Optimization of Biomass Pellet Composition for Enhanced Syngas Quality and Energy Efficiency Using Causal Discovery and Non-Dominated Sorting Genetic Algorithm II

Abstract The increasing demand for sustainable and renewable energy sources has driven significant interest in biomass as a viable alternative source, which also converts waste into energy. The work has focused on designing an experimental setup and then performing the gasification of three agricultural wastes, wheat straw pellets (WSP), rice straw pellets (RSP), and soybean straw pellets (SSP), by forming their pellets. These are then mixed in different proportions as per hypercube sampling data for the gasification process, and pellets are then subjected to elemental analysis, proximate analysis, and calculation of their high heating value. All the pellets are passed through a downdraft gasifier, and the syngas produced is assessed for its properties. After this process, the artificial intelligence (AI) models, like causal discovery networks (CDNs) for correlation study between the input and output parameters, and multiobjective Non-Dominated Sorting Genetic Algorithm II (NSGA-II) for optimum mix identification, are used. The algorithm predicted that a mixture of 13.3% WSP, 8.0% RSP, and 78.7% SSP in the pellets would produce 18.89% CO in the syngas, having a higher heating value (HHV) of 4.97 MJ/m3. The experimental validation with these mixes found that the algorithm had an error of less than 5% proving its efficiency. These analyses help refine the pellet mix and processing parameters, ensuring improved energy yield and operational efficiency.