Analysis of factors affecting the performance of an aviation system using selected models

The aim of this study is to quantitatively analyze the factors influencing energy consumption in unmanned aerial vehicles (UAVs) based on operational data from 177 flights of the DJI Matrice 30T drone. Battery consumption modelling was proposed using variables available at the UAV user level. A comparison of analytical methods (linear regression, LASSO) and machine learning algorithms (Random Forest, XGBoost) was performed. The models were then evaluated using the coefficient of determination R^2 and the root mean square error RMSE. Analytical methods show moderate effectiveness (R^2 = 0.425, RMSE = 14.87%), while machine learning models show significantly higher predictive accuracy: Random Forest achieved R^2 = 0.983 and RMSE = 0.328%, and XGBoost R^2 – 0.973 and RMSE = 3.26%. The analysis of variable significance shows that the greatest impact on energy consumption is exerted by: flight time, distance traveled, and discharge current. Seasonal factors also proved to be significant, indicating the impact of weather conditions on battery discharge dynamics. The results confirm the superiority of adaptive machine learning methods over classical analytical models in forecasting UAV energy consumption based on operational data and indicate the direction for further research taking into account detailed meteorological data. Unlike previous studies, this study is based on operational data from actual UAV missions and uses only variables available from the user’s perspective. It also provides a methodical comparison of analytical approaches and machine learning algorithms on a single real-world flight log dataset and additionally considers the impact of seasonality and operating conditions on battery consumption – an aspect largely overlooked in the literature.