Energy

Energy

This innovation area aims to develop deep learning models and applications for monitoring and detecting energy resources.

Motivation

Data from the Earth’s subsurface, such as borehole imagery and seismic data, is important for energy exploration. Automated interpretation of this complex data has significant potential for monitoring and detecting natural resources in a more efficient manner.

Our innovations

Visual Intelligence’s research efforts have resulted in more robust and reliable methods for automatic analysis of complex imagery from the digital subsurface, enabling more efficient and detailed energy exploration. Examples of innovations include methods for:

• detecting geological phenomena in seismic data using content-based image retrieval (CBIR).

• automatically detecting microfossils from microscope images using self-supervised learning,

• quantifying the uncertainty when identifying geological layers.

These methods were developed in close collaboration with user partner Equinor.

Addressing research challenges

The amount and quality of labelled training data is a significant challenge in this field. As existing interpretations are not made for machine learning purposes, the annotation quality for a particular task is unknown and interpretations are generally incomplete. Our developed methods within the energy field help address these research challenges in different ways.  

For example, a Visual Intelligence-developed framework for explaining embeddings extracted with self-supervised models was implemented in the seismic CBIR system. The pipeline for detecting and classifying microfossils offers a novel wat for extracting features automatically using self-supervision, making such models less reliant on labelled data. Our inclusion of uncertainty estimates in deep learning-based species classification applied to microfossil data has also shown to improve performance.

‍

Synergies within the innovation area and across other areas

The seismic CBIR system and the method for analysis and classification of microfossils are closely connected method-wise and benefit from each other. Both include components of a pretrained visual image transformer using self-supervised learning and a CBIR component for exploring and analyzing the resulting embedding space.

Self-supervised learning is a general topic that encompasses all four innovation areas. For example, the seismic CBIR system shares many similarities with a VI-developed framework for CT image retrieval using self-supervised learning. This illustrates how developed methodologies within this innovation area can be transferred across innovation areas.