基于RSN-GRU融合网络的锂电池荷电状态估计

State of charge estimation of lithium battery based on RSN-GRU fusion network

为了提高锂电池荷电状态(SOC)估计的精度,提出一种基于残差收缩网络(RSN)与门控循环单元(GRU)并行融合的锂电池SOC估计方法.RSN提取锂电池参数输入序列的局部特征并通过一个小型子网络去除多余的噪声,GRU提取锂电池多参数输入序列的时序信息,最终将RSN和GRU进行并行融合得到锂电池的SOC估计值.多种动态工况及不同温度下的实验结果表明:RSN-GRU并行融合网络能显著提高锂电池SOC估计精度,在25℃环境温度下估计结果的平均绝对误差为0.34%,均方根误差为0.51%,比单独的GRU和RSN估计精度分别提升了50%和61.7%.另外,将RSN-GRU与其他多种常用网络进行了对比,结果表明该网络SOC估计精度更高,具有明显的优越性.

To improve the accuracy of state of charge(SOC)estimation of lithium batteries,a parallel fusion network of residual shrinking network(RSN)and gated recurrent unit(GRU)was proposed for SOC estimation of lithium batteries.RSN extracts local features of input sequence of lithium battery and removes noise through a small sub-network.Meanwhile,GRU extracts history information of input sequence,and finally RSN and GRU are fused in parallel to obtain the SOC estimate of lithium battery.The experimental results under various dynamic working conditions and different temperatures conditions show that the RSN-GRU parallel fusion network can significantly improve the SOC estimation accuracy of lithium batteries.The mean absolute error(MAE)and root mean square error(RMSE)of the estimation results at 25℃are 0.34%and 0.51%,respectively.Compared with GRU and RSN,the estimation accuracy is improved by 50%and 61.7%,respectively.In addition,the results of comparison between RSNGRU and other commonly used networks show that the SOC estimation accuracy of this network is higher than others and it has obvious superiority.