Carbon Burnout
Carbon burnout is controlled by the char morphology which strongly depends on the rank and petrographic composition of the coal and the thermal and gaseous environment under which the coal devolatilises.  Minerals within the coal can have a slight influence by catalyzing the oxidation reaction.
The prediction of carbon burnout based on empirical curves derived by  Blake and Robin (1982) has been widely used in the power industry. This burnout model requires only one boiler design parameter, Furnace Heat Release rate (FHR) (fuel burn rate divided by furnace volume), and the operational parameter of excess air level used in the boiler. The coal parameters are the dry ash free volatile matter content and the fineness (percent passing 75  um) of PF.  
The figure below shows the predicted and actual burnout performance versus the volatile content of a wide range of coals at both full scale and pilot scale.  The predicted full scale burnout is based on the design of a typical Asian power plant using imported coals while the actual full scale data is from a wide range of power plants.  This is one reason for the spread in this actual data.  
The pilot scale data are from projects conducted in ALS's 150 kW Boiler Simulation Furnace (BSF) all these results were for coals fired at a fineness of about 70%.  The predicted pilot scale curve uses the FHR that best fitted the pilot scale results without adjusting for carbon loss and fineness, this FHR is close to the calculated FHR for the pilot furnace.  For lower volatile coals, the poorer burnout will reduce the calculated FHR of the BSF by as much as 20%.  When the FHR is adjusted for burnout there is better agreement between actual and predicted burnout for the pilot scale results.
As indicated in this figure the volatile matter content of a coal can be used as a general guide to carbon burnout. But, as shown by differences between pilot scale predicted and actual data for some coals, other factors, such as maceral composition of the coal, also influence burnout. Su and others (2001b) showed that a maceral index can give a slightly better fit to pilot scale data, though their correlation does not allow the scaling- up of pilot data to predict full scale performance.
Other empirical approaches to the prediction of burnout have been published. Wu and others (2006) have applied petrographic techniques to estimate char morphology which allows prediction of burnout. This has been successfully used by a UK power plant for the selection and quality control of imported coals.  Niksa has incorporated the work of Hurt in modeling char conversion with his FlashChain devolatisation model to predict burnout in PF power plants and model the performance of gasifiers ( Niksa & Hurt 2005).