Machine Learning in Geoscience – November SIG Meeting

We interpret seismic data in order to develop our geological understanding of a region, support play fairway analysis and build reservoir models.  The application of machine learning to seismic interpretation has the capacity to improve both the quality of our interpretation and the speed with which we interpret the data.  However, it is important that we apply the right machine learning at the right time and for the right reasons otherwise we run the risk of losing the advantages that it gives us.  This presentation will demonstrate the way that both unsupervised and supervised machine learning can improve seismic interpretation and look at how varying the parameters affects our results.

Unsupervised machine learning requires no significant input from the user beyond the basic parameters.  The computer uses machine learning algorithms to cluster data points into groups with similar properties.  This can lead to a greater understanding of the geology that we are seeing in the seismic.  Supervised machine learning requires the input of training data.  In this instance, the computer learns what a particular image contains and then looks for similar images.  This is used for facies and fault interpretation.

In both instances, the quantity and quality of the input data will affect our result.  In addition, there are basic parameters that affect the algorithms and have the capacity to modify the result.  Understanding how these parameters affect the result will have a significant bearing on the results that we get and will hence impact the ability of the process to affect the outcome of the reservoir model.  It is therefore important that the geoscientist understands more than an equation.

Kindly sponsored by Hoolock Consulting 

Tim Gibbons has over 30 years of experience in the oil industry.  He initially worked for two major operators as a geophysicist and worked in new ventures, exploration and production.  He then worked for three major service / consulting companies in a variety of managerial, sales and business development roles.  He has experience of selling computing technology, technical geoscience services and multi-client projects around the world, particularly Europe, Australia, Africa and North America.

In 2016, Tim founded Hoolock Consulting to provide sales consulting services to the upstream oil and gas industry, with a focus on geoscience.  He provides sales training and consultancy as well as selling products on behalf of a number of clients.