Petronor Trials Generative AI to Boost Efficiency and Profitability in Production Plants

Making Oil Production plants more profitable and efficient. Concept Testing: A solution that predicts operating conditions for Oil Refining Plants

A key component of this case acceleration program is the Gidabot tool, which was designed to act as a sophisticated “digital assistant” for chemical engineers at the Petronor Refinery. Gidabot helps engineers quickly access and utilize technical documents, such as manuals and data sheets, which are crucial for their tasks. Additionally, it predicts how various operating conditions will impact refinery processes, enabling engineers to make informed decisions and adjustments. This leads to improved performance and efficiency at the refinery. However, developing this tool was a challenge.

Begoña López, Key Account Manager (KAM) for Utilities Data at Ayesa explains that the challenge was to create a virtual assistant that engineers could use to quickly to find information using simple, natural language. The team developed a hybrid solution that combines various technologies. It uses generative AI to understand and process information, computer vision to handle complex figures, and predictive machine learning to analyse historical data from the refinery. This combination helps engineers get useful insights and make accurate adjustments to improve the refinery’s catalytic processes, making them more efficient and effective than before.

López, explains the project’s goals: “Engineers need to regularly check technical data from manufacturers in order to make adjustments. This impacts the quality of the final product and the efficiency of the process. These routine checks are extremely time-consuming because they involve extensive, highly technical manuals that are difficult to search. Additionally, these manuals often include complex figures, which apparently current generative AI models have struggled to address effectively.”

Addressing Generative AI Limitations

Generative AI models can have trouble providing accurate, consistent, or practical solutions due to their limitations, such as difficulty handling complex figures or inconsistencies in their outputs. However, this initiative has shown that combining AI models with careful management can lead to more reliable results. According to the Key Account Manager for Utilities Data, “the result is a strong solution that provides a simple and quick way to verify responses.”

Marian Aradillas, Head of Advanced Analytics at Ayesa, supports this approach: “By merging generative AI with traditional machine learning, we can analyze large amounts of data and identify complex patterns. We also use OCR (optical character recognition) and computer vision technology to find important information in technical manuals more easily, which improves data access and prediction accuracy. Understanding the business and its processes has been essential in creating an effective solution.” She concludes, “The integration of hybrid generative AI models with other AI technologies has led to the creation of an intelligent virtual assistant prototype that offers significant value to Petronor.

Here are the main benefits of the Gidabot initiative:

1. Enhanced Efficiency: The Gidabot tool helps engineers quickly access and utilize technical documentation, streamlining their workflow and reducing the time spent searching for information.
2. Improved Decision-Making: By predicting how various operating conditions will impact refinery processes, Gidabot enables engineers to make more informed decisions and adjustments, leading to better performance.
3. Increased Accuracy: The integration of generative AI with OCR (optical character recognition) and computer vision technologies allows for more accurate extraction and analysis of information from complex technical manuals.
4. Optimization of Processes: The ability to analyze large volumes of data and recognize complex patterns helps optimize refinery operations, enhancing overall efficiency and productivity.
5. Cost Reduction: By improving efficiency and performance, the Gidabot tool contributes to reducing operating costs at the refinery.
6. Scalable Solutions: The project demonstrates a scalable approach to digital transformation, with the potential for broader application across various industrial processes and settings.
7. Competitive Advantage: Leveraging advanced AI technologies gives Petronor a competitive edge by enhancing its operational capabilities and fostering growth through innovation.

The potential of AI in industry

• Accurately predicting catalytic cracking conditions has the potential to revolutionize industrial plant management by enhancing production efficiency and reducing operating costs.
• The solution developed represents a significant advance towards the complete digitalization of industry. By harnessing generative AI, companies can make more informed decisions, optimize processes, and improve operational efficiency. This not only boosts competitiveness but also drives growth.

Petronor and Ayesa will continue their partnership to explore new generative AI initiatives, aiming to drive industry transformation through innovation and creativity.