Capacity Estimation Method of Lithium Battery Based on Dynamic Working Condition Data Image and Deep Learning
毕贵红 1黄泽 1谢旭 2张文英 1骆钊1
扫码查看
点击上方二维码区域，可以放大扫码查看
作者信息
1. 昆明理工大学电力工程学院,昆明650500
2. 华能澜沧江水电股份有限公司糯扎渡水电厂,普洱665000
折叠
摘要
针对实际应用中基于动态工况下电池状态参数的片段数据进行电池健康状态(state of health,SOH)实时估计的问题,提出基于动态工况下锂离子电池状态参数(电压、电流、温度)实测数据二维特征图像和深度学习的锂离子电池容量估计算法.首先,将动态工况下电池状态参数监测量(电压、电流和温度)的片段数据转化为二维特征图像.其次,提出基于残差卷积神经网络(residual convolutional neural network,Res-CNN)和门控循环单元(gate recurrent unit,GRU)网络结合的多通道深度学习模型Res-CNN-GRU,以构建动态工况下电池状态参数特征图像和SOH之间的复杂非线性关系,其中电压、电流和温度的二维特征图像以三通道的方式输入到Res-CNN-GRU模型中,模型输出为对应电池的相邻参考充放电循环实验所获得容量的差值.研究结果表明:此方法在锂电池随机充放电工况下对电池健康状态估计效果更佳,且Res-CNN-GRU模型的泛化性和全局特征提取能力较强.论文研究为现实工况下电池健康状态估计的进一步深入研究提供了参考.
Abstract
Aiming at the problem of real-time estimation of battery state of health (SOH) based on the fragment data of battery state parameters in dynamic working conditions, we proposed a capacity estimation algorithm of lithium-ion bat-tery based on two-dimensional feature image and deep learning of measured data of state parameters (voltage, current and temperature) of lithium-ion battery in dynamic working conditions. Firstly, the fragment data of the monitoring quantity of battery state parameters (voltage, current and temperature) under dynamic working conditions were converted into two-dimensional feature images. Secondly, a multi-channel deep learning model Res-CNN-GRU based on the combina-tion of residual convolutional neural network (Res-CNN) and gate current unit (GRU) network was proposed to build the complex nonlinear relationship between battery state parameter characteristic image and SOH under dynamic working conditions. Finally, two-dimensional characteristic images of voltage, current and temperature were input into the Res-CNN-GRU model in the way of three-channel, and the model output was the difference of the capacity obtained from the adjacent reference charge-discharge cycle experiments of the corresponding battery. The results show that this method is more effective in estimating the battery health state under random charging and discharging conditions of Li-ion batter-ies, and the generalization and global feature extraction ability of the Res-CNN-GRU model is better. This study provides a reference for further research on battery health state estimation under realistic working conditions.
维普
万方数据