Estimation of properties distribution of C7+ by using artificial neural networks

In characterization of wide boiling range heptane plus (C7+) fractions in addition to bulk properties such as molecular weight (MW), specific gravity (SG), etc., properties distribution is also required. Bulk properties can be measured easily but determination of properties distribution is more costly and time consuming. In this work an artificial neural network (ANN) has been trained and tested with 62 samples (881 data points) of crude oil and gas condensate with complete characterization from all over the world. Inputs of the ANN are the bulk molecular weight (MWb), bulk specific gravity (SGb) and cumulative weight fraction (CXw) and the outputs include properties distribution for boiling point (Tb), molecular weight (MW) and specific gravity (SG). The estimated properties distribution is in a good agreement with the experimental results.

Research Highlights
► Properties distribution for C7+ fractions were predicted precisely.
► The bulk molecular weight and specific gravity were used as input parameters to a trained Artificial Neural Network (ANN) for prediction complete properties distribution.
► ANN can be used for prediction properties distribution by only two bulk properties.