Well Placement Optimization Using Hybrid Optimization Technique Combined with Fuzzy Inference System

Abstract

Decision on the location of new wells through infill drilling projects is a complex problem that depends on the reservoir rock and fluid properties, well and surface facilities specifications, and economic measures. Conventional approach to address this is a direct optimization that uses the numerical flow simulation. However, this is computationally very extensive. In this study the authors use a hybrid genetic algorithm (HGA) optimization technique based on the genetic algorithm (GA) with helper functions based on the polytope algorithm and the neural network. This hybridization introduces hill-climbing into the stochastic search and makes use of proxies created and calibrated iteratively throughout the run. It is emphasized that the numerical models are constructed based on scarce data, hence the simulation forecasts are uncertain and consequently the deterministic global solution is not achievable. To resolve this the authors used a Fuzzy Inference System (FIS). For economic evaluation they use net present value (NPV). Therefore, the FIS output is incorporated into the NPV, and a new objective function called corrected NPV (CNPV) is constructed. The authors validate the method by optimizing the placement of water injection wells in a section of an oil reservoir located in the west of Iran by maximizing the CNPV. It was observed that the number of simulations required to find the optimal well configurations were reduced significantly by using HGA.