Statistics

Proactive Early Warning for EV Battery Packs: A Concept Note on Reading the Recovery Speed of the BMS Voltage Signal

Authors: Prakash Vaithyanathan
Comments: 3 Pages.

The battery management system of an electric vehicle records cell voltage, current and temperature many times a second. Every current method that uses these signals to detect degradation is either reactive, acting after a fault code is raised or a capacity threshold is crossed, or trained, requiring labelled examples of degradation events to build a model. This note proposes a proactive layer that needs neither. It watches how quickly the small, cycle-level voltage fluctuations return to their normal level after each disturbance. In a healthy cell this return is fast. As the cell approaches a tipping point, the return slows, and this slowing can be seen in the signal before any threshold is crossed. The slowing is measured by the lag-1 autocorrelation of the voltage residual within a rolling window. When this statistic rises persistently towards 1, a graded shunt acts on the entities of the BMS domain, adjusting the charge rate, flagging the cell, or pausing the charge, before the cell crosses a danger line. The shunt is the core of the invention. The lag-1 autocorrelation as an early warning signal rests on well-established physics. The application of a graded shunt to the BMS domain, triggered by the statistic rising persistently towards 1, is what this note proposes. It is a concept note. No battery data have been analysed.
Category: Statistics