State-of-health estimation of lithium-ion battery based on convolutional-gated recurrent neural network with self-attention mechanism

The state of health (SOH) of lithium-ion batteries is a critical parameter of the battery management system. An accurate estimation of it can improve the battery's overall life. The data-driven method has the problems of low long-term prediction accuracy and difficult feature recognition. Therefore, this paper proposes a SOH estimation method based on the convolutional gated recurrent neural network with the self-attention mechanism. Firstly, the convolutional neural network (CNN) is applied to the input data, and its convolution operation is used to extract important local features. Then, the self-attention mechanism is added later to give greater weight to the more important features in the training process. Finally, the gated recurrent unit (GRU) recurrent neural network extracts the long-term correlation from the weighted output data. The experimental data of lithium-ion batteries from three sources are used to verify the correctness and effectiveness of this method. Experiments were based on whether the self-attention mechanism is present, different prediction starting points and ambient temperature, and analyzed and compared with the common machine learning methods. The experimental results show that the method exhibits good estimation accuracy, with a maximum mean absolute error below 1.78% and a maximum root mean square error below 2.3%.