Decision trees optimize workover program

Sameh Macary
Egyptian Petroleum Research Institute
Cairo

Amr El-Haddad
Gulf of Suez Petroleum Co.
Cairo

A rapid increase in well workover activity in the October field, Gulf of Suez, led to a decision tree approach for enhancing the chances for obtaining successful jobs.

Workover costs are a significant part of the expenditures made by petroleum producing companies. Designing workovers is a complex process that may involve material incompatibilities, chemical interaction, physical restrictions, and cost considerations.

The analysis demonstrates how simple risk-assessment techniques can be used to minimize the risk associated with workovers. The application of these decision tools has enhanced workover evaluation, and consequently improved oil production through 1997.

Gulf of Suez Petroleum Co. (Gupco) operates the October field, the largest oil producing field in Egypt.

The statistical evaluation of October field workovers involved all available data, over the entire life of the field. The workover history included about 600 different jobs performed in about 80 wells.

The analysis classified these workovers by field, reservoir, platform, well, cost, type, contractor (service company), rig/rigless jobs, and job duration. Workover jobs are subgrouped into mechanical, water shutoff, perforation/reperforation, acid stimulation, and tubing cleaning.

October field

The October field is in the north-central part of the Gulf of Suez. It covers about 7,000 acres and has an estimated ultimate oil recovery of 1,163 million bbl.

The field marked the first discovery of oil in a Nubia reservoir in the northern half of the Gulf of Suez. Another importance aspect of October is the discovery of the Asl oil-bearing formation, downthrown to the western boundary faults.

The field provides about 40% of Gupco's production, and about 15% of Egypt's total production.

October field was discovered in May 1977 and began producing in November 1979. Currently, it produces about 140,000 bo/d, and has a cumulative oil production of about 800 million bbl.

The Nubia formation is the main oil reservoir, with a 1,100-ft gross thickness. The Nezzazat group, Matulla-Wata-Raha, with a 500-ft gross thickness, and Nukhul clastic, 300-ft gross thickness, are secondary oil reservoirs.

October field is an elongated, northwest-southeast trending, northeast dipping pre-Miocene fault block. The field is bounded to the west by a large, down to the west fault and to the south by an east-northeast trending cross fault.

It is also separated into the Main (South) October block and North October block by a north-northeast trending cross fault. The North block is downthrown relative to the South block.

The field is bounded to the east by an oil/water contact, originally at 11,670-ft subsea in the Nubia.

The Nubia has an active water drive. Both bottom and edge water contribute to the water moving up structure.

Recently, a desire to accelerate production has increased the workover activity. In addition, the reservoir pressure has declined over 3,000 psi during the period covered by this study. This means that workover procedures that may have been successful 10 years ago, when water cuts were low and reservoir pressure was high, may now not be appropriate.

Data analysis

The Gupco management team initiated a project to investigate the optimization and improvement in decision making before any future workovers. Therefore, a statistical workover evaluation was done for the period November 1979 to mid-1997. The evaluation examined 574 workovers in 83 wells ( Table 1 [9,301 bytes]).

The survey includes all workover types such as mechanical, water shutoff, perforation/reperforation, stimulation, and tubing clean out. These workovers are classified by field, reservoir, platform, well, rig/rigless, job type, duration, and service company.

The data-gathering phase created a workover data base for 38 mechanical jobs, 107 water shutoffs, 51 perforation/reperforation jobs, 147 stimulations, 141 tubing clean outs, and 90 combined jobs.

Mechanical and perforation/

reperforation jobs were not included in the decision tree analysis because the jobs were complex and were combined with other job types. Also, there was an insufficient number of jobs to guarantee a fair statistical evaluation.

Data were collected from various sources such as daily production reports, monthly status reports, 6-month workover reports, and Cairo/Ras Shukeir field well files.

Table 2 [115,379 bytes]summarizes the workovers up to mid-1997. It is worth mentioning that the evaluation considers only unique jobs, not combined jobs.

Mechanical workover

A total of 38 mechanical jobs were done with a total production gain of 70,000 bo/d. Average percentage of success was 68%, and payout time was 13 days. The analysis confirmed that mechanical workover success is dependent on the completion fluid type.

Average losses of production from mechanical workovers, not combined with other jobs, was -19 bo/d/job for workovers using sea water (15 jobs), compared with +312 bo/d/job for workovers with diesel (23 jobs). The fact that damage is less for workovers using diesel than seawater suggests possible clay problems.

Therefore, diesel is recommended as a kill fluid because it is less dense and is considered to be a non-foreigner formation fluid. Also, mud fluid filtration and system cleaning are critical for reducing damage.

Average gain from mechanical workovers, followed by reperforation jobs, was 3,500 bo/d/job (10 jobs), compared with 800 bo/d/job for workovers followed by stimulation jobs (8 jobs). The four-fold production increase after reperforation rather than stimulation may suggest mechanical damage such as suspended solids, redeposited sand fill, etc.

More work is planned on the kaolinite clay problem and this may lead to further changes in October workover procedures.

Water shutoff

A total of 107 jobs (9 rig and 98 rigless) have been performed with a production gain of 210,000 bo/d.

Success averages 82% with a 3-day payout time. The time distribution ( Fig. 1 [358,745 bytes]a) reflects a decrease in rig activities during the last 5 years, accompanied by an increase in rigless activities. The main Nubia showed an excellent success ratio of 84% from 101 jobs.

The aerial distribution of these jobs (Fig. 1b) indicates the Nubia water movement began in 1983 from the C platform and continued moving in a linear direction to the A and B platforms in 1989.

A significant increase in shutoff activities took place in 1993 because water reached most platforms. This suggests the start of a radial-flow water drive (Fig. 2 [117,193 bytes]).

Planned simulation modeling will update water movement for all Nubia zonation.

Perforation/reperforation

Gupco has run 51 perforation/reperfo ration jobs (four rig and 47 rigless) and have realized a 35,000 bo/d production gain (Fig. 1c). The Nubia showed an average gain of 729 bo/d/job (42 jobs), compared with 333 bo/d/job (6 jobs) for the Nezzazat reservoir.

Average percentage of success was 56%, and payout time was 6 days.

The relatively low percent of success is probably due to perforating commingled zones that have different permeabilities (M1, MN, and TZ sub-zones). The flow from the most permeable zone (M1) creates a backpressure on the lowest permeability zone (TZ).

Because perforation/reperforation workovers have a relatively low success and low gain per job (680 bo/d/job), perforating commingled zones of different permeabilities should be limited.

Acid stimulation

Up to mid-1997, Gupco performed 147 acid stimulation jobs that resulted in an overall production gain of 121,000 bo/d.

Percentage of success averages 76% with 3-day payout time. The Asl reservoir showed an excellent response to acid stimulation (100% success). Average gain was 1,700 bo/d/job for Asl (16 jobs), compared to 778 for Nubia (92 jobs), 565 for Nezzazat (37 jobs), and 500 for Nukhul reservoir (2 jobs).

The wide range of stimulation results (Fig. 1d) between different areas in the Nubia reservoir led to the construction of a "probability of success map." This map contoured wells with the same percent of success for acid stimulation jobs throughout their life (Fig. 3 [50,993 bytes]).

Three different areas of success (low, medium, and high) were obtained. The new contour map matches reasonably well with the kaolinite map (Fig. 4 [66,357 bytes]).

Some questions arose concerning the effect of water cut on stimulation efficiency. Average production gain from 95 stimulated wells having less than 10% water cut (before job) was 1,055 bo/d/job. The success was 80%. This compares with a production gain of 385 bo/d/job and a 66% success for high water cut wells (48 stimulated wells having higher than 10% water cut).

Also, the comparison (Fig. 1e) of different contractors (service companies) showed a lower result for BJ Services, 318 bo/d/job (22 jobs and 59% success).

The second contractor Halliburton (HAL) had an average production gain of 470 bo/d/job (17 jobs-71% success), compared to 1,062 bo/d/job (32 jobs-77% success) for the third contractor, Schlumberger Dowell (DS).

These results led to a sensitivity analysis to evaluate the influence of additives used on stimulation results.

Tubing clean outs

A total of 141 tubing clean outs showed a production gain of 81,000 bo/d. Average percentage of success was 59% and payout time was 2 days. The time distribution of these jobs (Fig. 1f) shows scale deposition started in 1985 and gradually increases with time.

Average production gain (Fig. 1g) in the Nubia was 527 bo/d/job (108 jobs) compared to 875 bo/d/job for Asl (8 jobs), 296 bo/d/job for Nezzazat (21 jobs), and 312 bo/d/job for Nukhul reservoir (4 jobs). These results indicate that HCl acid is an excellent scale remover in Nubia and Asl wells

BJ (Fig. 1h) had the highest success, in terms of production rate, of 1,030 bo/d/job (37 jobs-72% success in terms of jobs), compared to 952 bo/d/job for HAL (24 jobs-76%) and 769 bo/d/job for DS company (16 jobs-58%).

Decision trees

By definition, the decision tree is a graphical representation of expected value calculations, consisting of the decision chance of terminal nodes connected by branches. It is nothing more than a pictorial summary of a series of events. ¹

This summary increases in importance as projects increase in complexity. Decision tree analysis evaluates current and contingent future decisions, and provides a logical, defensible basis for decisions.

Also, decision-tree techniques explore alternatives and modify them. Therefore, the decision tree is an updateable technique.

These trees are especially useful in representing joint and conditional probabilities. Decision trees are also additive models that accept new branching or details. By using a decision tree, one can determine the probability of certain outcomes and generate a density (mass) curve of cumulative probability.

Decision trees help organize and present information. They were introduced over 20 years ago, and although widely discussed in technical journals, are not widely applied.

To illustrate their use, three decision trees were constructed for water shutoff, acid stimulation, and tubing clean out jobs for the October field, Nubia reservoir. The output considers only the last 5 years, during which most of the workovers were done.

For water shutoff jobs, the branches of the decision tree introduce rig/rigless, duration for rig jobs, cost, and probabilities of success. The values of $493,000 for cost and 11 days for duration (Fig. 5a [113,840 bytes]), represent the mode value of all water shutoff jobs.

The probability of each outcome for each branch in the event tree is calculated by multiplying the probabilities at each node.³ For Example, the calculation for the probability P(x) of having a successful rigless water shutoff costing less than $67,0000 is as follows:

P _{rigless jobs} * P _{cost $67,000} * P _success = 0.920 * 0.753 * 0.828 = 0.574.

This compares with a probability of 0.036 to have a successful rig job at a cost less than $493,000 within about 11 days.

In decision trees, the sum of all branches having the same origin equals one. Also, the end product of the decision tree, for example, P(x) equals unity.

Figs. 5b and 5c contain another node that considers contractor probability of success. For acid stimulation, one gets probabilities of success of 0.074, 0.174, and 0.047 for BJ, DS, and HAL, respectively. For the same sequence of companies, the probabilities of success of tubing clean outs with the lowest cost are 0.309, 0.059, and 0.139.

These decision trees in conjunction with the probability of success maps enhance decision making, check the sensitivity of an option to different parameters, and highlight cost critical areas.

Study results

The initiated workover data base for October field has led to a complete statistical evaluation for these jobs. Decision trees were then constructed to illustrate the probability of different options.

Decision trees also serve for quality control. In other words, after monitoring the workover activities and optimizing their efficiency, probability of success should be recalculated and compared with the original one. If the difference is positive, it means an increase of success probability.

The application of such decision tools has enhanced the overall evaluation of October workover activities. Consequently, improved oil production has been observed in 1997 for most of these activities.

Table 3 [18,863 bytes] summarizes the 1996-97 gained oil production for each $1,000 of job cost.

Water shutoff and tubing cleaning jobs showed an excellent improvement of about 15% in oil production. Also, some improvement (3%) has been observed for stimulation activities (Fig. 6 [57,667 bytes]).

Mechanical and perforation/reperforation showed extremely bad results in 1997 and will require more efforts to minimize damage from normal workover operations. Workovers should be performed without killing the well. This includes using such equipment as coiled tubing, snubbing units, and through-tubing wire line.

One difficulty faced by the management team was the timing of the production tests before and after the workover. The evaluation of success is strongly influenced by the timing of the test.

Acknowledgment

We thank the Gupco engineering group for its invaluable support and assistance and the Gupco management for permitting us to publish this article.

References

1. Petroleum Risks and Decision Analysis, OGCI Training Course Manual, Houston, December 1995.
2. Campbell, John, Analysis and Management of Petroleum Investments: Risk, Taxes, and Time, Second edition, February 1996, CPS Publishers.
3. Holgate, N.J., et al., "Accounting for Negative Publicity in Project Economics," JPT, July 1997, pp. 719-23.

The Authors