The Prediction of the Performance of an Oil Reservoir by Proxy Model: A Case Study

The Prediction of the Performance of an Oil Reservoir by Proxy Model: A Case Study

Allahyar Daghbandan (Department of Chemical Engineering, University of Guilan, Rasht, Iran) and Seyed Mahdi Chalik (Department of Chemical Engineering, University of Guilan, Rasht, Iran)
DOI: 10.4018/IJCCE.2015070104
OnDemand PDF Download:
$30.00
List Price: $37.50

Abstract

Simulation model of an undeveloped oil reservoir is full of uncertainty. Assessing the effect of these parameters on the simulation results, is very important task in reservoir engineering. Making a proxy model is a method for forecast reservoir performance under different production scenarios. In this study, GMDH-type neural network is used as a proxy model and also Experimental Design theory is used to get the most information full data set which is applied to the input of the neural network. The traditional way needs to very large number of simulation but this method is very time consuming and costly. A sensitivity analysis was conducted to understand the most important initial parameters. In this study, proxy model is created to predict FOPR, Field Oil Production Rate, in a reservoir under immiscible gas injection scenario to 15 years.
Article Preview

1. Introduction

This study has been performed on an underdeveloped oil reservoir that situated in the central part of Iran, Fars state (Figure 1). Based on the field owner wanted, reservoir name is confidential. In recent years, EOR methods have been used in many reservoirs to maintain field pressure and enhance oil recovery. In this study, immiscible gas injection method as EOR method has been used. To predict the performance of an oil reservoir, there are many uncertainty parameters. The purpose of this study is to find the main uncertainty parameters with the most significant affecting on the behavior of the oil reservoir. This work traditionally done by many numbers of simulator runs that is very time consuming and in addition, in this way is ignoring Effect of uncertainty parameters on each other. Experimental Design method is used for two purposes to entrap the total uncertainty with minimum numbers of simulator runs and making proxy model. Experimental Design method and proxy model have been applied frequently in petroleum industry. For example, Damesleth et al (1991) used the experimental design theory to develop a model and to analyze available Uncertainty in reservoir. Dejean et al (1999) used the Experimental Design to study the effect of uncertainty parameters on behavior of the reservoir and acquire optimum product. AL-Fattah and Startzman (2001) used proxy model for Predicting natural gas production. Rise could make a proxy model by using Experimental Design that this model has been applied finally to study risk analysis (Rise, 2006). Yu et al (2008) making proxy model with help genetic programming and it has been applied for history matching. Asadisghandi and Tahmasebi (2011) compared back-propagation neural network and empirical correlation for prediction of oil PVT properties in Iran oil fields. Experimental design in reservoir simulation was introduced as an integrated solution for uncertainty analysis (Moeinikia & alizadeh, 2012). Proxy model resultant from an artificial neural network were used for predicting the recovery performance of surfactant polymer floods (Al-Dousari & Garrouch, 2013). In this study, our aim was to combine Experimental Design method and artificial neural network to make a proxy model which has been used to forecast Production of Undeveloped oil reservoir.

Figure 1.

Two views of the studied reservoir

2. Experimental Design

Experimental Design methodology is the most significant stage in making an efficiency proxy model. Experimental Design method is a table of set of tests and simulations which maximum information about system can be acquired with minimum numbers of tests and simulations. In fact, Experimental Design methods are effectively distributing Simulator Runs within ranges of uncertainty parameters, therefore are minimizing the numbers of runs required to study system. In this study, two type experimental design is used as one variable-at a time design and three-level full factorial design Which will be described below (Cox & Reid, 2000; Mason et al., 2003).

Complete Article List

Search this Journal:
Reset
Open Access Articles: Forthcoming
Volume 6: 2 Issues (2017): 1 Released, 1 Forthcoming
Volume 5: 2 Issues (2016)
Volume 4: 2 Issues (2015)
Volume 3: 2 Issues (2013)
Volume 2: 2 Issues (2012)
Volume 1: 2 Issues (2011)
View Complete Journal Contents Listing