Digital tool for predicting the quality indicator of diesel fuel at the hydrotreating unit of Volgogradneftepererabotka

The complex process of oil refining and production of petroleum products includes the stage of forecasting the quality of output products. In case of insufficiently accurate forecasting, there is a risk of increased rejects, loss of resources, reduction of quality output. To increase the efficiency of oil refineries, there is a demand for improvement of quality indicators of oil refining units by reducing the deviation of real values of product quality indicators from the specification boundary (acceptable quality values). Previously, models were used in predicting the quality of products (diesel fuel), which did not always contribute to the output of quality products. There is a risk in the production of petroleum products: - out of specification. This results in product rejects. - leaving a significant reserve up to specification limits specification boundaries, which leads to loss of resources and reduction of quality products output and increase of waste. Therefore, the forecasting stage in the production of petroleum products is very important. In addition, predictive quality models can reduce the number of complex and expensive laboratory tests, due to the improvement of the quality of the output product. In order to improve the accuracy of existing models, machine learning based models were developed. A digital tool, a virtual analyzer (VA), based on a machine learning model was formed, which allows predicting the flash point of hydrotreated diesel fraction in real time with an error of no more than 2°C (2°C better than the current solution) without additional laboratory tests. A VIRTUAL ANALYZER is a model that shows the current value of a product quality indicator based on refining process data such as temperature, pressure, and archived laboratory measurements. Improving forecast accuracy with AI-based models helps stabilize quality indicators of high-margin products, reduce costs and increase production volumes.