• Reviews direct and indirect battery health indicators (HIs) used in SOH and RUL prediction. • Identifies key challenges in HI selection across data, feature engineering, and variability. • Analyzes impacts of SOC, C-rate, temperature, and capacity regeneration on HI reliability. • Synthesizes 2021 onwards literature to map trends in HI extraction and data-driven methods. • Proposes a unified framework for smart data processing, feature fusion, and real-time HI adaptation. The existing literature extensively explores deep learning-based models for battery health forecasting, highlighting their advantageous capacity to effectively model the intricate and nonlinear nature of battery data. Due to the nature of data-driven models, the selection of input factors (in this case, battery health indicators) significantly affects the prediction accuracy of the models. Therefore, this review analyzes the selection of battery health indicators for forecasting State of Health and Remaining Useful Life of a battery in a three-step manner: (i) analyzing recent trends in battery health indicators, (ii) identifying the challenges faced in selecting suitable battery health indicators, and (iii) proposing future directions to address these challenges. By synthesizing the current challenges and future prospects, this review paper provides researchers and engineers with a roadmap for advancing battery health management and enabling the widespread adoption of battery technologies in a sustainable and efficient manner.
Soon et al. (Wed,) studied this question.