The modeling and design of an active battery cell balancing system using Multilevel Converter (MLC) for EV/HEV/PHEV is studied under unidirectional as well as reciprocating air flow. The MLC allows to independently switch ON/OFF each battery cell in a battery pack. The optimal policy (OP ) exploiting this extra degree-of-freedom can achieve both temperature and state-of-charge (SoC) balancing among the cells. The OP is calculated as the solution to a convex optimization problem based on the assumption of perfect state information and future driving. This study has shown that OP gives significant benefit in terms of reduction in temperature and SoC deviations, especially under parameter variations, compared to uniformly using all the cells. It is also shown that using reciprocating flow for OP gives no significant benefit. Thus, reciprocating flow is redundant for MLC-based active cell balancing system when operated using OP.