Journal of the Korean Society of Mineral and Energy Resources Engineers ISSN:2288-0291(Print) 2288-2790(Online) 한국자원공학회지

The major impediment to formal optimization of large producing fields is the cost of computing the state of the objects being optimized. Previous studies for prediction of reservoir performance have used local inflow performance relationships or material balance models. These approaches ignore flow interactions among wells during the optimization process, often resulting suboptimal operations. In this study, a new polynomial neural network with layer over-passing structure has been developed to replace a relatively time consuming reservoir simulator through robust and systematic search algorithm. The networks are trained from a representative samples of simulations, which form a re-useable knowledge base of information for addressing many different management questions. The proposed approach reduces significantly computational effort for optimization of the development scheme within reasonable accuracy and outperforms other neural network models.