Pyrolysis of several kinds of biomass was carried out in a fluidized bed at atmospheric pressure using nitrogen as fluidizing gas. A chaotic neural network model was set up to study biomass pyrolysis. Three schemes of network input were applied and compared in the simulation

and the most effective one which could make sound prediction about the pyrolysis products was determined. This scheme involved experimental operation parameters

biomass proximate analysis data and chemical component analyses data. A rational explanation for the three schemes was given based on the difference of the evolution and chemical structures between biomass and coal.