A dataset of Early fault diagnosis of electric vehicle batteries based on physics-informed DeepSeekV3 architecture 核心 · 已核验

mxdstz409feb8v19

Accurate early warning of battery failure is essential to ensure the safety of electric vehicles (EVs). However, the scarcity of early fault samples, the difficulty in capturing weak features, and the fact that pure data - driven models are prone to violate physical laws still pose challenges for early fault diagnosis. To address these issues, an unsupervised joint diagnosis framework combining generative pre-trained transformer and weighted one-class support vector machine (GPT-WOCSVM) is proposed for the first time in this paper. The charge data GPT (Charge-dataGPT) is constructed based on the physics-informed DeepSeekV3 architecture. It can use the charging data of the first 10 minutes to generate the charging sequence for the next 20 minutes in accordance with the electrochemical law through autoregression; WOCSVM is trained only with normal data, and early fault recognition is realized by judging whether the generated sequence deviates from the normal mode.

This dataset contains the implementation code and model of GPT-WOCSVM.

本卡描述如含来自官方页的原文片段,其版权归原作者,不在本站 CC-BY 4.0 许可范围(见关于与许可)。

落地页

https://data.mendeley.com/datasets/b9tryr388p/2

许可证

CC-BY-4.0 (置信:verified_official)

国内可访问性

国内直连:可达 (2026-07-11 检测) 非直连:可达 (2026-07-11 检测)
「直连超时」表示检测窗口内未完成,系慢或不稳定证据,不构成封锁证据。

发布年份

2026

发布方

Mendeley Data

设备类型

battery

PHM 任务

fault_diagnosis

故障工况