过度拟合
自编码
计算机科学
一般化
电池(电)
人工神经网络
深度学习
人工智能
噪音(视频)
机器学习
数据挖掘
数学
数学分析
功率(物理)
物理
量子力学
图像(数学)
作者
Marui Li,Chaoyu Dong,Xiangke Li,Xisong Dong,Huaguang Zhang,Hongjie Jia
标识
DOI:10.1109/ecce50734.2022.9947649
摘要
The lithium-ion battery is an important energy storage means, so the monitoring of its state is very important. In particular, temperature variations of lithium-ion batteries seriously affect their performance and safety. It is necessary to predict the temperature change in advance to take corresponding strategies to prevent the danger occurrence. Both long and short-term memory networks and temporal convolutional networks are effective methods to deal with sequence regression problems. However, the method based on deep learning will reduce its generalization ability when there is overfitting or data change. Therefore, a sequential network-model alliance module (SNMAM) is established in this paper, which combines a thermal model and advanced neural network to improve generalization ability and prediction accuracy. In addition, SNMAM applies an denoising autoencoder to reduce the impact of measurement noise. Complete experiments are presented to verify the effectiveness of the SNMAM designed in the paper.
科研通智能强力驱动
Strongly Powered by AbleSci AI