International refinery blending multiple feedstocks of fuel oil.
Project Information
Client Type
Refinery
Products Used
Machine Learning
Geography
International
International refinery that blends fuel oil from multiple input feedstocks that desires high prediction accuracy for density, sulphur, viscosity, flash point, and p-value. There is incomplete feedstock data due to process uncertainty and measurement error.
Problem
Problem
International refinery that blends fuel oil from multiple input feedstocks that desires high prediction accuracy for density, sulphur, viscosity, flash point, and p-value. There is incomplete feedstock data due to process uncertainty and measurement error.
Solution
Validere’s data scientists produced various machine learning models which delivers real-time estimates of previously unavailable data. Providing the ability to predict final blend quality and integrate the information across multiple client systems.
Results
Validere’s predictive machine uncovered$120k /month in total expected value, and 40-65 monthly hours of client team hours saved (Process engineering & optimization technical team).