Predicting the Onset of Asphaltene Precipitation in Heavy Crude Oil Using Artificial Neural Network

Abstract

In flow assurance issues, asphaltene precipitation in crude oil tends to block the wellbore, the production tubing, the flowlines and as well as surface facilities, thereby reducing the quantity of crude oil that could be recovered during recovery, hence there is need to predict the onset conditions under which asphaltene would precipitate. Previous models (thermodynamic/colloidal) attempt to predict the onset of asphaltene precipitation using the solubility parameter, crude oil/n-alkane mixture and the refractive index of asphaltene. However, due to the constraint in handling numerous and complex data set, this work attempts to predict the onset of asphaltene precipitation (onset solvent to bitumen/asphaltene ratio as a function of temperature and pressure) using artificial neural network (Neurosolution 6).

The results obtained show that the onset solvent bitumen ratio obtained using the neural network was close to the experimental (desired) onset solvent bitumen ratio (MSE of 0, Err% 0.0553 for the training set and MSE of 0.006581, Err% of 3.343 for the testing set) with an average absolute deviation of 3.56.

Artificial neural network is a robust predictive tool for predicting the onset of asphaltene precipitation in heavy crude oil.

Keywords: Asphaltene, Precipitation, Structure of asphaltene, Neural Network, Onset and amount