The PSO-ANN modeling study of highly valuable material and energy production by gasification of solid waste: an artificial intelligence algorithm approach

Gasification technology is an effective way to achieve high value-added utilization of solid waste. In this work, a typical fluidized bed reactor was used to investigate the gasification of pinewood particles. The influence of temperature, equivalence ratio (ER), and steam/biomass ratio on product yield was investigated. Experimental results show that higher temperature benefited the yield of syngas while inhibited the yield of biochar. Increasing ER benefited the gasification reactions but too much air would cause over oxidation and lower the concentration of H-2 and CO. Artificial neural network (ANN) model and particle swarm optimization (PSO) coupled with ANN model were built to predict the product yield during waste gasification. The optimized transfer function for hidden layer is logsig, with optimized neuron number of 12. Network weight analysis shows that hydrogen content, gasification temperature, carbon content, and ER are the largest impact variables for H-2 concentration, CO concentration, biochar yield, and syngas yield, respectively. PSO-ANN model results show that the PSO algorithm could be used to optimize the weight of ANN model and significantly improve the prediction accuracy. The largest deviation for CO concentration reduced from 13.93 to 8.39%. Results proved that ANN model can be used as an effective tool to predict the product distribution in solid waste gasification.