基于元素注意门控复用的锂离子电池荷电状态估计

为了提高荷电状态(state-of-charge,SOC)估计精度，提出一种基于元素注意门的电池荷电状态递归神经网络，为输入向量的每个特征元素分配不同的重要程度，验证并分析不同神经元数量和隐藏层层数下的测试结果，利用确定的最优参数设置进行不同温度下的电池SOC估算，在不同电池特征参数下对SOC估计任务的重要性进行可视化分析。相同数据集的SOC估计精度表明，提出的网络模型在SOC估计任务中精度有明显提升。

State-of-charge estimation of lithium-ion batteries based on gated recurrent unit with element-wise-attention gate

To enhance the precision of state-of-charge (SOC) estimation for lithium-ion batteries, we introduce a novel recurrent neural network leveraging element attention gates. Different importance levels are assigned to each feature element of the input vector. The test results under different numbers of neurons and hidden layers are verified and analyzed. The optimal parameter settings determined are used to perform SOC estimation under different temperatures. Visualization analysis is conducted on the importance of SOC estimation tasks under different battery feature parameters. The SOC estimation accuracy of the same dataset shows that the proposed network model has significantly improved accuracy in SOC estimation tasks.