Machine learning on Crays to optimize petrophysical workflows in oil and gas exploration

Abstract

Summary The oil and gas industry is awash with sub‐surface data, which is used to characterize the rock and fluid properties beneath the seabed. This drives commercial decision making and exploration, but the industry relies upon highly manual workflows when processing data. A question is whether this can be improved using machine learning, complementing the activities of petrophysicists searching for hydrocarbons. In this paper, we present work using supervised learning with the aim of decreasing the petrophysical interpretation time down from over 7 days to 7 minutes. We describe the use of mathematical models that have been trained using raw well log data, to complete each of the four stages of a petrophysical interpretation workflow, in addition to initial data cleaning. We explore how the predictions from these models compare against the interpretations of human petrophysicists, and numerous options and techniques that were used to optimize the models. The result of this work is the ability, for the first time, to use machine learning for the entire petrophysical workflow.

Multiple imputation via chained equations for elastic well log imputation and prediction

“…The fluid content of a rock is an important aspect of well logging and fundamental to downstream subsurface analyses. Similar results were found by Brown et al (2020), where the downstream petrophysical analysis workflow was bypassed by training gradient boosted tree algorithms to predict petrophysical logs directly such as water saturation and porosity. Training data in this case must include expertly curated petrophysical logs with special attention paid to preprocessing of the input logs.…”

“…Future tests may consider using these data as input if available. Similar to the work of Brown et al (2020), this would also result in imputed petrophysical products.…”

“…Unlike Brown et al (2020), we have not included derived petrophysical logs in this study. From a machine learning perspective, petrophysical logs such as water saturation, porosity and clay volume can be viewed as engineered features which augment or extend our view of the raw input data.…”

Foreword to the Special Issue of the Cray User Group

“…Recent years have seen the introduction of machine learning techniques into traditional simulation‐based workflows and also their use to supplement human interpretation. In this special issue, Brown et al demonstrate the power of machine learning to optimize petrophysical workflows in the oil and gas industry. The authors compare the results predicted by trained models against the interpretations of human petrophysicists and find that machine learning can be effective in all stages of a typical petrophysical workflow.…”

A novel marble recognition system using extreme learning machine with LBP and histogram features

“…ANN, one of the traditional methods, is used as a classification and clustering algorithm with its input and output nodes 29 . ANNs are often used on batch processing or real‐time applications 30 . In studies 31,32, the authors carry out a real‐time study of marble image classification using ANN structures embedded on the Programmable Logic Controller.…”