Filtering Geological Realizations for SAGD

Steam Assisted Gravity Drainage (SAGD) process is one of the most viable thermal recovery methods for heavy oil reservoir. In order to assess uncertainty in such reservoirs, many different equally-probable models, called realizations, are generated. Each realization is used with a flow modeling process to obtain the degree of uncertainty in reservoir performance parameters such as the production rate. Clearly, this can be a cumbersome process, since many of the realizations may exhibit nearly equal flow performance and also simulation process is very time consuming for such thermal models. In this research, we introduce a visual analytics framework for filtering of realizations for SAGD and select the ones that potentially represent different flow performance. This framework is based on calculating (dis)similarity distance between all pairs of models and then clustering models using the computed distance. Each cluster contains models that have potential similar flow performance. The results on the case studies used in our framework, show that our technique is much faster than and well matched with the brute force approach, which is based on running the flow simulations for all realizations. In addition to that, the proposed framework represents results in visual and interactive way.