.png){kind=icon}
.png){kind=icon}
Return
A CMMOG-based lithium-battery SOH estimation method using multi-task learning framework
.png){kind=icon}
PRE
AI
DOI：10.1016/j.est.2024.114884.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Accurately diagnosing the state of health (SOH) of lithium-ion batteries (LIBs) is essential for effective battery management systems (BMS). However, the prevailing deep learning-based state estimation methods necessitate the recalibration of model weights under varying operating conditions or across different battery samples. This requirement results in significant consumption of computational resources and diminished generalization capabilities. To address these challenges, this paper proposes a multi-task learning (MTL) framework based on a Convolutional Neural Network-Multi-gate Mixture of Gated Recurrent Units (CMMOG), which is capable of concurrently managing multiple SOH estimation regression tasks. The framework integrates high-dimensional health factor information through the convolutional neural network (CNN), maps the health states using GRU, and orchestrates the weights through a multi-gated network (MMN). Furthermore, it incorporates incremental capacity analysis and fuzzy entropy into the feature extraction process to enhance the robustness of the model. Experimental results demonstrate that the proposed model achieves faster convergence, with an average reduction of 16.95 % in computation time, while maintaining at least the same level of accuracy as single-task learning methods. Additionally, the average accuracy improves by 39.92 % in generalization experiments involving non-training samples.
Keywords:
Battery health state estimation
CMMOG
HIs extract
MTL
Homoscedastic uncertainty
Journal
IF:
9.8
Papers:
2.2W
Citations:
10.1W
Organization
No organization information available