Online Learning and Management of Battery Systems

PI: Nicolas Bambos; Co-PI: Ram Rajagopal; Student: Ilai Bistritz

With better management of the battery pack at the heart of each electric Vehicle (EV), we can both improve the range and the health of every EV. The main task of the battery management system (BMS) is to balance the state of charge (SOC) of the different cells in the battery pack. To date, state of the art cell balancing algorithms assume highly accurate SOC estimation which can only be achieved using model-based control solutions. Unfortunately, electrochemical models of batteries are too costly and are still not reliable enough to be used for cell balancing in practice. As a result, the prevalent method today is passive cell balancing, where cells with an excess of charge just release it as heat. Passive cell balancing is not energy efficient and complicates the thermal management of large battery packs, risking a thermal runaway, which compromises safety. 

Our project proposes a simple consensus-based cell balancing algorithm that does not assume any knowledge of the electrochemical model of the battery. Instead, our algorithm only uses SOC estimates from online measurements to make small and careful sequential balancing decisions. Our plan is to establish mathematically that our algorithm balances the battery even with very noisy SOC estimates. For the second step we will consider the more challenging case where balancing is done while the battery is charging or discharging. Balancing while charging can accelerate the charging rate, relying on the balancing mechanism to protect the cells from over-charging. The third step is to study the safety and reliability of our algorithm in simulations, that should also support the theoretical findings.

Our model-free cell balancing method only requires a simple circuit and much cheaper sensors compared to existing approaches. As such, it has the potential of making active cell balancing practical. In contrast to passive cell balancing, active cell balancing is much faster and energy efficient. This enhanced energy efficiency both increases the range of the EV and allows for larger SOC margins that protect the health of the battery.