Modelling coal gasification with a hybrid neural network

Abstract Gasification of two coals was carried out in a batch feed fluidized bed reactor at atmospheric pressure using steam as fluidizing medium. A model of coal gasification was developed, incorporating a first-principles model with a neural network parameter estimator. The hybrid neural network was trained with experimental data for the two coals and gave good performance in process modelling. A parameter for the overall reactivity of char, namely ‘active char ratio’ (ACR), was identified by the neural network, as a function of gasification time and temperature. The ACR profile showed a strong dependence on coal type. Other parameters estimated by the neural network also reflected distinct characteristics of the two coals.