Automatic Estimation of Fracture Properties in Multi-Stage Fractured Shale Gas Horizontal Wells for Reservoir Modeling

Abstract

Abstract Owing to low productivity of vertical wells in shale gas formations, operators have started to drill horizontal wells in these ultra-tight gas formations. Wells that are drilled in shale formations have undergone massive hydraulic fracturing treatments to improve the productivity. Since the introduction of horizontal drilling in shale formations, wells have been hydraulically fractured at several stages. Each stage is located a few hundred feet away from the other stages. Multi-Stage fracturing of horizontal drilling in shale gas wells increases the production significantly. Integration of detailed fracture network mapping in numerical simulation of these reservoirs is impractical because of high resolution time-space seismic data required to map the fracture surfaces, which in most wells are not acquired because of economic considerations. Multi-million cell numerical models are required for each individual well due to massive refining of gridblocks around the fractures. Conditioning the refined reservoir models to performance data adds to the complexity of the problem, since any adjustments made to the fracture properties can change the gridding of the model. Hence, automatic history matching tools are not well suited for application in grid-refined, detailed geological models representing shale gas reservoirs and wells. In this study, we propose a new multi-stage compartmentalized numerical modeling technique for shale gas reservoirs, which is an extension to dual porosity modeling. The methodology is simplistic, is not built using detailed high resolution fracture maps, and requires less CPU time. Ensemble Kalman Filter (EnKF), a minimum mean square error estimation tool, is used for estimating the reservoir properties. Combined EnKF/compartmentalized modeling technique is implemented in this paper to automatically estimate the fracture properties of the fracturing stages of a horizontal well. The algorithm only requires well’s performance data, such as gas production rates, flow back volumes, bottom-hole pressures, and tracer concentrations. Numerical models that are conditioned to observations are shown to accurately estimate the production rates and are useful for reserves booking and field development.