Advancements in battery state prediction methods: A comprehensive review of data-driven approaches and degradation mechanisms

Abstract

Predicting the state of lithium-ion batteries (LIBs) plays a vital role in proactively identifying aging and damage, thereby enabling timely maintenance and repairs to extend their lifespan and ultimately reduce waste while advancing the development of clean energy. This comprehensive review aims to provide an extensive overview of state prediction methods and characterization techniques, emphasizing the need for specific data-driven approaches to characterize battery degradation and establish correlations between inputs and outputs. Furthermore, the evaluation of primary anode, cathode, and separator materials in batteries is discussed. The study delves into the significance of both cycle degradation and calendar degradation, providing insights into degradation phenomena and parameters perspectives. Moreover, the roles and characteristics of physical parameters for degradation characterization and state parameters for battery operation status are analyzed. The review extensively examines the strengths and weaknesses of classic and state-of-the-art physics-based models, as well as data-driven models, highlighting critical issues and challenges related to degradation mechanisms, datasets, performance trade-offs, and optimization methods. Additionally, innovative ideas are proposed, and future development directions are discussed.