Neural Networks for Field-Wise Waterflood Management in Low Permeability, Fractured Oil Reservoirs

Abstract

Abstract An optimal water injection policy maximizes oil recovery per barrel of injected water while minimizing formation damage and maintaining reservoir pressure. Optimal water injection into low permeability, fractured oil reservoirs is problematic because of highly nonlinear and complex reservoir dynamics. Likewise, current first principle models of fluid movement in fractured, low permeability rock systems are insufficient to design, operate, and predict the performance of large scale waterfloods Historically, the conflict between prudent reservoir management and meeting field injection-production targets has resulted in reservoir and well damage, injectant recirculation and irreversibly lost oil production. Here we present the next generation of "intelligent" field surveillance and prediction software based on neural networks and implemented on a PC. We demonstrate a new approach to field-wise performance prediction and optimization of waterfloods that recognizes an oil field as a coupled, highly nonlinear system of injectors and producers. With lease-wide historical data from a waterflood in the Lost Hills Diatomite (Kern County, CA), we construct several neural networks which recognize that individual well behavior may depend on well history and the injection-production conditions of surrounding wells. Some of our neural networks accurately predict wellhead pressure as a function of injection rate, and vice versa, for all injectors. Other networks history-match oil and water production on the well-by-well basis, and predict future production on a quarterly or half-year basis. Finally, our neural networks recognize and suggest water injection policies that lead to the minimum injected water and the best oil recovery. Introduction This paper outlines a new, field-wise approach to managing large fluid injection projects in tight hydraulically fractured reservoirs. We use neural networks to analyze the past performance of waterflood projects and to predict future oil recovery, and water injection and production. Neural networks are useful in that a structural model between injection and production need not be specified in order to predict performance. The neural network approach recognizes that individual well behavior may depend on the well history and the injection/production conditions of surrounding wells. Also, lease-wide production is the result of injection and production at many wells and their interactions. Our approach discerns injection policies that lead to the minimum injected water and the best oil recovery. We focus on water injection in tight reservoirs because significant quantities of crude oil remain in them, and state-of-the-art understanding of fluid movement in low permeability rock systems is not sufficient for design and operation of large fluid injection projects. Water injection is also important for mitigating reservoir compaction and surface subsidence. In tight rocks, project operation is problematic because reservoir dynamics are highly nonlinear and complex. An optimal injection policy (i.e., the schedule of injection rates chosen to produce a field) for tight fields minimizes formation damage while maximizing oil production per unit volume of injectant. Fluid injection into low permeability reservoirs (diatomites, chalks or carbonates), either for pressure maintenance or secondary oil recovery is very difficult. On one hand, injection rates must be low enough to prevent reservoir damage from overpressuring and inducing unwanted fractures. On the other hand, these rates must be high enough to make the costly fluid injection process economic. P. 681