Basic HTML Version
OH – ET – VA - LL: Analysis of Dynamic Data in Shale Gas Reservoirs – Part 1 – Version 2 (December 2010)
p 24/24
Multiphase flow
All the above is considering single phase gas flow, but the creation of fractured horizontal wells
requires the injection of massive amounts of water, and quite evidently we must question
ourselves about the necessity to take multiphase flow (with at least gas and water) into
account in the numerical formulation. In the shale gas context, relative permeability curves
and capillary pressure curves can be very different in the matrix and in the fractures. Our
model allows one set of curves only, and should be extended. This may be of particular
importance to correctly simulate the clean-up phase, and account for potentially large volumes
of pumped water trapped in the fracture network. Water loss can be due to combined effects
such as gravity and very strong capillary forces, as shale can initially be in a sub-irreducible
water saturation state. Ultimately, this may lead to a strong well productivity reduction, as it
reduces the contact surface available for flow. Particular attention should hence be paid to the
simulation of two-phase flow in this context.
Too many unknowns and too little data
As we will see in Part 2, when you look at real life data over a relatively short period of time
(one or a few years), very little occurs because of the very low speed of the diffusion. As a
consequence it is pretty trivial to match the observed response with a simple straight line.
There is no problem with this, except that we cannot use such technique to accurately predict
the longer term production. The matches with the analytical and numerical models were not
better nor worse. However the resulting parameters and the forecast were indeed very
different.
We suspect that all additional effects we are describing above will help if we want to be
realistic in our prediction and assess when and how the production will behave in the long
term. However these represent a large number of unknown parameters if the observed data
we have are reduced to a single apparent linear flow. Therefore the challenge will not only to
implement these mechanisms, but also to find a way to assess the related parameters.