History matching of petroleum reservoirs employing adaptive genetic algorithm

N C Chithra Chakra,Deoki N Saraf

doi:10.1007/s13202-015-0216-4

Abstract

History matching is an important phase in reservoir modeling and simulation process, where one aims to find a reservoir description that minimizes difference between the observed performance and the simulator output during historic production period. For the automatic history-matching problem through reservoir characterization, a global optimization method called adaptive genetic algorithm (AGA) has been employed. AGA is a relatively new optimization technique which has adaptive genetic operators that dynamically update the crossover and mutation probabilities in each generation according to fitness of population to reach optimal solutions. Only critical parameters such as porosity and permeability distributions have been found by the optimization route, the rest being adjusted manually, if necessary, in the present study. History-matching results from AGA were also compared to those from conventional simple genetic algorithm (SGA). The AGA and SGA techniques were utilized to determine permeability map that resulted in a good match for past field history. The methodology was tested and validated by implementing it on a known 2D synthetic black-oil reservoir, which was subsequently used for a real-field reservoir situated in Cambay Basin, Gujarat, India. AGA methodology was able to outperform the SGA in terms of reduced computation load and improved history match.

Highlights

History matching, a process in which certain input parameters to the reservoir simulator such as porosity, permeability, thickness, saturations, depth of oil/water contact, connate water saturation, relative permeability, etc. are altered singly or collectively to obtain a match between simulator prediction values and observed historic data relating to flow rates of oil, gas, water, pressures, gas–oil ratio (GOR), WOR, and their variations as a function of time
The adaptive genetic algorithm was subsequently used for the history matching of the same 2D synthetic reservoir
The successful application of genetic algorithm in extracting a realistic permeability map of a 2D synthetic reservoir showed the technique as a promising optimization tool toward automatic history matching

Summary

Introduction

A process in which certain input parameters to the reservoir simulator such as porosity, permeability, thickness, saturations, depth of oil/water contact, connate water saturation, relative permeability, etc. are altered singly or collectively to obtain a match between simulator prediction values and observed historic data relating to flow rates of oil, gas, water, pressures, GOR (gas–oil ratio), WOR (water–oil ratio), and their variations as a function of time. A process in which certain input parameters to the reservoir simulator such as porosity, permeability, thickness, saturations, depth of oil/water contact, connate water saturation, relative permeability, etc. The spatial inhomogeneity and anisotropic nature of the reservoir rocks result in very large dimensionality of the reservoir model which make this task complex. Reservoir history matching is considered to be an inverse problem, where one seeks to back calculate the system parameters from a given system output. The reservoir production data are available, but the reservoir static parameters (permeabilities and porosities) are unknown which need to be estimated. The spatial variation of these properties due to rock heterogeneity makes it an ill-posed problem since a very large number of permeability maps may lead to the same output, where most of these may be unrealistic.

Objectives

Results

Conclusion