Development and Testing of Artificial Neural Network Based Models for Water Flooding and Polymer Gel Flooding in Naturally Fractured Reservoirs
Author | : Mohammed Alghazal |
Publisher | : |
Total Pages | : |
Release | : 2015 |
ISBN-10 | : OCLC:927777178 |
ISBN-13 | : |
Rating | : 4/5 ( Downloads) |
Download or read book Development and Testing of Artificial Neural Network Based Models for Water Flooding and Polymer Gel Flooding in Naturally Fractured Reservoirs written by Mohammed Alghazal and published by . This book was released on 2015 with total page pages. Available in PDF, EPUB and Kindle. Book excerpt: The increasing demand for energy and accelerated consumption of hydrocarbon fuel have made it a necessary objective for the oil and gas industry to continuously search for ways to improve and maximize recovery from oil reservoirs, to meet this growing global demand. Water flooding is one of the most common secondary recovery practices used in the petroleum industry to maintain reservoir pressure and improve oil displacement efficiency and recovery. Nonetheless, water flooding could pose several production problems in certain types of naturally fractured reservoirs, jeopardizing the overall sweep efficiency and oil recovery in the field. The presence of these heterogeneous natural fracture systems highly influences and complicates fluid flow process in the reservoir's transport media. These fractures provide easy conduits and fluid pathways for the injected water, causing early premature water breakthrough, excessive water production and rapid decline of oil rate.The implementation of polymer gel treatments is one of the viable solution commonly used in the industry to mitigate sweep conformance problems and improve oil recovery from naturally fractured reservoirs. Water-soluble polymer solutions are combined with cross-linking agents to form an in-situ gel that can be injected with water into the reservoir media. This polymer gel not only improves the overall mobility ratio of injected fluid, but also provides a mean to plug the conduit fractures and subsequently improving overall volumetric sweep efficiency and oil recovery from the reservoir matrix.Reservoir simulators are commonly used to build reliable reservoir models for the purpose of history matching, production forecasting and evaluation of various design scenarios. Nonetheless, reservoir simulation can become very computationally demanding and time- consuming process. This problem could be overcome by the development of Artificial Neural Network (ANN) models that could be used to generate various possible scenarios at a muchefficient time pace compared to reservoir simulation.The main objective of this research is to develop neuro-simulation proxy models for theimplementation of water flooding and polymer gel flooding in naturally fractured reservoirs. Three main ANN models, one forward and two inverses, were developed for each scenario, water flooding and polymer gel flooding.The first ANN, Forward ANN, provides a forward solution to predict the production profiles of oil rate, water cut and recovery factor for a given set of reservoir and design data. Forward results were matched within a desired tolerance of l0%. The second ANN, Inverse ANN- 1, provides an inverse-looking solution to estimate the project design parameters required to produce a given production profile for a given set of reservoir properties. Five design parameters were investigated, including: reservoir's drainage area, injection rate, producer bottom-hole pressure, polymer concentration and cross linker concentration. The last ANN, Inverse ANN-2, can be used as a tool for history matching and estimation of reservoir properties given a production profile and project design parameters. The reservoir properties predicted by this model include: matrix and fracture porosity, matrix and fracture permeability, fracture spacing, reservoir thickness and initial water saturation. The results from inverse ANN models were produced with an average error of 5 to 10%, per design parameter, and an average error of 8 to 28%, per reservoir property. Collectively, a total of six ANN tools were developed for the purpose of this research and were all encapsulated in a user-friendly Graphical User Interface (GUI) environment, to allow the end users for an easy access and utilization of these expert tools.