Geospatial Internet of Things: Framework for fugitive Methane Gas Leaks Monitoring

GIScience 2016 Short Paper Proceedings Geospatial Internet of Things: Framework for fugitive Methane Gas Leaks Monitoring L. J. Klein, R. Muralidhar, F. J. Marianno, J.B. Chang, S. Lu, H.F. Hamann IBM T. J. Watson Research Center, Yorktown Heights, NY 10598 Abstract We present a framework for wireless sensor network monitoring and detection of methane leaks from natural gas well pads. The wireless sensor network can measure methane concentrations across a well pad and combined with advanced analytics it can locate and determine the leak rate. Simulations of the inverse and forward modeling problems indicates that methane leaks can be localized within a 1 m distance from their original locations. The wireless sensor network and real time analytics can be extended to monitor multiple methane leaks and methane background levels. 1. Introduction Methane has a much larger global warming potential compared to carbon dioxide. Methane gas is emitted by agricultural and waste management sources, however more than 30% of emitted methane gas is coming from energy exploration sites (natural gas and petroleum system and coal mining). With more than half a million natural gas well pad sites developed in USA, understanding the impact of methane gas on human health and long term climate impact became important. In the past, standalone high precision sensors were used to measure methane gas leaks. These measurements can offer a very precise methane concentration assessment but the spatial coverage is limited. Current methane measurement and modeling techniques are lacking the capabilities to localize leaks on a well pad (Zavala-Araiza 2015; Lyon 2015; Foster-Wittig 2015). An alternative method to detect methane leaks is using satellite observations. While the satellite methods offer a large scale geospatial observation, the spatial resolution of the detection method is too coarse for single leak detection (Turner 2015; Veefkind 2012). There is certainly a need to combine the high accuracy local wireless sensor measurements with large scale satellite observations for (1) accounting all methane leaks over a regional area and (2) attribute methane leaks to emission sources. Here we present a novel methane monitoring solution based on wireless sensor network. Methane sensitive sensors are distributed on a 10 m grid and are measuring in real time the methane concentration. In addition, the wind direction and speed is measured as well. We note that each well pad have construction on their perimeters and an associated infrastructure (storage tanks, well heads, etc). These structures will cause turbulence to wind flow pattern. The well pad layout can be extracted from high resolution satellite or drone imagery. The layout is used for three dimensional reconstruction of a gas well pad and generate a Computer Aided Design (CAD) models. The CAD model is a necessary input into CFD for dispersion modeling. The advantage of our proposed method is that each methane leak can be identified and localization on a well pad. The concept for large area methane measurements, across multiple well pads, in presented in Figure 1. Each of the 4 well pads have its own Wireless Sensor Network that measure methane and wind for that specific locations. Inter well pads communication is enabled using a Wide Area Network (WAN) to transfer the sensor data and computational load between well pads. Each well pad may have one or more Raspberry Pi computers that act as a computational platform and data gathering device (edge devices). If no leak is present on a well pad, the sensor values can be sampled every hour, however in case of large methane leaks the sampling rate may have to be increased to a measurement each second. Since significant amount of data may be generated on each well pad, data processing needs to be carried out on the edge device and only aggregated and processes sensor values are sent to the cloud platform. Locations of the sensors on the well pad as well as the site layout are geospatially located. Measurement on a single well pad may be affected by a leak on a different well pad. This scenarios can