Estimation of aircraft fuel consumption and emissions during taxiing processes
利用欧洲枢纽机场的真实滑行数据，比较了传统线性模型与机器学习模型在预测飞机滑行燃油消耗上的准确性，发现机器学习更优，但仍有改进空间。
The aviation sector faces growing challenges in reducing emissions as global passenger numbers continue to rise. While most research focuses on flight operations, this research investigates aircraft emissions during taxiing. Therefore, existing taxiing fuel models are compared and evaluated on a theoretical and empirical basis. Using an extensive dataset of real-world taxiing data from a European hub-airport, we evaluate linear models widely used in practice based on ICAO fuel burn indices and more sophisticated models that account for operational factors and ambient conditions. Additionally, we estimate machine learning models to identify relationships not captured by current models. Results show that machine learning approaches provide more accurate fuel consumption predictions than traditional methods. Notably, empirically calibrated thrust settings are higher than literature defaults. However, unexplained variance remains, suggesting potential for further improvements. Our findings offer valuable insights for integrating emission models into airport operations, helping stakeholders implement effective emission reduction strategies.