SOC estimation method for lithium-ion batteries considering pressure signals

Accurately estimating the State of Charge (SOC) is a crucial measure to prolong the battery's service life, which in turn helps extend the driving range of vehicles. Therefore, how to precisely and quickly estimate SOC has become a topic of extensive research. This paper employs various signals, including stress, as estimation signals to assess the battery's SOC. The lithium iron phosphate battery is selected as the research subject. Its characteristics are analyzed, and electrochemical and stress modeling are conducted on the battery. A coupling relationship between stress and the battery is established, exploring the impact of stress on the battery. Using data collected from condition tests, an LSTM neural network is employed, with voltage, current, temperature, and stress as inputs and SOC as the output. SOC is calibrated through ampere-hour integration. With 90% of the BJDST condition experimental data serving as the training set for the lithium iron phosphate battery and 10% as the test set, high-precision estimation of SOC is achieved, with a Root Mean Squared Error (RMSE) of 0.627%, Mean Absolute Error (MAE) of 0.502%, and Mean Error Squared (ME) of 2.321%.