Data driven fuel consumption prediction model for green aviation using radial basis function neural network

Abstract

Abstract In response to the growing demand for sustainable aviation, a fuel consumption prediction model based on Radial Basis Function (RBF) Neural Networks was proposed. Using high-resolution onboard Quick Access Recorder (QAR) data, which contains richer flight parameters and higher accuracy, RBF models were constructed based on the extracted key influencing factors for different flight phases, including takeoff/climb, cruise, and descent/approach. The model provides a lightweight and computationally efficient solution for high-dimensional, nonlinear flight data, ensuring accuracy with lower computational burdens. It is suitable both for pre-flight ground-based fuel consumption prediction and deployment in resource-constrained onboard environments, enabling real-time prediction during flight operations. Experimental results showed that the RBF model’s prediction errors for the takeoff/climb, cruise, and descent/approach phases were 5.73%, 3.36%, and 14.04%, respectively, significantly outperforming the comparison models. The error variances from ten-fold cross-validation were 0.31%, 0.15%, and 0.29%, respectively, confirming the robustness of the model. Further analysis indicated that the model can be employed to evaluate the “fuel penalty for carrying additional fuel” patterns and enhance fuel efficiency. This study provided valuable insights and theoretical support for airlines in optimizing flight planning and minimizing fuel consumption, thereby contributing to the sustainable development of green aviation.