Reservoir characterization plays a critical role in appraising the economic success of reservoir management and development methods. Nearly all reservoirs show some degree of heterogeneity, which invariably impacts production. As a result, the production performance of a complex reservoir cannot be realistically predicted without accurate reservoir description. Characterization of a heterogeneous reservoir is a complex problem. The difficulty stems from the fact that sufficient data to accurately predict the distribution of the formation attributes are not usually available. Generally the geophysical logs are available from a considerable number of wells in the reservoir. Therefore, a methodology for reservoir description and characterization utilizing only well logs data represents a significant technical as well as economic advantage.

One of the key issues in the description and characterization of heterogeneous formations is the distribution of various zones and their properties. In this study, several artificial neural networks (ANN) were successfully designed and developed for zone identification in a heterogeneous formation from geophysical well logs. Granny Creek Field in West Virginia has been selected as the study area in this paper. This field has produced oil from Big Injun Formation since the early 1900's. The water flooding operations were initiated in the 1970's and are currently still in progress. Well log data on a substantial number of wells in this reservoir were available and were collected. Core analysis results were also available from a few wells. The log data from 3 wells along with the various zone definitions were utilized to train the networks for zone recognition. The data from 2 other wells with previously determined zones, based on the core and log data, were then utilized to verify the developed networks predictions. The results indicated that ANN can be a useful tool for accurately identifying the zones in complex reservoirs.


Advanced computer models are available for simulating the fluid flow in increasingly complex reservoirs for the purpose of determining hydrocarbon recovery and appraising the economic success of reservoir management and development methods. However, an accurate description of the reservoir is necessary for the model to predict the performance of the reservoir reliably.

Nearly all reservoirs contain several zones due to the existence of contrasting lithologies, diagenesis, or sedimentological complexity. These zones usually influence the hydrocarbon movement, distribution, and production. Therefore, recognition of the zones has economic implications in reservoir management. As a result, the reservoir description cannot be realistically determined without accurate prediction as to how the zones are distributed in the reservoir. The prediction and identification of the zones are complex problems which require integration and interpretation of various geological and engineering information. The major problem in determining the distribution of the zones stems from the fact that the core data that are required for zone identification are available only from a few wells in a reservoir.

The objective of this study is to investigate the feasibility of using an Artificial Neural Network as a tool for zone recognition and identification in a heterogenous reservoir utilizing geophysical well logs. Neural networks, a biologically inspired computing scheme, is an analog, adaptive, distributive, and massively parallel system that has been used in many disciplines and has proven to have potential in solving problems that require pattern recognition. Known as sixth' generation computing, neural networks are widely used in many disciplines from Wall Street to airport security devices. Since they process data and learn in a parallel and distributed fashion, they are able to discover highly complex relationships between several variables that are presented to the network. As a model-free function estimator, neural networks can map input to output no matter how complex the relationship might be. There are several paradigms (supervised, unsupervised, and reinforced) that can be used to generate neural networks.

The main interest in neural networks has its roots in the recognition that the brain processes information in a different manner than conventional digital computers. Computers are extremely fast and precise at executing sequences of instructions that have been formulated for them (algorithm).

P. 27

This content is only available via PDF.
You can access this article if you purchase or spend a download.