Seismic inversion is technically challenging, time consuming, expensive and requires significant manual input and strong priors. Traditional inversion methods are limited by small number of drilled wells within the study area and are ineffective in data poor frontier exploration settings. We offer an innovative solution of applying supervised machine learning for seimic inversion that aims to use a global well database, improving accuracy and reducing cycle time through automation. The challenge in most ML based projects is how to generate “training data”. We propose a unique solution of utilizing all available well data for creating a Global AVA Database. We aim to train a set of models using this database, to predict lithology, porosity and fluid type from selected set of features or seismic attributes. In this paper we demonstrate results from two case studies- first using synthetic dataset (SEG Seam Model) and another field case study (Gulf of Mexico).
Presentation Date: Wednesday, October 14, 2020
Session Start Time: 9:20 AM
Presentation Time: 11:00 AM
Location: Poster Station 1
Presentation Type: Poster
