Predicting distillation properties of fuel blends using Machine Learning

Distillation properties of gasoline are regulated strictly to assure proper performance of the final product. The evaporated volume percentage at 70, 100, and 150 °C (E70, E100, E150) and the final boiling point are regulated in Europe by the standard EN 228. Blending various gasoline components affects the distillation values in a non-linear fashion, making the prediction of these properties challenging. Furthermore, new renewable components will require ways to predict their effects on gasoline specifications. We used Machine Learning to create a model to predict the distillation points of gasoline blends based on the blending recipe. Real refinery data was used, together with blends created for R&D purposes. The values predicted were evaporated volume at 70 and 100 °C (E70 and E100). Altogether 11 different machine learning models were trained, cross-validated and tested using nine different pre-processing methods. It was found that Support Vector Regression (SVR) was the most effective at predicting the distillation points. It achieved a mean absolute error (MAE) of 1.29 and 0.66 respectively for E70 and E100. Compared to a widely used linear model which had MAEs of 12.2 and 3.27, the SVR model was found to be substantially more effective. In addition, the SVR model maintained its performance when used on oxygenated blends, where the linear model was highly inaccurate. It was also found that higher distillation points behave more linearly and therefore there's less need for complex non-linear models near the higher end of the distillation curve. The methodology presented in this work can be used to predict other distillation points.