A form of artificial intelligence (AI) is now being utilized to make the global oil market more transparent.
Using a form of AI called convolutional neural networks (CNN), Palo Alto, Calif.-based Orbital Insight can detect oil storage tankers previously unknown to existing data sources. The company gathers imagery from a number of satellite partners, then utilizes cloud computing to analyze images and identify not only oil storage tankers, but traffic activity in retail store parking lots, global water reserves and agriculture. It is also used to map poverty on a regional basis.
Tracking oil storage tanks allows for greater understanding of the global supply chain. This understanding, in turn, enables better decision-making, Kevin O’Brien, chief business officer at geospatial Big Data company Orbital Insight, told Rigzone.
Last year, the company rolled out backtest data for its China Oil Storage and U.S. Oil Storage products. Once it finishes processing data for the entire world, it will roll out additional products based on customer feedback, including Europe, the Middle East, Russia, Brazil, India, Venezuela, Angola, Nigeria and Iran. Products for countries such as Iran, where little to known data exists on oil storage supplies, are particularly in demand, Jason Lohn, head of core engineering at Orbital Insight, said.
Founder’s Vision: Monitoring Entire World on Near-Time Basis
The company’s technology was developed by Dr. James Crawford, who at one time headed up robotics research at NASA’s Ames Research Center in Sunnyvale, Calif. After leaving NASA, he ran Google’s books division, where AI was used to image more than 20 million books. Watching the growth in the space industry, Crawford was pleased to see the massive increase in the number of satellites being launched. However, he knew that nobody could process the massive pipeline of imagery that would flow from these satellites to earth. So he launched Orbital Insight to meet this need.
Crawford’s vision was to find a way “of looking at the entire planet on a near-time basis for the first time ever, then use AI to interpret changes or identification of different types of objects on the planet,” Kevin O’Brien, chief business officer at Orbital Insight, told Rigzone in an interview. The idea was to help people globally make better decisions, whether they were technologists, government entities, non-government organizations, charities or business professionals.
Neural networks are designed to operate the way the human brain operates, Shriram Ramanathan, senior analyst at Lux Research, told Rigzone. This type of AI has been around since the 1950s, but only in recent years have computers become fast enough to run such simulations on a large scale.
“A human brain won’t recognize a car when it sees it for the first time; however, with repeated exposure, the brain gets trained and learns to identify cars. Likewise, neural networks are trained using training images and subsequently learn to recognize images,” Ramanathan explained.
While neural networks look at the entire image, convolutional neural networks only look at small parts of the image at any given time, Ramanathan said. When identifying a car, a human brain just focuses on major features such as the outline of the car and not details such as the color of the mirror or door handle.
“Likewise, by focusing on small parts of the image at any given time, CNNs emphasize important features such as outlines and edges. This also makes these networks computationally more efficient compared to traditional neural networks. In general, artificial intelligence algorithms such as these are increasingly becoming important in predictive maintenance as well as operational intelligence in the oil and gas space,” Ramanathan told Rigzone.
China Results Highlight Problem of Global Oil Tracking
Orbital Insight initially focused on testing its oil tracking product in the United States, then China. The United States is generally accepted as the most transparent in terms of information on oil inventory tanks and offshore tanks, Barclays’ analyst Michael D. Cohen told Rigzone. Other countries such as China take longer to release information, provide data in an opaque format, or are not necessarily trusted by the marketplace.
The results of examining China’s onshore storage tanks was stunning, Lohn said. Existing estimates said approximately 500 onshore floating roof oil storage tanks existed in China. Instead, Orbital Insight found 2,100.
“Analysts have long suspected that the oil supply numbers China’s government published underestimated the actual amount in storage,” Lohn said.
Instead of the 200 to 300 million barrels that the Chinese government has reported, Orbital Insight’s findings put that number in the 500 to 600 million barrel range.
The tracking of global oil supplies is definitely an issue, Cohen said. For some countries, supply estimates are made based on tanker tracking and indigenous consumption. The supply figure is implied from the other two variables, for which some countries have poor data quality. Other countries do report a supply figure usually based on an aggregate of companies, or some just report a number without any detail.
“These are also prone to misinformation and are sometimes not retroactively reviewed in case some companies did not report in the survey,” Cohen stated. “Even with those agencies that do track tankers, one only knows if the AIS [automatic identification system] beacon is on, and sometimes those are shut off,” said Cohen. “Sometimes the tankers that carry oil products, rather than just crude, which makes the correlation between the number of tankers and crude exports break down.”
Greater federal regulation and reporting rules over aspects of oil market infrastructure, including marine tankers, is needed, Tyler Slocum, a member of the Commodity Futures Trading Commission’s Energy and Environmental Markets Advisory Committee and director of the energy program at Public Citizen, Inc., told Rigzone.
“While there are several, quite expensive, proprietary databases and satellite tracking firms that provide deep-pocketed access to all shipping info, including some elements of the Chinese market, there is very little public domain data available,” Slocum said. “And even the proprietary databases don’t have universal access to all shipping info, including some elements of the Chinese market.”
Greater Accuracy, Speed Benefits of AI Technology
Orbital Insight expects the number of satellites to increase as providers such as DigitalGlobe launch inexpensive, next generation satellies, which will boost the available supply of low, medium, and high-resolution images the company uses. A few years ago, one to two dozen of these satellites orbited the earth. By the end of 2017, O’Brien forecasts that number to grow to over 150 or 175. In 2018, O’Brien expects that number to double or triple.
Besides accuracy, O’Brien said the company’s technology, available through subscription service, allows not only better decisionmaking, but faster decisionmaking. While the U.S. Energy Information Administration (EIA) will remain the gold standard for oil tracking, the agency’s voluntary survey-based process to track oil supplies takes five days. By comparison, Orbital Insight can gather information on supply in 60 minutes. The speed of gathering data can provide an advantage to people trading in the oil futures market as a compliment or hedge to EIA numbers.
The company’s initial rollout of its oil products is focused on floating roof oil tankers, the most common way to store oil. Orbital Insight is initially focusing on land storage tanks, but Lohn said the technology does have the capability to help track oil supplies in offshore tankers in port or at sea. Lohn said the company will work with partners to determine of supply on an offshore tanker is oil or another product.
Determining the location of a cargo ship is possible, but determining exactly what’s on the ship is a more complicated matter, Sandy Fielden, director of research for commodities and energy at Morningstar, told Rigzone. Doing so requires obtaining information not only on whether a tanker is carrying crude or crude products, but what type of crude and crude products are on board. This information about what is loaded on an offshore tanker cannot be gleaned by looking at satellite imagery, but by obtaining information from brokers, shipping agents or customers.
“Satellite technology can tell you a lot, but not everything,” Fielden stated.
Lohn admitted that data processing is intensive, and not perfect. To ensure analyses are correct, Orbital Insight also has employees examine data and filter out objects such as water towers, which look like oil storage tanks.
Financial services and government customers are two of the markets on which Orbital Insight focuses. The company is also expanding into infrastructure and asset monitoring, as energy flows into this area as well, O’Brien said. For example, a refinery owner might want to find new sources of oil for its facility. The company’s customer roster includes some of the world’s major oil companies, refiners, and service providers.