Email Address: Haishan.Luo@utexas.edu
Google Scholar: https://scholar.google.com/citations?user=RM5rZpYAAAAJ&hl=en
Reservoir simulation code is trending heavier and far more complicated in modern times as the requirement for modeling various complex problems is increasing. Needs for more phases, more components, more reactions, better phase behavior models, fractures, geomechanics, geophysical/petrophysical integrations, foaming/emulsifying, to list a lot more, although crucial at certain conditions, are overwhelming and exhausting both developers and users, whereas in contrary, they bring down the original purpose and lead to more uncertainties associated to the processes parsed and optimized by engineers and decision makers. I am particularly interested in breaking the ceiling in this area and addressing these challenges through integrating with machine-learning or deep-learning methods. One potential application could be to train certain models with the help of extensive data, which then inherently have predictive capabilities underneath the main pipe of the simulator. In the absence of data, another resolution could be to train a reliable model itself with millions of solutions corresponding to selected features. A critical step for the success of this application is to set up features with a physical and meaningful perspective that requires expertise of the domain knowledge. The same idea could be applied to refine certain processes in a more simplified way such as the data-driven models or surrogate models. The automated history-matching process should be guided by the artificial intelligence in order to get around some impasses caused by overwhelming and secondary-effect models and parameters. Currently, our research team has developed and tested a few predictive models through training extensive data collected by our lab in past several decades, which are coupling to the reservoir simulator. With a small step forward, we expect to expand this idea to address more problems in reservoir simulation and reservoir engineering.
UTCHEM: In charge of development and maintenance of the chemical flood reservoir simulator UTCHEM for the UT chemical EOR consortium. Focused on developing predictive modules for polymer rheology and surfactant phase behavior with cutting-edge models such as the structure-property model refined from thousands of lab data.
Fingering Problem: Seek for effective and practical methods to scale-up fingering behavior in fluid flow processes displaced by gas or polymer and surfactant solutions.
Modeling Foam in Tight Rock: Develop new gas EOR simulator to model foam in fractured tight rock coupled to gas compositional model.