Multi-model deep learning-based state of charge estimation for shipboard lithium batteries with feature extraction and Spatio-temporal dependency

The state of charge (SOC) is a key indicator for measuring the remaining battery capacity. However, the different charging and discharging behaviors, as well as the topology connections, may lead to undesirable estimation errors, affecting the battery output performance. This paper proposes a novel...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Journal of power sources Jg. 629; S. 235983
Hauptverfasser: Qiu, Yanxi, Wen, Shuli, Zhao, Qiang, Zhang, Bing, Wu, Guodong, Zhu, Miao
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Elsevier B.V 15.02.2025
Schlagworte:
Electric ship
Shipboard lithium battery
State of charge estimation
STGCN-Transformer
State of charge estimation
Shipboard lithium battery
STGCN-Transformer
Electric ship
ISSN:0378-7753
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:The state of charge (SOC) is a key indicator for measuring the remaining battery capacity. However, the different charging and discharging behaviors, as well as the topology connections, may lead to undesirable estimation errors, affecting the battery output performance. This paper proposes a novel method to estimate the SOC of shipboard lithium batteries. This method combines spatial-temporal graph convolutional networks (STGCNs) with Transformer deep learning networks. Furthermore, the STGCN is utilized to construct graph structures, simulating the complex interconnectivity of battery energy storage systems. The Transformer is used for estimating the SOC sequence windows. Additionally, a recurrent neural network is employed as an additional encoder and decoder architecture to enhance multiscale time series processing and capture the nonlinear battery characteristics to reduce fluctuations in the SOC. A principal component analysis is used to eliminate redundant battery features to improve the computational efficiency of the estimation model further. The experimental results indicate that the proposed deep learning network model can effectively estimate the SOC of shipboard lithium batteries with high accuracy and stability. •Proposed an ensemble deep learning framework for accurate battery SOC estimation.•Developed an STGCN-based method for shipboard SOC estimation with pack connections.•Integrated long and short time series data to enhance model performance.
ISSN:0378-7753
DOI:10.1016/j.jpowsour.2024.235983