A State of Health Estimation Method for Lithium-ion Power Batteries Based on Incremental Capacity Curves and Transfer Learning

The estimation of the State of Health (SOH) of lithium-ion batteries is crucial in Battery Management Systems (BMS). Data-driven approaches are one of the effective methods for SOH estimation. However, the complex cycling conditions of batteries and limited data, combined with the fragmented nature of data collection due to real working conditions, make estimating the SOH of lithium-ion batteries a challenging task. To address these issues, an estimation method based on partial incremental capacity curves and transfer learning is proposed. First, only a portion of the voltage segments is used for incremental capacity analysis, which is then input into a Stacked Bidirectional Gated Recurrent Unit (SBiGRU) network. Finally, transfer learning is employed to address the challenges posed by limited data and differing data distributions. The proposed method is further enhanced with hyperparameter optimization to improve estimation accuracy. The accuracy of the proposed method is validated on two public datasets. Experimental results indicate that SOH can be predicted effectively using only partial incremental capacity curves, and the method also performs well when training and testing with different voltage ranges. These results highlight the promise of the proposed SOH estimation method for lithium-ion batteries.