UAV Battery Prognostics and Flight Time Estimation

No Thumbnail Available
Journal Title
Journal ISSN
Volume Title
Sähkötekniikan korkeakoulu | Master's thesis
Date
2023-01-23
Department
Major/Subject
Control, Robotics and Autonomous Systems
Mcode
ELEC3025
Degree programme
AEE - Master’s Programme in Automation and Electrical Engineering (TS2013)
Language
en
Pages
80
Series
Abstract
Lithium-based battery systems are extensively used in the electric mobility industry. The safety, prognostics, and longevity of the batteries are assured by battery management systems. One of the features of these management systems is to accurately determine the charge in a battery which is used to estimate the remaining run time of the electric vehicle, in this case, a drone. It is currently not possible to evaluate the charge of a battery by simply measuring the external parameters such as voltage or current. This problem is also known as the state of charge estimation in scientific literature. \\In this thesis, a highly accurate battery state of charge estimation method is developed and the result from this estimation is further used to predict the remaining flight time of the drone. This is done by developing an estimation algorithm based on data-driven approach. An Artificial Neural Network takes the voltage, current, and temperature information as input to predict the State of Charge. Since this is a time-series forecasting problem, the estimation algorithm specifically utilizes a type of Neural Network called the Recurrent Neural Network. This can capture long-term dependencies and model sequential data without requiring any accurate physics-based system modeling knowledge. Using the battery charge estimation, the remaining battery charging or discharging time can be predicted based on the current consumption of the drone. The performance of the proposed model is compared to existing methods that use various variations and combinations of Recurrent Neural Networks and other types of neural networks to predict the state of charge of a lithium-based batteries. The results showed that the proposed model achieved superior accuracy for state of charge prediction in UAV batteries.
Description
Supervisor
Zhou, Quan
Thesis advisor
Suvitie, Arto
Keywords
deep learning, state of charge, time-series forecasting, LSTM, estimation
Other note
Citation