Characterization of reservoir fluid properties, especially bubble point pressure (Pb) and formation volume factor (Bo), is one of the most significant steps in the process of reservoir studies and management. Furthermore, the reservoir parameter estimation is strongly dependent on these two parameters in the reservoir simulation and material balance calculations. These PVT properties are predicted through two different methods, including equations of state and empirical correlations. In this work, a novel approach is proposed to develop reliable optimized models for estimation of PVT properties at various reservoir conditions. This paper introduces Particle Swarm Optimization (PSO) and Genetic Algorithm (GA) as population-based stochastic search algorithms to optimize the weights and biases of networks, and to prevent trapping in local minima. GA and PSO were used to minimize the neural network error function. A total of 569 data sets were used to develop the models to estimate the values of Bo and Pb. The results of the proposed models are compared to the results of conventional correlations to enable the model predictions to provide a good level of accuracy for the results in training and testing stages. The results show that artificial neural networks (ANNs) remarkably overcame the inadequacies of the empirical models where PSO-ANN improves the performance significantly. Additionally, the regression analysis releases the efficiency coefficient (R2) of 0.96 for FVF and 0.99 for Pb, which can be considered very promising.
نت نشــرهای ویژهمشاهده ی همه »
لطفا نظر خود را با دیگران به اشتراک بگذارید