基于RSN-GRU融合网络的锂电池荷电状态估计OA北大核心CSTPCD

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 RSN-GRU and other commonly used networks show that the SOC estimation accuracy of this network is higher than others and it has obvious superiority.