Over the past decade, the use of machine learning, predictive analytics, and other artificial intelligence-based technologies in the oil and gas industry has grown immensely. These technologies have advanced over the last 18-24 months as the drop in oil price has driven companies to look for innovative ways to improve efficiency, reduce costs and minimize unplanned downtime.
Leveraging Big Data
The use of advanced software to analyze data and provide recommendations for improving operational performance is not new to the oil and gas industry. Since the early 1990’s, operators in the upstream segment have been using a wide range of technologies designed to ingest and analyze information pertaining to downhole conditions (for example temperature, pressure and geophysical makeup), drill bit performance and reservoir dynamics.
For many years, however, the implementation of these tools required a significant up-front investment, which meant that they could only be financially justified on wells with high production levels.
(Read “Technology Helps Ease Oil Patch Pain.”)
This has changed over the past few years as advances in computer hardware and software, and the number of providers offering data-driven solutions, have both increased. The relative success of AI-based technology in other fields such as healthcare, finance and manufacturing has also grabbed the attention of the oil and gas industry, leading many companies to experiment with it in their own operations.
Perhaps the most surprising demand driver for advanced analytic tools has been the worldwide drop in oil prices, which has tightened margins and forced operators to shift their focus away from increasing production to optimizing it.
Intelligence at the Wellhead
Wireless networks and remote sensors that can collect and transmit data from a variety of measuring points are becoming increasingly common in the oilfield. So is the use of advanced software solutions featuring complex machine learning algorithms, which have the capability to sift through terabytes of data to identify patterns, make predictions and provide recommendations to operating personnel regarding how to optimally control and manage their assets.
(Subscribe to the free Unconventional Oil & Gas newsletter.)
In operations featuring a high level of automation, software bypasses humans entirely and makes micro-adjustments to parameters such as individual pump strokes, rate of penetration and chemical injection rates to maintain optimal production and efficiency.
Collected data can also be used to track equipment performance, helping to predict mechanical failure and alert operators of potential disruptions so unplanned equipment downtime can be minimized. As the software ingests more data, its ability to make predictions improves, adapting and becoming more intelligent over time via a learning method similar to human intellect.
Powered by hundreds of computer servers and able to process billions of data points in real-time, machine learning algorithms have the ability to layer information associated with multiple variables on top of one another. This allows for quick identification of trends and patterns that otherwise would be difficult and time-consuming to detect, even to the most trained human eyes.
For instance, instead of monitoring a single variable like pressure differential in a reservoir, machine learning software can take into account a number of factors important to overall drilling strategy. These include equipment ratings, seismic vibrations, strata permeability and thermal gradients. When layered, these data can be used to determine not only the optimal direction of the drill bit, but also how it should be controlled (that is, rate of penetration) as it bores through the ground.
Information is logged, contextualized and visualized via human machine interfaces, allowing personnel to monitor the overall performance of wells and make more intelligent decisions aimed at improving operations.
Predictive software can also be used to analyze data to determine if downhole conditions are conducive to potentially catastrophic events such as a lost circulation, stuck pipe or blowouts. By leveraging data to understand what contributes to the likelihood of such an event, algorithms can provide recommendations to the control system and operating personnel to minimize the odds of such an occurrence. This has become critical given the fact that the cost of non-productive time can be as high as millions of dollars per day in some cases.
Another technology that exploration and production companies are leveraging alongside machine learning algorithms is case-based reasoning (CBR). CBR is a subset of AI that works by scouring databases of documented problem cases in real-time to attempt to identify cases similar to the issue being encountered. Once a case that has a matching description is retrieved, the system digs deeper to see what actions were taken to address the issue.
CBR systems can be used by themselves or in conjunction with other analytic solutions. They have become more prevalent throughout the drilling sector and provide operators with recommendations on how to deal with a variety of downhole-related issues. The technology itself can be an effective tool for developing best practice guidelines, and has become a useful aid for personnel tasked with making difficult decisions. CBR-based systems are limited, however, in that they require a large searchable case library/database and a succinct description of current problem situations in order to be effective.
AI-based technology has also had an impact on the way that exploration and production companies discover and model oil reserves. Advanced imaging technologies continue to play the most important role in identifying drilling locations. However, in most instances, specific pressure, volume, temperature and permeability characteristics of the reservoir can be obtained only by drilling test or pilot wells.
Fuzzy logic is an AI-based technique used for prediction and reasoning when information is unreliable and/or incomplete. This makes it particularly useful for filling information gaps during reservoir characterization. It also has become increasingly common in other engineering control applications in the energy industry where input parameters are highly variable. These uses include enhanced recovery, well stimulation and infill drilling.
The use of AI-related technologies in oil and gas does not stop with exploration and production, as many operators in the petrochemical refining sector now rely on predictive analytics and model predictive control to continuously improve the overall performance of their facilities and more effectively maintain equipment.
Some companies are even applying it to streamline the transport, refining and distribution of oil and gas. These firms use advanced algorithms to analyze data such as economic conditions and weather patterns to forecast demand, which allows for better allocation of resources and optimal pricing.
Although an increasing number of oil and gas companies are embracing AI-based technology, the industry as a whole still lags many others when it comes to fully leveraging its data. Breaking down information silos and making data more available to key decision-makers across various disciplines will be integral to improving this situation in coming years.
In the short-term, however, the use of machine learning, predictive analytics and other data-driven solutions will continue to help companies improve efficiency and remain profitable in the current economic environment.
And although their hand may have been forced by low prices for end products, those who have made the investment in new technologies may be able to emerge from the downturn stronger than ever.