Well representation in simulation models of naturally fractured reservoirs considering the impact of DFN upscaling

Abstract

The study of naturally fractured reservoirs (NFR) is complex and has many challenges in reservoir modeling and simulation, requiring non-conventional techniques. The use of Discrete Fracture Networks (DFN) to build dual permeability (DK) simulation models has been one of the processes applied. The main objectives of this work are (1) to evaluate the impact of DFN upscaling methods on new wells placed in the model and (2) develop a proposal for wellrepresentation for new wells in simulation models of NFR. Three DFN permeability-upscaling methods at three model fidelities (high, medium-high, and medium fidelity) and the concept of Characteristic Flow Unit (CFU) were used. The upscaling methods applied were the Oda, Oda Corrected and Flow-based with four boundary conditions. The work was focused in three main steps: (1) to observe the differences in static and dynamic well parameters when changing the upscaling method and fidelity scales; (2) to test well calibration methodologies in a coarse grid by considering the previous step conclusions, and (3) to test the different DFN upscaling methods at a field scale to evaluate the impact on predicted field production. The main results for the first step suggested more uncertainty of well dynamic data in medium fidelity models and showed that the Oda method is the most sensitive to the grid-block scale change. Otherwise, combining the high-fidelity scale with the flow-based upscaling (linear pressure) tended to provide less variation in well flow rate and BHP, which led to the use of this model as the reference case for the second step analysis. Among the tested methodologies, the producer and injector well calibration was achieved with the substitution of their well index (WI) in the matrix and fracture systems by the reference model's WI. The field scale application showed a significant impact of the DFN upscaling method in the short-term production forecast, decreasing in impact as the production time increases. The results show that if the DFN permeability upscaling method variabilities are not considered when placing new wells in DK models, the reservoir development and management phase can be significantly affected. The proposed well representation utilizing a WI based on the flow-based method showed promising results to reduce the impact of DFN upscaling methods in the well's production and injection forecast for medium-fidelity models. This work presents an investigation of a possible wellrepresentation approach for simulation models of naturally fractured reservoirs, providing for real field cases a better representation of injection and production forecast related to fault and fracture inclusion, when utilizing DFN permeability-upscaling methods.