The modeling of complex hydrocarbon mixtures is a current issue. The presently available analytical techniques are insufficient alone to fully characterize the molecular details of heavy oil fractions to the level for new development of a molecular-level kinetic model. Stochastic reconstruction (SR) methods which build a set of molecules that mimic the properties of complex mixtures by using partial analytical data help to overcome this drawback. Although the classical SR algorithm produces reasonable molecule sets for light and medium fractions, performance degrades for heavier fractions. The main reason for this is the lack of structural parameters needed to define the variations in side-chain and ring configurations. As an extension, a novel structural parameter set including specific parameters for ring and chain configurations was implemented to the SR algorithm. In addition to this, in order to ensure an extensive structural connection between the generated molecules and the experimental data, the 1H NMR spectrum was divided into six different regions, and these hydrogen types were used in an objective function. In order to validate the SR with an extended parameter set, it has been applied to six different petroleum asphaltenes. The extended parameter set resulted in a decrease in the objective function value between 45% and 85% compared to the basic parameter set. Moreover, the extended parameter set increases the fitting ability of the SR algorithm without sacrificing the compositional space of the generated molecules.

No abstract

An approach for the stochastic reconstruction of petroleum fractions based on the joint use of artificial neural networks and genetic algorithms was developed. This hybrid approach reduced the time required for optimization of the composition of the petroleum fraction without sacrificing accuracy. A reasonable initial structural parameter set in the optimization space was determined using an artificial neural network. Then, the initial parameter set was optimized using a genetic algorithm. The simulations show that the time savings were between 62 and 74% for the samples used. This development is critical, considering that the characteristic time required for the optimization procedure is hours or even days for stochastic reconstruction. In addition, the standalone use of the artificial neural network step that produces instantaneous results may help where it is necessary to make quick decisions.

Stochastic reconstruction (SR) methods are used to generate a series of molecules that mimic the properties of complex mixtures using partial analytical data. Determining a quantitative composition using these methods is limited by the property prediction methods used. This paper addresses the use of two key measurements in the characterization of petroleum fractions, namely density and boiling point distributions. It is known that the different methods used in estimating these two basic properties have varying error rates. Boiling point prediction performances of the various group contribution methods were tested via the molecular library established for molecules that can be found present in the petroleum fractions. It has been observed that the combined use of these methods results in close to a 50% reduction in sum of squared errors than any single method. The predictive performances of the density calculation methods were similarly tested. The best-calculated density results were found via the Yen–Woods method with support from the linear mixing rule based on molar fractions.