Predição de vida útil remanescente em baterias lítio-íon aplicando a metodologia de metalearning

Abstract

Lithium-ion batteries are present in several devices and machines, and the unexpected end of their life cycle can cause the suspension of equipment operation, which can lead to failures, damage, or accidents. There are several studies about remaining useful life prediction for these batteries, usually based on empirical models or models trained by machine learning algorithms. This work seeks to approach a new proposal for the estimation of the useful life, through the usage of an image search system based on its content to classify spectrograms of the variables such as the current and terminal vol- tage, along with the metalearning methodology assisting in the classification the best descriptor responsible for extracting the information from these spectrograms, consequently improving the accuracy of the estimation.