Estimation of SoH and internal resistances of Lithium ion battery based on LSTM network

State of Health (SoH) and internal resistances, including the solid electrolyte interphase (SEI) resistance and charge transfer resistance, are important parameters that change in the long-term representation of the aging state of Lithium-ion batteries. Using long short-term memory (LSTM) network, a neural network with the ability to remember long-term data features, this paper presents a method for estimating SoH and internal resistances of Lithium-ion batteries using LSTM network with deep learning mechanism. Based on experimental data including voltage, current, temperature with 03 charge/discharge scenarios and measuring impedance, input/output data structure is set up to reflect aging features used for estimating SoH and internal resistances by LSTM. The first LSTM network is designed to estimate SoH, then the data including current, voltage, temperature and estimated SoH will be used to estimate the SEI resistance and charge transfer resistance by the second LSTM network. With this structure, the estimation of internal resistances in practice will become simpler as it does not require measuring capacity and impedance spectroscopy. Comparing the estimation errors using LSTM and FNN with 03 performance metrics including mean absolute percentage error (MAPE), mean percentage error (MPE) and root mean square error (RMSE) shows that the estimation results of SoH and internal resistances of the cell by LSTM have higher accuracy than the estimation by Feedforward Neural Network (FNN).