Evolving robust intelligent model based on group method of data handling technique optimized by genetic algorithm to predict asphaltene precipitation

摘要

Precipitation of asphaltene during primary production of hydrocarbon reservoirs leads to formation damage and well bore plugging. Therefore, proposing an accurate model to estimate asphaltene precipitation under various operating and thermodynamic conditions are crucial. In this study, a new mathematical model based on the integrating group method of data handling (GMDH) with genetic algorithm has been developed to predict asphaltene precipitation as a function of reservoir pressure and temperature, crude oil API, bubble point pressure, Saturated-Aromatic-Resin-Asphaltene (SARA) fractions and mole percent of non-hydrocarbon gases. Genetic algorithm technique has been applied to optimize the most appropriate network structure of GMDH model. In order to accomplish modeling, asphaltene precipitation of different crude oils from a number of Iranian reservoirs at wide ranges of operating conditions have been measured experimentally and applied for network construction. The accuracy of developed model has been evaluated by both statistical and graphical error analysis techniques. The average absolute relative deviation of the proposed model is 3.65%, which indicates model predictions are in excellent agreement with experimental data. Also, the comparison of developed GMDH model with scaling equation and least squares support vector machine (LSSVM) reveals the superiority of the proposed GMDH structure in prediction of asphaltene precipitation over scaling equation and LSSVM technique. In addition, the Leverage approach has been applied to detect suspected data. The results show that all experimental data are reliable and located within the applicable domain of developed model. Finally, a comprehensive sensitivity analysis based on the relevancy factor has been carried out which shows that percentages of resin and saturated components have the largest direct and inverse impacts on asphaltene precipitation, respectively.