“…In addition, there are other related studies on SoC estimation, such as equivalent circuit model (ECM)-based estimation with noise compensation [46], OCV error compensation based on DNN [47], the open circuit voltage-charge amount (OCV-Q) curve fitting method using a convolutional neural network (CNN) [48], the event-driven Coulomb counting method (CCM) algorithm for unbalanced SoCs [49], and CCM based on modified parameters [50]; however, DNN-and KF-based methods require high computational power and an additional learning process. The OCV and CCM are primarily used to indicate the charging state of a battery [51,52]; however, because OCV is used when the internal battery state is stabilized, it is not sufficiently stable for a nonlinear battery [9]. Furthermore, because another CCM calculates the SoC by accumulating the charge current, the CCM has the disadvantage of increasing the SoC if an error occurs in the initial current measurement value [53].…”