Uncertainty management in data-driven state estimation: Taxonomy, methods, and battery applications

Gwanpil Kim

doi:10.1016/j.inffus.2026.104642

Uncertainty management in data-driven state estimation: Taxonomy, methods, and battery applications

2026-07-21

0

PRE

AI

DOI：10.1016/j.inffus.2026.104642

Original

Original request for help

Save

Abstract

En 中文

• The survey organizes uncertainty along two axes, its source (Aleatoric and Epistemic) and its management approach (memory-based and model-based). • Memory-based approaches mitigate noise, irregularity, and incomplete data before modeling, whereas model-based approaches quantify, propagate, and calibrate predictive uncertainty within the model. • The dominant uncertainty source shifts along the prediction horizon, from Aleatoric-dominated current-state estimation (SOC and SOH) toward Epistemic-dominated future-state prediction (RUL). • Open problems and research directions are identified for trustworthy predictive maintenance beyond batteries.

Keywords:

Uncertainty management

Battery state estimation

Aleatoric and epistemic uncertainty

Battery management systems

Uncertainty management in data-driven state estimation: Taxonomy, methods, and battery applications

Abstract

Journal

Organization