神经网络模型在锂离子电池表面温度预测中的应用研究  

Application of Neural Network Model to Predicting Surface Temperature of Lithium-Ion Battery

作　　者：陈实[1] 方凯正[1] 穆道斌[1] 吴伯荣[1] 

机构地区：[1]北京理工大学化工与环境学院,北京100081

出　　处：《北京理工大学学报》2013年第4期421-424,共4页Transactions of Beijing Institute of Technology

基　　金：国家"九七三"计划项目(2009CB220100);国家科技部前沿技术研究项目(2010DFA72760)

摘　　要：基于人工神经网络构建了锂离子电池表面温度的预测模型.该模型为3层网络结构,其中输入层中有4个节点,隐含层中有9个神经元,输出层中有1个节点.训练结果表明,模型具有较快的收敛速度和优秀的训练质量,从而保证了预测的精确度.模型的预测值与实验值吻合程度高,说明了模型工作的有效性.模型预测电池在较高环境温度(80℃)下以10C倍率放电结束时的表面温度为86.71℃,仅比环境温度高出6.71℃.该模型有助于电池热管理系统的研究与开发.The surface temperature of lithium-ion battery during charging/discharging is predicted by a model based on artificial neural network （ANN）. The model is a three layers network, and there are four nodes in input layer, nine neurons in hidden layer and one node in output layer. Training results show that the model is of fast convergence and excellence training quality, which guarantees the prediction accuracy of surface temperature. The results of predicted temperature accord well with the experimental data, indicating that the constructed model is effective. Under ambient temperature of 80 ℃, battery＇s surface temperature is predicted as 86.71 ℃ at the end of 10C magnifying rate discharging, only 6.71℃ higher than the ambient temperature. The presented model could facilitate the research and development of battery thermal management system.

关 键 词：锂离子电池 神经网络模型 电池表面温度 预测 环境温度 

分 类 号：TM911[电气工程—电力电子与电力传动]