Batteries are one of the cornerstones of the energy transition that we are facing today. The recent years have seen a rapid rise in the penetration of Li-ion batteries in our day-to-day lives, from electric vehicles to stationary storage solutions. However, this rise has been accompanied by numerous battery fires being reported, despite various standards in place for battery safety. This calls for dedicated techniques to detect a potential thermal runaway (TR) event with sufficient time ahead to warn the user.

In this light, first, a systematic literature review is performed to identify the different paths that a Li-ion cell may take to TR, and shortlist the best sensors and models that could be used to predict a TR event. From this review, temperature and voltage were identified as two key parameters for such a prediction. Of the two parameters, temperature is more relevant since it accompanies all types of TR incidents regardless of the type of abuse (electrical, mechanical, thermal or chemical). Hence, temperature estimation of Li-ion cells is focused on in the rest of the thesis.

A temperature estimation method using electrochemical impedance spectroscopy (EIS) is explored. The phase of the cell's impedance at a certain frequency is correlated with its temperature. The frequency chosen is the one attributed to the relaxation of the Li ions within the solid-electrolyte interface (SEI) layer. This is because the phase of the impedance at this frequency is strongly dependent on the temperature of the cell and almost independent of the state-of-charge. The distribution of relaxation times technique is used to identify this frequency. This method is then validated on a cell under several normal operating conditions by comparing it against a temperature sensor placed within the cell.

Given this estimation method, a final validation to predict the onset of TR using this method is performed by the electrical abuse of a cell through overcharging and the thermal abuse by overheating. While the cell is under thermal abuse, the temperature estimation method is accurate until 60 °C, which is the upper limit of the operating temperature of the cell. Thereafter, the decomposition of the SEI layer begins which results in fallacious temperature estimations. With the cell under electrical abuse, the method is accurate until 4.4 V, after which rapid gas formation in the cell drastically changes its impedance, introducing inaccuracies in the temperature estimation. Despite these inaccuracies under TR condition, several markers from EIS data have been identified which can be used to detect a thermal or electrical abuse to the cell.