AI to Reduce Carbon Emission in Oil and Gas Industry
The oil and gas industry is being subjected to increasing pressure from consumers, investors, and regulators alike to abate its CO2 emissions. While the task is not easy, Artificial Intelligence can play an important role in the process.
Calls for Greener Oil and Gas
On 22nd April 2016, 174 countries and the European Union have signed The Paris Climate Agreement including China and the United States. The agreement aims to decrease the rise of global temperature below 2 degrees. The parties pledged to perform actions to reduce national Greenhouse Gases emissions by legally binding themselves to nationally determined contributions (NDCs). To achieve their targets countries are imposing new regulations including increased taxes or incentives on key CO2-emitting sectors. The Oil and Gas industry is certainly one of them.
According to McKinsey & Company, the Oil and Gas industry contributed directly 9% of global Greenhouse Gases emissions in 2015, and 33% indirectly through its value chain. This adds up to 42% of global emissions. Luckily Artificial Intelligence (AI) comes up to the rescue. AI will help in decreasing global Greenhouse Gases emissions by up to 4% in 2030 relative to the business-as-usual baseline, a report by PwC in 2019 concludes.
For AI to achieve its full potential in every industry it needs a significant amount of data. That is why it is necessary, for Oil & Gas companies that want to harness the power of AI, to build an Internet of Things (IoT) and sensors infrastructure to collect data. It would be beneficial to collect data from operations, activities, equipment, transport, and everything that would be of value. It is to build what comes to be called a digital twin of the physical system. This will allow companies to accurately estimate and forecast, using AI, CO2 emissions. It will ultimately reveal the aspects of the system that generates high carbon emissions.
Upstream operations account for the process of producing raw oil or gas materials. Upstream operators can reduce CO2 emissions using AI in several ways. First, they can use AI for well-placing analytics to avoid drilling dry wells and in turn spare the earth all associated CO2 emissions. Second, they can develop predictive maintenance systems that can sharply reduce CO2 emissions. For instance, it can reduce the number of unplanned flaring. Flaring means burning off flammable gas released because of unplanned over-pressuring in equipment. According to a McKinsey and Company report, one upstream operator found that nonroutine flaring accounts for 70 percent of all flaring emissions. They largely occur due to weak reliability. By replacing equipment at the right time, the number of such flares can be significantly reduced. Third, AI can be used to optimize CO2 storage. Geologic carbon sequestration is the process of pressurizing CO2 until it becomes a liquid and then injecting it into geologic basins. It is considered an effective technology to reduce emissions. The same procedure can be utilized to increase the production of oil. Researchers in the field are proposing AI algorithms to find the optimal solution for achieving both objectives, increased oil production, and CO2 sequestration, at the same time.
Midstream and Downstream
Downstream operations account for the process of processing raw oil and gas materials and distributing them to customers. Midstream, on the other hand, accounts for the transportation, storage, and marketing of both downstream and upstream products.
On the midstream front, AI is used to detect leaking pipelines to be replaced. Downstream operators adopt AI to increase their energy efficiency. Downstream operations consume a lot of energy to process raw materials, largely in the form of heating. By accurately monitoring and forecasting heat requirements, it is possible, using AI, to find out adjustments that reduce consumed energy and recover wasted energy to use it elsewhere.
Several Oil and Gas companies have already started investing in using AI to reduce CO2 emissions. BP and Shell both vowed to reach net-zero carbon emissions in 2050. ExxonMobile, on the other hand, vowed to reduce its upstream operations emissions in the range of 15 to 20 percent in 2025.
According to their website, BP has ventured $5 million in a Houston technology start-up called Belmont Technology which has produced a cloud-based geoscience platform that utilizes AI. The platform feeds on data related to geology, reservoir, and historic information. It then processes them and links them together into a knowledge graph. BP experts can then ask the platform questions in natural language. The platform uses AI neural networks to interpret results and perform rapid simulations. The technology aims at reducing the time invested in data collection, analysis, and simulation by 90 percent whether in exploration or reservoir modeling. BP has also invested $20 million in Beyond Limits. This investment sights deploying AI technology that assist in exploration missions offshore and in accelerating operational insights and automation.
As part of their longstanding relationship, Shell and Microsoft are joining forces to develop decarbonization AI-based solutions. Shell uses Microsoft’s Azure cloud alongside digital twins technologies by Kongsberg and AI technologies by C3.ai to generate a virtual picture of the system and to aggregate and analyze data. Just recently, Shell, Microsoft, C3.ai, and Baker Hughes declared the Open AI Energy Initiative (OAI). The initiative works as an ecosystem of AI solutions to help transform the energy industry.
AI and CO2
AI is set to play a major role in Oil and Gas industry decarbonization. Several companies including international ones have already invested in AI to reduce their CO2 emissions. Investment in AI not only accomplishes its purpose in reducing emissions but also increases productivity and accounts for a notable value in return through additional revenues and cost savings.
However, the CO2 footprint of AI itself has been under investigation recently. Researchers at MIT have published a report that estimated that training and searching neural network architectures consumes power that emits what is equivalent to five times the lifetime of average U.S. car emissions. The community is increasing its effort to propose greener neural networks.