An intelligent fusion estimation method for state of charge estimation of lithium-ion batteries

Cheng, Xingqun; Liu, Xiaolong; Li, Xinxin; Yu, Quanqing

doi:10.1016/j.energy.2023.129462

Cited by 11 publications

(1 citation statement)

References 41 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Additionally, advanced techniques like transfer learning and model compression can be employed to mitigate the impact of increased complexity in certain scenarios. In many studies, the combined methods consist of deep learning neural networks and physical model-based method with filter design for initial SOC identification and real time estimation, respectively [28][29][30][31]40]. These approaches not only provide accurate SOC estimation but also overcome the estimation errors during the initial stages of operation.…”

Section: Introductionmentioning

confidence: 99%

An Improved AhI Method With Deep Learning Networks for State of Charge Estimation of Lithium-Ion Battery

Chang,

Kung

2024

IEEE Access

View full text Add to dashboard Cite

Accurately estimating the state of charge of lithium-ion batteries has become a crucial challenge in the development of battery management systems within the electric vehicle industry. To address this issue, numerous studies have explored various solutions, which can be categorized into two main approaches: 1) battery model derivation coupled with filter design, and 2) data collection for battery system identification. Ensuring high accuracy and real-time applicability are vital requirements for successful battery management systems development. The paper introduces a novel approach for estimating the state of charge, which includes two key components. The first one is the neural network utilization: this involves a neural network architecture that incorporates a one-dimensional convolutional neural network and long short-term memory and a multi-layer perceptron. The second one is the Ampere-hour integration compensation: it employs Ampere-hour integration to refine the predicted state of charge estimates obtained from the neural network, thereby improve the accuracy of the estimations. To validate the effectiveness of the proposed method, the researchers applied it to well-established lithium-ion battery datasets from the University of Maryland. The experimental results, when compared with previous studies, demonstrate that the proposed method significantly improves the accuracy of state of charge prediction across various environmental and operational conditions.

show abstract