We investigate an applicability of Bayesian-optimization (BO) to optimize hyperparameters associated with support-vector-machine (SVM) in order to classify facies using elastic properties derived from well data in the East Central Graben, UKCS. The cross-plot products of the field dataset appear to be successfully classified with non-linear boundaries. Although there are a few factors to be predetermined in the BO scheme such as an iteration number to deal with a trade-off between the prediction accuracy and the computational cost, this approach effectively reduces possible human subjectivity connected to the architecture of the SVM. Our proposed workflow might be beneficial in resource-exploration and development in terms of subsurface objective technical evaluations.@en

One of major technical competitions in energy industry relates to how optimally deep-learning architectures we can design. Optimization of hyperparameters is treated as labor-intensive. However, it is important to tune the parameters especially when we deal with relatively small targets, yet high-impact consequences can be resulted. In this study, we adapt Optuna, the global optimizer, for tuning the hyperparameter of the deep-learning scheme of the extended long-short term memory with forget gates. We apply this framework for predicting lithological facies. Although the macro difference with and without Optuna is not significant in this study, our results indicate that Optuna could make large commercial impacts when targets are small yet difficult to be captured.@en

Semi-supervised deep-learning architectures provide a multi-layer, pattern recognition, approach that is powerful and ideally suited to the data rich environment that exists at the heart of the oil and gas industry. In this study we apply this technology in order to classify facies using elastic impedances from UK North Sea well and seismic data. The semi-supervised deep-learning method in this study uses a self-training strategy that combines both labelled and unlabelled data during the training phase so that classified data subsequently becomes part of the training dataset in the next iteration. This approach is ideal when the availability of labelled data is limited by practical constraints, which is often the case in subsurface geoscience. The resulting outputs of classified facies were visualised using elastic impedance cross-plots after application to a single training well from a North Sea oil discovery. To validate the result we upscaled the classification model to equivalent seismic data in order to compare the learning from the training well with two blind wells. The results indicate that semi-supervised deep-learning has the potential to accurately determine facies, including hydrocarbon distributions, in subsurface data at a field scale.@en