Integrated neural network and AspenPlus model for entrained flow gasification kinetics investigation

Entrained flow gasification is a well-established technology, however, the main obstacle in process design is the complex gasification mechanism, since numerous phenomena at extreme process conditions take place simultaneously. This study is focused on integrated thermodynamic and artificial neural network approach (ANN) for entrained flow gasification kinetics investigation. Data on 102 feedstock materials composition was used in the AspenPlus gasification simulation, where sensitivity analysis was performed for different equivalence ratios (0.1?0.7) and gasification temperature (1200?1500?C) values. For analyzed materials, an optimal equivalence ratio range exists (usually 0.3?0.4), maximizing gasification efficiency. The obtained results were used in ANN development for each output variable (syngas composition, efficiency, heating value, and carbon conversion). Matlab algorithm was used for the determination of the optimal number of neurons (1?20 range) in each ANN. High R2 values (>0.99) for all models suggested good agreement between simulated and predicted values. Genetic algorithm-based optimization studies for maximization of hydrogen content and cold gas efficiency result in mean ER values of 0.35 and 0.41, respectively, at a temperature of 1200 ?C. Yoon interpretation method was used for quantifying the relative impacts of each input variable on syngas content and gasification efficiency. The proposed approach represents a powerful tool that can facilitate the investigation of the entrained flow gasification and process design.