Abstract
Increase in global natural gas production over the last 15 years has led to the use of new and untapped reservoirs including high pressure-high temperature ones in order to meet the consumer demands. As flow characteristics of gas in various environments such as porous media, wells, and pipelines are influenced by its viscosity, it is an important parameter for petroleum engineers. In this work, an immense gas viscosity dataset consisting of 3017 laboratory data points was used for properly implementing two smart techniques: radial basis function neural network with two training algorithms as well as multilayer perceptron neural network with four training algorithms. By using these techniques, various models with high accuracies were developed for viscosities estimation of gas mixture, pure methane, and pure nitrogen at high pressures (5000–25000 psia) and high temperatures (100–1880 °F). The radial basis function (RBF) neural network with ant colony optimization (ACO) (namely RBF-ACO model) is considered as the best model. Average absolute relative errors of the aforementioned model for estimating pure methane, pure nitrogen and gas mixture viscosities are 0.36 %, 0.49 %, and 1.76 %, respectively. RBF-ACO model provides better results comparing with other presented empirical models. Also, RBF neural network optimized by particle swarm optimization (PSO) shows a high error for estimating the viscosities of methane and gas mixture and RBF neural network optimized by genetic algorithm (GA) yields a high error for estimating the viscosity of gas mixture. Afterwards, effects of input parameters on the viscosity value obtained by RBF-ACO model, were investigated using the relevancy factor. Finally, based on numerical simulation, a sensitivity analysis was conducted for measuring the uncertainty of cumulative gas production resulted from gas viscosity estimation for a high pressure-high temperature gas reservoir. This process indicates that accurate estimation of gas viscosity plays an important role in reliable estimation of cumulative gas production.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have