• Editorial Board +
• For Contributors +
• Journal Search +
Journal Search Engine
ISSN : 1598-7248 (Print)
ISSN : 2234-6473 (Online)
Industrial Engineering & Management Systems Vol.18 No.1 pp.104-115
DOI : https://doi.org/10.7232/iems.2019.18.1.104

# Applications of Clustering Models in Offshore Drilling for Crude Oil and Natural Gas

Hesham K. Alfares*, Katta G. Murty, Ahmed A. AlSaaty
King Fahd University of Petroleum & Minerals, Dhahran, Saudi Arabia
University of Michigan, Ann Arbor, USA
Saudi Aramco, Dhahran, Saudi Arabia
Corresponding Author, E-mail: alfares@kfupm.edu.sa
March 9, 2018 August 25, 2018 January 14, 2019

## ABSTRACT

An important problem in offshore drilling for crude oil and natural gas is to find the optimum allocation of wells to drill, to multiple drilling rigs; to minimize the total cost of drilling. In practice, the drilling cost of each well is a function of both the distance to the drilling rig and the rig’s individual cost. In this paper, we consider this problem and present both an exact and a heuristic algorithm under two cases. Case 1 considers the problem in which rig locations are given, while Case 2 considers the problem in which rig locations are to be determined optimally. The new models and algorithms are unique because each drilling rig has a different cost rate per mile. Using data from an offshore oil field under development, the new models are used to minimize the total drilling cost for the two cases. In addition, a computational study is conducted, confirming the effectiveness of the proposed heuristic algorithm.

## 1. INTRODUCTION

Currently, crude oil and natural gas are the most valuable commodities for international trade. For governments of countries with access to the sea, there is a great incentive to discover the crude oil and natural gas reserves in their offshore areas. This has led to the development of offshore drilling, which has now become a thriving multi-billion-dollar industry in several areas of the world. However, offshore drilling is much more expensive compared to conventional drilling on land. This paper uses optimization techniques to reduce the cost of offshore drilling.

In offshore crude oil production, drilling is conducted with flow lines connected to processing platforms (Murty, 2014). Today we are not limited to drilling a well vertically all the way to the bottom, but we can drill in a horizontal direction. Therefore, it is not necessary to move the rig to each well location in order to drill. It is quite possible to group (cluster) a set of wells to be drilled into one group, and then drill all of them one after the other with one rig fixed in a central location. Selecting the locations of different drilling rigs, and determining the set (cluster) of wells to be drilled by each rig, are complex optimization decisions that significantly impact the total drilling cost for a given offshore field.

This paper introduces new models and solution algorithms to optimize the locations of rigs and the assignment of wells to different rigs. The problem addressed here is unique because the drilling cost rates per mile are not equal for the different rigs. Therefore, the drilling cost of each well depends on both the distance to the drilling rig and the cost rate of that particular rig. We first consider the case in which the locations of the rigs are fixed, and then the case in which these locations of the rigs are considered as decision variables. New integer programming (IP) models are formulated, and an effective heuristic solution method is proposed. Computational comparisons are performed to evaluate the performance of the proposed heuristic solution method.

The remaining sections of this paper are organized as follows. The relevant literature is reviewed in Section 2. The specific offshore drilling problem is described, and its background, costs, and parameters are presented in Section 3. Case 1, assuming known and fixed rig locations, is discussed in Section 4. Optimum and heuristic solutions for Case 2, assuming unknown and variable rig locations, are discussed in Section 5. Finally, conclusions and suggestions for future research are given in Section 6.

## 2. LITERATURE REVIEW

In this section, we focus on previous work relevant to the optimum locations of offshore platforms (rigs) and the optimum allocation of the wells in the field to the different platforms. Sahebi et al. (2014) provide a widerscope literature review of optimization models and applications in crude oil supply chains, including facility location problems in offshore oil fields.

Traditional models assume that the locations of the wells to be drilled are given. Devine and Lesso (1972) develop an iterative two-stage algorithm to minimize the cost of developing offshore fields. First, the wells are assigned to different platforms, and then the optimum locations of these platforms are determined. Then, these platform locations are fixed, and the wells are optimally reassigned to the platforms. Hansen et al. (1992) present a model for the optimal location and well assignment of offshore drilling platforms in Brazil. To minimize the total cost, an exact mixed-integer programming (MIP) solution and an approximate Tabu Search solution are proposed. Rosing (1994) discusses the practicality of several assumptions in the model of Hansen et al. (1992). Applying clustering, integer programming and network flow models, Kabadi et al. (1996) propose a two-stage iterative approach to minimize the total drilling cost. First, locations for the drilling platforms are determined, and then each well is assigned to a given fixed-location platform.

In several papers, the locations of the wells to be drilled are not given, but must be determined together with the locations of the drilling platforms. García et al. (2012) determine the locations of the wells and platforms locations and also the connections between the wells and the platforms. Rodrigues et al. (2016) present a binary linear programming (LP) model for minimizing the development costs of an offshore oil field. The model is used to determine the number, location and capacities of the platforms; the number and locations of the wells; and the interconnections between the platforms and wells. Cunha et al. (2016) combine cellular genetic algorithms and Monte-Carlo sampling to optimize well locations in oil and gas fields. Karkevandi-Talkhooncheh et al. (2018) integrate neuro-fuzzy inference with particle swarm optimization and genetic algorithms to determine the optimum placement of wells for drilling.

Several studies consider a wider-scope problem, involving not only drilling locations, but also the optimal planning and scheduling of offshore oil field development and operations. Iyer et al. (1998) formulate a mixedinteger programming (MIP) model for this problem, and use a sequential decomposition algorithm to solve it. Van Den Heever and Grossmann (2000) extend the model of Iyer et al. (1998) by including exact nonlinear functions instead of linear approximations. Carvalho and Pinto (2006) extend the work of Iyer et al. (1998) by adding investment constraints and the effect of reservoir pressure on well production and revenues.

Other variations of offshore oil field optimization problems have been considered. Rosa and Ferreira Filho (2012) optimize the location of offshore drilling rigs, assuming the distances between the wells and the rigs affect not only pipeline length and cost, but also the wells productivity, and hence revenue. Rosa et al. (2018) use a mixed-integer LP (MILP) model to determine the number and locations of the platforms and their well assignments, as well as the pipeline network layout and the pipeline diameters. Sahebi and Nickel (2014) develop an MIP model that determines platform locations, allocates wells to platforms, and optimizes transportation and pipeline planning decisions.

Next, we describe a new problem for locating offshore oil drilling platforms and assigning wells to each one. The unique feature of the problem is that the cost of drilling each well depends not only on its distance to the platform, but also on the choice of platform used.

## 3. CLUSTERING PROBLEMS IN OFFSHORE DRILLING OF CRUDE OIL WELLS

### 3.1 Problem Description and Background

Clustering problems and their applications are discussed by Kabadi et al. (1996). This paper presents an important application of clustering models in offshore oil field development. Given the number and locations of the production well targets in an offshore oil field that need to be developed, the problem is to partition the set of production well targets into clusters or groups ranging up to 25, where each group is to be drilled by a single drilling platform (rig). If the rig locations are not already fixed, we also have to determine the best location to position the different rigs in order to minimize the total cost of developing the field. This is a clustering problem involving huge sums of money. The cost of drilling a production well depends on the distance of the well from the position of the rig and several other factors. Arriving at an exact cost is very difficult, thus it has been estimated by using information from past drilling jobs.

This section is based on a case study for developing an offshore oil field located 60 miles off the Eastern coast of Saudi Arabia. The offshore oil field where this study was done covers a 15×8 mile area located off the east coast of Saudi Arabia, and it has 90 potential target locations to be drilled. Currently, there are four jacked-up drilling platforms of different types assigned to this field. Initially, however, the project development plans will consider these four rigs and an interest area of 5×5 miles around the existing rigs. In this 5×5-mile area, there are 36 target well locations to be drilled, which are all within drilling distance to at least one of the 4 rigs in their present position. These 36 target well locations are numbered 1 to 36.

### 3.2 Costs and Parameters

In order to make the right decisions, we need to estimate all the relevant costs. Given four rigs and 36 wells, let:

• i= well number, i = 1 to 36:

• j = rig number, j = 1 to 4

• Lij = length (distance in miles) between the well at target location i and rig in position j

• cij = cost (in $) of drilling the well at ith target location by rig in position j • D = maximum distance between the well and the drilling rig = 5 miles • wj = maximum number of wells that rig j can drill. • Rj = daily rate of rig j, i.e. rig cost ($ per day)

• W = total number of wells

• G = total number of rigs

Drilling cost cij is estimated for feasible (i, j) pairs, i.e. pairs in which the distance Lij does not exceed the maximum feasible distance D. Using data from past drilling jobs around this field, there are several different cost components that contribute to the cost cij. All of these components can be estimated as linear functions of Lij. Based on past drilling data from the same fields, these various components are listed below for a general well of length Lij from a feasible (i, j) pair.

(i): Daily working rates paid to the drilling contractor: in crude oil production, the daily rate is the amount that a drilling contractor is paid by the oil company for operating each rig per day. In offshore drilling, the daily rates of drilling rigs vary by their capability, and the market availability at the time the contract for the rig is signed. The four rigs on location in this field belong to different service companies, and have different rates, Rj, as shown in Table 1.

The number of days required to drill a single well is directly proportional to the distance between the rig and the well. From the available well drilling history, the number of days required to drill a well from a distance of Lij miles is estimated to be given by the following equation:

$Number of ​ days = 77.616 L i j + 10.081$
(1)

(ii): Cementing: cement is used to hold the casing in place and to prevent fluid migration between subsurface formations. The price of cementing jobs varies based on additives used in mixing, well depth, well geometry, and rock characteristics in the field.

(iii): Drilling Fluid: drilling fluids, also referred to as drilling mud, are added to the wellbore to facilitate the drilling process by controlling the pressure, stabilizing exposed rock, providing buoyancy, and both cooling and lubricating while drilling. Their cost also depends on the type of fluids used and the rock characteristics.

(iv): Transportation, overhead, and other services: these expenses refer to the transportation of equipment, tools and personnel to and off the rig, spare parts and iron tubes used to run the completion; and the company’s drilling overhead such as office, personnel, insurance, fuel, water, logistics, third party vendor services, etc. The bigger the operation, the higher the overhead coast which is incurred and the more third party vendors that are involved.

Adding all the costs of: (ii) cementing, (iii) drilling fluid, and (iv) transportation, overhead, and other services, the drilling cost as a function of the distance Lij is approximated by the following equation:

$c i j = 3 , 054 , 121 L i j + 471 , 562$
(2)

The approximate total cost of drilling, as a function of the distance Lij in miles, is obtained by adding equations (1) and (2) as shown below:

$c i j = R j 77.616 L i j + 10.081 + 3 , 054 , 121 L i j + 471 , 562$
(3)

In the following sections, we will consider two cases for the optimal drilling of wells by rigs in the offshore oil field. In the first case, we assume that the rigs are fixed in their current locations, and we simply divide the wells into four groups (clusters) and assign each group to one of the four rigs. In the second case, we assume that the rig locations are not fixed, and we determine the optimum rig locations along with the optimum assignment of wells to rigs.

## 4. CASE 1: FIXED RIG LOCATIONS

Here we consider the problem of drilling all the wells that can be drilled by rigs in their present positions, without moving them, at minimum total cost.

We will refer to the pair: (well at ith target well loca-tion and the jth rig in its present position), as the (i, j)-pair, and refer to this pair as a feasible pair if the distance between them is within the maximum, Lij < D. The objective of Case 1 is to find the optimum allocation of target wells to rigs fixed in the field, in order to drill at minimum total cost.

In this case, all 4 rigs are fixed in their present posi-tions, and there are 36 target locations where wells have to be drilled. We want to allocate each of these target well locations to one of the rigs used to drill, without moving any rig from its present position. We define the set of feasible pairs as:

F = {(i, j) : i = 1 to 36, j = 1 to 4, and (i, j)} is a feas-ible pair, i.e. Lij < D}

The decision variables in the model for this case are: xij defined only for (i, j) ϵ F, where:

$x i j = 1 , i f w e l l i i s d r i l l e d b y r i g j , ( i , j ) ϵ F 0 , o t h e r w i s e$

Then the objective function for this problem in this case is:

(4)

The above objective function is subject to con-straints (5)-(7) below. Constraint (5) allocates each well to only one rig. Constraint (6) limits the number of wells that can be drilled by each rig. Finally, constraint (7) imposes the binary value restrictions on the decision variables.

$∑ j = 1 G x i j = 1 , i = 1 , … , W$
(5)

$∑ i = 1 W x i j ≤ w i , j = 1 , … , 4$
(6)

(7)

The model has |F| decision variables, and it is a linear minimum-cost bipartite network-flow problem. Therefore if the constraint (7) is replaced by xij ≥ 0 for all (i, j) ϵ F, every basic feasible solution of the resulting model will be a 0-1 solution. Hence the model can be optimally solved by any software program designed for minimum-cost network-flows problems, or by any linear programming (LP) software program. In this paper, all the optimum LP solutions were obtained by OpenSolver for Excel (Mason, 2012).

The input data for case 1 is shown in the tables in Appendix A, where W = 36 and G = 4. Figure 1 shows the locations of the 36 wells and the current locations of the 4 rigs. Table A1 shows the distances in miles between the four rigs (in their current locations) and the 36 target wells. Since all the distances are within the limit D = 5 miles, then each of the 36 wells can be drilled from any of the four rigs, and the set F includes all 4×36 xij variables defined by the range i = 1,…, 36 and j = 1, …, 4. Since the distances in Table A1 are given in miles, then equation (3) and the rates in Table 1 are used to compute the drilling costs shown in Table A2.

Constraint (6) expresses the requirement that the total number of wells allocated to each rig cannot exceed the rig’s handling capability (wj). Since offshore rigs can handle a big number of wells that may go up to 50 wells, two scenarios were considered. In the first scenario, the 36 target wells are equally divided among the four existing rigs. This means that each rig is allowed to work on 9 wells, i.e. wj = 9. In the second scenario, the capacity for each rig is assumed to be unlimited. This means that a single rig is capable of drilling all the 36 wells, i.e. wj = 36.

### 4.1 Case 1a: Equal Number of Wells Per Rig

#### 5.2.4 Case 2b: Heuristic Solution with Unlimited Number of Wells Per Rig

In scenario b, we remove the limits on the number of wells per rig and set the value (w = 36) in constraint (19). Solving the simplified model defined by (16)-(21), the optimal solution selects the sites of wells numbered (3, 20, 23, and 32) as rig locations, with corresponding distances (7.25, 4.5, 7.13, and 4.01 miles), respectively. Assigning rigs in the reverse order of distance to cost, rig numbers 1, 2, 3, and 4 are assigned to the sites of wells numbered 20, 3, 32, and 23, respectively. The number of wells assigned to each of the four rigs ranges from 8 to 11.

The partial cost of drilling is Z1 = $86,885,062, and the sum of the distances from all of the wells to their corresponding rigs is 22.89 miles. Using equation (23), the total drilling cost is given by Z = 458,294,439. The final solution of the heuristic method is shown in Table 7. This is only a 1% increase in total cost above the optimum solution obtained in section 5.1.2, whose total cost is Z =$452,962,073.

#### 5.2.5 Case 2: Evaluation of the 2-Stage Heuristic Solution

In the drilling location case study presented above, the optimum solution was relatively easy to find due to the manageable problem size (36 wells, and 4 rigs, and 5328 variables). For larger real-life problems, optimum solutions are too difficult, making it necessary to use the proposed two-stage heuristic procedure. In order to confidently apply this procedure, its performance must be thoroughly evaluated. Therefore, extensive numerical experiments were conducted to determine the quality of the heuristic solutions.

As shown in Table 8, 30 test problems were solved by two methods: the optimum (Opt.) solution, and the two-stage heuristic (Heur.) solution. These problem are divided into 3 sizes: 40, 50, and 60 wells, respectively corresponding to 6,560, 12,750, and 21,960 binary variables. Each size has 5 type-a problems with equal number of wells per rig, and 5 type-b problems with unlimited number of wells per rig. Different problem sizes also differ in the number of wells W, number of rigs G, and the oil field size. For each problem, uniform distributions were used to generate different rig daily rates and different well locations within the field.

Table 8 shows that the heuristic procedure has an impressive near-optimal performance. The heuristic solution’s costs were optimal in 9 problems, and differed from optimal on average by 0.38% and at most by 1.92%. The heuristic’s performance is better for type-a problems, with an average cost increase of 0.12% above optimum. However, its performance for type-b problems is still quite impressive, with an average cost increase of 0.64% above optimum. These results confirm that the heuristic procedure is a practical and effective solution technique for large-size real-life drilling location problems.

## 6. DISCUSSION & CONCLUSIONS

In this paper, two practical cases of allocating wells to drilling rigs in an offshore oil field have been considered. In Case 1 the rig locations are assumed to be fixed, while in Case 2 the rig locations are assumed to be decision variables to be optimally determined. For each case, two different scenarios were considered, one with and another without a limit on the number of wells assigned to each rig. The unique feature of the problem is the difference in drilling costs per mile among the available rigs. This makes the cost of drilling each well dependent on both the distance to the rig and the cost per mile of the specific rig used for drilling.

New integer programming models have been formulated to represent the two cases for this location and allocation problem. In addition to the optimum solution by integer programming, a heuristic method has been developed to handle larger model sizes, especially for Case 2, where the rig locations are decision variables. Based on the practical case study and numerical experiments with 30 test problems, the heuristic method has proved its ability to produce near-optimum solutions for Case 2. The heuristic solution’s cost exactly matched the optimum solution in many problems, and was only 0.4% above optimum on average.

There are several interesting directions for extending the work presented in this paper, especially by integrating the offshore drilling problem with other relevant problem in offshore oil fields. For example, the problem considered here can be integrated with the offshore transportation problem (Sahebi and Nickel, 2014), or with the oil field employee scheduling problem (Alfares, 2014).

## ACKNOWLEDGMENTS

This work was completed during a Fulbright Scholar visit by Prof. Katta Murty to King Fahd University of Petroleum & Minerals (KFUPM). The authors are grateful for the generous research support by both KFUPM and Fulbright.

## Figure

Distance between each well and the current location of the four rigs.

## Table

Daily cost rates of rigs in \$/day

Distances in miles between the 4 fixed rigs and the 36 wells

Cost in US Dollars of drilling from fixed-location rigs to each well

Case 1a: Optimum solution with 9 wells per rig

Case 1b: Optimum solution with no limit on number of wells per rig

Distances in miles between the 36 well locations

Case 2a: optimum solution with a limit of 9 wells per rig

Case 2b: optimum solution with no limit on number of wells per rig

Case 2a: Heuristic solution with a limit of 9 wells per rig

Case 2b: Heuristic solution with no limit on number of wells per rig

Computational evaluation of the heuristic solution

## REFERENCES

1. Alfares, H. K. (2014), Multiobjective scheduling of remote- area employees with minimum cost of transportation, Journal of Industrial Engineering, 2014, 11.
2. Carvalho, M. C. A. and Pinto, J. M. (2006), An MILP model and solution technique for the planning of infrastructure in offshore oilfields, Journal of Petroleum Science and Engineering, 51(1-2), 97-110.
3. Cunha, A. A. L. , Duncan, G. , Bontempo, A., and Pacheco, Springer.M. A. C.(2016), Optimum wells placement in oil fields using cellular genetic algorithms and space efficient chromosomes, Proceedings of SAI Intelligent Systems Conference, Springer, 15-38.
4. Devine, M. D. and Lesso, W. G. (1972), Models for the minimum cost development of offshore oil fields, Management Science, 18(8), B-378–B-387.
5. García, A. , Guerra, B. , Bentes, C., and Simonetti, L.(2012), An intelligent design model for offshore petroleum production elements layout. In: Pavón J., Duque-Méndez, N. D., Fuentes-Fernández, R. (eds), Advances in Artificial Intelligence–IBERAMIA 2012, IBERAMIA 2012. Lecture Notes in Computer Science, 7637, Springer, Berlin, Heidelberg, 702-711.
6. Hansen, P. , de Luna Pedrosa Filho, E. , and Ribeiro, C. C.(1992), Location and sizing of offshore platforms for oil exploration, European Journal of Operational Research, 58(2), 202-214.
7. Iyer, R. R. , Grossmann, I. E. , Vasantharajan, S., and Cullick, A. S.(1998), Optimal planning and scheduling of offshore oil field infrastructure investment and operations, Industrial & Engineering Chemistry Research, 37(4), 1380-1397.
8. Kabadi, S. , Murty, K. G. , and Spera, C.(1996), Clustering problems in optimization models, Computational Economics, 9(3), 229-239.
9. Karkevandi-Talkhooncheh, A. , Sharifi, M. , and Ahmadi, M.(2018), Application of hybrid adaptive neurofuzzy inference system in well placement optimization, Journal of Petroleum Science and Engineering, 166, 924-947.
10. Mason, A. J. (2012), OpenSolver – An open source addin to solve linear and integer progammes in excel. In Klatte, D., Lüthi, H. J., and Schmedders, K. (eds), Operations Research Proceedings 2011, Springer, Berlin, Heidelberg, 401-406.
11. Murty, K. G. (Ed.).(2014), Case Studies in Operations Research: Applications of Optimal Decision Making (Vol. 212), Springer.
12. Rodrigues, H. W. L. , Prata, B. D. A. , and Bonates, T. O.(2016), Integrated optimization model for location and sizing of offshore platforms and location of oil wells, Journal of Petroleum Science and Engineering, 145, 734-741.
13. Rosa, V. R. , Camponogara, E. , and Ferreira Filho, V. J. M.(2018), Design optimization of oilfield subsea infrastructures with manifold placement and pipeline layout, Computers & Chemical Engineering, 108, 163-178.
14. Rosa, V. R. and Ferreira Filho, V. J. M. (2012), Optimizing the location of platforms and manifolds. ASME 2012 31st International Conference on Ocean, Offshore and Arctic Engineering, American Society of Mechanical Engineers, 813-818.
15. Rosing, K. E. (1994), Considering offshore production platforms, European Journal of Operational Research, 72(1), 204-206.
16. Sahebi, H. and Nickel, S. (2014), Offshore oil network design with transportation alternatives, European Journal of Industrial Engineering, 8(6), 739-761.
17. Sahebi, H. , Nickel, S. , and Ashayeri, J.(2014), Strategic and tactical mathematical programming models within the crude oil supply chain context: A review, Computers & Chemical Engineering, 68, 56-77.
18. Van Den Heever, S. A. and Grossmann, I. E. (2000), An iterative aggregation/disaggregation approach for the solution of a mixed-integer nonlinear oilfield infrastructure planning model, Industrial & Engineering Chemistry Research, 39(6), 1955-1971.