Professional Documents
Culture Documents
A Practical Guide To Unconventional Petroleum Evaluation: Boyd Russell
A Practical Guide To Unconventional Petroleum Evaluation: Boyd Russell
Petroleum Evaluation
Boyd Russell
Petroleum Club
November 28, 2012
Outline
Declines
bias
Percentile type wells
Grouping Issues
Recap
Declines
How Much Production Data is Needed?
100%
Mean Well
Percent Variation
0%
Individual Wells
-100%
0
Time, Years
Unconventional Road Show
Declines
Which Decline Method to Use?
Declines
Auto Forecasting Why Its Critical
Forecasts are
implicit to the
ISP method
Implicit
forecasts are
usually
inaccurate
Type well
quality is
compromised
Better forecasts
yield better
type wells
Profit Optimization
Best wells drilled first
Implicit forecasts for the newer wells created from older,
better wells
Type wells are optimistic
Technical Play
Wells improve as technology develops
Implicit forecasts for the newer wells created from older,
poorer wells
Type wells are pessimistic
200
History 70%
cutoff
History 50%
cutoff
150
100
50
0
0
50
100
150
200
250
Type Wells
Percentile Type Wells
2.
Selected Wells
. Select wells with EUR similar to the desired percentile ranking
. Build a type well using the selected wells
P75
Time Slice uses of wells and full probability range
Red Dots
EUR distribution for
Hugoton wells
Blue Dots
Time slice EUR
Deviates significantly
from desired EUR
Not reliable, even
when prediction is
used
The Time Slice method will not reliably return the desired EUR
Type Wells
Grouping Issues
Recap
Resource plays are statistical.
Resource plays are game changers that require new technologies.
Decline Curve Analysis requires a minimum two to five years of
historical data to be reliable.
Individual well forecasts are not reliable when based on early
production data.
Stretched exponentials, were developed to replace Arps and
address transient flow issues.
Stretched exponential characteristics:
Complex and difficult to solve.
Data manipulation is convoluted.
Do not handle multiple flow regimes.
Not better at forecasting than using Arps with a five year or other
suitable transition.
Recap
Easy and accurate auto-forecasting is required to perform large
resource studies.
Auto forecasting algorithms must handle multiple flow regimes,
bad data, operational issues and any other eventuality and select
the most appropriate segment for the forecast.
Auto fit algorithm must converge and provide the best fit.
Building analogs requires statistically valid groups and proper
type curves.
Always combine history with an accurate forecast when creating
type curves.
When creating percentile type wells, the selected well approach
should be used and the time slice method should be avoided.