EssencePS CEO and chief developer Nigel Goodwin has brought to our attention a paper
presented at the recent SPE Reservoir Simulation Symposium which offers
a technical description of methods and algorithms used in EssRisk,
Essence’s software flagship for history matching, uncertainty-based
prediction and production optimization. Goodwin argues that brute force
Markov chain Monte Carlo methods cannot be applied exhaustively to the
complex field of fluid flow simulation. Even fast proxy models may fail
to represent the full range of uncertainty. Moreover, the ‘black box’
nature of proxy models make their evaluation hard. Engineers generally
prefer straightforward deterministic models.
Goodwin advocates Hamiltonian MCMC techniques,
along with efficient proxy models which lead to reliable and
uncertainty quantification and also generate an ensemble of
deterministic reservoir models. The technique is claimed to be the
foundation of a new generation of uncertainty tools and workflows.
The paper (50
pages and 178 equations ) is not for the faint hearted. But unusually,
as Goodwin observes, ‘unlike most vendors, service providers and oil
company research departments we are completely open about the
algorithms used. Our added value lies in efficient implementation and
in our user interface.’ More from EssencePS.
© Oil IT Journal - all rights reserved.