Sensing Elevated Carbon Dioxide Levels for Detecting Battery Cell Venting in Packs

Abstract

Li-ion battery thermal runaway is a critical safety issue for large-scale battery packs, and gas venting of a battery cell is a precursor of an imminent thermal runaway. The proposed global technical regulation No.20 on Electric Vehicle Safety requires an advance warning 5 minutes prior to hazardous conditions caused by a thermal runaway event. Voltage-based detection methods are less effective in large packs with many parallel connected cells, and temperature-based methods provide slow and sparse information when fewer than 10% of the cells are instrumented. To achieve immediate detection for thermal runaway, vent-gas detection methods are promising due to their fast response and easy implementation in a pack. The composition of battery vent-gas during faults that can lead to a thermal runaway event include CO2, CO, H2, and volatile organic compounds (VOCs). The best gas species to target for a detection strategy is still debated. The composition of the vent-gas depends on cell chemistry, abuse condition, and cell state of charge. For robustness, the detection strategy needs to be sensitive to all possible failure mechanisms, be inexpensive, and require no maintenance. A Non-Dispersive Infrared (NDIR) CO2 sensor can meet these requirements, because the detection can be made immediately when the presence of CO2 is observed from the first venting event, and the sensor type has low cost and with some sensor’s lifetime over 15 years.To check the NDIR sensor responsiveness, an overcharging test for a prismatic 4.9 Ah cell was performed and the gas venting is observed before thermal runaway is triggered. Quickly after the occurrence of gas venting, the NDIR CO2 sensor showed a strong response with maximum gas concentration exceeding 30,000 ppm as shown in the figure. In a battery pack, the fast velocity of the ejected gas from cell venting ensured the fast response of the gas sensor. To apply the gas detection strategies for large battery packs, the volume and average CO2 concentration can be modeled to set the gas detection threshold. Figure 1