(554b) Ensuring Environmental Friendliness of Horizontal Shale Gas Wells Through Zonal Isolation: A Model-Based Approach

Shale gas horizontal wells face
problems related to gas leakage from various zones of a well into the air and
water reserves. Protecting the environment and improving well productivity is
one of the biggest challenges for shale gas production. This problem can be
solved by improving zonal isolation. Zonal isolation is necessary to avoid any
communication between rock formations penetrated by a well. Zonal isolation
mainly depends on the quality of well cementing, namely the tightness of the
seal created by the cement placed between a metal casing and well walls.

Cement bond quality is governed
by various physical factors and conditions. Casing data (internal casing
diameter, number of centralizers and casing-hole relationship), cement system
data (slurry type, density, yield, additives composition), rheological
properties of drilling mud & cement and operating temperature and pressure
conditions. Cement bond quality is measured in terms of a bond index (BI) value,
with zero value indicating poor cementing and value at 1 representing perfect
cementing.  The BI value can be calculated using Cement Bond Log (CBL) data,
following a method given by (Fitzgerald
1985).

As it has been well established
that a large number of physical factors affect the overall cement bond quality
and corresponding BI value, a model capturing the effect of such factors on BI
would be useful for the design of effective and efficient cementing jobs.  Such
a model can be developed from available field data.  Because data are
relatively scant, Partial Least Square Regression analysis (PLS) (De
Jong 1993) is used for model development.

A PLS model was built using as
inputs dimensionless quantities related to all physical factors considered to
affect BI, and as outputs the average and standard deviation of BI along the
length of a cemented well casing. It was found that out of all quantities taken
in the analysis, few dimensionless quantities played a major role in affecting
the BI.  Cross-validation (Leave-one-out) results obtained from the PLS model
are presented in Figure 1. Figure 2 shows
the latent variable loadings for major contributing dimensionless factors.

Figure
1  Cross validation VS
Experimental BI data

Figure
2  Latent variable
loadings for various dimensionless quantities

The above as well as additional
results on the average and standard deviation of the BI are discussed in the
presentation.

In conclusion, the PLS model
presented here can be used to qualitatively determine the significance of
physical factors affecting the quality of well cementing (captured by BI value). 
More importantly,  it points to the need for additional data such that
model-based tools can be built that will aid in ensuring consistent and
predictable quality of well cementing jobs. References

De Jong,
S. (1993). "SIMPLS: An alternative approach to partial least squares
regression." Chemometrics and Intelligent Laboratory Systems 18(3):
251-253.

Fitzgerald,
D. D., B.F. McGhee, and J.A. McGuire (1985). "GUIDELINES FOR 90% ACCURACY
IN ZONE-ISOLATION DECISIONS." Journal of Petroleum Technology 37(12):
2013-2022