A temporal convolution and gated recurrent unit network with attention for state of charge estimation of lithium-ion batteries

The state of charge (SOC) is one of the most critical indicators of lithium-ion batteries, which can directly reflect the remaining power of the battery. It is related to a variety of external factors and has characteristics such as time-varying and non-linear. Traditional machine learning models cannot capture the highly complex dynamic characteristics of batteries for SOC estimation. In this work, a deep learning model is proposed to estimate the SOC of lithium-ion batteries, which combines the advantages of temporal convolutional network, gated recurrent unit network, and attention mechanism. The model combines temporal convolutional networks with gated recurrent unit networks to extract temporal information from input time series over a longer time scale. The attention mechanism is added to the network, which can view all hidden states from the network sequence, and assign weight values to highlight important information. The model is evaluated using lithium-ion battery data under various driving conditions, and the results showed that the mean absolute error and root mean square error were less than 0.3 % and 1 % under different test and training conditions. To further evaluate the performance of the model, we conduct experiments on two public datasets and compare the proposed model with various fundamental deep learning methods. The experimental results show that the proposed hybrid model has the best estimation accuracy and robustness for different temperatures and different battery types.