Notice: The data I thought would be out today: EIA Crude Oil Production by State will not be out until Thursday, Feb. 27th. However I will have another post coming out later today anyway.
This is a guest post by Ovi Colavincenzo
There is considerable discussion on this site regarding when the North Dakota portion of the Bakken will peak. Having looked at the monthly Bakken data that the State publishes, it raised the question of whether it was possible to do a reverse analysis of the data and then use it to develop a model that would replicate the ND Bakken production, exactly. The objective being to provide further insight on what is happening in the ND Bakken.
In order to do this, the following conditions and information were required:
- A monotonically increasing number of new producing wells
- A typical/average decline curve for the ND Bakken field
- Not too many wells being shut/reworked each month
The last bullet is a preferred condition because if a number of low producing wells are shut and replaced by newer high producing wells, then the estimated flow rate of the new wells will be on the high side.
From 1999 to mid 2005, approximately 200 wells were in production in every month. The addition of an increasing number of new wells began to occur in mid-2005, so start date for the analysis was set at the beginning of 2008 to address the first bullet point above.
Figure 1: Source: The Shale Revolution” by J.D. Hughes
For the decline curves, two were used and are shown in Figures 1 and 2. One came from “The Shale Revolution” by J.D. Hughes, November 19, 2013 shown in Figure 1. The other came from North Dakota’s Directors cut, “Tribal Leader Summit” 09 05 12 (PDF), Figure 2.
Figure 2: Source: North Dakota Director’s Cut “Tribal Leader Summit”
Note that the ND data is yearly and the Hughes data is monthly. While the Hughes data represents the average decline rate for a number of wells, it is not clear what data was used to derive the “Typical Bakken Well Production” curve shown in Figure 2. Note the significant difference in the initial production rates, 904 b/d vs 580 b/d. Also a post in “The Oil Drum #9954”, by Rune Likvern, raises questions of how typical it is.
Since it is not clear from the monthly ND data, whether wells are being closed, either temporarily for maintenance/re-fracking or permanently, the ND production and wells data were averaged using a three month moving average prior to being used in the development of the model. The intent of using the moving average was to minimize the effects of closing and reopening wells. There is some indication that the opening and closing of wells does impact the variability in the results and is shown below in a comparative chart.
To conduct this analysis, a normalized Bakken decline curve was required. The decline curves shown above were normalized for this analysis. Two curves were used in order to assess the impact on the results from using two different curves. The two normalized decline curves are shown in Figure 3. The major difference between the two occurs in the first two years. After the first year, the Hughes decline curve is down by 71% vs. 53% for the ND Director.
Note that there are only four years of original Hughes data. Beyond four years, the data is extrapolated and parallels the ND Director data. The original North Dakota Director’s Cut data was yearly and the monthly data was interpolated from the yearly data using a sliding quadratic function.
Figure 3: Normalized Bakken decline curves
To assess the affects of the differences in the two data sets, the cumulative production associated with each decline curve was compared. The curves in Figure 4 show how the production rate increases due to drilling one identical well each year in the first month of each successive year. Looking at the Hughes curve, after drilling 1 well each year for 7 years, the production rate from the 7 wells is only 1.9 bbls/day in 72th month. After 7 years of production, adding one new well at the beginning of the 8th year only increases the production rate of the field by 0.07 bbls/d, to 1.97 b/d, i.e., 7% of the initial production rate of the new well. The Red Queen has shown up.
Figure 4: Cumulative yearly production from adding one well each year
While the ND Bakken data published monthly shows the overall increase in the field’s production, it does not provide information on the productivity of the new wells completed in that month. As indicated above, the objective of this analysis is to calculate the amount of additional new production required each month to use in a model that would exactly match the ND Bakken production data, after accounting for depletion from all previous months. Below are three charts that summarize the results of the analysis and the model’s output.
Figure 5: Monthly new production
Figure 5 shows the amount of production added by the new wells each month to match the ND Bakken data that was averaged over three months.
Figure 6: Monthly addition of new wells
In Figure 5 note the dramatic increase in additional production from roughly 30 kb/d to 60 kb/d in mid 2011. This sharp increase is related to the sharp monthly increase in new well completions from 70/Mth to 130/Mth during the same period shown in Figure 6. The trend in new crude production parallels the number of new well completions to the beginning of 2013 and into 2014. What is surprising is the drop in the average number of new well completions beginning in mid 2013. Is this a sign of things to come?
Figure 7: Comparison of Hughes and ND Director initial production rate
The average initial production rate of the new wells added each month is shown in Figure 7. One can see a roughly constant production from the new wells beginning in mid 2008 up to 2013 for both curves, except for the two spikes. Looking at the Hughes data, the new wells from mid 2008 to the beginning of 2013 were averaging 500 b/d, increasing to approximately 650 b/d at the end of 2013. A similar trend appears in the ND Director’s Cut graph but the growth in 2013 is less pronounced. This raises the question of whether the increasing well production rates reflect the increasing length of the new wells being drilled or are they being drilled in sweeter areas. I have read that some wells are now 2 miles long and think that initial production would be proportional to the length of the well.
The dip in well production in early 2009 is associated with the dip in total ND Bakken production. The sharp increase in Q1-10, I believe is associated with the temporary closing of older, low production wells, which are then brought back on line after maintenance a few months later when the average well production settles into a more typical stable range.
The decline rate that the new wells must offset to keep production increasing is shown in Figure 8 below. It appears that the decline rate at the end of 2013 is somewhere between 55,000 b/d/Mth and 75,000 b/d/Mth. If the new wells are supplying approximately 500 b/d and the average decline rate is around 65,000 b/d, then a minimum 130 wells/month are required to offset the decline.
Figure 8: Total monthly decline rate from all of the previous months’ wells
It was noted earlier that the original ND Bakken data showed high variability in the number of wells drilled and this had a significant impact on the results when analysing the original data set.
Figure 9: Original data set of monthly new well completions
To provide the readers with an idea of the variation in the original data, the original monthly well data is shown in Figure 9 and its impact on the initial production rate is given in Figure 10. Compare the variability in Figure 9 with Figure 6 and Figure 10 with Figure 7. One can see how the monthly variation in well completions affects the variation in the initial production rates shown in Figure 10. However note that the four month moving average is similar to the Hughes curve in Figure 7.
Figure 10: Impact of high variation in original wells data on initial production rate
In Figure 10, even though only 40 wells were completed in December 2013 (Figure 9), the initial production rate of the wells was still the same as in November and December, close to 800 b/d, based on the original data. In other words the drop in the production rate from November to December, from 911,292 b/d to 862,987 b/d was due primarily to the low number of well completions.
This model is capable of projecting future production rates or scenarios by providing production estimates/guesstimates for future months. However, I believe the model’s greater value is in showing/evaluating how the growth in production in the ND Bakken has increased through a combination of the addition of new wells and drilling wells that have a higher initial production rate. By continuing to track the initial production rate, we should see either a drop in number of wells or a drop in the initial production rate or both as we approach the peak. This behaviour should manifest itself in both decline curves.
I will keep these charts updated on a timely basis or when something significant occurs.