Enhanced Prognostics for Lithium-Ion Batteries in eVTOL Aircraft Using CatBoost and LightGBM Algorithms

Objectives: The goals of this study are as follows: The comprehension of advanced and efficient E-VTOL techniques is required to control batteries better as well as to increase their performance, safety, and lifespan. The data-driven machine learning models using CatBoost and LightGBM are developed to predict the battery charge, health, and RUL based on flight conditions. This paper aims to assess the impacts that charging/discharging rates as well as temperature have on batteries and determine characteristics that influence discharging. Methods: In the framework that has been proposed, the models used for the analysis and prediction of battery charge, health, and RUL include CatBoost and LightGBM. These algorithms define charging and discharging profiles and the effects of temperature on the battery. Different comparative studies on machine learning algorithms are performed to find an optimal solution for battery prediction on E-VTOL vehicles. It is analyzed in real-time using flight profiles to predict the takeoff, landing, and cruising phases of the flight. Findings: The suggested framework effectively estimates the battery charge, health, and RUL, outcompeting all previous algorithms. CatBoost is known for its ability to handle categorical data and its efficient training process, making it suitable for real-time applications. LightGBM is recognized for its speed and efficiency, particularly in handling large datasets, and its robustness in making accurate predictions. Novelty: Drawing from these methods, this has proposed a new battery control in E-VTOL vehicles using a machine learning framework that incorporates handling categorical features, robustness to overfitting, automatic feature scaling, built-in cross-validation, and robustness to noisy data. Keywords: CatBoost, Electric Vertical Take Off and Landing, Gradient Boosting Algorithm, LightGBM, RUL, SOC, Battery Lifespan, Battery Management, SOH