Enhancing Battery State of Charge Estimation Through a Combination of Extended Kalman Filter and Artificial Neural Network
摘要
In the field of battery management and state estimation, accurately determining the initial state of charge (SOC) is a necessary yet difficult task. The Extended Kalman Filter (EKF) is an effective technique for estimating the battery state of charge; nevertheless, its accuracy can be considerably impacted by the initial battery charge state (initSOC), especially if it differs significantly from the real state of charge. This disparity has the potential to cause problems such as sluggish convergence or, in more extreme cases, full divergence in the estimation process. To solve this, we propose an innovative approach in this paper that uses 1D Convolutional Neural Networks (1D CNN) to improve the EKF’s initial state estimate.