Efficient Sensing Matters a Lot for Large-scale Batteries

A large-scale battery pack that consists of hundreds or thousands of battery cells must be carefully monitored. Due to the divergence of cell characteristics, every cell should be monitored periodically and accurately. There are two important issues in monitoring large-scale packs. First, sensing the health condition of battery cells must be timely to capture the turning point at which the battery condition abruptly changes. Failure to capture such an important event can cause irreversible damage to the battery, especially when its State-of-Charge (SoC) is very low. Second, the more the hardware components are used, the higher the failure rate the system will suffer. The frequency of monitoring battery cells, thus, should be adjustable to the underlying load demand, considering the fact that a low load demand has a minute impact on the battery condition. We propose to address these issues via an adaptive monitoring architecture, called ADMON. ADMON lowers the sensing latency effectively, making it effective to enhance the tolerance of physical cell failures. ADMON consists of sensing, path-switching, and computing systems. The sensing system collects data from a battery-cell array. The path-switching system effectively connects a specific sensor and a micro-controller that is part of the computing system. The path-switching system is characterized by three exclusive types of topology: n-tree-based, cascaded, and parallel. The computing system is synergistically combined with the other two systems while three policies specified in the computing system are applied. The ADMON architecture is shown to outperform a non-adaptive monitoring system with respect to the battery life by 67%.