Estimation of SoC, SoH, and RUL in battery management systems using artificial intelligence: A tertiary mapping study

The growing demand for efficient energy storage systems, driven by the electrification of mobility and the global energy transition, has made the battery management system a critical component. This study presents a tertiary review that maps, analyzes, and synthesizes the secondary literature on the application of artificial intelligence (AI) techniques to estimate the state of charge (SoC), state of health (SoH), and remaining useful life (RUL) of lithium-ion batteries. Based on a systematic literature review, 43 secondary studies were selected and analyzed. The results indicate a dominance of machine learning techniques, such as artificial neural networks (ANN) and support vector regression (SVR), with a growing trend toward deep learning (DL) architectures. Among these, long short-term memory (LSTM) variants, particularly bidirectional LSTM (Bi-LSTM) and hybrid CNN-LSTM models, are frequently employed to capture complex temporal dynamics. For hyperparameter tuning, bioinspired algorithms such as particle swarm optimization (PSO) and genetic algorithms (GA) are the most prevalent optimization methods. The NASA, CALCE, and Oxford datasets are the most widely used for model validation. Key challenges identified include data quality and availability, computational complexity for implementation in embedded systems, and model generalization. Future directions point toward the hybridization of physical and data-driven models, developing lightweight algorithms (TinyML), adopting of eXplainable AI (XAI), and the creation of standardized benchmarks to promote open science. This tertiary mapping study consolidates the state of the art, harmonizing perspectives and providing a unified panoramic view to guide future research and industry professionals.