I read the paper on "diffusive growth", and wanted to comment on your Bakken wells figures. First the NDIC typical well does not match well with the actual data from the NDIC on Bakken output. That is if we assume all wells added each month are "NDIC typical wells" output would be considerably higher than the current 820 kb/d. Any of the three hyperbolic well profiles in my recent blog post match the NDIC data closely. I matched an OU diffusive model to the medium TRR well profile, and get the following:

C0=11,165,000, D=6.63E-06, S=0.00596,
where
C(t)=C0/(1+1/sqrt(D*((1-exp(-S*t))/S))),

and C(t) is cumulative barrels of output in month t, and t is month from first output from a given well.
The Arps Hyperbolic being matched is below:
Med TRR- qi=10570, b=0.81, di=0.0883, EUR(10)=259 kb, EUR(20)=311 kb, EUR(30)=339 kb

Note that using data directly, to estimate an OU Diffusion Model does not work well for the Bakken or Eagle Ford because we only have 24 to 36 months of data to work with and the OU Diffusion model matches the data poorly over the short term. I have found that a hyperbolic model must be used and then an OU Model can be fit to that hyperbolic. The best match is a hyperbolic over the first 30 months and an OU model for Months 31 to 720. You may have a more elegant solution.

Charts of OU Diffusion Model vs Hyperbolic at links below, note that x axis is months from first output and y axis is cumulative barrels of oil produced.

test image

Hi WHT, looks interesting. ðŸ™‚

i'm bookmarking your blog - you have good stuff. Rgds WeekendPeak (from TOD)

test

http://twitter.com/WHUT/statuses/381476066495176704

Test one

Hi WHT,

This blog is great!

I read the paper on "diffusive growth", and wanted to comment on your Bakken wells figures. First the NDIC typical well does not match well with the actual data from the NDIC on Bakken output. That is if we assume all wells added each month are "NDIC typical wells" output would be considerably higher than the current 820 kb/d. Any of the three hyperbolic well profiles in my recent blog post match the NDIC data closely. I matched an OU diffusive model to the medium TRR well profile, and get the following:

C0=11,165,000, D=6.63E-06, S=0.00596,

where

C(t)=C0/(1+1/sqrt(D*((1-exp(-S*t))/S))),

and C(t) is cumulative barrels of output in month t, and t is month from first output from a given well.

The Arps Hyperbolic being matched is below:

Med TRR- qi=10570, b=0.81, di=0.0883, EUR(10)=259 kb, EUR(20)=311 kb, EUR(30)=339 kb

Note that using data directly, to estimate an OU Diffusion Model does not work well for the Bakken or Eagle Ford because we only have 24 to 36 months of data to work with and the OU Diffusion model matches the data poorly over the short term. I have found that a hyperbolic model must be used and then an OU Model can be fit to that hyperbolic. The best match is a hyperbolic over the first 30 months and an OU model for Months 31 to 720. You may have a more elegant solution.

Dennis Coyne

Charts of OU Diffusion Model vs Hyperbolic at links below, note that x axis is months from first output and y axis is cumulative barrels of oil produced.

DC