A novel semi-supervised learning approach for maritime lithium-ion battery monitoring

Abstract

We propose a novel semi-supervised learning method to monitor the State of Health of lithium-ion batteries, a prominent technology for the electrification of the transport sector. Our approach enables State of Health monitoring of batteries with no labeled data, starting from a minimal set of labeled data from another similar battery. This can be achieved by exploiting the relation between a pseudo-capacity measure and the total capacity of the labeled data. Our results with operational data from maritime batteries show that the approach is valid and can lead to significant progress in failure prevention, operational optimization, and for planning batteries at the design stage.

A novel semi-supervised learning approach for maritime lithium-ion battery monitoring