Accurate state of health (SOH) is a key indicator to evaluate the lifespan of lithium-ion batteries (LIBs). However, complex nonlinear capacity degradation and unmeasurable internal chemical reactions pose great challenges to accurate SOH estimation of LIBs. Therefore, this article proposed a novel bidirectional long short-term memory (BiLSTM) data-driven model to further improve the precision of battery SOH prediction. First, The incremental capacity (IC) curve was obtained by the battery aging cycle test. Subsequently, the health factors related to SOH were extracted from the IC curve. Moreover, the health indicators are taken as the input to the BiLSTM model, and the calculated SOH is the output of the model. Finally, the BiLSTM model was trained by the CS2-36 battery dataset provided by the CALCE team and the model prediction accuracy was verified by another CS2-37 battery dataset. The results indicate that the mean absolute error (MAE) of the proposed BiLSTM data-driven model is kept within 1%. Moreover, the root mean square error (RMSE) was reduced by 18.4% and 48.8%, respectively, compared with the gated recurrent unit (GRU) and LSTM model.
An accurate BiLSTM data-driven model for SOH prediction of lithium batteries based on health indicators
2024-10-10
1158747 byte
Conference paper
Electronic Resource
English