Abstract
The growing urgency for sustainable energy alternatives has accelerated research into biofuel utilization in gas turbine systems, where Artificial Intelligence (AI) offers transformative capabilities for performance assessment and emission prediction. This study investigates the AI-based assessment of biofuel performance parameters in gas turbine combustion systems with a focus on emission characterization. The primary objectives include evaluating the predictive accuracy of machine learning models for NOx and CO emissions and comparing the combustion efficiency of various biofuel blends against conventional Jet-A fuel. A quantitative secondary data analysis methodology was adopted, employing published experimental datasets (2019–2025) from the UCI Machine Learning Repository and peer-reviewed combustion studies, analysed using ANN, Random Forest, SVM, and LSTM algorithms. Two hypotheses were formulated: H1 states that AI-driven machine learning models achieve statistically significant prediction accuracy (R² > 0.90) for NOx and CO emissions from biofuel-fueled gas turbines; H2 states that biofuel blends (B20–B50) produce statistically significant reduction in CO emissions compared to conventional Jet-A fuel while maintaining thermal efficiency above 26%. The results confirm that the Random Forest model achieved R² = 0.9792 for NOx prediction, validating H1 (t = 14.82, p < 0.001). Biofuel blend B20 demonstrated 20.20% CO reduction with thermal efficiency of 27.80%, supporting H2 (t = 8.67, p < 0.001). It is concluded that AI-based predictive frameworks provide a reliable pathway for optimizing biofuel combustion and achieving emission reduction targets in gas turbine power generation.
