An Artificial Neural Network Implementation for Evaluating the Performance of Cyclic CO2 Injection in Naturally Fractured Black-oil Reservoirs
Author | : Mehmet Yavuz |
Publisher | : |
Total Pages | : |
Release | : 2017 |
ISBN-10 | : OCLC:1005105675 |
ISBN-13 | : |
Rating | : 4/5 ( Downloads) |
Download or read book An Artificial Neural Network Implementation for Evaluating the Performance of Cyclic CO2 Injection in Naturally Fractured Black-oil Reservoirs written by Mehmet Yavuz and published by . This book was released on 2017 with total page pages. Available in PDF, EPUB and Kindle. Book excerpt: Increasing oil and gas consumption of countries has increased demand for oil for their energy needs.Therefore, enhanced oil recovery (EOR) methods have become attractive for these countries toreach their required production since only a small portion of the original oil in place (OOIP) canbe produced with primary and secondary production techniques. Gas injection, specifically cyclicCO2 injection, has been found to be an effective EOR method to improve oil recovery in naturallyfractured reservoirs. Fractured reservoirs favor this process because the fractures in this systemprovide an extensive contact area for injected gas to penetrate into the reservoir and diffuse intothe low-permeable matrix. These fractures contribute to an easy delivery not only for injecting thegas but also for producing the oil. Therefore, in cyclic CO2 injection, a single well is used for bothinjection and production. Designing such a process requires optimization of the design parameters,which are the amount of gas injected, injection rate, and the duration of injection, soaking andproduction periods. Commercial reservoir simulations require heavy computational time in orderto go through a large number of possible scenarios to determine the optimal design parameters.Thus, an artificial neural network (ANN) that is able to solve non-linear and complex relationshipsin a remarkably short time has been used as an alternative to overcome these limitations in recentyears.In this research, artificial neural networks (ANNs) are developed to determine the performanceindicators and the optimum design parameters of a cyclic CO2 injection process in naturally fracturedreservoirs. The developed models in this study comprise a wide range of reservoir characteristicsas well as the relative permeability and reservoir fluid compositions. The first developedmodel is forward-looking and related to the reservoir performance. This model basically mimicsthe commercial reservoir simulation to predict the cumulative oil production but provides resultsin seconds when compared to commercial reservoir simulators. The second model, inverse-looking,goes beyond the numerical simulators and provides not only the design parameters of a cyclic CO2injection process but also opportunities for engineers to evaluate a huge number of possible scenarioswith at least an 85 percent confidence margin of error to select the best one in the shortest time. The last developed model is also an inverse model and is used for history matching. The objectiveof this model is to predict some of the uncertain reservoir properties by using the field data andproduction history.