The very first EAGE / PESGB Workshop on ML, November 2018, focused on the latest developments in digital transformation, AI / ML, analytics, cloud computing, approaches, challenges and business impact in upstream. Link

https://www.eage.org/site...-learning?sc_lang=en


Specifically of interest here are Deep Learning applications -

1. Reservoir properties prediction - compare deep neural network with multi-variate regression and probabilistic neural network approaches.

2. Real-time fractures monitoring - unsupervised clustering with Gaussian models to automate detection and analyze microseismic data.

3. Production forecast - deep learning of history matching using recurrent neural network (RNN), and specifically long-short term memory (LSTM) implementation, resulting in "reliable prediction of production rates*".

4. Beyond full waveform inversion (FWI) - applying deep neural network to synthetics (based on original work by Tarantola) of pseudo-spectral FWI. Another step toward physics-based ML model.


*jive with my own experience on experimenting with straight NN vs. LSTM on type curves, automating decline curve analysis (DCA) and quick-n-dirty remaining recoverable reserves (RRR).