Resilience grant strand 2 for IUK project SHAIR - automated seismic interpretation using neural networks

点击翻译按钮获取中文摘要

SHAIR will create a new software service that will deliver highly accurate interpretation of seismic data to determine the likely presence of oil or gas.The main focus of this project is to develop new technology for increasing the accuracy and automation for imaging the subsurface using AI/ML within an industry-leading algorithm known as XWI. The computational power needed to run these algorithms is massive -- exceptionally 1 million virtual CPUs have been used.Technology from S-Cube's partner in SHAIR, RagnaRock Geo from Norway, currently contains an AI driven horizon interpretation with minimal human interpretation. It takes in subsurface volumes and uses their improved Geophysics-guided Neural Networks (GNNs) innovation of Convolutional Neural Networks (CNNs) to detect discontinuities in the rock structure, and interprets these as 'horizons'.S-Cube's technology, XWI, spun out from a world leading research group from Imperial College London is the result of 20 years of development, covered by multiple patents. It is an advancement of traditional Full-Waveform Inversion (FWI) producing high quality 3D subsurface models useful in assessing subsurface features ahead of drilling.The main result will be a novel suite of tools available to energy companies as a SaaS product via which users will pay to more accurately analyse their data. The new technology will enhance existing XWI algorithms generating superior velocity models constrained using automatically-generated seismic horizons.This will be deployed as a new service that makes it more cost effective for oil companies to grow reserves and minimise emissions with more accuracy than the traditional method of analysis used to discover oil.Energy companies will become more profitable via better analysis and a higher rate of success and be able to diversify from petroleum to new sources of lower emissions subsurface energy. This improved interpretation of data should reduce the likelihood of drilling for hydrocarbons which results in none being found and therefore can help the environment with fewer wells drilled.