基于卷积-双向长短期记忆网络的电池SOC预测  

Battery SOC prediction based on convolution bidirectional long term and short term memory network

作　　者：陈继斌 李雯雯 孙彦玺 许静 张单 CHEN Jibin;LI Wenwen;SUN Yanxi;XU Jing;ZHANG Dan(College of Building Environmental Engineering,Zhengzhou University of Light Industry,Zhengzhou Henan 451002,China)

机构地区：[1]郑州轻工业大学建筑环境工程学院,河南郑州451002

出　　处：《电源技术》2022年第5期532-535,共4页Chinese Journal of Power Sources

基　　金：国家自然科学基金青年基金(51804278)。

摘　　要：锂电池的预测性维护是电池应用的重点,实现它的关键是有效地预测锂电池的荷电状态(SOC)。随着信息技术和深度学习网络算法的发展,深度学习法在SOC预测方面显示出很好的潜力。提出了一种基于卷积-双向长短期记忆网络(CNN-BiLSTM)的SOC预测方法,利用卷积神经网络(CNN)模型提取局部特征,双向长短期记忆网络(BiLSTM)充分利用双向SOC的时间信息。在公共数据集上的对比实验表明,该模型提高了SOC预测的准确性。Predictive maintenance of lithium battery is the focus of battery application,and the key to achieve it is to effectively predict the state of charge(SOC)of lithium battery.With the development of information technology and deep learning network algorithm,deep learning method shows great potential in SOC prediction.A new SOC prediction method based on convolution bi-directional long-term and short-term memory network(CNN BiLSTM)was proposed.Convolution neural network(CNN)model was used to extract local features,and bi-directional longterm and short-term memory network(BiLSTM)made full use of the time information of bi-directional SOC.The experimental results on public data sets show that the model can effectively improve the accuracy of SOC prediction.

关 键 词：荷电状态 卷积神经网络 双向长短期记忆 锂电池 深度学习 

分 类 号：TM912[电气工程—电力电子与电力传动]