Abstract Facies classification from well logs is an indispensable part of seismic interpretation and is important in the determination of sequence stratigraphy and ultimately reservoir characterization. Although there have been improvements in the tools used to perform this task, it remains laborious, subjective, and error-prone. Achieving a proper classification is complicated by increasing dataset sizes as well as the need for correlated multidisciplinary models. Recent developments in machine learning provide an opportunity to assist interpreters in accomplishing this task while also improving the accuracy of classification results. Applications of machine learning methods for automating facies classification from well logs have previously been explored, however these have largely focused on evaluations or comparisons of individual algorithms or of ensembles of homogeneous agents. The proposed method combines heterogeneous agents to enhance prediction accuracy. Specifically, supervised learning, which provides a direct mapping between the data domain and the solution domain while introducing bias to generalize the mapping, is combined with unsupervised learning, which does not depend on similar generalization bias or training data but also does not provide a direct mapping between the data and solution domains. The combination is accomplished via the joint probability density function (PDF) of the supervised classification, which is used to guide identification of clusters delineated by unsupervised learning. This multi-agent approach can reduce bias introduced during training and provides a basis for generating a probability distribution for each sample rather than a discrete classification. The distribution, in turn, can be used to more accurately model the continuous nature of well log signals, which reflects continuity in lithological regimes.
Paper
Full text
Improving facies prediction by combining supervised and unsupervised learning methods
Semantic Scholar · Engineering · 2021
Abstract
Abstract Facies classification from well logs is an indispensable part of seismic interpretation and is important in the determination of sequence stratigraphy and ultimately reservoir characterization. Although there have been improvements in the tools used to perform this task, it remains laborious, subjective, and error-prone. Achieving a proper classification is complicated by increasing dataset sizes as well as the need for correlated multidisciplinary models. Recent developments in machine learning provide an opportunity to assist interpreters in accomplishing this task while also improving the accuracy of classification results. Applications of machine learning methods for automating facies classification from well logs have previously been explored, however these have largely focused on evaluations or comparisons of individual algorithms or of ensembles of homogeneous agents. The proposed method combines heterogeneous agents to enhance prediction accuracy. Specifically, supervised learning, which provides a direct mapping between the data domain and the solution domain while introducing bias to generalize the mapping, is combined with unsupervised learning, which does not depend on similar generalization bias or training data but also does not provide a direct mapping between the data and solution domains. The combination is accomplished via the joint probability density function (PDF) of the supervised classification, which is used to guide identification of clusters delineated by unsupervised learning. This multi-agent approach can reduce bias introduced during training and provides a basis for generating a probability distribution for each sample rather than a discrete classification. The distribution, in turn, can be used to more accurately model the continuous nature of well log signals, which reflects continuity in lithological regimes.