Uncertainty analysis and optimization in cyber-physical systems of reservoir production

Abstract

Reservoir closed-loop production optimization management, which shows promising validity in oilfield exploitation engineering, is a typical cyber-physical system. At present, many scholars mainly optimize single reservoir model by some algorithms. However, compared to normal strategy of conventional numerical simulation model, single model of reservoir fails to reflect the reliability in probability. Because the realizations of single model are limited by geological uncertainty, reservoir closed-loop production optimization management based on multi-model was proposed. Furthermore, through ensemble Kalman filter (EnKF), multi-model can be updated on real-time. The expectations of net present value (NPV), based on the updated reservoir models, are calculated as objective function, which are optimized by simultaneous perturbation stochastic approximation (SPSA) to perform optimization production procedures. Example applications demonstrate that the reasonable geological model estimations can be obtained from the updated multi-models which obviously decrease the uncertainty of geology. Furthermore, multi-model optimized control strategy improves both injection effects, economic benefits and decreases the risk of development compared to the normal strategy.