Development of Artificial Neural Networks Applicable to Single Phase Unconventional Gas Reservoirs with Slanted Wells PDF Download

Author: Chris Raymond Affane Nguema
Publisher:
ISBN:
Category :
Languages : en
Pages :

Get Book Here

Book Description
The increasing demand in energy has strengthen the dependence on fossil fuels. On the other hand, the conventional hydrocarbon reservoirs are depleting rather quickly which prompted an important study of the hydrocarbon reservoirs from unconventional reservoirs. Over the last two decades, the improvement in technology and recovery methods has allowed the industry to extract hydrocarbon from unconventional reservoirs. There are been important advancements in drilling and reservoir engineering technologies.In order to overcome some of the costs associated with the exploitation of those reservoirs, an extensive use of techniques such as directional drilling to has been largely recommended and has proven to be more efficient. Directional drilling allows to control the direction of the wellbore to increase the contact with the target or pay zone location among other significant benefits.Reservoir simulation refers to constructing computer models to gain a better understanding of reservoirs. It is mostly used to predict the flow of fluids or to match the properties of the reservoir. However, it has shown to have be limited when not enough information about the reservoir is available.Artificial neural network (ANN) is a technique used in many fields that has been able to compensate for some of the limitations associated with other approaches such as reservoir simulation. It relies on observed data to build highly non-linear and strong links among them that make it possible to obtain a more accurate prediction of the missing information.The main goal of this study is to develop an ANN tool for a single phase unconventional gas reservoir that can predict reservoir properties such as porosity, permeability and compressibility. The tool applicability has been extended for a large range of data. It provides predictions from two network structures, cascade forward backpropagation and radial basis function with an option to compare them. Each of the ANN model, therefore, differs by the type of networks used, the porosity system (single or dual), the well inclination (0 to 90), and the transient data available (pressure or production rate). The performance of each network was evaluated using the average percent error, the mean bias error (MBE), and the root mean square error (RMSE).