29th October 2020
Speaker: Ali Moradi Tehrani – CGG
Topic: Hybrid Deep Machine Learning Inversion
Machine Learning techniques have been used for many years in geophysics for different applications including reservoir property prediction. The method suffering two weaknesses, one was the fact that it was mainly data processing based and not physics based. The other one was restrictions caused by limited hidden layers of the neural networks. Those networks could model non-linear systems, but not just any non-linear systems. Therefore reservoir properties that have highly non-linear relation with seismic could not be modelled properly.
Hybrid means adding physics to the system. Downton and Hampson (2019) developed a hybrid theory-guided data science (TGDS) model (Karpatne et al., 2017) to augment the amount of data. The outputs of the theory-based component are used as the inputs to the data science component. Rock physics theory and statistics are used to model the elastic parameter to generate a large number of pseudo wells. These pseudo wells are then used to model synthetic seismic gathers, which are then used to train a Deep Neural Network (DNN). The trained DNN is then applied to the real dataset. This work shows the application of this workflow to a field in the North Sea producing commercial volumes of oil.
Ali Moradi Tehrani has more than 20 years of experience in development and application of different geophysical techniques around the world. He has a wealth of technical knowledge and commercial expertise focused on advanced quantitative interpretation and reservoir characterization including modeling, deterministic and geostatistical inversion, seismic attribute analysis and classification.
Ali holds a Doctorate in Geophysics from TUDelft. He previously worked at Schlumberger as a Lead Geophysicist and Workflow Consultant. He joined CGG in 2011 and he is currently CGG GeoSoftware Regional Technical Manager for Europe and sub-Sahara Africa.