Reservoir characterization, well log analysis and interpretation are key enablers for the optimization of field development and the maximization of recovery. Key challenges in well log analysis are that the current interpretation practices could be subjective, discrete and demanding. Additionally, comparing and analyzing well log data from multiple wells requires constraints to the geological setting. Leveraging on big data well log availability in developed areas, we present an innovative fourth industry revolution (4IR) method for the reconciliation of well logs to automatically assist the process of well log interpretation.

The training set of the method consists of reconciled sonic, resistivity and density well logs together with their target set of saturation and porosity. A multi-layer neural network was trained on selected data sets and validated on a synthetic reservoir case as well as on a reservoir box model.

The results are very promising, paving the way towards an innovative approach for the artificial intelligence assisted interpretation of well logs.

Keywords:
machine learning, neural network, log data, reservoir box model, log analysis, upstream oil & gas, katterbauer, intelligent reconciliation, artificial intelligence, reconciliation
Subjects:
Formation Evaluation & Management, Open hole/cased hole log analysis
Copyright 2020, Society of Petroleum Engineers
You can access this article if you purchase or spend a download.