Skip to main content

IBM Uses Low-Cost Sensors to Help Address Climate Change Issues

Natural gas is considered by many to be a cleaner energy source. The downside to natural gas is the methane that is a primary component. If it leaks into the air—estimates suggest approximately 1-2 percent leakage is common at production sites—it absorbs the sun’s heat and warms the Earth’s atmosphere. Methane is estimated to be the second-largest contributor to global warming, after carbon dioxide.

To address this issue, the Advanced Research Project Agency for Energy (ARPA-E) is partnering with a number of universities, industrial research labs and technology companies to take part in its Innovative Technology to Obtain Reductions (MONITOR) program. IBM has taken a leading role in the program, developing low-cost, highly accurate sensors to detect methane leaks.

As the appropriately last-named William Green, senior manager of the Photonic and Nanoscale Systems Department at IBM Research, describes here, not only do these sensors pick up signs of methane leaks, but they also work in tandem with machine learning and advanced analytics. They help oil and gas companies lessen unintended methane emissions and increase the amount of natural gas they capture and put to market.

IBM Systems Magazine (ISM): What is IBM’s contribution to the ARPA-E MONITOR program’s effort to identify and reduce fugitive methane?

William Green (WG): Overall, the integrated sensing solution we’re developing at IBM Research will ultimately use novel sensor technology and advanced physical analytics to localize and quantify methane leaks within certain applications in the oil and gas industry.

Why would we want to do that? Well, we hear a lot of about carbon dioxide as a greenhouse gas, but if you look at methane and how it traps heat in terms of its greenhouse warming potential, it’s actually about 35x more powerful than carbon dioxide over a hundred-year period. So methane can actually rapidly accelerate the greenhouse gas effect when it makes it into the atmosphere. While there are other significant sources of atmospheric methane around today, including agriculture and landfills, the oil and gas industry is really responsible for roughly a third of the total fugitive methane emissions into the environment.

So with that in mind, ARPA-E, which is part of the U.S. Department of Energy, funded a program a few years ago challenging the technical community to build low-cost, continuously monitoring end-to-end systems that can detect and localize leaks occurring on natural gas production well pads. By quickly detecting and pinpointing where they are, well pad owner-operators can repair and eliminate leaks, particularly the largest ones, known in the industry as “super-emitters,” before they can have a significant impact.

Academic groups and government agencies have conducted a number of estimates of the total U.S. production of natural gas escaping into the environment. Some of these estimates indicate that as much as 5 percent of the product is actually just leaking into the air. Not only does this have a significant environmental impact, but is also inefficient in industry terms. Companies can leverage an investment into new methane monitoring technologies to help them reduce leakage and recover more product. Being informed about leaks in real-time as they occur will also allow well pad owner-operators to be more aware of potential safety issues.

ISM: What is your team’s specific role in this program?

WG: We proposed the development of an end-to-end leak detection system to ARPA-E that provides 24-7 autonomous and continuous monitoring. The system acquires methane concentration data from a network of point sensors distributed in an optimal way over the well pad. Using accurate physical models and various statistical analysis techniques running both locally at the sensor nodes as well as in the cloud, this data is processed to not only detect the presence of a leak, but also pinpoint its location and the leak rate. These are important pieces of information that eventually go into a cloud-based asset-level interface that an end user, such as a natural gas well pad operator at an oil and gas company, could use to monitor all of their well pads nationwide, understand which ones are operating within limits and learn which ones require a repair crew to address a leak.

Of course, sensors that can detect methane already exist. They’ve been around for a long time. But the problem is that they aren’t cost effective. They’re priced anywhere from $10,000 to $125,000 per individual unit depending upon the specific technology and sensitivity level. Natural gas producers can’t afford to deploy multiple sensors like this to conduct the spatial-location analytics needed to discover the source of a leak. Nor can they afford to deploy them across about a half-a-million active wells in the U.S. that are producing methane and natural gas today.

So our program target was trying to build an end-to-end system that would include not only all of the sensors, but could also cap the cost of ownership and operation over a given year to less than $3,000. We’re not there yet, but we’re developing the prototypes. In fact, we’ll be deploying sensor networks, with a couple of field tests this and next year, at functional well-pad sites that will allow us to evaluate our technology and demonstrate its use. This is one of the key ARPA-E program milestones: performing functional field testing.

ISM: Could you explain the technology involved in these sensors?

WG: Well, going back to that point about the current technology being very expensive, a technique can be used to detect methane and other types of molecules in the air. It’s called “optical spectroscopy” and allows you to detect a spectral “fingerprint” that’s unique to each different type of molecule. You use a tunable laser, scan the laser wavelength over a small range and pass that beam through a path where it can interact with the trace gas you are trying to detect while photodetecting the light at the end of the path. The fingerprint shows up in the detected signal, and you can use that to estimate the concentration of the gas, for example, methane, within the optical path sampled. Commercial sensors that operate on this principal are available, but cost at least tens of thousands of dollars per unit.

At IBM Research, we’re developing a gas-sensing platform that uses a wafer-scale, or integrated optics technology called “silicon photonics.” We’ve been developing and using silicon photonics at IBM for over 15 years, with the primary focus to date being on high-speed optical interconnects for systems and data centers. Based on continuing research in this field, we felt motivated to get into the new area of sensors and began working on photonic chips for methane detection a few years ago.

We’ve now used our expertise and background in silicon photonics to build a device for a very different application than what we’ve been historically doing. The photonic chip sensor we’ve designed uses a long optical waveguide so we’re not projecting the tunable laser beam out into the air. It’s confined to the silicon photonic chip, but part of the light actually sticks out into the air around the waveguide. That’s where the sensing takes place. The photodetectors and laser source reside on a separate small III-V chip, which is assembled onto the silicon photonic sensor chip using advanced photonics packaging techniques that we’ve also developed. Altogether, this unique methane sensor is a standalone chip that’s approximately 5-by-5 millimeters or smaller in size, which includes the light source, photodetectors and the sensing mechanism. We’re targeting a price somewhere in the ballpark of a few hundred dollars for this sensor in volume production, including packaging, electronics, assembly and, of course, the semiconductor die themselves.

ISM: Aside from the sensor hardware itself, are there any analytics or machine learning involved in this?

WG: Yes. The idea is that as we collect data from a given site, machine learning models can be applied to enhance the speed of the prediction and capture the leaks as they occur. That can, for example, occur through refinement and improved modeling of local wind conditions. Natural gas leaks, or “plumes” will disperse and be carried along by the local winds. As we get more wind data at a specific well pad, we can have access to on-site historical trends rather than prevailing wind data that’s available publicly but on very low-resolution satellite imaging. That’s one way that learning can occur. Then a variety of physical models can actually be applied to extract information about where leaks occur and how the methane is distributed in space over time. Those varieties of models, including full computational fluid dynamics, can be implemented and combined in a way to blend each prediction with specific proportions and weights that are learned through a cognitive-type algorithm.

ISM: I understand cattle are big producers of methane. Is any work being done to address that?

WG: That’s another big topic because they create roughly 30 percent of the overall methane emissions that get into the air. There may be less that we can do about that in terms of mitigation. However, the technology we’re developing for the oil and gas industry could certainly be adapted to monitoring methane levels within an agricultural setting.

ISM: Are there any other applications for this type of technology?

WG: Methane is the main application today, but this platform can be extended to other types of molecules and trace gases, some of which can be related to environmental pollution. They can also be related to healthcare, such as breath analysis where you can perform noninvasive measurements of various internal conditions. If you have a platform that has significant sensitivity, you can find molecules that correlate to conditions that might otherwise need exploratory biopsies to diagnose. The silicon photonic trace gas-sensing platform itself is multipurpose. It can be used for a wide variety of applications.