Data-Driven Battery Health Prognostics of Lithium-ion Battery Using Auto Regressive Models and Hybrid Physics-Inspired Methods

Abstract

Accurately modeling lithium-ion batteries (LiBs) allows for improved harnessing of their capabilities without compromising their safety and lifespan. Ships that use batteries for power, whether partially or fully electric, are a versatile solution for marine systems to decrease fuel consumption and pollution. Observing the state of health (SOH) and state of charge (SOC) of the battery makes it possible to control the available power, thereby guaranteeing safer propulsion and maneuvering that depends on battery power. There is presently a significant push to reduce emissions and transition to more environmentally friendly technologies for maritime transport. The use of data-driven diagnostics with the help of machine learning can improve the prediction of maritime battery health. These diagnostics allow for the development of models for estimating the SOH and SOC, as well as predicting Remaining Useful Life (RUL), utilizing battery degradation or aging data. The objective of this thesis is to develop a Data-Driven Methods (DDM) for studying battery degradation and thereby predict the RUL of batteries with precision in order to determine the available service time left before the battery’s performance degrades to an unacceptable level. Investigation of degradation will let users know when to replace the battery to prevent any failure and what measures to take to maximize its lifespan