Fig.2: Particle Mass-loss profile Robert Nachenius, Thomas van de Wardt, Frederik Ronsse, Wolter Prins Department of Biosystems Engineering, Ghent University, Coupure Links 653, 9000 Gent, Belgium Tel.: +32 (0) ; Validation of a torrefaction reaction model for a continuous screw conveyor reactor Introduction This work falls within a larger project wherein different reactor technologies for biomass torrefaction are to be evaluated and compared. A model is necessary to test hypothetical reactor schemes and biomass feed types. This model must be reflective of actual thermochemical conversion processes and must therefore be validated. The purpose of the model is to predict the torrefied biomass yield for different particle sizes under various thermal histories. Experimental Method Torrefaction experiments have been carried out in a continuous screw conveyor reactor (Fig.1) based on a feed of pine chips (d p < 6 mm). Residence time and reactor temperature were found to be most important parameters. Cold-flow experiments performed to determine residence times have indicated plug-flow. Thermal histories were obtained by combining axial temperature measurements from the shell of the screw with residence time measurements (determined by cold-flow experiments) and used as model inputs. The explicit, numerical model combines the reaction kinetics (obtained by non-isothermal thermogravimetric analyses (TGA) at 2, 10, 20 °C/min heating rates) and expressed through isoconversional analysis) with transient heat transfer from the conveyor shell to the pine chips in order to predict the achieved torrefied biomass yield. The model was run for the different particle sizes present in the feed and summed proportionally, to predict the yield achieved by the bulk. Fig.1: Continuous Screw Reactor Model Details The model was built on the basis of several simplifications: Drying is determined from sorption data (Hailwood-Horrobin equation). Only desorption is calculated and adsorption is excluded Pressure changes within the particle are ignored and mass transfer effects are neglected. The kinetic data are only valid for mass yields above 19% (dry basis) but this is well within the typical torrefaction range. The thermal properties of pine, water, and torrefied biomass are taken from literature as far as possible. Particle sizes remain constant (no shrinkage or attrition). Results A modeled mass loss profile for a spherical, 3 mm pine chip is given in Fig.2. The initial drying stage can be distinguished from the torrefaction stage A parity plot comparing the yields predicted by model and those obtained experimentally is given in Fig.3. Experimental temperatures ranged from 275 ° C to 375 ° C and residence times varied from 210 s to 650 s. Conclusions The torrefaction model is able to indicate the transient mass loss due to drying and torrefaction at short residence times only and overestimates conversion at longer residence times. It appears that the model overestimates the rate of heat penetration into the particles. The model is more sensitive to changes in external heat transfer (Biot numbers are less than 0.17 for these particle sizes). Further Work Experiments are planned in order to measure external heat transfer rates. If this is insufficient in addressing the model discrepancies, further complexity may need to be incorporated into the model (anisotropy, mass transfer limitations). The model will also be validated against results obtained from other feedstocks (switchgrass and miscanthus). Kinetic data will also be taken from TGA experiments performed at higher heating rates such as those seen in this work. Acknowledgements Dr Marco Frank from Sasol Technology is gratefully acknowledged for his support. Fig.3: Model-Experimental Parity Plot