Stochastic Optimization of Gas Cycling in Gas Condensate Reservoirs

Abstract

Abstract A major challenge in gas condensate reservoirs is to minimize the condensate drop-out which is mostly heavier fractions representing the rich part of the gas accounting for most of its heat capacity and consequently its price. Most widely used technique to improve gas condensate reservoir performance is gas cycling. In this study, gas cycling in the form of produced lean gas re-injection was considered. Parameters which impact the effectiveness of gas cycling processes are often evaluated by performing limited sensitivity studies on all or some parameters. A more efficient technique that uses a stochastic optimization algorithm to automatically optimize the parameters governing the gas cycling process is proposed in this paper. Stochastic optimization of well placement and gas cycling operational parameters has not been reported previously in the literature. Global optimization techniques such as Covariance Matrix Adaptation Evolution Strategy (CMA-ES) and Differential Evolution (DE), both of which are evolutionary algorithms, were used to optimize the producer and injector well locations, fraction of produced gas to be re-injected, and production and injection rates. A modified reservoir model from the Third SPE Comparative Solution Project was used in this study. The optimization was done on several realizations and utilizing a numerical compositional simulator. Net Present Value (NPV) of the gas cycling project and field average oil saturation (FOSAT) were used as the objective functions separately. The optimum parameters and well placement would maximize NPV and minimize FOSAT. The results show that NPV can be significantly increased if the parameters are optimized simultaneously rather than sequentially suggesting the importance of optimizing well placement alongside the operational parameters. The results also show that optimization of well placement, sales volume and production rates remarkably reduces the condensate saturation in the reservoir. Furthermore, DE was seen to be more robust and performed better than CMA-ES. The presented technique provides a base for oil companies to consider stochastic optimization in designing field development plans for gas condensate reservoirs.