Abstract
In order to improve the estimation level of lithium batteries and promote the accurate control of the battery management system, accurate state of charge (SOC) estimation is very important. The CNN algorithm and the two-dimensional CNN (2DCNN) algorithm have been studied in the SOC estimation, but it is a technical difficulty to apply the three-dimensional CNN (3DCNN) algorithm to the SOC estimation. This paper firstly designs two-dimensional and three-dimensional datasets to describe the aging degree and SOC. The time and space dimensions of the three-dimensional dataset are used to memorize the short-term data and the long-term of the battery. Then, this paper proposes a fused convolutional neural network (FCNN) algorithm, which consists of two layers of neural networks in series. The FCNN algorithm can consider the aging degree of the battery, and is based on the definition of the SOC estimation. The results show that the fused 3DCNN has advantage over the 2DCNN in battery capacity estimation. In addition, the FCNN algorithm considering the battery capacity can improve the SOC estimation accuracy, and has also been verified by the comparison of the mean absolute percentage error.
Funder
National Natural Science Foundation of China
the Qinglan Project of Jiangsu Province of China
Publisher
The Electrochemical Society
Subject
Materials Chemistry,Electrochemistry,Surfaces, Coatings and Films,Condensed Matter Physics,Renewable Energy, Sustainability and the Environment,Electronic, Optical and Magnetic Materials
Cited by
5 articles.
订阅此论文施引文献
订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献