Machine learning-driven prognostics for lithium-ion batteries: enhancing RUL prediction and performance in smart energy storage systems

Abstract

In the evolving landscape of energy systems, batteries play a critical role in enabling hybrid and stand-alone renewable energy storage solutions. Precisely estimating battery life and remaining useful operational life will go a long way in enhancing the efficiency of the system with assured reliability in smart power storage devices. This report comprehensively surveys advanced approaches in the management of batteries through state-of-the-art artificial intelligence tools-support vector machines, relevance vector machines (RVM), long short-term memory (LSTM) models, and bayesian filters-that are being used with a view to enhancing remaining useful life (RUL) estimates and making real-time system health monitoring capabilities possible. Modeling approaches surveyed include state estimation, capacity, and thermal management, while discussing their applicability to lithium-ion batteries. The review also explores publicly available battery datasets, feature engineering strategies, and hybrid diagnostic frameworks. A technoeconomic perspective is provided to assess system performance in renewable-integrated power grids. This paper aims to consolidate current knowledge, provide comparative insights into the strengths and limitations of different approaches, and highlight open research challenges to guide future developments in smart AI-enabled battery systems that support sustainable and resilient energy infrastructure.