Optimization of a reservoir development plan using a parallel genetic algorithm

Abstract

A parallel genetic algorithm has been applied successfully to design a production plan that is substantially superior to that obtained using a conventional engineering approach. The reservoir, a dipping structure, was expected to yield optimum production using a rolling line drive from downdip to updip positions. The simulation allowed for 3800 positions for each of 11 wells, giving a total of 1.3×10 31 options. The genetic algorithm sampled 1650 of these and was able to identify seven solutions that would increase production by over 30% compared with the rolling line drive. In contrast, a random search using 850 samples managed to find only two plans that improved production; in each of these cases the improvement was less than 1%.

Process systems engineering perspective on the planning and development of oil fields

“…Using proxy models Parallelizing simulations on multinode computers Using grid/distributed computing …”

Perspectives on the Design and Planning of Oil Field Infrastructure

“…The simulation cost is reduced by proxy modeling [4,6,16], parallelization and applying distributed computing [21,22], or using the streamline simulators, rather than finite difference simulator [23].…”

Contrastive analysis of three parallel modes in multi-dimensional dynamic programming and its application in cascade reservoirs operation

“…Therefore, we have to put the emphasis on the second approach that is to reduce the computational time of DP by using advanced computer technology. In recent years, the rising popularity of multi-core processors provides the necessary hardware for the implementation of parallel computing, and the CROO problem not only contains a large amount of calculations, but also has a certain degree of parallelism which is conducive to the implementation of parallel computing (Chen et al, 2008(Chen et al, , 2009Carter and Matthews, 2008;Tsai et al, 2009;Pinto et al, 2013). So, it has become an inevitable trend to improve the computational efficiency of DP by combining parallel processing technology (Zheng et al, 2013;Bruno et al, 2013;Zhan et al, 2013).…”