Artificial Intelligence (AI) techniques have been successfully applied in the characterization of oil and gas reservoirs especially in the prediction of porosity, permeability, water saturation, dew point pressure, PVT properties and lithofacies. As good as each technique might be, it has its limitations in terms of ability to handle uncertainties, ability to learn and certain data size requirements. The performance of each technique is also limited by the nature and complexity of the problems to be solved.
The concept of combining two or more AI techniques to reduce the impact of their respective limitations and to enhance their performance is becoming increasingly popular. This hybridization effort attempts to complement the weaknesses of one technique with the strength of another.
In this paper, the basic requirements for hybridization are presented and a number of successful hybrid techniques are also reviewed. The architecture of some of the popular hybrid techniques such as Adaptive Neuro-Fuzzy Inference System, Functional Networks-Type-2 Fuzzy and Functional Networks-Support Vector Machines are concisely discussed.
This is meant to be a guide for those who might want to dig deeper into this increasingly-popular concept of hybrid computational intelligence for the possible realization of the future AI-assisted reservoir simulation for better exploration, production and exploitation.