully reproduced by the extended CPD model. (C) 2017 Elsevier B.V. All rights reserved.