Executive Summary : | Adoption of electric vehicles (EVs) is globally being viewed as one of the major routes of reducing the carbon footprint of the transportation sector by phasing out the use of fossil fuels and thus mitigating the global warming. India has adopted an aggressive but realizable approach towards achieving net zero status on carbon emission. However, safe operation of EVs requires an efficient battery management system (BMs), which is still at an early stage of development. This is particularly true for electric 2-wheelers and 3-wheelers where compulsions of low cost often prevent implementation of BMs. However, the Indian EV market is mainly driven by this sector, where absence of proper BMs has led to a large number of EV accidents, which have slowed down its adoption by the users. The proposed investigation aims to develop computationally efficient and robust strategies for early detection of abnormal conditions like overvoltage, overcharging and over-discharging in single cells and battery packs of electric vehicles that could lead to thermal runaway. The techniques will be based on tools of dynamical systems analysis applied to time series data acquired from on-board sensors for voltage, current or temperature measurements. The methods to be developed needs to be computationally simple and have low data requirement so that they can be implemented in real life EVs with limited on-board data acquisition or computational capabilities. Additionally the techniques would need to be robust to minimize mispredictions and false alarms. In the proposed investigation, a laboratory-scale experimental setup will be designed and developed for generating time series data for parameters like current, voltage and temperature in single Li-ion cells and battery packs under normal operating conditions and abnormal conditions like over voltage, over charging and over discharging. Features of the time series will be identified and tools of dynamical systems analysis applied to develop methods for early warning of abnormal conditions that can potentially lead to thermal runaways. Different methods based on statistical, graph-based and symbolic time series analysis will be used and compared to identify the most suitable strategy for development of early warning system for battery faults. |