A New Update Rule of RLSEKF-based Joint-estimation Filters for Real-time SOH SOC Identification

In order to accurately estimate the SOC and SOH of a lithium-ion battery used in an electric vehicle (EV), we propose an Adaptive Diagonal Forgetting Factor Recursive Least Square (ADFF-RLS) for accurate battery parameter estimation. ADFFRLS includes two new proposals in the existing DFF-RLS; The first is an excitation tag that changes the behavior of the DFFRLS and the EKF according to the dynamics of the input data. The second is auto-tuning that automatically finds the optimal value of RLS forgetting factor based on condition number (CN). Based on this, we proposed a joint estimation algorithm of ADFF-RLS and Extended Kalman Filter (EKF). To verify the accuracy of the proposed algorithm, we used experimental data of hybrid pattern battery cells mixed with dynamic and static patterns. In addition, we added a current measurement error that occurs when measuring at EV, and realized data that is closer to actual environment. This data was applied to two conventional estimation algorithms (Coulomb counting, Single EKF), two joint estimation algorithms (RLS & EKF, DFF-RLS & EKF) and ADFF-RLS & EKF. As a result, the proposed algorithm showed higher SOC and SOH estimation accuracy in various driving patterns and actual EV driving environment than previous studies.