A Genetic Algorithm-based Artificial Neural Network for Prediction of Oil PVT Properties in the Upstream Industries

At the level of preservation of oil wells in upstream oil industries, complicated experimentations, called PVT, are done for the recognition of reservoir fluid properties. Problems such as probable dangers, time consuming, and the inaccuracy of samples and limitations in temperature and pressure have led to tend to increase the use of intelligent methods in this field. In this study, in order to avoid the mentioned problems and find the complex and nonlinear relationships between data and PVT experiments, artificial neural network has been used. Because the suitable choice of the initial weights increases the neural network efficiency, genetic algorithm is used in order to adjust the initial weights. For evaluating the proposed approach, Iran oil reservoir fluid properties are implemented. The results of research showed that the use of genetic algorithm-based artificial neural network, in contrast to the classical methods, predicts the reservoir fluid properties in a shorter time and with high accuracy. Therefore, the proposed neural network can be seen as a powerful approach toward the prediction of Iran oilfield oil PVT properties.