Modeling and optimization of ammonia/hydrogen/air premixed swirling flames for NOx emission control: A hybrid machine learning strategy

In the face of escalating climate change concerns, the quest for sustainable energy solutions is more pressing than ever. This study delves into the potential of hydrogen and ammonia as alternative fuels, with a focus on ammonia's promise due to its zero-carbon emissions and high energy density. Employing machine learning techniques, specifically XGBoost and SVR, this study presents a comprehensive analysis of ammonia and hydrogen fuel blends to predict NOx emissions and flame temperature with high accuracy, achieving R2 values predominantly above 0.97. The model's precision is particularly noteworthy compared to other machine learning techniques, where it consistently outperforms with the lowest MSE of 3508.31 and an impressive R2 value of 0.97653. A detailed feature importance analysis underscores the significance of NH3 mole proportion, equivalence ratio, and total mass flow rate in influencing nitrogen emissions. Furthermore, the proposed XSN optimization framework has proven effective in reducing nitrogen compounds, achieving a substantial decrease in N-gases concentration by 51.91 %, from 69.81 ppm to 33.57 ppm. The hybrid model developed in this study demonstrates exceptional capability in managing multiple optimization objectives, thereby offering advantages in reducing the overall harmful emissions while maintaining stable operation in practical applications of NH3/H2 combustion. This research enhances the accuracy of emissions prediction under diverse conditions and provides valuable insights into effective strategies for controlling nitrogen emissions from NH3/H2 combustion.