Linguistics

Engineering

Philosophy

Materials science

Social psychology

Petroleum engineering

Data mining

Computer science

Geology

Reservoir computing

Physics

Nanotechnology

Reservoir modeling

Psychology

Mathematics

Optics

Feature (linguistics)

Characterization (materials science)

Domain (mathematical analysis)

Focus (optics)

Mathematical analysis

Artificial intelligence

Deep learning

Artificial neural network

Feature engineering

Recurrent neural network

Popularity

Machine learning

Computers & Geosciences

Nowadays Machine Learning approaches are getting popular in almost all the domains of Engineering Applications. One such widely used approach is Artificial Neural Networks (ANN), that has been successfully applied in many disciplines and becoming popular in the domain of Reservoir Characterization too (Verma, A.K. 2014). A considerable number of neural network papers have been published till now in this domain, and its application is still on the way. The main motive of application of ANN in this domain is to use acquired data from different geological and geophysical sources in determining the characteristics of a reservoir by analyzing the correlation of various data sources. When properly trained, ANN can predict the reservoir properties, by identifying the complex nonlinear relationship associated with the input data. Different ANN models have been used in the domain of reservoir characterization starting from shallow to deep models with the progress over time. In some scenarios, ANN is combined with other soft computing methodologies resulting in hybrid models. Recently deep learning models of ANN are gaining popularity in many fields including oil exploration. The popularity of deep models is due to its automatic feature extraction capability, ability to handle high dimensional data and above all the ability to solve a problem like our human brain does, learning with multiple levels of abstractions. In this survey, we focus on different evolution of ANN in Reservoir Characterization over time. The evolution is in terms of architecture, learning, as well as to combine with other models of machine learning to improve its modeling capability, that now extends towards recent advanced techniques of ANN called deep learning. From the survey, it is apparent that the application of ANN is very vital in this field and its application will continue even in the future in making intelligent interpretation of oil reservoir.

Artificial Neural Networks in the domain of reservoir characterization: A review from shallow to deep models