Development and Application of Inverse Model Using Genetic Algorithm for Reservoir Characterization

This article presents the development of the inverse model and its application for the characterization of heterogeneous reservoir. Until now, in a area of reservoir characterization, researches that reduce their uncertainty have been performed by integrating static data of cores, logs, seismic, experiment, etc., and dynamic data of production and 4D seismic data. In order to integrate dynamic data into reservoir characterization models, an optimization algorithm must be used in order to minimize the difference between observed and calculated data. In this study, we developed the inverse model using a genetic algorithm, and the optimization method for the inverse calculation was the real-coded genetic algorithm, which is not sensitive for initial value and is possible to search the global optimum. By utilizing the developed model, we performed the characterization to estimate the distributions of porosity and permeability. Firstly, in order to determine the optimal constraint for this system, the inverse calculations were carried out by increasing the maximum values of initial constraints estimated by Kriging method. To obtain results, we determined the case for 2.5 times the maximum permeability as the optimal constraint. From the results of inverse calculations to determine the optimal parameters, such as population size, operator and operation probability for this system, it was found that the selection operator was more sensitive than the crossover, and the roulette wheel operator was more suitable than the tournament as the selection operator. Also, the optimal population size and crossover probability were determined to be 450 and 0.85, respectively. Finally, as the result of characterization for this reservoir system, it is proved that the developed model have been generated the favorable porosity and permeability distributions.