Determination of Crude Oil Saturation Pressure Using Linear Genetic Programming

Saturation pressure is a PVT property predicted either by equations of state or empirical correlations. Both are known for their limited accuracy. In this work, a new model is developed to estimate crude oil saturation pressure using linear genetic programming (GP) technique. A total of 131 crudes covering wide ranges of composition and reservoir temperature and different geographic origins were used to build and test the model. The database used consists of reservoir fluids composition (N-2, CO2, H2S, methane to hexane, and heptane plus components), sepecific gravity, and molecular weight of the heptane plus fraction, reservoir temperature, and the experimentally measured saturation pressures. The data were randomized to exclude the effect of crude geographical origin and divided into three data sets (training, validation, and testing). The training and validation data sets were composed of 41 crude oils each, and they were used to build the model. The rest of the data were used to blind test the model developed. The proposed model efficiency was tested against two known equations of state (Soave Redlich Kwong and Peng-Robinson) in addition to Elsharkawy empirical correlation. The testing process indicates the superiority of the proposed model in term of average absolute relative error. The proposed model is much simpler than the emprical correlation predicting saturation pressure as a function of only three input variables, namely the methane mole fraction, molecular weight of the heptane plus component, and the reservoir temperature. In addition, it eliminates the need for splitting and characterizing the heavy fraction necessary for the equation of state models.