Enhancing the energy yield of Amaranthus hybridus-derived biofuel using alkaline pretreatment : experimental and data-driven investigation

Abstract

he global quest for sustainable and eco-friendly fuel alternatives has spurred interest in biofuels. This study presents an experimental and data-driven framework for investigating the impact of alkaline-pretreatment on the combustion properties of solid-biofuel derived from Amaranthus hybridus. The aim of the study is to enhance the energy content of the biofuel using NaOH-pretreatment and provide data-driven insight into its energy drivers. The experimental analysis involved ultimate and proximate analysis, calorific value (CV) determination, and structural characterization. Advanced data analytics, including correlation analysis, feature importance analysis (FIA), dimensionality reduction using Principal Component Analysis (PCA), and neuro-fuzzy modelling, were developed to explore relationships among biofuel properties. A Grid Partitioning (GP)-clustered Adaptive Neuro-Fuzzy Inference System (ANFIS) tuned with Particle Swarm Optimization (PSO) was developed for CV prediction. NaOH pretreatment increased the CV from 11.38 MJ/kg to 12.79 MJ/kg (a 12.9 % improvement). The FTIR analysis revealed a C–O stretch difference of 4 cm−1, while the SEM analysis revealed morphological restructuring. The correlation-based parameter profiling revealed fixed carbon (FC) as the only positively correlated parameter with CV. The FIA revealed FC as the most influential predictor of energy content with a Gini index of 0.821, while PCA further confirmed FC’s dominance in driving calorific performance. The GP-clustered ANFIS-PSO model with triangular membership functions outperformed other configurations with high accuracy (RMSE = 0.1070, MAPE = 7.354 %, MAD = 0.0828, and MAE = 0.0895). This research contributes valuable insights into optimizing solid biofuel combustion properties through pretreatment strategies, supported by advanced computational, data-driven and neuro-fuzzy techniques. HIGHLIGHTS • Alkaline pretreatment (0.5 % NaOH) improved CV by 12.9 % from 11.38 to 12.79 MJ/kg. • SEM and FTIR confirmed NaOH pretreatment-induced fiber delignification and porosity. • Fixed carbon (FC) emerges as the dominant driver of energy with a GI-value of 0.821. • PCA shows that 98% of the data variance is captured in PC1 with FC as the dominant feature. • GP-ANFIS-PSO model achieved the best CV prediction with RMSE of 0.1070 and R2 of 0.932.