Modeling of ash agglomerating fluidized bed gasifier using back propagation neural network based on particle swarm optimization

Recently, back propagation neural networks (BPNNs) have been applied to various gasification processes. Although the aforementioned BPNNs obtain relative satisfying prediction results, most of these models have some obvious drawbacks such as over fitting problem which can greatly reduce the prediction accuracy of BPNNs. In this study, a back propagation neural network optimized by particle swarm optimization (BPNN-PSO) was strongly proposed as a novel model to solve these problems. More importantly, this new BPNN-PSO approach was used to predict gas composition, gas yield, lower heating value of syngas and gasification temperature for coal gasification in a pilot-scale pressurized ash agglomerating fluidized bed gasifier. Three variables that affected the prediction results were selected as input parameters which were oxygen flow rate, steam flow rate and coal flow rate. The prediction accuracy of the newly-employed method was compared with pilot experimental data. The results convincingly denoted that the proposed approach possessed high accuracy with highest correlation coefficient, lowest mean squared error and mean absolute percentage error. Additionally, effects of oxygen to coal ratio and steam to coal ratio on gasification performance were precisely and detailedly studied using the newly-proposed BPNN-PSO approach. (C) 2017 Elsevier Ltd. All rights reserved.