During long-term operation

the complex chemical reactions and physical processes of Li-ion batteries (LIBs) may induce faults

thereby threatening system safety. Therefore

it is crucial to accurately and efficiently detect and locate early faults in battery packs. This paper proposes a method for early fault detection and localization of LIB packs based on improved local gravitation analysis. Firstly

based on the local gravity characteristics of the battery cell

this study proposes the temperature abnormality index (TAI) to quantitatively evaluate the temperature field inconsistency of the battery system. Subsequently

this paper adaptively caclulates the threshold of the TAI using non-parametric kernel density estimation (KDE) to achieve rapid fault detection

and designs the local gravitation contribution function to achieve accurate localization of the faulty cell. This paper designs 9 groups of internal short circuit faults with different parameters on a 6s4p lithium-ion battery pack to experimentally verify the proposed method. Experimental results show that this method can quickly detect and accurately locate early internal short circuit faults

with a false alarm ratio no more than 1.5%

a detection delay of no more than 90s

and a fault location accuracy of up to 100%.