This preview shows page 1. Sign up to view the full content.
Unformatted text preview: s for shell 1 and shell 8. On shell 1, the fluids are well solved for and each
fluid volume and total fluid show little statistical variation. Shell 8 however displays large statistical variations. Looking
more closely, one can observe that the large statistical variations are mainly on bound water and gas but that the oil volume is
fairly constant and accurate. This is expected; bound water has a fast decay signal and only a limited number of echoes carry
the information. Gas signal is small due to under-polarization and low hydrogen index, comparatively SNR is much lower for
a gas volume than for a fluid. CSUG/SPE 147305 3 Conversely, oil is fully polarized with the acquisition sequence used and its signal decays slowly, so the number of echoes
useful for processing is high and contributes to an improved SNR.
The poor determination of bound water volume on shell 8 is easy to correct. As it is bound water, it should be the same
whatever the depth of measurement, so shell 8 bound water volume can be replaced by the one of shell 1. An alternative is to
perform a 4D inversion where the NMR inversion is performed on all the shells simultaneously and using heavier weights for
shell 1 measurements sensitive to bound water.
Gas volume measurement is more challenging to improve and would require undesirable levels of stacking. Hence an
alternative processing solution is proposed. Instead of solving for an apparent porosity, enable MR Scanner’s fluid typing
capability to compute an apparent fluid density. This apparent fluid density can then be used to compute porosity from
This technique is not new (Minh, 2007). The computed fluid volumes from the three shells were used to compute a fluid
density and invasion profile. This was then combined with the radial density response to compute an apparent fluid density in
the density measurement volume. However, as observed previously, the gas volume is not accurate on the deep shell so an
alternative workflow was developed to avoid propagating the large error of the gas volume to the final porosity computation.
Fig. 10 displays the ideal NMR response for a 30 pu sand. The bound water should be the same on the three shells, the OBM
filtrate volume should be equal to or less for subsequent deeper reading shells if a normal invasion profile is assumed and the
total porosity measured on each shell is constant. Unfortunately deeper reading shells also have much more statistical
variations in fluid volume. Hence only the fluid volumes with good accuracy are used in this processing; that is the shallower
shell bound fluid volume and the oil volumes for each shell.
The NMR gas volume is not used as its uncertainty is relatively high, rather the gas volume computed is the difference
between the total porosity and the combined bound water and oil volumes.
Total porosity is the unknown we want to solve for, so we need to perform an iterative processing with the following steps:
View Full Document
- Fall '12