Reservoir Modeling and Simulation for the Basalt Oil Reservoir Accomplishing Gas Injection IOR

The gas injection IOR project for the basalt reservoir started in April 2014 in Yurihara oil field, Japan. In this reservoir, water cut increase in the borehole at the structurally high level had been observed. This was interpreted as a phenomenon relating to the characteristics of the inhomogeneous basalt reservoir. It was difficult to explain these using the previous reservoir model based on the simple geological concept, in which the lithofacies distribution was drawn into a single concentric circle form whose center corresponds to the crater area of a volcano. In order to improve the history match of the reservoir simulation, the new geological study started and the geological model was revised overall. In this concept, multiple volcanic craters are located dispersively in this field. Then, the reservoir model was renovated by multi-point geostatistical approach utilizing the geological training image. In facies modeling, the basalt was classified in three types, which are sheet flow, pillow lava and hyaloclastite. Since the most productive zone seems to be pillow lava, the production wells have been completed in the zone, where the pillow lava were dominantly observed in well logging. The problem here is that the productivity is much different in the pillow lava. In other words, the content of pillow is not proportional to productivity. Consequently, we needed to set the much variation of permeability in the pillow facies. So, the permeability distribution was estimated by Gaussian simulation, where we added the seismic attribute as soft data. Then, the distribution was modified by gradual deformation method, where the objective function was calculated using the residual error with the observation and the simulated value of bottom-hole pressure and the water cut. A several realizations matching to production history were extracted, and are currently utilized to gas injection optimization. Not only that, but also the prediction for the optimal timing of starting WAG is required.