Low-temperature separation with a monoethylene glycol (MEG) injection process is a common dehydration technique for natural gas processing. However, the MEG-based dehydration system frequently suffers significant glycol loss during plant upset conditions, causing double penalties of economic loss and air emissions. Thus, it is very important to minimize MEG loss in the dehydration process. In this paper, an efficient and effective methodology to reduce MEG loss under upset conditions of a natural gas dehydration plant has been developed. First, a plant-wide steady-state simulation model is developed and validated at normal operating conditions. Next, the root cause analysis for MEG loss is performed by introducing various process upsets to the simulation model, which indicates that most MEG loss is due to the fluctuating temperature of the stripper column overhead. After that, a plant-wide dynamic simulation model is developed to help generate a new control strategy to regulate the stripper column operation and cope with other plant upset conditions, so as to minimize the MEG loss and improve natural gas product quality and plant operability. Simulation results indicate that plant MEG loss can be reduced by 37%.