American Oil and Gas Reporter - August 2017 - 74

SpecialReport: Horizontal Wellbore Construction
reservoir or fluid characteristics.
The models are built with predictive
capabilities that can calculate well performance estimates based on well architecture, measured reservoir characteristics,
well spacing, completion parameters, frac
job practices and operational constraints.
Data from blind wells that were not used
in developing the predictive models are
used to validate the results. The predictive
models generate type curves for different
areas of a field that summarize well performance as a function of multiple reservoir characteristics and design parameters,
and they can be combined with Monte
Carlo simulation to:
* Quantify uncertainties associated
with well productivity;
* Measure and compare the quality
of historical frac jobs performed in a
field, area or particular zone;
* Determine the amount of reserves
and production that potentially have been
missed as a result of sub-optimal completion practices; and

("production index") is plotted against
four measured well parameters: the number
of stages, the amount of proppant pumped
per foot of lateral, net thickness and the
stimulated lateral length.
Looking at this figure, it is very hard
to detect any patterns or trends from the
raw data. WQA incorporates fuzzy set
theory to classify and plot wells based
on the fuzzy membership function of the
classifications (performance levels of the
individual wells). Although the techniques
used are extremely simple and classifications are intuitive (i.e., poor, average
or good wells), the results are quite revealing of the nature of the oil and gas
production from shale. In many cases,
clear trends and patterns can be extracted
from seemingly chaotic data.
Using the fuzzy membership functions
and classifications, the data demonstrated
in Figure 1 were plotted for each class of
wells to search for patterns in how poor
wells, average wells and good/very good
wells behaved as a function of several

* Measure and rank the success of
previous completion/stimulation practices.
Moreover, the predictive model can
be combined with evolutionary optimization algorithms to identify the optimum
frac designs for new wells, and map the
natural fracture network as a function of
well and completion design or other operational constraints, and identify refrac
candidate wells and recommend the most
appropriate completion design.
WQA And KPIs
As part of the pre-modeling analysis,
the WQA algorithm performs analysis on
raw data collected from the field. While a
priceless treasure in shale plays, raw data
from well construction, logging, completion,
stimulation, production and other operations
does not reveal much about storage and
transport phenomena in shale. Figure 1
shows an example of raw data from a
grouping of some 140 Marcellus horizontal
wells. Pressure-corrected, 30-day cumulative production in barrels of oil equivalent

FIGURE 2
Well Quality Analysis of 140 Marcellus Wells
Classiﬁed by 30-Day Cumulative Production Performance
11

Number of Stages

128
126

Poor
Wells

124
122
All Data

Poor Wells

Avg. Well

25

Good Well

3,160

3,100

3,000
2,950

Average of
All Wells
Poor
Wells

2,800
2,750

2,715

2,700
All Data

Poor Wells

Average
Wells

2,905

Good Wells

3,050

2,850

30

Avg. Well

All Data
3,150
3,100
3,050
3,000
2,950
2,900
2,850
2,800
2,750
2,700

74 THE AMERICAN OIL & GAS REPORTER

Excellent Wells

Good Wells

Very Good Wells

Average
Wells

Very Good Wells

Good Wells
Good Well

Poor
Wells

1,300

All Data

Very Good Wells

Poor
Wells
Poor Wells

Avg. Well

Good Well

Poor Wells

Avg. Well

Good Well

130

Excellent Wells

Very Good Wells

Very Good Wells Excellent Wells

Net Thickness

135

Average of
All Wells

125

Poor
Wells

120

All Data

Very Good Wells

Lateral Length

3,100

Poor Wells

Avg. Well

Good Well

Very Good Wells Excellent Wells

Lateral Length

3,050
3,000

Average of
All Wells

2,650
2,600
Good Well

1,350

140

Average of
All Wells

20

Lateral Length

3,150

Avg. Well

Net Thickness

121

120

Poor Wells

1,400

Excellent Wells

Average of
All Wells

All Data
40

1,450

1,250

35

Good Wells

134

2,900

Poor
Wells

1,500

Good Wells

Good Wells

137

136

130

1,400

Proppant per Foot
Average of
All Wells

All Data

Poor
Wells
Poor Wells

Avg. Well

Good Well

2,950
2,900
2,850

Average of
All Wells

2,800
2,750
2,700

Poor
Wells

2,650
2,600
Very Good Wells

All Data

Poor Wells

Avg. Well

Good Well

Excellent Wells

138

132

1,450

1,350

142

1,600
1,550

1,500

Good Well

Very Good Wells Excellent Wells

Very Good Wells

Avg. Well

Good Well

Very Good Wells

140
Average Net Thickness (ft)

Poor Wells

Net Thickness

Avg. Well

Good Wells

All Data
142

Very Good Wells

Proppant per Foot
Average of
All Wells

1,440

1,450

Good Well

Poor Wells

Good Wells

Poor
Wells

1,550

Poor
Wells
All Data

1,650

Average
Wells

1,500

Average
Wells

Average of
All Wells

Average
Wells

Proppant (lbs/ft)

1,610

Avg. Well

1,650
1,600

1,600

Poor Wells

8.50
8.00

Average
Wells

Proppant per Foot

All Data

Average of
All Wells

9.00

Average
Wells

1,700

9.50

Average
Wells

Good Well

Very Good Wells

Avg. Well

Very Good Wells

Poor Wells

1,650

1,550

Number of Stages

7.50

Very Good Wells

All Data

Poor
Wells

8.00

Good Wells

8.5

Good Wells

Poor
Wells

8.50

Good Wells

9.00

9.00

Average
Wells

Average of
All Wells

Average of
All Wells

Average
Wells

9.50

9.50

Average
Wells

9.6

Good Wells

10.00
10.00

8.50

Average Lateral Length (ft)

10.50

Number of Stages

10.50

10.00

Average
Wells

Average Lateral Length (ft)

10.50

Very Good Wells Excellent Wells
American Oil and Gas Reporter - August 2017

Table of Contents for the Digital Edition of American Oil and Gas Reporter - August 2017
