METHANE DETECTION AND CHARACTERIZATION WITH AI SENSOR FUSION AND DECISION-ANALYTIC PLACEMENT OF RAPIDLY DEPLOYABLE SENSORS
Immediate action is required to mitigate greenhouse gas emissions and its impact on climate change. Methane emissions have been estimated to produce 80 times the warming effect of carbon dioxide and are responsible for a third of anthropogenic warming. Recent legislation in inspections and penalties for leaks has motivated new efforts to identify sources of fugitive methane emissions and remediate them. However, there are major challenges, costs and safety concerns in identifying and characterizing leaks in remote and hard-to-reach production sites. New international satellite data can be used to flag general areas of large emissions, but do not have the resolution to identify the exact source or faulty equipment, which is needed to develop a plan for remediation. This paper illustrates an integrated ground-aerial smart sensor approach built around a machine learning framework for methane inspections and characterizations. With the ability to be rapidly deployed by sUAS (small unmanned aerial systems) to gather a more localized and ground-level assessment of the leak, a dynamic optimal placement of sensors can be used to for improved remediation decision making. A case study is presented using the methane GasVid leak video dataset.
