For many unconventional reservoirs, the initial oil saturation is extremely low (around 30 %). In such circumstances, it is necessary to use alternative geological and hydrodynamic simulators to forecast production levels. A novel adaptive modeling approach based on cascades of fuzzy logic matrices was proposed and implemented for a mature carbonate oil reservoir in the Permian basin of Texas where next generation waterflooding was considered to revive development process and increase oil recovery.
The proposed approach is a variant of machine learning to solve the classical analysis and synthesis problem. At the analysis stage, the input and target parameters are normalized in order to equalize their significance. For each pair of input parameters, a fuzzy-logic matrix is formed and populated with actual values of the target parameter. The set of matrices forms a cascade in which every component is characterized by its own membership function. At the synthesis stage, forecasted results of the target parameter are calculated taking into account all membership functions and correspond to the maximum values of their superposition.
By means of the proposed approach, geological and hydrodynamic models of the unconventional reservoir were successfully created, which made it possible to estimate the distribution of its initial and remaining oil reserves for all reservoir formations. The options for further reservoir development were also considered including the reactivation of non-operating wells, drilling new wells, and initiation of waterflooding. The obtained results confirmed that waterflooding was able to enhance the cumulative oil production of the reservoir compared with the base case without waterflooding, and its role became more significant as the number of operating production wells increased. The calculated results also showed that the proposed approach was quite sensitive to the changes of input parameters, for example, to the number of production and injection wells. The cumulative oil production varied several times depending on the considered option. The calculated results justified the ability of the proposed approach to forecast the development results and to choose the proper strategy for the reservoir reviving. A distinctive feature of the proposed approach was its ability to adapt to any of geological and field conditions, such as the extremely low initial oil saturation that most reservoir simulators do not take into consideration.