State of Health estimation of lithium-ion batteries based on energy features and temporal convolutional network

Lithium-ion batteries (LIBs) need to maintain high energy efficiency or power level in several application scenario. Accurate state of health (SOH) forecast is essential for designing a safe and reliable battery management systems (BMS). In recent years, the flexibility and adaptability of data-driven approach in SOH prediction have been demonstrated, which poses a challenge for the health indicators (HIs) selection. In contrast to prior approaches for feature extraction, a multi-category HIs fusion strategy with a wide range of applications, which combines the energy features of discharging voltage curves and constant current/constant voltage (CCCV) charging curves, is presented in our article. Specifically, a temporal convolutional network (TCN) model is proposed to estimate SOH of LIBs using energy-based features extracted within equal charge voltage interval (ECVI) and equal discharge voltage interval (EDVI). On this basis, we carry out experiments on the aging data of CALCE public datasets. The experimental results systematically validate the superiority of the proposed method, which covers high estimation accuracy and satisfied universality to different battery.