Oil and Gas Data Mining AI Agent

Welcome to our 3rd edition of Upstrima AI Agents Newsletter

In this edition we will showcase our next Agent - Data Mining and Insight Retrieval Agent.

But first - why Data Agents are like dogs fetching sticks ?

Instead of thinking about AI as something magic, superhuman or alien, it’s easier to analogize them to animals, intelligent nonhumans we have experience training. We are not expecting dogs to understand complex concepts, such as "self-consciousness" or "attitude". Instead, we are expecting them to learn a behavior, such as retrieval of a tennis ball or a stick. Same with AI Agents, and Data Retrieval agent in particular. This agent, like a dog, needs to know what to look for and how to find a particular stick among hundreds of other sticks lying on the ground in a forest. Dogs use tools - smell, vision, shape recognition. Agents also use tools - algorithms (such as RAG), meta data indexing and user instructions.

Challenge

Manual data search and extraction is a task that drilling engineers routinely facilitate. This task involves extracting data from a variety of sources such as daily operational data, standards and guidelines, equipment manuals, and catalogues. Given the scattered nature of the sources, this task becomes challenging, time-consuming and often frustrating.

To ensure timely troubleshooting and continuous improvement in constrained work environments, the ability to extract the correct information fast, subsequently translating it into meaningful insights while constantly validating against the latest industry standards and regulations, is a critical component of personal productivity.?

Although Generative AI solutions can provide answers to basic Oil and Gas queries, they are not capable of building knowledge in the long term, nor are they customizable to navigate through drilling specific documents, often lacking consistency and accuracy in providing the answers. Thus, currently available Generative AI models cannot be used as the dependable source of data and insights, but rather remain an analogue of Wiki, holding and providing general knowledge without application.

Solution

In order to come closer to a practical, useful and dependable AI solution, a dedicated AI Agent was developed and trained on a wide set of oil and gas knowledge to understand drilling natural language and to know how to navigate drilling specific documents and datasets. By leveraging specialized indexing, namespace mechanisms, and metadata structures, the AI Agent was found to efficiently retrieve the most relevant data, including well information, data types, and timestamps with high accuracy. Furthermore, by incorporating domain-specific knowledge and industry-specific logic from the oil and gas industry, the AI Agent processes and refines the data to enhance its relevance and usability, ultimately delivering more precise and useable data insights.

Case Study

The following examples below demonstrate Data Mining AI Agent output on prompts in natural language to extract drilling specific information from different types of data source e.g. directional reports, daily drilling reports and geological reports (Figures 4, 5 and 6). User specified the desired well name, date and type of report to query the data from. The answers provided by the Agent were accurate, complete and consistent when asked differently. The agent supplemented each answer with relevant details implied to the answer although not directly asked. Such as directional data in Figure 1, prognosed vs actual depths and deltas in Figure 2, overpulls and intervals of tight spots in Figure 3.

Conclusion

• Is this Data AI Agent a ground breaking technology that will re-invent Oil and Gas ? Of course not.
• Will it help drilling engineers to save time and increase accuracy on a daily basis? Absolutely!
• Is it going to be useful for other AI Agents to perform data analytics and task automation? Hell yeah!!!

The key is not only to have all your data at the fingertips, but also to be able to use it to perform any further actions you want. Maybe use it as an input to other agents or applications. Or to detect trends and extract insights from hundreds of wells in minutes without having to switch to any other software.

It only takes minutes to extract data from hundreds of offset wells. 
And the best part: users don't have to look for or open any of the original files.         

Read in our next edition about AI Agents of higher complexity that are used to automate routine tasks and streamline DE workflows.

Disclaimer: no robot-dogs are offered as part of the Upstrima AI solution.

although... ?? would be cute to have a dog like this walking around the rig site and aggregating data from different sources. Need to think about it...

Follow as at Gain.Energy

Sign up for Beta Testing here https://www.gain.energy/