Digital twin for electric vehicle battery management with incremental learning

The current Industry 4.0 revolution promotes the use of cyber-physical systems to enhance manufacturing and other industrial processes via automation, real-time analysis, etc. Data communication between individual systems plays an important role in this revolution's success. As defined by researchers, Digital Twin is the digital representation of a physical system that enables predictive maintenance. Due to the increase in environmental pollution, battery-powered electric vehicles (EVs) are regarded as the urgent solution to internal combustion engines in the transportation business, despite obstacles such as safety concerns and range estimation. State of Health (SoH) and State of Charge (SoC) are two battery metrics that, when precisely anticipated, permit safer and longer battery use. Predicting these parameters online is computationally and financially expensive. Alternately, some of these factors could be predicted in the cloud rather than on the vehicle, hence cutting costs. Consequently, the EV business is one example where cloud-to-vehicle data connection saves total costs. A digital twin for an EV battery would aid in the estimate of battery parameters for predictive maintenance. This paper presents a Digital Twin paradigm for EV battery management in which SoH is predicted in the cloud and SoC is estimated on-vehicle. A continuous learning method is also proposed for forecasting SoH, whereas the Kalman filter is used to estimate SoC. The proposed framework predicts the SoH with a mean square error of 0.022.