Towards machine-learning driven prognostics and health management of Li-ion batteries. A comprehensive review

Prognostics and health management (PHM) has emerged as a vital research discipline for optimizing the maintenance of operating systems by detecting health degradation and accurately predicting their remaining useful life. In the context of lithium-ion batteries, PHM methodologies have gained significant attention due to their potential for enhancing battery maintenance and ensuring safe and reliable operation. Among the various approaches, data-driven methodologies, particularly those leveraging machine learning (ML) models, have gained interest for their accuracy and simplicity. To develop an optimized data-driven PHM system for batteries, a comprehensive understanding of each step involved in the PHM process is crucial. This review paper aims to address this need by providing a thorough analysis of the different phases of battery PHM, encompassing data acquisition, feature engineering, health diagnosis, and health prognosis. In contrast to previous review papers that primarily focused on battery health diagnosis and prognosis methods, this work goes beyond by encompassing all essential steps necessary for developing a tailored PHM methodology specific to lithium-ion batteries. By covering data acquisition methods, feature engineering techniques, as well as health diagnosis and prognosis methods, this paper fills a significant gap in the existing literature. It serves as a comprehensive roadmap for researchers and practitioners aiming to develop PHM systems for lithium-ion batteries using ML techniques. With its in-depth analysis and critical insights, this review paper constitutes a substantial contribution to the field. It provides valuable guidance for designing effective PHM methodologies and paves the way for further advancements in battery maintenance and management.