State of charge estimation by multi-innovation unscented Kalman filter for vehicular applications

Vehicle to grid technology is no longer a fiction but rather a reality. Due to recent technological advances in bidirectional power transfer, EVs could serve not just as transportation tools but also as electric storage units for the grid. As a result, battery management is now more crucial than ever to control the energy transfer to and from the battery pack, while keeping all battery parameters within a safe and optimal region. In order to do so, accurate knowledge of SOC is of significant importance, since it reflects the inner state of the battery. This paper proposes, a multi-innovation theory to enhance the estimation accuracy of the unscented Kalman filter. By expanding a single innovation voltage value to multi-innovations consisting of the previous and current values of the battery's output, the accuracy of the UKF is drastically improved. All the design aspects of the proposed MIUKF are detailed. Moreover, a hybrid Levenberg Marquardt approach is implemented for battery internal parameter identification. Finally, the experimental results indicate that the proposed MI-UKF is robust against unpredicted operational conditions, and it can enhance the accuracy of the UKF with more than 1%.