Thursday, May 16, 2013

Updated Well Profile for the North Dakota Bakken and the Effect on future scenarios

Note that all of the "Bakken" Scenarios I have presented, in this post and previous posts only cover the North Dakota portion of the Bakken/Three Forks play.

A recent post by Rune Likvern at the Oil Drum entitled, Is the Typical NDIC Bakken Tight Oil Well a Sales Pitch? has provided some new information which I will use to update my recent scenario for future Bakken output.

Another post by Mr. Likvern from which I gathered considerable knowledge is entitled Is Shale Oil Production from Bakken Headed for a Run with “The Red Queen”? . In addition a post by Heading Out at the Oil Drum had a number of helpful comments by both Mr Likvern and Webhubbletelescope, the entire thread has comments of interest. There are also a number of posts on the Bakken at The Oil conunDRUM  blog(written by Webhubbletelescope) from which I have learned much. I am indebted to both Webhubbletelescope and Mr Likvern for sharing their knowledge.

Previous research by Mr. Likvern suggested that average well productivity had been declining by as much as 20 % per year. Recently this decline has stabilized at around 85 kb for the first year of output from the average well, the second year of output is about 45 kb for a 2 year EUR of 130 kb. The well profile is below (note that any chart can be clicked to enlarge):


I attempted to match this profile for Jan 2008 to Jan 2013 below:

Figure 2- Model 1 Avg Well Profile

Note that the left scale is different on my chart and the monthly output rather than yearly average output is shown, only the highest monthly output curve and cumulative output (Jan 08 to Jan 13) should be compared with Mr. Likvern's curves.  The agreement between the two cumulative curves is fairly close, with the 20 year EUR about  380 kb for each well profile.

For comparison, the well profile I use is somewhat less optimistic:

Figure 3 - Model 3 Avg Well Profile

The 20 year EUR is about 284 kb. A concise summary of the Arps equations used for modelling the decline of an oil or gas well by Dr Steven W. Posten is entitled Decline Curves.

The hyperbolic decline curve matching Mr Likvern's average well has qi=14,700 barrels/month, b=1.48, and Di=0.28, note that this curve violates the rule suggesting b should be between 0 and 1, such curves are unrealistic in the sense that the cumulative output will approach infinity over long time periods.

My well profile has qi=14,925 b/month, b=0.95, and Di=0.19.

The following charts compares the NDIC well profile (qi=21,500 b/m, b=1.094, and Di=0.185), with the NDIC Typical Well presented in the Emmons Presentation at the NDIC website (slide 29), Rune Likvern's data (read from figure 1), the hyperbolic (Model 1) which closely matches it, my model (Model 3), and an intermediate model (Model 2).

Figure 4

Figure 5
The chart above is for illustration only(I do not think any wells will produce beyond 50 years of first output and most will be shut in before 30 years). Note that Model 1 does not level off much with time and at 45 years is still producing 12 barrels per day and at 1000 years output is still 1.5 barrels per day. Model 3 is not a lot more realistic as output is 1.5 barrels per day at 113 years and 3.8 b/d at 45 years.

Using the well profiles presented in figures 2 and 3 and NDIC data for Bakken/Three Forks oil output in North Dakota, we can check how well these well profiles match the data. For Model 1 we have:

Figure 6
For Model 3 we have:

Figure 7
Recent USGS Mean Estimates (released April 30, 2013) suggest 7.4 BBO (billion barrels of oil) for the entire Bakken/Three Forks (slide 16) with a range of 4.4 to 11.4 BBO at the F5 and F95 levels respectively. For the North Dakota portion of the play the mean estimate is 5.8 BBO(slide 18).

I have adjusted my medium scenario to try to match this mean estimate by adjusting the rate that average well productivity decreases and not allowing the rate of increase in producing wells to increase beyond 2000 net wells added per year (currently the rate is 1769/year). For Model 3 we have:
Figure 8
This scenario ignores economics and assumes oil prices will rise enough to keep breakeven oil prices below market prices so that new wells remaine profitable, essentially it assumes all technically recoverable reserves (TRR) will be economically recoverable (profitable).  If we leave all assumptions  the same as in Figure 8 with the exception of using Model 1 for our well profile we get the following scenario (Figure 9):

Figure 9
Note that both these scenarios (in figures 8 and 9) follow the NDIC prediction of 42,500 producing wells and that the expected number of wells added per year will be 2000 wells/year (see slide 28 in the Emmons Presentation ).  Both scenarios start at about 1800 wells/year in 2013 and remain at that level to 2015, then ramp slowly to 2000 wells per year by the end of 2018.  Also note the increase in output over the 1953 to 2073 period for Model 1 (9.6 BBO) vs Model 3 (5.9 BBO), a 63 % increase.

In order to match the USGS Mean estimate for TRR while also following the NDIC's future scenarios for rate of producing wells added per year and total number of producing wells the productivity of the average new well had to decrease as follows:

Figure 10

 By 2028 average well productivity falls to 8 % of the Jan 2013 level.  To make the scenario more realistic, we now attempt to consider profitability.  We initially assume a fairly rapid rise in real oil prices (2013 $) of 7.57 % per year from 2013 to 2037.  We also assume that new wells are only added if the real break even oil price of the average well is lower than the real price of oil.  To calculate break even oil prices at the refinery gate, we assume well drilling and fracking costs are $9 million per well, OPEX plus financial costs are $7/barrel, transportation cost is $12/barrel, and royalty and taxes are 25 % of the well head revenue (refinery gate revenue minus transportation costs).  Then we use a discount rate of 10 % to find the net present value of the net revenue (revenue-cost) stream from oil output over an assumed 30 year well life.

Figure 11
Note that the breakeven prices are based on Model 3 and average well productivity decreases as shown in figure 10 above.  Based on the assumptions for oil prices, break even oil prices, and well productivity decreases and using the scenario presented in figure 8 above, we expect that new wells will no longer be added after early 2018 because they would not be profitable.  The following scenario is the result:

Figure 12
Replacing the Model 3 well profile with Model 1 (and no other changes) we have:

Figure 13
The introduction of break even prices lowers the output from 1953 to 2073 to 4.7 BBO from 5.9 BBO in the initial scenario (TRR) for Model 3.  Model 1 output from 1953 to 2073 would be 7.7 BBO if oil prices rise at a slower rate than I assumed such that the breakeven price of $129/barrel is above the real oil price by early 2018.  Note that the higher well profile of Model 1 would lead to a lower break even price of $129, if this well profile is correct then the USGS mean estimate would be likely be surpassed unless prices rise much more slowly.

The Model 1 well profile may be too high and Model 3 may be more realistic, future data will help to sort this out.  I created a scenario using the Model 1 well profile which results in output similar to the USGS mean estimate of 5.8 BBO over the period from 1953 to 2073. The scenario assumes that prices remain below the Jan 2017 break even price of $107/barrel so that new wells are no longer added after early 2017 and that 1500 wells are added per year from 2013 to 2016.  Total wells are 11,000 and output from 1953 to 2073 is 5.9 BBO, well productivity decreases as in figure 10.  See figure 14 below:

Figure 14
If we use the scenario above, but change the average well profile to Model 3 we get:

Figure 15

In this case prices would need to rise to $112/barrel by July 2016 in order for new wells to be added profitably.  Output from 1953 to 2073 falls to 3.6 BBO from 5.9 BBO using Model 1(figure 14).

In a future post I will look at scenarios which would match the USGS high and low estimates using Model 3 and assuming that F5 and F95 estimates for the total Bakken/Three Forks can be used to estimate F5 and F95 for the North Dakota portion.

This is a revision of a very rough draft I posted on May 16, 2013, I apologize to anyone who read it, I will do a better job of editing before posting in the future.



1 comment:

  1. Did you know you can shorten your long urls with Shortest and earn money from every visitor to your shortened urls.