Predicting the effect of bed materials in bubbling fluidized bed gasification using artificial neural networks (ANNs) modeling approach

Abstract

The effect of different bed materials was included a as new input into an artificial neural network model to predict the gas composition (CO2, CO, CH4 and H2) and gas yield of a biomass gasification process in a bubbling fluidized bed. Feed and cascade forward back propagation networks with one and two hidden layers and with Levenberg-Marquardt and Bayesian Regulation learning algorithms were employed for the training of the networks. A high number of network topologies were simulated to determine the best configuration. It was observed that the developed models are able to predict the CO2, CO, CH4, H2 and gas yield with good accuracy (R2 > 0.94 and MSE < 1.7 × 10−3). The results obtained indicate that this approach is a powerful tool to help in the efficient design, operation and control of bubbling fluidized bed gasifiers working with different operating conditions, including the effect of the bed material.