You are on page 1of 2

Oil Field Development Using Neural Network Model

Fariba Salehi1, Ronak Azizi2, and Arnoosh Salehi3


1
Department of Computer, Faculty of Science, Islamic Azad University, Karaj branch, Iran
salehi60@yahoo.com
2
Department of Mathematics, Faculty of Science, Islamic Azad University, Karaj branch,Iran
3
Department of Reservoir Engineering, Pars Oil and Gas Company (POGC), Tehran, Iran

Abstract. This paper applies a methodology to structure the field development


schemes using artificial neural network in conjunction with numerical
simulation for Sirri oil field. In this method, a few field development scenarios
are studied using a numerical simulator. The results of these studies are used to
train the neural network. The trained neural network model is then used as a
predictive tool for field development purposes.

Keywords: artificial neural network, numerical simulation, oil field


development.

1 Introduction

Reservoir studies are performed to predict the future performance of a reservoir based
on its current state and past performance and to explore methods for increasing the
ultimate recovery of hydrocarbons from a reservoir.
Reservoir simulator is a sophisticated computer program which solves a system of
partial differential equations describing multiphase fluid flow in a porous reservoir
rock. The bases for reservoir simulators are: reservoir engineering principles, a set of
partial differential equations to describe the flow of fluids through porous media,
finite difference techniques to obtain numerical solutions for the partial differential
equations for fluid flow, and computer programming to perform the calculations
electronically.
The reservoir is first divided into segments, or blocks, using X, Y, and Z axes.
Rock and fluid properties are then assigned to each block to describe the reservoir
system. Computations are carried out for all phases in each block at discrete time
steps. The results, or output, usually consist of production volumes and rates, pressure
and saturation distributions, material balance errors, and other process specific
information provided at selected time steps.
Some of the main challenges faced during a field development study include: The
presence of large degrees of freedom to be examined; the possibility of having
extensive interaction of parameters the sparse nature of the field data; the time
constraints encountered.
Conventionally, in field development studies, the existing production data is
utilized to perform a history match. The optimum locations of the wells are then

D.-S. Huang, L. Heutte, and M. Loog (Eds.): ICIC 2007, CCIS 2, pp. 29–38, 2007.
© Springer-Verlag Berlin Heidelberg 2007
30 F. Salehi, R. Azizi, and A. Salehi

decided by running the reservoir simulator by placing the wells at various locations in
the reservoir. In field development studies, a large number of scenarios which result
in a time consuming and expensive process must be considered.
Artificial neural network are predictive tools capable of solving highly non-linear
problems. A neural net consists of a large number of simple processing elements
called neurons. Each neuron of the neural network is connected to other neurons by
means of directed communication links, each with an associated weight, which
represents information being used by the net to solve the problem. At the beginning,
these connections are assigned independent weighting factors. The input to each node
is multiplied by its associated weighting factor and then summed with the product of
each of the other input nodes and their respective weighting factors. An activation
threshold is then added to this sum, and the result is processed by a non-linear
activation function within the node. A typical multilayer neural-network model is
shown in Figure 1. As shown, the artificial neural-network model has an input layer,
and output layer and one hidden layer. The input layer contains an array of variables
into which the input data of the system are read from an external source. Similarly,
the predicted data or results, which can be multiple vectors, are written in the output
layer [1-3].

Input Hidden Output


Layers Layer(s) Layer(s)

Input
Output
Data
Result

Fig. 1. Schematic of multilayer neural network model

In recent years, there are increasing applications in the oil and gas industry. Neural
network has been used in various petroleum engineering areas, such as geology,
geophysics, drilling and completion, formation evaluation, production and
stimulation, reservoir engineering and economic etc [4-6].
This paper applies a methodology to structure the field development schemes, a
few field development scenarios are studied using a numerical simulator. The results
of these studies are used to train the neural network. The trained neural network is
then used as a predictive tool for field development purposes. The use of neural
network models significantly reduces the number of computationally demanding
numerical simulation runs. In addition, the potential well locations are not restricted
solely to the centers of the grid blocks.

You might also like