Performance prediction of fluidised bed gasification of biomass using experimental data-based simulation models

Abstract

Wood sawdust is gasified in air-fluidized bed with steam injection for the enrichment of product gas with hydrogen. A gasification experimental setup with sand as bed material is designed and developed for this purpose with a biomass feed rate of 10.3 kg/h. Air and steam flow rates are varied between 0.0042–0.0063 and 0–0.0072 m3/h, respectively. Axial variations of temperature and pressure inside the reactor shell are investigated. The data for the product gas composition from the experiment are utilised to develop two models. One is a feedforward artificial neural network (ANN) model for the prediction of gasification temperature and product gas composition. The second is a Redlich–Kwong real gas equilibrium correction model incorporating tar (aromatic hydrocarbons) and unconverted char to predict the product gas composition, heating value and thermodynamic efficiencies. Good accuracy of ANN prediction with experimental results is achieved based on the computed statistical parameters of comparison such as coefficient of correlation, root mean square error (RMSE), average percentage error and covariance. The corrected equilibrium model developed by introducing correction factors for real gas equilibrium constants shows satisfactory agreement (RMSE = 5.96) with the experimental values. Maximum concentration of hydrogen achieved in the experiments is 29.1 % at the equivalence ratio (ER) = 0.277 and steam to biomass ratio (SBR) = 2.53. The corresponding predicted values are 28.2 % for ANN model and 31.6 % for corrected equilibrium model. The corrected equilibrium model for wood sawdust is validated with major air–steam gasification experimental results of other biomass materials and is found to be 95.1 % accurate on average. It is revealed from the study that the ANN model (RMSE = 2.64) is a better predictor for the product gas composition than the corrected real gas equilibrium model (RMSE = 5.96). The study proposes a more comprehensive ANN model capable of simulating various process conditions in fluidised bed gasification applicable to variety of biomass feedstocks.