The neighborhood approach to prediction of permeability from wireline logs and limited core plug analysis data using backpropagation artificial neural networks

Permeability is a key parameter associated with the characterization of any hydrocarbon reservoir. In fact, it is not possible to have accurate solutions to many petroleum engineering problems without having accurate permeability knowledge. Until now, the petroleum industry tried to acquire reliable permeability values via laboratory measurements on cores or well test interpretation, which are both accurate but not adequate methods for complete reservoir description. Wireline log data and core plug analysis correlation have also been used to estimate permeability, but due to the correlation methods available, this approach does not always yield accurate and adequate results. A method using artificial neural networks (ANNs)—simple mathematical implementations of human brain activity—shows promise, but this method also has its own disadvantages, like failing to succeed when limited core plug analysis data are available. The neighborhood approach to the use of ANNs to predict reliable permeability values is proposed as a solution to the problem. This paper also contains a review of conventional ANN architecture and tries to reveal some interrelations of ANN design parameters. A case study is included to present the conventional and proposed designs of backpropagation ANNs as tools to predict permeability where limited and heterogeneous core data are available. The results of this new approach are accompanied by a technical discussion that includes comments about why results are difficult to repeat and how it is possible to improve the ANN architecture.