PL EN
Energy consumption and efficiency degradation predictive analysis in unmanned aerial vehicle batteries using deep neural networks
 
More details
Hide details
1
Mechanical Engineering Department, University of Technology-Iraq, Baghdad, Iraq
 
2
Institute of Mechanical Engineering, Faculty of Mechanical Engineering, Bialystok University of Technology, ul. Wiejska 45C, 15-351 Bialystok, Poland
 
3
Department of Chemical Engineering and Petroleum Industries, Al‐Mustaqbal University College, Hillah, Iraq
 
4
Department of Computer-Aided Design Systems, Lviv Polytechnic National University, 79013 Lviv, Ukraine
 
5
Faculty of Automatic Control, Robotics and Electrical Engineering, Institute of Robotics and Machine Intelligence, Poznan University of Technology, Poland
 
 
Corresponding author
Luttfi Ahmed Al-Haddad   

Mechanical Engineering Department, University of Technology-Iraq, Baghdad, Iraq
 
 
Andrzej Łukaszewicz   

Institute of Mechanical Engineering, Faculty of Mechanical Engineering, Bialystok University of Technology, ul. Wiejska 45C, 15-351 Bialystok, Poland
 
 
Adv. Sci. Technol. Res. J. 2025; 19(5):21-30
 
KEYWORDS
TOPICS
ABSTRACT
The growing reliance on Unmanned Aerial Vehicles (UAVs) necessitates efficient energy management to ensure optimal performance and longevity. This study investigates the energy consumption patterns and efficiency degradation in DJI Mini 2 drone batteries by the utilization of a deep neural network (DNN) for predictive analysis. The experimental work conducted repeated flight tests and monitoring battery discharge from 100% to 27% over 20 trials. The testing conditions, including flight duration and environmental factors, were controlled to ensure repeatability and to minimize any external influences on the recorded data. These data were stored onto AIRDATA and then recollected for new labeling. The initial flights demonstrated stable performance, while subsequent flights showed a gradual reduction in flight time which indicated performance degradation. To ensure consistent power usage and minimize external influences, hover mode was selected for all flights. The DNN was trained on this data and employed metrics of Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), Coefficient of Variation of the Root Mean Squared Error (CVRMSE), and determination coefficient (R²). The proposed model achieved an MSE of 0.352%, RMSE of 0.593%, MAE of 0.324%, MAPE of 0.857%, CVRMSE of 0.743%, and R² of 0.981. These results demonstrate the DNN's ability to accurately predict future power consumption that in turn provides insights for energy management and extending battery life. This research contributes to the development of sustainable UAV operations by enhancing understanding of battery performance under real-world conditions.
Journals System - logo
Scroll to top