Artificial Intelligence in Upstream Oil and Gas: Top Use Cases and Benefits

Exploring oil and gas reservoirs requires 3D imaging of the field and processing the data acquired from petrophysical and geophysical studies, including seismic surveying at the scale of the reservoir. AI reduced the time taken to process these 3D images, an otherwise time-intensive process.

Furthermore, the 3D images (or seismic cubes) are studied and segmented by experts, and the whole process can take more than a year to conduct an accurate seismic study. Using modern deep learning-based pattern recognition techniques makes it possible to accelerate the interpretation process by a factor of 10-1000.

Potential cracks or corrosions in the exploration tank can have disastrous repercussions on the environment and society. To prevent this from happening, industries are now deploying AI-powered crack detection robots that inspect the welding site for cracks by moving a pipe connected through a welding joint in the tank. These robots are also being deployed to inspect the pipeline for intelligent pipeline monitoring, even though they need experienced engineers to manipulate them.

Reinforcement learning (a form of semi-supervised ML) is incredible new way organizations are leveraging AI to control their drilling equipment. It works by training a machine learning model based on simulations of steering the drill into the subsurface and historical drilling records. Furthermore, it brings the sensor data from seismic surveys and other important factors like pressure, temperature, etc.

Later, the drilling machine engineer can input arguments to penalty or reward functions to help the machinery adapt to changing operating conditions, thus reducing gas extraction costs.

Unmitigated salt depositions in hydroprocessing effluent trains and FCC and crude fractionator overheads can cause numerous operational hiccups, including safety and economic risks. Manually formatting and analyzing data from these unaligned systems can take hours. One solution to this problem is to employ deep learning for semantic segmentation while interpreting salt deposits on seismic images.

This makes it easier to calculate salt deposition temperatures and compare them with safety margins. Various AI tools are now available that provide this information in real-time.

By leveraging technologies like the internet of things (IoT) and AI, we can now address the growing challenges in optimizing oil and gas production. For instance, AI is highly resourceful when weeding out false alarms and making accurate predictions based on various modeling approaches using recent trends.

Furthermore, AI comes in handy while forecasting electrical subversive pump (ESP) operational issues and developing a predictive ESP maintenance model. For demonstration purposes, real-time production data can be transmitted into an ML-based predictive maintenance model from SCADA systems using secure IoT connections over the cloud.

One of the most common methods for non-destructive testing of oil and gas pipelines is the magnetic flux leakage (MFL) technique which looks for defects and anomalies in the extraction pipe wall. However, this method is quite difficult and involves highly complex MFL image analysis. Modern machine learning techniques using kernelization, support vector regression, principal component analysis, and methods for reducing feature space dimensionality offer a highly promising outlook in this respect. The models still use MFL data, but the performance is adequate for detecting defects and assessing their severity.