How AI Narrows the Planning-Operations Gap at Oil Refineries

Oil refineries are increasingly turning to industrial artificial intelligence to reconcile the persistent disconnect between meticulously crafted production plans and the unpredictable realities of the plant floor. Challenges such as fluctuating feedstock quality, varying operating rates, and equipment conditions often cause units to deviate from initial planning models, impacting crucial metrics like yields, product quality, and energy consumption.

Hussein Zein, Emerson’s vice president for Saudi Arabia and Bahrain, explained that these advanced AI models enhance, rather than replace, traditional planning systems. By integrating engineering expertise with actual plant data, the hybrid models offer a more accurate representation of the nonlinear behavior inherent in refinery units. This approach allows for better adaptation to changing market conditions, feedstock characteristics, and operating levels, which often simplify or distort linear models.

“

Conventional models often rely on linear or semi-linear representations of unit behavior.

Conventional planning models, often relying on linear or semi-linear representations, struggle to capture the intricate interactions within refinery units like reactors and separators. When these simplified plans move to the operational stage, discrepancies between predicted and actual performance can emerge, necessitating time-consuming manual adjustments by engineers. The effects of these gaps can ripple through the entire refinery, influencing everything from feedstock blending and production targets to margin forecasts and inter-unit coordination.

Emerson and Aramco reported that these hybrid models have demonstrated remarkable predictive accuracy, reaching up to 98.5% in specific units such as continuous catalyst regeneration units and catalytic reformers. While this high accuracy was achieved in select areas and is not an overall average, it signifies a major step forward in optimizing refinery operations. Teams are actively working to extend this advanced modeling approach to other critical units, like hydrocracking, further refining operational efficiency and predictive capabilities.

“

The approach does not replace the conventional models that refineries have relied on for decades. Instead, it enhances them to better represent the nonlinear behavior of units under different feedstocks and operating conditions.