Robustness and Reliability of Model-based Sensor Data Fusion in a Lithium-Ion Battery System

Abstract

In recent years, battery monitoring systems with sensors on cell level have been introduced to enhance robustness and safety. With model-based sensor data fusion the uncertainty of cell voltage and current measurement is diminished. Within this contribution approaches are explored to strengthen the robustness of the sensor data fusion method. In particular, two Kalman filters are investigated that are based on Student's t instead of Gaussian noise. Furthermore, the underlying parameter estimation is improved by adaption of measurement noise and taking the estimator windup into account. Simulation results show that each of the presented methods increases the robustness of the sensor data fusion framework and may be combined for best performance. Moreover, with sensor data fusion the reliability of each cell measurement is enhanced, which is also investigated within this contribution.