Abstract:

Identification of dominant flow channels is a critical step in oilfield development. Conventional studies on interwell dominant flow channels require data that are difficult to acquire, in addition to complex operations and high costs. Moreover, conventional numerical simulation methods employ grid-based complex geological models which feature burdensome data loading and long computation cycle. To quickly identify dominant flow channels, an interwell connectivity inversion model was established based on a capacitance resistance model for production (CRMP), and the injection-production data transmission relationship was constructed using connectivity coefficient and time constant. The ensemble Kalman filter (EnKF) was used as an optimization algorithm for automatic history match of production data, enabling quantitative characterization of interwell connectivity. The average permeability between wells and oil saturation at well points were calculated using the principle of hydroelectric similarity principle, fractional flow equation, and relative permeability curve. An oil saturation field map was plotted to characterize the dynamic production process of oilfield. The standard for dividing dominant flow channels was developed using the K-means algorithm, and multidimensional clustering analysis was conducted on connectivity coefficient and production data to achieve intelligent identification of dominant flow channels. Finally, the proposed model was applied to actual oil reservoirs. In this application, the noise effects of data and the model were considered, and multiple sets of parameters were updated simultaneously through data assimilation during the model training. The average value was taken as the optimal solution, thereby avoiding the ambiguity in results. The results indicate that the CRMP- and EnKF-based model accurately delineates the spatial distribution of dominant flow channels, and also reduces the uncertainty of reservoir parameter inversion. Actual reservoir applications demonstrate basically identical oil saturation fields inverted by CRMP and obtained by numerical simulation. The effects of data update and parameter convergence are significant, and the obtained connectivity coefficient agrees well with the tracer results, reaching 83%. The findings prove that the proposed technique is of great significance for decision-making in future waterflooding, profile control and well selection in oil reservoirs.

Key words: oilfield development, waterflooding, dominant flow channel, dynamic and quantitative characterization, ensemble Kalman filter, capacitance resistance model