Over the past two decades, Lithium-ion (Li-ion) batteries have become ubiquitous in society. Their relatively high energy density, and recharge efficiency have resulted in their mass adoption. With falling prices, batteries have moved from single cell applications to increasingly larger systems comprised of many cells, such as electric vehicles (EVs), and fixed systems used in grid-scale load-leveling and uninterruptible power supplies (UPS). These systems typically have thousands, or tens of thousands, of cells in order to meet the desired capacity and power draw capabilities. Li-ion energy storage systems (ESS) are generally safe. However, Li-ion cells can fail under abnormal conditions potentially resulting in a catastrophic event known as thermal runaway (TR). During this failure process, exothermic reactions within the cell can result in excessive temperatures (∼ 800-1000°C), and produce highly flammable, toxic vented gases. As has been done for legacy battery technologies, such as Lead-acid, it is important to characterize these hazards, and to develop prevention, detection, mitigation strategies for TR and TR propagation. Experiments were conducted with arrays of Li-ion, pouch-format cells of two cathode chemistries: LCO and NMC. These experiments represent pseudo-modules to characterize TR failure propagation. This work makes the assumption that a single cell can and will fail for a specific system. This failure is represented using an external heater applied to one cell on one side of the array. From the initial failure, this work analyzes the failure propagation process, discusses the patterns seen, and evaluates different experimental techniques. One of the key hazards resulting from TR and TR propagation is the previously mentioned vent gas production. This vent gas is primarily composed of hydrogen, carbon dioxide, carbon monoxide, and various hydrocarbons. Altogether the produced vent gas can result in fire if ignited, or an explosion if allowed to accumulate. Furthermore, Li-ion cells have been found to produce sparks and eject relatively hot soot particulates during TR that can result in vent gas ignition. This work characterizes the vent gas release process for the two cell chemistries tested, including the total volume production, rate of release, and gas composition. While more toxic gases have also been detected in trace amounts in Li-ion vent gas, the primary goal of this work is to characterize the TR failure propagation process as it relates to fire and explosion related hazards. Thus, toxic gas characterization is beyond the scope of this work and is left to other studies in the literature. The second hazard from Li-ion TR propagation is the resulting elevated cell temperatures from the internal exothermic reactions. This work discusses the internal reactions responsible for TR proposed in the literature. Furthermore, this work evaluates the effect of a cell’s state of charge (SOC) on these reactions. Experimentally, thermocouples (TCs) were used to measure the temperatures between cells. These experimental data were then used to create and calibrate a computational model for predicting TR propagation in cell arrays of varying SOC. The effect of SOC is significant from the modeled internal kinetics. However, the heat transfer process between cells was found to be the primary challenge in modeling cell-to-cell TR propagation. Finally, the computational model was used to evaluate the performance of thermal separation materials between cells to inhibit cell-to-cell TR propagation. This work found that insulation, and thermal sink layers, i.e. aluminum plates between cells, could slow TR propagation, but may still lead to TR propagation via other modes of heat transfer to the cell array, including heat convection from the produced hot vent gases. Additionally, this work does not fully study the effect of external flaming combustion which was found to accelerate TR propagation in preliminary tests. Thus, external flaming combustion could reduce the efficacy of thermal separation materials between cells. Regardless, the computational model is capable of accurately predicting TR propagation in these arrays without external flaming combustion, and can be used to quickly evaluate several potential mitigation strategies. These modeled mitigation strategies can then be further validated with additional experimental testing.