Battery State-of-Health Estimation for Autonomous Aerial Vehicles Using Features from Constant-Power Cruise Phase

Autonomous aerial vehicles (AAVs) powered by lithium-ion batteries (LIBs) offer promising solutions for urban air mobility. An accurate estimation for LIBs state-of-health (SOH) ensures the safe operations of AAVs. In this work, we investigated a battery degradation dataset simulating the AAVs working conditions. The features were extracted from the constant-power cruise phase for the first time. A K-nearest neighbors (KNN) model was optimized by Bayesian hyperparameters optimization algorithm. Integrating a transfer learning (TL) layer, the KNN-TL model achieves good performance across various working conditions of AAVs using 1 % labeled target domain samples, with RMSE ranging from 0.532 % to 1.216 % and MAE ranging from 0.396 % to 0.967 %.