Experimental research on cellulosic biomass pyrolysis and BP neural network prediction

Three types of cellulosic biomass which were rice straw, cotton stalk and pine sawdust were used and the contents of cellulose and lignin in the biomass were analyzed chemically. It was observed that the characteristic of biomass pyrolysis was dependent on its components of cellulose on the basis of TGA experiments. The higher the cellulose content, the faster the pyrolysis rate. The most probable mechanism function obtained by the Malek method. It was analyzed that biomass pyrolytic process should be divided into two stages to establish dynamic model respectively, the front stage for D1 model, and the second stage for F1 model. It established BP neural network with momentum added which given an efficient prediction on activation energies. It shows that the diversion of forecasting value from the tested is not more than 1.23 kJ.mol^-1 with a relative error within ±2.5%. It proves that the BP neural network has a better forecasting ability.