Bringing techniques and expertise from other disciplines to bear on a new area of application can be important to making step-change advancements in any area, and a recent example was provided at NETL "Mastering the Subsurface" Meeting in Pittsburgh (13-16 Aug 2018). A panel session (with presentations) was devoted to the results of a recent workshop on 'Machine Learning' sponsored by NETL and hosted at Carnegie Mellon University in Pittsburgh on the 17-18 July. The formal title of the workshop was "Real-time Decision Making for the Subsurface", and it brought together experts in CO2 storage and unconventional oil and gas with data scientists from academia, national labs, and industry, who were experts in machine learning, artificial intelligence and the subsurface.
In general, geoscience data is amongst the largest in the sciences. In the CO2 storage area, there is a need to integrate large data sets of different data types, eg from multiple sensors with discrete data and from the continuous data that we are beginning to collect from new monitoring technologies such as distributed acoustic sensing (DAS) and the Modular Borehole Monitoring system (MBM).
Bringing machine learning to such applications can enable autonomous monitoring and fast decision-making for well control and reservoir management, and potentially management of seismicity. One application would enable the extraction of useful signals from seismic noise; another would be to validate existing models; the list goes on.
The National Risk Assessment Partnership's (NRAP) approach has been to use conventional physics-based prediction with empirical results to describe complex system behaviour. Similar work within NRAP and at DOE's national labs is now looking at machine learning to potentially discover signals that can be used to predict leaks or identify states of subsurface stress that could cause induced seismicity, ie earthquakes.
The slide below describes how these data techniques can come together for management of the subsurface.
This obviously all has great potential for assisting safe and cost-effective CO2 geological storage. Carnegie Mellon University will produce a report of the workshop in due course, and we at IEAGHG will be reviewing this topic in more detail for its application and potential benefits to CO2 storage.
