[V8N1] – Is AI/Machine Learning Model for the Fuel Blending System A Viable Alternative? – Part III

OMS Newsletter AIMachine Learning Models min

In the last article, we discussed the Gold Standard First Principle Blending Model (FPBM), its pros and cons, and the reason for its failure in many refineries due to many factors. Some of the factors for the failure of FPBM are management’s disinterest in updating and sustaining the technology, ignoring the tangible loss of 25-40 million per year, and attrition, transfers, and inadequate training of operational staff and engineers.

The impact of these controllable factors is shown in the chart below.

We can see that two areas, namely, non-linear blend models and non-functional or non-existent online analyzers, can and do affect the refinery’s bottom line. Let us not forget that the refinery’s profit is generated from 80-90% of blended products manufactured in the offsite operations area.

1711126399234?e=1716422400&v=beta&t=oA026TBqcyjOaVkmGriklRSIAh Edh3FTCmyIAql1Mg

Having said that, what is our option to minimize the impact of the above very high factors on the performance of the Fuel Blending System?

Let us discuss the potential role of AI and machine learning to minimize the impact of these factors.

In a blending system, there are three components required for its success. They are the component qualities, predicted blend qualities at the blend header, and finally, the final blend qualities of the product tank analyzed by the lab analysis. Each of these components can use the help of AI and machine learning algorithms. The most substantial impact is at the blend header, as it receives all components individually and not in aggregated form as at inlet of the the product tank.

1711126414117?e=1716422400&v=beta&t=e3FwWsZynF2 lE0CRIXns8hDOKDZ2IIr6kBkoLlFe2I

The potential use of AI (Artificial Intelligence) and machine learning to replace online analyzers in fuel blending is a topic of growing interest in the refining industry. Online analyzers are crucial in the fuel blending process, providing real-time data on the composition and properties of the blend, ensuring product quality and meet compliance with specifications. The integration of AI and machine learning can complement or, in some scenarios, potentially replace certain aspects of online analyzers, but there are several considerations to take into account.

We can use predicted blend qualities using linear or non-linear blend models and build an AI model using the historical data to compare the analyzer value. There is a catch. We can not build an AI model to predict analyzer values if the refinery has not installed online analyzers or digitized the online analysis data to build the historical database.

However, we can estimate the expected blend header qualities by back-casting the qualities of the product tank by lab analysis, composition, and predicted component qualities. But this approach is less accurate and further discussion is outside the scope of this article.

The chart below demonstrates how AI/machine learning can be used to estimate and compare against the analyzer values using the non-linear blend models.

1711126427765?e=1716422400&v=beta&t=odxQDgGm9cnxAgbwKtUfwuabS jPL tNbusIWV9 8W0

Let us discuss further the project of using AI / Machine in the fuel blending system to squeeze that extra drop of the barrel.

Advantages of AI/Machine Learning

  • Predictive Analytics: AI can analyze historical data and predict the outcomes of various blending strategies, potentially optimizing the blend more efficiently than traditional methods.
  • Real-time Optimization: Machine learning models can continuously learn from the blending process, improving blend recipes in real time to enhance quality and reduce costs.
  • Anomaly Detection: AI systems can quickly identify anomalies or deviations in the blending process that might indicate equipment malfunctions or errors, potentially before traditional methods detect them.

Limitations & Challenges

  • Sensory Data: Online analyzers provide direct, real-time measurements of physical and chemical properties. While AI can predict and optimize, it relies on data from some form of physical measurement. Completely replacing online analyzers would require AI systems capable of accurately predicting these measurements without direct sensory input, which is a significant challenge.
  • Complexity of Fuel Blending: Fuel blending involves numerous variables, including the properties of blend components, environmental factors, and regulatory requirements. Capturing and accurately modeling this complexity is challenging for AI and machine learning.
  • Regulatory and Safety Considerations: The fuel industry is highly regulated, and products must meet specific standards. Relying solely on AI predictions without direct measurement may pose regulatory and safety challenges.

Integration Rather Than Replacement

  • Complementary Role: Instead of outright replacement, a more feasible approach is integrating AI and machine learning with existing online analyzers. AI can enhance decision-making, provide advanced predictions, and optimize processes based on data from online analyzers.
  • Hybrid Systems: Implementing hybrid systems where AI and machine learning provide advanced analytics and predictions while online analyzers perform real-time measurements could offer the best of both worlds.

Conclusion

While AI and machine learning offer promising opportunities to enhance fuel blending operations, completely replacing online analyzers is not currently feasible due to the direct measurement capabilities and regulatory compliance requirements.

The future likely lies in a collaborative approach where AI supplements and optimizes online analyzers, leading to more efficient, accurate, and cost-effective blending processes. As technology evolves, the role of AI in this domain is expected to grow, further enhancing the capabilities and efficiency of fuel blending operations.


Disclaimer: OMS eLearning Academy and ChatGPT collaborated as Humans and AI to generate this article for you.


In Part IV of this series, we will discuss the hybrid model of the First principle and AI/Machine Learning to make the best of both approaches.


Stay tuned for more groundbreaking publications and enrich your expertise with OMS Academy. 🌐✨

Sign up today to start your learning journey.


➡️ Sign Up Now 🚀


Thank you for being a part of the OMS eLearning Academy community.

If you have any questions or feedback, please contact us.

“At OMS eLearning Academy, excellence in education isn’t just our goal; it’s our promise – a promise to fuel minds and forge futures in Downstream Refining.”

– Is AI/Machine Learning Model for the Fuel Blending System A Viable Alternative? – Part III" sku=""]
Explore Our Courses