Digital Twin Empowered Dynamic State of Health Prediction with Extended Kalman Filter for Lithium-ion Batteries

The state of health (SOH) estimation of Li-ion batteries has received extensive attention to ensure the safety and reliability of power supply systems for new energy vehicles. The traditional data-driven approach for Li-ion batteries SOH estimation requires the complete discharge sequence as input. However, the complete sequence is impossible to obtain in practical applications. In this work, a real-time SOH estimation with a data-driven way based on digital twins is proposed. The digital twin model uses an extended Kalman filter (EKF) to reconstruct the real-time partial discharge sequence into a complete discharge sequence by assigning the weight between model predictions and the observations from the sensor. The output of EKF is then fed into a long short-term memory network with the discrete wavelet transform to obtain SOH estimation. Verified on the NASA dataset, more than 95% accuracy can be achieved with only 30% of the discharge data obtained.