Improved particle swarm optimization–long short-term memory model with temperature compensation ability for the accurate state of charge estimation of lithium-ion batteries

The nonlinearity and time-varying features of lithium-ion batteries, including temperature, pose a challenge for accurate state of charge (SOC) estimation. The SOC estimation of lithium-ion batteries is tackled in this paper using an improved particle swarm optimization-long short-term memory (IPSO-LSTM) model with temperature compensation ability that combines the advantages of the two algorithms, the PSO and LSTM, towards robust modeling and enhanced SOC estimation accuracy. The PSO algorithm optimizes the hyperparameters before input into the improved LSTM model for accurate SOC estimation. Experimental datasets are obtained from two ternary lithium-ion batteries for model training, testing, and validation at various temperatures under complex working conditions. According to the results, the proposed hybrid model achieves faster convergence and more accurate results in a shorter time compared to the conventional LSTM. The optimal values for the root mean square error, mean absolute error, and mean absolute percentage error for the LSTM are 0.308%, 0.189%, and 0.242%, and those for the proposed IPSO-LSTM model are 0.063%, 0.041%, and 0.048%, respectively. These outcomes demonstrate that the proposed model performs better compared to the LSTM and other existing methods in terms of accuracy and robustness for reliable SOC estimation of lithium-ion batteries.