Battery Safety: Machine Learning-Based Prognostics

Abstract

Lithium-ion batteries play a pivotal role in a wide range of applications, from electronic devices to large-scale electrified transportation systems and grid-scale energy storage. Nevertheless, they are vulnerable to both progressive aging and unexpected failures, which can result in catastrophic events such as explosions or fires. Given their expanding global presence, the safety of these batteries and potential hazards from serious malfunctions are now major public concerns. Over the past decade, scholars and industry experts are intensively exploring methods to monitor battery safety, spanning from materials to cell, pack and system levels and across various spectral, spatial, and temporal scopes. In this Review, we start by summarizing the mechanisms and nature of battery failures. Following this, we explore the intricacies in predicting battery system evolution and delve into the specialized knowledge essential for data-driven, machine learning models. We offer an exhaustive review spotlighting the latest strides in battery fault diagnosis and failure prognosis via an array of machine learning approaches. Our discussion encompasses: (1) supervised and reinforcement learning integrated with battery models, apt for predicting faults/failures and probing into failure causes and safety protocols at the cell level; (2) unsupervised, semi-supervised, and self-supervised learning, advantageous for harnessing vast data sets from battery modules/packs; (3) few-shot learning tailored for gleaning insights from scarce examples, alongside physics-informed machine learning to bolster model generalization and optimize training in data-scarce settings. We conclude by casting light on the prospective horizons of comprehensive, real-world battery prognostics and management.