Massively Distributed Bayesian Analysis of Electric Aircraft Battery Degradation

Abstract

Electric vertical takeoff and landing (EVTOL) aircraft have high power and energy requirements that must be understood throughout their batteries’ life. Using a fast electrochemical model and high throughput computing, we independently and simultaneously sample the degradation modes for 21392 cycles of an EVTOL aircraft battery data set, obtaining a posterior probability distribution for each degradation mode throughout the life of the aircraft. The model shows a median error of 32.5 mV (or 1.9%) across all cycles. We conduct an identifiability analysis on the generated mode distributions. We analyze correlations between the modes and cycling characteristics, finding that depth of discharge, temperature, and charging current all have significant impacts on degradation. We introduce a protocol for early identification of active degradation mechanisms, identifying electrolyte oxidation, active material dissolution, and growth of a solid-electrolyte interphase as the most likely causes of battery degradation. Finally, we discuss other applications of large-scale sampling of battery degradation modes.