The challenge of discovering new reserves coupled with the current dwindling oil price has necessitated the need to generate and sustain long-term production from existing fields through improved or enhanced oil recovery (IOR/EOR) processes. There is however, no established mechanism to match the thousands of candidate reservoirs worldwide to the subtle and critical variations in reservoir properties that control the success of the many different EOR options. We present a neuro-fuzzy approach to screening potential hydrocarbon reservoirs for enhanced oil recovery (EOR) applications.First, reservoir field data from multiple successful thermal, miscible, chemical and biological EOR projects across different petroleum systems worldwide were trained to establish knowledge pattern and represent it using fuzzy rules. This is achieved by combining fuzzy technique with neural network learning capability to deduce knowledge from the EOR data in a form akin to linguistic rules. Then, the extracted knowledge pattern was validated and used to determine the combination of reservoir properties which could best characterise the key heterogeneities that control EOR success. The model output can be used in screening potential reservoir for EOR application.
|Number of pages||14|
|Journal||Advances in Petroleum Exploration and Development|
|Early online date||28 Sep 2016|
|Publication status||Published - Sep 2016|
- reservoir screening
- artificial intelligence
- training and learning