Assessing the Impact of Onsite Maintenance on Offshore Wind Farms: The Case of the Repurposed Oil Platform

(1)

Assessing the Impact of Onsite Maintenance on

Offshore Wind Farms: The Case of the Repurposed Oil

Platform

Master Thesis

By

(2)

ii

Assessing the Impact of Onsite Maintenance on Offshore Wind Farms: The Case of the Repurposed Oil Platform

Groningen, December 2019 Word Count: 12,939

Student Information

Name: Megan Ambrose

Groningen Student Number: S3908216 Newcastle Student Number: B7074393

Groningen Email: m.c.ambrose@student.rug.nl Newcastle Email: m.ambrose2@newcastle.ac.uk MSc Technology & Operations Management (University of Groningen) MSc Operations & Supply Chain Management (Newcastle University) Supervisors

Dr. Onur Kilic Dr. Ying Yang

University of Groningen Newcastle University

o.a.kilic@rug.nl ying.yang2@newcastle.ac.uk

Universities

University of Groningen

Faculty of Economics and Business Nettelbosje 2, 9747 AE

Groningen, Netherlands Newcastle University Business School (NUBS) 5 Barrack Road

(3)

iii

Abstract

There is an energy transition occurring in the Dutch North Sea. The traditional offshore oil and gas industry is declining and facing a costly mass decommissioning. Meanwhile, the upcoming offshore wind industry is rapidly growing but must overcome challenges with accessibility before it is able to reach its full potential capacity. In order solve the challenges that both industries are facing, an opportunity for collaboration arises. This thesis aims to conduct an investigation into the use of repurposed offshore oil platforms as a maintenance support site for offshore wind farms. Thereby increasing the accessibility of offshore wind farms and reducing the need for the decommissioning of offshore oil and gas platforms. This investigation is conducted by a quantitative modelling approach executed via a simulation. The results provide insights into how an onsite maintenance can be optimized for offshore wind farms. It is concluded that cost savings of up to 33% and availability increases of up to 45% can be achieved compared to servicing from a harbor.

Keywords: energy transition, operations and maintenance, onsite maintenance, maintenance support organization, quantitative modelling

Acknowledgements

(4)

iv

Table of Figures

Figure 1: Main turbine components that experience minor failures. Adapted from Carroll et al (2016). ... 12

Figure 2: Visual representation of the service from harbor scenario... 13

Figure 3: Visual representation of the platform refurbishment scenario ... 14

Figure 4: Visual representation of the platform refurbishment + relocation scenario ... 14

Figure 5: Research design for the development of an offshore maintenance operations system ... 15

Figure 6: Wind farm operations process diagram ... 16

Figure 7: Sensitivity of the total cost to variation in failure rate ... 27

Figure 8: Sensitivity of the time based availability to variation in failure rate ... 28

Figure 9: Sensitivity of the total cost to cost of energy ... 28

Figure 10: Sensitivity of the total cost to the distance from shore ... 29

Figure 11: Sensitivity of the availability to the distance from shore ... 30

List of Tables

Table 1: Summary of Travel Times ... 20

Table 2: Weather Data ... 22

Table 3: List of simplifications and assumptions ... 24

Table 4: Generic Case Inputs ... 25

Table 5: Generic Case Outputs ... 26

Table 6: Impact of different shift lengths on the key outputs per scenario ... 30

List of Abbreviations

KPI Key Performance Indicator

OO&G Offshore Oil and Gas

OW Offshore Wind

OWF Offshore Wind Farm

(6)

2 of 46

1 Introduction

The world is currently in the midst of an energy transformation. While the renewable sector flourishes in terms of new investments and increased capacity, there is a noticeable decline in the traditional, non-renewable sector (Ruoff, 2016). This transition is largely driven by the rising cost of fossil fuels, the threat and impacts of climate change, and the resulting government initiatives (Shafiee and Sørensen, 2017). These government initiatives, such as the Paris Agreement, create incentives for the continued investment into renewable energy and a declining interest in non-renewables (McCollum et al., 2018). Similarly, these initiatives set goals reaching years into the future, which suggests that this energy transition will only continue to gain momentum in the future (Falkner, 2016). This indication of continued change coupled with the new experiences such a transition will bring, leads to unique challenges within the various energy industries. These challenges are exemplified in the offshore industries, where both the Offshore Oil and Gas (OO&G) industry and the Offshore Wind (OW) industry operate within the same environment.

As with most non-renewables, it is theorized that the OO&G industry has reached its peak and will only decline in the future as demand decreases (Ruoff, 2016). This is further evidenced by the fact that new installations and investments in the OO&G industry were at a record low in 2018 (Wind Europe, 2019). As this decline in the industry continues, more and more OO&G structures, such as offshore platforms, will become obsolete (Leporini et al., 2019). As many government policies dictate the removal of unused structures, the OO&G industry is facing a massive decommissioning of their offshore structures (Sommer et al., 2019). It has been estimated that with the current trends, the decommissioning process could cost the worldwide OO&G industry approximately €200 billion by 2040 (IHS Markit, 2016). This expensive decommissioning process will result in a major financial burden for an already declining industry. Because of this, the OO&G industry is challenged with reducing costs throughout the decommissioning process.

(7)

3 of 46

technical and logistical challenges which hinder the growth of this promising industry (McAuliffe et al., 2018). As a result, the operational availability of offshore wind farms (OWFs) falls between 60% and 70% (Shafiee, 2015a). This low availability represents a large loss in revenue and increase in operations and maintenance (O&M) costs for the OWFs. Until this availability can be increased, the OW industry will not be able to increase its capacity fast enough to meet the goals set by government initiatives and prove to be a profitable investment. Because of this, the OW industry is challenged with determining a way to increase availability in order to reach its full potential.

These circumstances in both industries within the same environment introduces the potential for working together to alleviate their individual challenges. For the OO&G industry, the mass decommissioning process has begun to be researched (Fowler et al., 2014; Leporini et al., 2019; Sommer et al., 2019). Research has unanimously shown that it is in the best interest of the OO&G industry to repurpose these offshore structures in order to avoid the expensive removal and destruction process (Sommer et al., 2019). While the possibilities for such a repurposing have been discussed, their practicality has not yet been investigated. For the OW industry, methods for increasing availability of been extensively researched (Hofmann, 2011; Shafiee and Sørensen, 2017; Seyr and Muskulus, 2019). Initial research has shown that onsite (i.e. offshore) maintenance can have a significant impact on the overall availability of an OWF (De Regt, 2012; Besnard et al., 2013; Maples et al., 2013). However, due to the high cost of installing an onsite maintenance accommodation, the previous research focused mostly on non-permanent methods such as mother vessels (Maples et al., 2013) or accommodations which could be shared by multiple wind farms such as service islands (De Regt, 2012). Both methods result in drawbacks, as mother vessels are not permanent and service islands require a very large investment and therefore must be shared (De Regt, 2012; Maples et al., 2013).

(8)

4 of 46

What effect would utilizing a repurposed offshore oil and gas platform for onsite maintenance have on the maintenance operations of an offshore wind farm?

The contribution of this research is, to the best of the author’s knowledge, the first study that presents an alternative, permanent method of onsite maintenance. By determining what the effect would be, the OO&G industry may have a possible solution to the mass decommissioning dilemma while the OW industry may have a viable method for implementing onsite maintenance. Understanding the impact of this solution will allow for more informed decision making within both the OO&G and OW industries and will add to the limited literature about onsite maintenance for offshore wind farms.

This thesis will explore the above-mentioned question by developing a quantitative model. The model simulates the OWF maintenance operations under multiple configurations of the proposed onsite maintenance. The model combines data on weather, failures and maintenance policies to closely mimic the real-world operations of an OWF. This model can be used to assess the impacts of repurposing an OO&G platform to serve as onsite maintenance on the availability of an OWF.

The remainder of this thesis will be explained as follows. Following this introduction, Chapter 2 discusses the relevant existing literature on the topic and establishes the foundation for the research. Chapter 3 explains the methodological approach used to conduct this research. Chapter 4 details the simulation model. Chapter 5 presents the numerical results of the model. These results and their implications are discussed in Chapter 6. Lastly, Chapter 7 provides concluding remarks.

2 Theoretical Background

This section discusses the existing literature on the topic and establishes a foundation for the conducted research. This is done by reviewing the O&M phase, maintenance policies, and maintenance support organization for OWFs. Following this, the benefits and challenges of onsite maintenance are discussed. Lastly, this review is summarized in a conceptual framework that sets out the approach being used in this thesis.

2.1 Operations & Maintenance

(9)

5 of 46

encompasses the total useful life of a turbine (Snyder and Kaiser, 2009). During this phase, the turbine is actively generating electricity and will require maintenance to ensure it is operating efficiently (Shafiee, 2015b). The costs associated with these activities are referred to as O&M costs. As the name suggests, O&M costs are a combination of the cost of operations and the cost of maintenance. Operations costs refer to the cost of day to day operations such as rent, leasing fees, and insurance (Shafiee et al., 2016). As these costs are not directly impacted by the chosen maintenance policy, they will not be considered in this thesis. Maintenance costs refer to the cost of conducting maintenance and can be broken down into indirect and direct costs. Indirect costs refer to the fixed and variable costs associated with being prepared to conduct maintenance such as port fees, onshore maintenance support and vessel fees (Shafiee et al., 2016). Direct costs refer to the costs of conducting repairs such as the cost of spare parts, transportation and maintenance technician labor (Shafiee et al., 2016). These direct costs will be the primary focus of this research.

The O&M phase accounts for 20-30% of the cost of energy, making it the second most expensive phase, next to the acquisition phase (Poulsen and Lema, 2017). During this phase, OW experiences much higher costs than onshore wind where O&M only account for 5-10% of the cost of energy (Shafiee and Sørensen, 2017). This discrepancy is largely due to the inaccessibility of the OWFs and will only continue to grow as new wind farms are located further and further offshore (Van Bussel and Schöntag, 1997). As the O&M costs continue to increase, there is an impact on the potential of OW. Because of this, the O&M phase has a large impact on the continued viability of OWFs.

(10)

6 of 46

et al., 2016; Shafiee and Sørensen, 2017; Stålhane et al., 2019). The primary focus was to optimize the current maintenance strategies in order to improve the availability of the OWF (Seyr and Muskulus, 2019). While the current literature has created an extensive of the existing conditions, there is still room for improvement. It may be beneficial to take a more exploratory route in order to identify and evaluate maintenance strategies which may have been previously overlooked. As such, this thesis takes an exploratory approach to improving the O&M phase.

2.2 Maintenance Policies

In order to improve the O&M phase of OWFs, it is vital to understand how wind farm operators conduct their O&M activities. According to Walford (2006), the main components of O&M activities are preventive maintenance, corrective maintenance, and operations. Preventive and corrective maintenance refer to repairs completed on the turbine components while operations refers to the day-to-day activities needed to be able to complete maintenance. These components are of interest in this thesis as they have the greatest potential to be impacted by a repurposed OO&G platform serving as onsite maintenance. Because of this, each of these three components will be discussed in detail below.

2.2.1 Preventive Maintenance

(11)

7 of 46

offshore wind industry until better condition monitoring technology are developed (Ding and Tian, 2012). Because of these challenges in applicability, preventative maintenance is commonly excluded from OWF research (Scheu et al., 2012; Rinaldi et al., 2017; Nguyen and Chou, 2018). In line with this precedence, it will not be considered in this thesis.

2.2.2 Corrective Maintenance

Corrective maintenance is conducted to repair a component following the occurrence of a failure event (Seyr and Muskulus, 2019). Due to the complexity of wind turbines and the harsh offshore environment, a high amount of corrective maintenance is needed for OWFs (Ding and Tian, 2012). Because of this, the majority of models consider corrective maintenance in some capacity (Seyr and Muskulus, 2019). This thesis follows this precedence and considers only corrective maintenance. The downside of corrective maintenance is that it typically results in more downtime and higher costs when compared to preventative maintenance (Swanson, 2001). Due to the nature of corrective maintenance, when a failure occurs the turbine is non-operable until the repair can be completed (Besnard et al., 2013). This means that it is in the best interest of the organization to complete corrective maintenance as efficiently as possible (Besnard et al., 2013). This aspect of corrective maintenance makes it particularly interesting to this thesis as it has high potential to be impacted via an onsite platform.

2.2.3 Operations

(12)

8 of 46

the autumn and winter where poor weather conditions are characteristic of the season (Van Bussel and Bierbooms, 2003). In order to reduce the scheduling impact from these weather delays, it is important to consider the way resources such as vessels, spare parts and maintenance technicians are organized (Hameed and Vatn, 2012). The organization of these resources is referred to as the maintenance support organization (Besnard et al., 2013). The maintenance support organization controls the maintenance operations and as such is very important to this research. Because of its importance, the maintenance support organization is further discussed in the following section.

2.3 Maintenance Support Organization

(13)

9 of 46

delay in terms of revenue loss and the total cost of maintenance. The aim of this thesis is to address this trade-off through the concept of onsite maintenance.

2.4 Onsite Maintenance

The location of the maintenance accommodation is a vital aspect of the maintenance support organization (Besnard et al., 2013). Onsite maintenance refers to the accommodation of maintenance support including technicians and equipment at the location where the maintenance is performed, in this case offshore (Hofmann, 2011). In this thesis, onsite maintenance is considered a location offshore which has the capacity to house an inventory of spare parts, maintenance crews, and all necessary equipment for completing repairs. For OW, the advantage of being onsite is that repair times could be shortened, meaning smaller weather windows would be needed (Tavner, 2012). This, in turn, should lead to fewer delays in maintenance activities and less turbine downtime. Shafiee (2015b) stressed the importance of having onsite maintenance sites for OWFs in order to decrease costs and increase the efficiency, availability and reliability of the turbines. However, very little research has been conducted on the possibility of onsite maintenance. Initial research has begun to be conducted into what such an onsite maintenance would look like. Besnard et al. (2013) determined the optimal number and size of technicians and vessels that would be needed if onsite maintenance on a mother vessel was implemented. The mother vessel concept was further investigated by Maples et al. (2013) who determined that it has the potential to improve the performance of OWFs. A mother vessel refers to a larger ship that may remain at sea for longer periods of time (i.e. two weeks) before returning to the harbor (Maples et al., 2013). While their research has shown this type of onsite maintenance can be beneficial, a mother vessel must return to the harbor eventually and this can result in delays. De Regt (2012) developed a methods to determine the optimal location of a service island which was used to service multiple OWFs. A service island referred to any offshore structure which could be used to house the necessary resources (De Regt, 2012). This service island would be very expensive so it must be shared by multiple OWFs in order to be cost effective. This type of shared resources approach can lead to improvements in the accessibility of multiple wind farms, however, the improvements are generally smaller than those without shared resources.

(14)

10 of 46

costs while retaining the benefits to reliability and availability. In this thesis, the concept of onsite maintenance is applied to a repurposed OO&G platform to test this theory.

2.5 Refined Research Questions

The main research question of this research is as follows:

What effect would utilizing a repurposed offshore oil and gas platform for onsite maintenance have on the maintenance operations of an offshore wind farm?

This question will be answered by investigating the below sub-questions:

RQ1: What is the impact of utilizing a repurposed offshore oil and gas platform for onsite maintenance on the key performance indicators of an offshore wind farm? RQ2: How does the location of the platform affect this relationship?

RQ3: How can the components of the maintenance support organization be managed to balance the trade-off between maintenance costs and service quality? RQ1 looks at the how the implementation of onsite maintenance on the platform will impact the performance of the OWF. This question builds on the previous research on onsite maintenance because it incorporates a permanent structure which will only be used by one OWF. Answering this questions will determine whether this can be seen as a viable solution that is equivalent or better than the previous methods of onsite maintenance. RQ2 delves deeper into the relationship and looks at different configurations of the harbor, OWF and offshore platform. By answering this question, trade-offs can be identified that will aid OWF operators in making decisions regarding the implementation of such an offshore platform. Lastly, RQ3 looks at the trade-offs within the maintenance support organization as a whole. By answering this question, it can be understood how the different components of the maintenance support organization are affected by onsite maintenance and what the trade-offs are between the cost and the service quality. Answering these three questions will create a better understanding of what happens when an obsolete OO&G platform is repurposed to serve as onsite maintenance for and OWF.

3 Methodology

(15)

11 of 46

3.1 Research Method

A quantitative modelling approach via a simulation study was used to assess the research questions. The nature of the research questions calls for a quantitative approach as opposed to a qualitative one because of the focus on identifying causal relationships and comparisons (Bertrand and Fransoo, 2016). In operations research, quantitative models are built to signify the behavior of real-life operational processes (Meredith et al., 1989). The value of quantitative modelling lies in its ability to identify the causal relationships between control and performance variables (Bertrand and Fransoo, 2016). By modelling the OW environment, the causal relationships between the presence of an onsite maintenance accommodation on an obsolete platform and the performance of an OWF could be identified.

Quantitative modelling is typically conducted via mathematical analysis. However, when a system is complex, a simulation approach is typically used (Bertrand and Fransoo, 2002). Simulation research aims to recreate a simplified version of reality in a series of models which can then be used to provide insights into how to operate and make decisions within reality (Ben-Daya et al., 2016). By utilizing a simulation approach, the OW environment could be recreated and manipulated to see how it would react to the implementation of onsite maintenance on an offshore platform. In addition, the use of a simulation gives the unique ability to evaluate multiple scenarios at once, allowing for a greater breadth of results (Robinson, 2014). This use of simulation is also less costly and allows for greater control than experimental research (Budnick et al., 1977). However, simulation is time consuming and can lead to problems with the external validity of the results as a simulation is inherently a simplified version of reality (Robinson, 2014). While a simulation is a good estimate of what could happen, is should be kept in mind that the complex reality may differ. To combat this, the assumptions made in this research followed the assumptions made in previous research and the inputs were tested via a sensitivity analysis.

3.2 Research Setting

The developed model creates a virtual representation of the OW environment. This virtual environment incorporates the following parties:

 1 offshore wind farm  1 onshore harbor  1 offshore platform

(16)

12 of 46

maintenance equipment, an inventory of small spare parts and maintenance technicians. It will also serve as a base of operations for all repair activities at the OWF. Only one OWF and harbor is considered because it is assumed that because of the cost savings resulting from utilizing an obsolete OO&G platform, only one OWF is necessary to cover the costs of implementation. Only one platform is considered because only one platform is needed to meet the needs of the OWF. However, this study could be expanded in the future to incorporate more platforms and OWFs.

The OWF considered in this thesis is theoretical, based on an average OWF. This theoretical OWF is located 50km from shore, as this is the average distance from shore for offshore wind farms (Ho et al., 2016). The OWF is made up of 100 turbines which generate 3.6 MWh on average (Wind Europe, 2019). The cost of energy is €170/MWh (Wind Europe, 2019).

Within this environment, only minor repairs are considered as they are able to be completed via the offshore platform. Minor repairs are defined as repairs which can be carried out by a maintenance technician via a workboat (Martin et al., 2016). Minor repairs can affect all parts of a turbine. This thesis focuses on the 19 components which experience minor repairs the most often, as identified by Carroll et al. (2016). An overview of the main components can be seen in Figure 1. In addition to those seen in the figure, there are also smaller components including service items, sensors, safety items, grease/oil/cooling liquids, various electrical components and a miscellaneous components category. Additionally, there components related to the processing of power including the power supply/convertor, transformer and circuit breaker/relay. These components are included in this thesis but cannot be seen in Figure 1.

(17)

13 of 46

It is assumed that all major repairs will be coordinated from the harbor. Following the example of several other papers, a corrective maintenance strategy is used, meaning that repairs are carried out following a failure (Sagarna et al., 2016).

3.3 Considered Scenarios

Due to the uncertainty in the offshore industry and the identified trade-off between maintenance cost and accessibility creating the possibility for multiple configurations of the onsite maintenance, a scenario analysis approach is used. In order to account for the different possible configurations of the harbor, OWF and platform, three possible scenarios have been developed. These three scenarios were selected as they clearly differentiate between the cost of implementation and the potential for improvements in service quality. These scenarios are described below.

Scenario 1: Service from Harbor – current practice

This most closely reflects the current practice of servicing OWFs where repairs are conducted from the harbor. In this scenario, all maintenance support including maintenance technicians, equipment, and spare parts comes directly from the harbor to the wind farm on an as needed basis. Figure 2 shows a visual representation of this scenario. Due to the distance between the harbor and the OWFs, this method is heavily dependent on weather conditions. As the distance increases, the necessary window of time for repairs also increases. This in turn leads to an increased time to repair and a decrease in the overall wind turbine accessibility.

Figure 2: Visual representation of the service from harbor scenario Scenario 2: Platform Refurbishment

(18)

14 of 46

However, this will also affect the transportation time from the platform to the wind farm. The platform would be supplied from the harbor according to a schedule and repairs will be conducted from the platform as needed.

Scenario 3: Platform Refurbishment + Relocation

Scenario 3 introduces the possibility of relocating the refurbished platform. A visual representation of this scenario can be seen in Figure 4. Under this scenario, there is a single platform that is relocated to within 5km of the OWF site and refurbished to serve as maintenance support for a single wind farm. This proximity to the wind farm decreases the necessary transport time from the platform to the OWF thereby decreasing the time needed for repairs. However, this scenario also results in a longer transport time from the harbor to the platform which may lead to delays in receiving needed resources. This scenario comes at an increase in refurbishment costs due to the additional expenditure that comes

with relocating an offshore platform.

Figure 4: Visual representation of the platform refurbishment + relocation scenario

3.4 Research Design

The research design that was used in order to develop the quantitative model takes into account the research setting and scenarios in order to ensure that all research questions were answered (Bertrand and Fransoo, 2016). The aspects in this research design represent the inputs, system dynamics and measured outputs of the simulation (Robinson, 2014). An overview of this design can be seen in Figure 5.

(19)

15 of 46

Figure 5: Research design for the development of an offshore maintenance operations system

This research includes the components of a maintenance support organization (inputs), the constraints and relationships that they face (system dynamics), and the performance measures (outputs) that can be used analyze the effects of the onsite maintenance. The following sections describe these components in detail.

3.4.1 Inputs

(20)

16 of 46

3.4.2 System Dynamics

The system dynamics of the model represent how the inputs are processed. This is the portion of the model that mimics the O&M of an OWF. The process that the simulation follows in order to achieve this can be seen in the below Figure 6. The modelling of the system dynamic components; failure occurrence, resource availability, weather availability and maintenance operations; provide the information needed to complete this process.

Figure 6: Wind farm operations process diagram

(21)

17 of 46

3.4.3 Outputs

The outputs of the system, or key performance indicators (KPIs), are broken into five measurable indicators of how the system is performing. In this thesis, the time-based availability (TBA), mean time to repair (MTTR), cost of delay (COD), total maintenance cost (TMC) and total cost (TC) have been identified as KPIs to be assessed (Gonzalez et al., 2017). TBA is defined as the percentage of time that a wind turbine is available to generate electricity (International Electrotechnical Commission, 2010). Equation 1 provides the mathematical calculation for TBA. TBA is an indicator of the success of maintenance actions as a low TBA measure is typically an indication of poor maintenance actions (Dinwoodie et al., 2015).

𝑇𝐵𝐴 = 𝑇𝑜𝑡𝑎𝑙 𝐴𝑣𝑎𝑖𝑙𝑎𝑏𝑙𝑒 𝑇𝑖𝑚𝑒

𝑇𝑜𝑡𝑎𝑙 𝑇𝑖𝑚𝑒 (1)

MTTR is the average time it takes a wind turbine to recover from any failure (Faulstich et al., 2011). Equation 2 provides the mathematical formulation for MTTR. Excessive delays, due to resource or weather, will lead to an increase in the MTTR. MTTR is a direct reflection of the speed of the maintenance support organization and should be minimized (Tavner, 2012).

𝑀𝑇𝑇𝑅 = 𝑇𝑜𝑡𝑎𝑙 𝐷𝑜𝑤𝑛 𝑇𝑖𝑚𝑒

𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝐹𝑎𝑖𝑙𝑢𝑟𝑒𝑠 (2)

COD can be defined as the loss in revenue that results from turbine downtime (Dinwoodie et al., 2015). Equation 3 provides the mathematical expression for COD. Similar to MTTR, excessive delays will lead to an increase in COD. Therefore, COD is a reflection of the service quality and should be minimized when possible.

𝐶𝑂𝐷 = (𝑇𝑢𝑏𝑖𝑛𝑒 𝐶𝑎𝑝𝑎𝑐𝑖𝑡𝑦 ∗ 𝑇𝑜𝑡𝑎𝑙 𝐷𝑜𝑤𝑛𝑡𝑖𝑚𝑒) ∗ 𝐶𝑜𝑠𝑡 𝑜𝑓 𝐸𝑛𝑒𝑟𝑔𝑦

𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑌𝑒𝑎𝑟𝑠 (3)

TMC can be defined as the annual cost of operating the wind farm including all planned and unplanned maintenance costs (Scheu et al., 2012). This is expressed in Equation 4. TMC is a reflection of the cost of the maintenance organization and should be minimized when possible.

𝑇𝑀𝐶 =𝐼𝑛𝑣𝑒𝑛𝑡𝑜𝑟𝑦 𝐶𝑜𝑠𝑡𝑠 + 𝐶𝑟𝑒𝑤 𝐶𝑜𝑠𝑡𝑠 + 𝑉𝑒𝑠𝑠𝑒𝑙 𝐶𝑜𝑠𝑡𝑠

𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑌𝑒𝑎𝑟𝑠 (4)

(22)

18 of 46

economic viability of the scenarios. All together, these five outputs can be used to compare the difference in impact between the three different scenarios.

4 Quantitative Model

This section details the quantitative model that was developed in order to assess the research questions. This focuses the system dynamics mentioned in the previous section to illustrate the relationships between the inputs and the outputs. This is done by discussing how each of the components of the system dynamics; failure occurrence, resource availability, weather availability and maintenance operations; were implemented into the quantitative model. In order to create a realistic model, data from the Dutch North Sea offshore environment was used. The Dutch North Sea exemplifies all the characteristics previously discussed and is therefore an adequate source for data.

4.1 Failure Occurrence Modelling

A failure acts as the initiator for the maintenance activities. In order to have an accurate representation of the OW environment, the occurrence of these failures must be implemented into the quantitative model. In this research, a failure rate of 6.2 failures per turbine per year was used. This failure rate was determined based on historical data from 350 offshore turbines located in Europe (Carroll et al., 2016). Using historical data to estimate the failure behavior is an accepted practice that has been used in previous research (Eecen et al., 2007; Stock-Williams and Swamy, 2019). This annual failure rate translates to an occurrence probability. In reality, as the turbine ages the failure rate will increase. However, this thesis chooses to omit degradation modelling. The assumption that the failure rate is constant over time is a made to simplify the modelling process. This assumption is common within OWF modelling and can be seen in previous research (Rademakers et al., 2003; Eecen et al., 2007; Scheu et al., 2012; Dinwoodie et al., 2015; Joschko et al., 2015). Based on this constant annual failure rate, it can be determined a given turbine has a 1.69% chance of failing on any given day.

(23)

19 of 46

When a failure has been recorded, the failed component is then assigned. As previously mentioned, 19 components were considered in this thesis. For each failed turbine, a random number between 1 and 19 is assigned to the failure. This number represents the specific component which has failed. As all 19 components are assumed to have the same probability of failure, this random number generation follows a uniform distribution. This is a valid assumption because the failure rate used is an aggregate of the failure rates for all components. The failed component determines the repair time and spare part needed to complete the repair. The repair time varies per component, ranging from two to ten hours (Carroll et al., 2016). The repair times for each component, their description, and the failure rates used for aggregation can be seen in the table in Appendix B.

4.2 Resource Availability Modelling

In order to conduct repairs, crews, vessels, and spare parts are required. Crews comprised of adequately trained maintenance technicians are crucial to the completion of a repair. As only minor repairs are considered, workboats are needed for every repair to transport crews to the turbine. Spare parts are required for all repairs to replace the failed component. In the case of an offshore platform, both spare parts and maintenance technicians must be transported to the platform via a transfer vessel. The resource availability simulates the availability of these crews, vessels and spare parts. In combination, these components can be used to determine the occurrence and duration of resource delays which have an impact on the performance of the OWF.

4.2.1 Crew Availability

(24)

20 of 46

technician offshore is paid €175/hour (Maples et al., 2013). This increase in hourly wage is due to the harsher conditions of living offshore for an extended period. Each technician is stationed on the offshore platform for a period of 14 days. This is typical for the offshore environment (Mette et al., 2018). Every 14 days a transfer vessel delivers maintenance technicians to the platform and returns technicians to the harbor. If this vessel is delayed, then it is assumed that no technicians are working, and no repairs can be completed.

The availability of the crew was simulated based on the quantity of failures, shift length and number of crews. Each day, the maximum number of repair hours that can be completed is equal to the number of crews times the shift schedule. For example, if 3 crews are working on a 12/7 schedule, then a maximum of 36 hours can be worked cumulatively on a given day. If a repair cannot be completed within this time, it is placed on a waitlist until the next day there are enough crews available.

4.2.2 Vessel Availability

Two types of vessels, workboats and transfer vessels, are considered in this thesis. Workboats are used to transport crew and equipment to the repair site (Besnard et al., 2013). The crew and workboats are mutually inclusive, meaning that one cannot operate without the other. This relationship indicates that there should be 3 workboats to accompany the 3 crews. Transfer vessels are used for the transportation of supplies from the harbor to the offshore platform (Seyr and Muskulus, 2019). If there is no offshore platform implemented, the transfer vessel is not considered. Transfer vessels are needed much less frequently than workboats, so only 1 is considered in this thesis. This is logical because the transfer vessel travels to the offshore platform on a schedule and it is assumed that the vessel has the capacity for all necessary resources, therefore, only one is needed at a time. Both vessels incur a cost of €1200 per day they are utilized, regardless of the amount of time they are utilized for (Besnard et al., 2013). Both vessel types travel at a constant speed of 37 km/hr (Sperstad et al., 2017). This speed, in combination with distances, can be used to calculate the travel times between different points. These travel times can be seen in the below Table 1.

Route Travel Time (Hours)

Harbor  OWF 3

Harbor  Refurbished Platform 1

Refurbished Platform  OWF 2

Harbor  Relocated Platform 3

Relocated Platform  OWF 0.5

(25)

21 of 46

The availability of a vessel is constrained by the weather. A wave height threshold of 1.5 meters is in place for both vessels (Sahnoun et al., 2015). A wind speed threshold of 10 m/s also constrains both vessels (Dinwoodie et al., 2015). This means that when wave heights exceed 1.5 meters or the wind speeds exceed 10 m/s, the vessel cannot safely travel. For the transfer vessel, unsuitable weather conditions will result in a delay in delivery of spare parts and personnel until the weather is suitable for an adequate period. For the workboat, unsuitable weather conditions will result in a delay in the repair activities until the weather is suitable for an adequate period.

The availability of the vessels was simulated using the same methods as the crew availability. Each day, the maximum number of repair hours that can be completed is calculated by the number of workboats times the shift schedule. The transfer vessel is always available to complete a delivery, subject to the weather conditions.

4.2.3 Spare Part Availability

The spare part availability is controlled by the inventory system. Each of the 19 components has a designated spare part held in inventory on the offshore platform. In the event that an offshore platform is not present, this inventory is held at the harbor. In this thesis, the inventory is controlled following a T,S policy. A T,S policy is conducted by placing orders at a given time interval to bring the inventory to a designated order up to level (Fung et al., 2001). In this thesis, the time interval was set at 14 days. This length was chosen to align with the schedule of the transfer vessel delivering maintenance technicians to the offshore platform. The order up to level was set at 5 parts for each component. This is adequate to meet the demand of a typical OWF (Lindqvist and Lundin, 2010).

The inventory system incurs holding costs, procurement costs, and ordering costs. The holding cost is €100/unit/day and is calculated for each component according to how much inventory is being held. The procurement cost is calculated according to the number of components ordered and the cost of each component. The procurement cost varies per component, a detailed list can be seen in the table in Appendix B. The ordering cost for the harbor is €100/order. The ordering cost for the platform is €150/order. This difference can be attributed to the increased cost of delivering the spare parts to the offshore platform. The total inventory cost is a sum of these components which were derived from (Lindqvist and Lundin, 2010).

(26)

22 of 46

requires 1 spare part of the respective component inventory. Every 14 days the inventory is assessed, and orders are placed to bring each components inventory to 5 units. It is assumed that there is no lead time for the spare part orders. The inventory for each day is calculated as the previous days inventory, plus any orders, minus the demand for that day.

4.3 Weather Availability Modelling

The accessibility of offshore wind farms is largely driven by the offshore weather conditions. Because of this, modelling of the weather data is necessary in the development of a simulation model of the OW environment. In general, this is done by using simple ocean parameters including wave height and wind speed to determine “weather windows” (Gintautas and Sørensen, 2017). Weather windows refer to the period of time where both wind speed and wave height are below the limits set forth by the vessels and therefore considered safe for operations (Gintautas and Sørensen, 2017). These windows vary day-to-day but have a clear seasonality. They can be modelled based on the distributions from historical data (Seyr and Muskulus, 2019).

In this thesis, the weather modelling is based on data from FINO 3, an offshore platform managed by the BMWi (Bundesministerium fuer Wirtschaft und Energie) and made available by the PTJ (Projekttraeger Juelich). FINO 3 is an offshore platform located 80km from shore in the North Sea. The analyzed data was from the period 1.1.2010 to 31.12.2018 and consisted of wind speeds (m/s) measured every 10 minutes from a height of 80km and wave heights (m) measured every 20 minutes from sea level. As wind speed and wave height are correlated, the weather window was calculated considering both. The weather was simulated by determining the weather window for each day. This weather window was modelled using the probability of good weather per season, the average weather window length per season and the standard deviation of weather window length per season of the FINO 3 data. A summary of this data can be seen in below in Table 2. Here, good weather is defined as both wind speed and wave height being below the vessel thresholds defined in the previous section. The weather window refers to the number of hours per day that met these requirements.

Season Probability of Good Weather Average Weather Window Length Weather Window Standard Deviation

Spring 63% 15.01 hours 9.45 hours

Summer 71% 16.95 hours 8.95 hours

Autumn 43% 10.31 hours 9.95 hours

Winter 36% 8.55 hours 9.42 hours

(27)

23 of 46

For each day of the simulation, a random number between 0 and 1 was generated, following a uniform distribution. This number was then compared to the probability of good weather for the respective season. If the random number was greater than the probability then it was assumed bad weather occurred and no repairs can be completed on this day. If the random number was less than the probability then it was assumed there was good weather. If good weather occurs, the weather window is then determined. The weather window is the average weather window length for the respective season, plus a random number generated between the positive and negative standard deviation. For example, for good weather in summer a random number is randomly generated between -8.95 and 8.95. This randomly generated number is than added to 16.95 to result in the weather window for the given day. The result represents the number of workable hours in a day. As only whole hours are considered for simplicity, the result is rounded to the nearest whole hour. This methodology for modelling the weather conditions was adapted from previous literature and ensures that the variation in weather is adequately incorporated into the model (Santos et al., 2015).

4.4 Maintenance Operations Modelling

Maintenance operations refers to the modelling of the repairs. All inputs and the above-mentioned subsystems come together to create the O&M system that dictates whether a repair is completed. This aspect models the process that was discussed in the previous Figure 6. For each failure that is recorded, the failed component and respective repair time is identified. Next, it must be determined if the spare part is in stock by looking at the daily component inventory. Next, it must be determined if there is enough time in the day to complete the repair. The total available time per day is a combination of the weather window and crew/vessel availability. For example, if there are 3 crews working a 12/7 shift and a 16-hour weather window, then each crew can work a full 12 hours resulting in a total of 36 total repair hours available on the given day. Based on this total availability, the repairs can be scheduled. In the result of a day that has more repair hours than hours available, the repairs will be completed in a manner which allows the most repairs to be completed in the available time. Any uncompleted repairs will simply be carried over to the next day and the process repeats. The combination of these steps results in the culmination of the maintenance operations process.

4.5 Simplifications and Assumptions

(28)

24 of 46

and assumptions are not expected to have a detrimental impact on the results. However, several will be tested to ensure their validity.

Simplifications Justifications

1. Only minor repairs will be considered.

Major repairs often require the use of specialty vessels and the shipment of large parts (Shafiee and Sørensen, 2017). As such, they are not likely to be improved through onsite maintenance support and will therefore not be considered.

2. Installation is not considered in this case.

The installation phase makes up a small portion of the wind turbine lifecycle and requires specialty crews and vessels (Poulsen and Lema, 2017). Because of this, it will not be considered.

3. The cost of energy is fixed for the duration of the simulation

The cost of energy affects the profitability of the wind farm (Laura and Vicente, 2014). This creates a more accurate comparison between all cases.

4. The repurposed oil platform is not at the end of its operational life.

For simplicity, it will be assumed that the platform chosen for repurposing is operational.

5. All minor repairs can be completed on site or at the maintenance platform.

The majority of minor repairs can be completed on site by technicians with a workboat (Shafiee, 2015b). 6. When a repair is

conducted, the component returns to a “good as new” state.

For simplification in the simulation of failure data, all components will be considered good as new after a repair, following the suggestions of Sahnoun et al. (2015) and Rinaldi et al. (2017).

7. Only corrective

maintenance is considered.

As discussed in the theoretical background, implementing only corrective maintenance is an acceptable practice in offshore wind farm modelling. 8. There is no lead time for

spare parts orders.

While a lead time could delay orders, it would occur in all scenarios and is therefore irrespective of the

application of onsite maintenance. 9. Vessels are assumed to

travel at a constant rate

This allows for simplification in the calculations of travel times and is a commonly accepted practice (Besnard et al., 2013).

10. Failure rates of

components are assumed to be constant over time.

This simplifies the failure data modelling and can be found in some other models (Seyr and Muskulus, 2019). 11. Component failure causes

the entire turbine to fail.

As production capacity is not considered in the scope of this thesis, it is acceptable to assume that the turbine is not producing when a failure has occurred.

12. Wind farm operators can accurately forecast the weather on a given day.

This allows for simplification in determining the daily weather window.

13. Repair times are assumed

to be fixed parameters. Each component requires a certain number of hours to repair based on the average of historical repair time data. 14. The distance between

turbines in a wind farm is assumed to be negligible.

This allow for simplification in the calculations of travel times and is a commonly accepted practice (Besnard et al., 2013).

(29)

25 of 46

5 Results

In this section, the results of the model are presented. First, the results of the scenario analysis are presented. Following this is a sensitivity analysis to evaluate the trade-offs, investigate the uncertainties and gain a deeper understanding of the results.

5.1 Scenario Analysis

The scenario analysis uses the developed quantitative model to investigate the three identified scenarios: service from harbor, refurbished platform and relocated platform. The results of this analysis will contribute to a better understanding of the impacts of an offshore platform for onsite maintenance and under what trade-offs are present in the relationship (Robinson, 2014).

5.1.1 Inputs

A generic case based on a typical OWF was used to examine the three scenarios. The characteristics of this wind farm can be seen in the below Table 4. These characteristics, along with the inputs outlined in the previous chapters make up the inputs of the model. All scenarios; harbor, refurbished platform and relocated platform; will be examined with these characteristics. The simulation took into account 10 years, with the first year being excluded from the results. This exclusions creates a warm-up period and ensures the accuracy of the calculated results (Bertrand and Fransoo, 2016).

Input Value Unit

Duration 10 Years

Number of turbines 100 Turbines

Failure Rate 6.2 Failures/Year/Turbine

Wind Farm Distance from Shore 50 km

Refurbished Platform Distance from Shore (Scenario 2) 30 km

Relocated Platform Distance from Shore (Scenario 3) 45 km

Energy Price 170 €/MWh

Turbine Capacity 3.6 MWh

Cost of Energy 100 €/MWh

Table 4: Generic Case Inputs

5.1.2 Outputs

(30)

26 of 46

Scenario MTTR _(days) TBA _{(millions/year)}TMC _{(millions/year)}COD _{(millions / year)}Total Cost Cost Benefit

Harbor 2.85 63% €50.4 €95.7 €146.1 N/A

Refurbished Platform 0.76 88% €70.1 €33.3 €103.5 29.16%

Relocated Platform 0.54 92% €73.6 €24.2 €97.8 33.06%

Table 5: Generic Case Outputs

MTTR and TBA reflect on the performance of the maintenance organization while the remaining outputs reflect on the costs of the maintenance organization. As expected, the performance of the maintenance organization increases as the refurbished platform is located closer to the OWF. Based on this results, there is a clear improvement in the performance of an OWF when any offshore maintenance accommodation is introduced. The MTTR was reduced by 73.33% with the refurbished platform when compared to the service from harbor scenario. Similarly, the TBA was reduced by 39.68%. This can be attributed to the increased accessibility of the OWF that comes with operating offshore. However, the improvement is less drastic when comparing the two offshore accommodations. When comparing the refurbished platform with the relocated platform, only a 29% decrease in MTTR and a 4.35% difference in TBA can be seen.

The costs associated with the various scenarios reflect the trade-off between cost of maintenance and cost of delay as discussed in previous chapters. As the platform is located closer to the OWF, the cost of delay drastically reduces while the total cost of maintenance increases. Similar to the performance measures, the cost of delay sees a sharp decrease with the introduction of an offshore platform. There is a 65% change when comparing the cost of delay in the harbor scenario to the scenario with a refurbished platform. There is an additional 27% decrease in the cost of delay when factoring in the relocated platform. The savings from the decreased cost of delay are hindered due to the increase in total maintenance cost that is incurred by implementing onsite maintenance. Operating offshore on the refurbished platform incurs a 39% increase in the total maintenance cost. This increases to 46% when comparing the harbor scenario to the relocated platform scenario.

Based on the above information, it is clear that conducting maintenance activities from an offshore platform has a positive impact on the KPI’s of an OWF, with the benefit increasing the closer the platform is to the OWF.

5.2 Sensitivity Analysis

(31)

27 of 46

different scenarios. A number of inputs were tested including the failure rate, cost of energy, distance from shore, and shift length. This analysis is discussed below.

5.2.1 Failure Rate

The failure rate has the highest level of uncertainty and as such, was tested with a variation of +/- 50%. This level of variation is common in operations research and incorporates (Maples et al., 2013). The results of this variation on the total cost for each scenario can be seen in the graph in Figure 7.

Figure 7: Sensitivity of the total cost to variation in failure rate

Here, for all variations of the failure rate the offshore accommodation is still the most cost beneficial policy. The cost benefit of the offshore accommodation increases with the increasing failure rate. At the lowest measured failure rate, the three scenarios are very close in cost. This can be attributed to the increase in availability that comes with a lower failure rate. The impact of the variation in failure rate on TBA can be seen in the graph in Figure 8. The increased availability in the harbor scenario causes the lower total cost due to the minimized cost of delay that comes with a higher availability. Similarly, the variation between the two offshore accommodations is both total cost and availability is minimal. This suggests that the offshore accommodation is better able to handle the uncertainty of the failure rate. This adds to the validity of the model as change in failure rate do not largely impact the results.

€80,00 €100,00 €120,00 €140,00 €160,00 €180,00 €200,00 -50% -40% -30% -20% -10% 0% 10% 20% 30% 40% 50% To ta l C os ts (€ /y ea r) in m ill io ns

Variation of Failure Rate

(32)

28 of 46

Figure 8: Sensitivity of the time based availability to variation in failure rate

5.2.2 Cost of Energy

The cost of energy is a highly variable input as it changes over time and can be hard to predict. For simplification, this research assumed that the cost of energy was constant over time. This assumption is now tested. It is important to test this input because the cost of delay is largely drive by the cost of energy. Significant changes in the cost of energy will impact the cost of delay which may impact our scenarios. Due to this importance, the cost of energy was tested with a variation of +/- 50%. The results of this analysis can be seen in the below Figure 9.

Figure 9: Sensitivity of the total cost to cost of energy

In this analysis we see, for the first time, an instance where the harbor scenario is the most cost beneficial. For low energy prices, the cost of delay in the harbor scenario is too low to justify the increased total maintenance cost that comes with operating from the offshore platform. However, this is only the case in very limited circumstances as the harbor scenario seems to be highly sensitive to variations in electricity price. Again, this can be attributed to the low availability in the harbor scenario and the energy price having a large impact on the cost of delay. The offshore accommodations are much less sensitive to the variations with only slight

40% 50% 60% 70% 80% 90% 100% -50% -40% -30% -20% -10% 0% 10% 20% 30% 40% 50% Ti m e Ba se d Av ai la bi lit y

Variation of Failure Rate

Harbor Refurbished Platform Refurbished + Relocated Platform

€50,00 €70,00 €90,00 €110,00 €130,00 €150,00 €170,00 €190,00 €210,00 -50% -40% -30% -20% -10% 0% 10% 20% 30% 40% 50% To ta l C os t ( €/ ye ar ) i n m ill io ns

Variation of Electricity Price

(33)

29 of 46

increases in the total cost. This can be attributed to the higher availability seen in these scenarios. As expected, the relocated scenario has the lowest total costs, but only slightly when compared to the refurbished platform.

5.2.3 Distance from Shore

The distance from shore is a factor that greatly affects the feasibility of implementing this plan industry wide. OWFs are located in varying distances from shore and are expected to be placed further and further from shore in the future. Because of this, it is essential to explore how the scenarios react to varying distances from shore. The distances were tested using values ranging from 10km to 100km. The results of this analysis on total cost can be seen in the graph in Figure 10. In order to keep in line with the initial scenarios, the refurbished platform is assumed the be located 60% of the distance between the OWF and harbor. For example, for a OWF 50km of shore the platform is assumed to be 30km from shore and 20km from the OWF. This assumption is made for simplicity due to the fact that there are many offshore platforms in the Dutch North Sea and it can assumed that at least one will fall within these parameters. The relocated platform is assumed to be located the same distance from shore as the OWF.

Figure 10: Sensitivity of the total cost to the distance from shore

This graph shows that in all cases the offshore accommodations are the most cost beneficial. However, for wind farms located very close to shore, the cost benefit between the refurbished and relocated platform is very minimal. This is due to the fact that it was assumed that the refurbished platform would be located 60% of the distance to the OWF. In cases where the wind farm is very close to shore, this distance is negligible when compared to the relocated platform. While neither offshore accommodation is very sensitive to the difference in distance, the relocated platform is particularly constant. This is because for this scenario, the distance from shore does not affect the distance needed to travel for repairs. The relocated platform is always

€80,00 €100,00 €120,00 €140,00 €160,00 €180,00 €200,00 10 20 30 40 50 60 70 80 90 100 To ta l C os t ( €/ Ye ar ) i n m ill io ns

Distance from Wind Farm to Shore

(34)

30 of 46

located at the OWF with a minimal distance to travel. This leads to a constant availability which can be seen in the graph in Figure 11.

Figure 11: Sensitivity of the availability to the distance from shore

The constant availability in the relocated scenario creates a bigger advantage as wind farms are located further and further offshore.

5.2.4 Shift Length

The shift length for the harbor scenario most closely reflect the current practice of OW maintenance which is 12-hours per day. The offshore scenarios most closely reflect the current practice of OO&G maintenance which is 24-hours per day. This variation in work shift may account for some variation in the performance and costs. Because of this, the shift length was analyzed for all scenarios. The results of this analysis can be seen in the below Table 6.

Shift Length

Harbor Refurbished Platform Relocated Platform

TBA Total Cost

(millions/year) TBA Total Cost (millions/year) TBA Total Cost (millions/year) 12-Hour 63% €146.1 66% €152.6 74% _€127.4 24-Hour 82% €128.3 88% €103.5 92% _€97.8

Table 6: Impact of different shift lengths on the key outputs per scenario

Here it can be seen that utilizing a 24-hour shift at the harbor can increase the availability of the wind farm the same amount as the offshore maintenance, but at a much higher total cost. Otherwise, the relocated platform is still the most beneficial under a 24-hour shift. When considering a 12-hour shift, the harbor scenario has better availability and lower costs than that refurbished platform scenario. However, the relocated scenario is still the most cost beneficial.

40% 60% 80% 100% 10 20 30 40 50 60 70 80 90 100 Ti m e Ba se d Av ai la bi lit y

Distance from Wind Farm to Shore

(35)

31 of 46

6 Discussion

This chapter highlights the main findings that can be derived from the results and their implications, both academically and practically. Additionally, the limitations of this study and proposed future research are explained.

6.1 Main Findings

The results of this developed model highlight the impacts of onsite maintenance on a repurposed OO&G platform on the performance of an OWF. It was derived that implementing onsite maintenance has a beneficial impact on the key performance indicators of indicators of an OWF. Furthermore, it was determined that while any offshore maintenance has an impact, it is most beneficial when located at the OWF.

These results are in line with the previously existing literature on onsite maintenance for OWFs. Besnard et al. (2013) studied an unidentified offshore accommodation to house maintenance technicians. This study found availabilities ranging from 96-98%. However, this study also had a focus on improved access systems which could be the cause of some of the additional availability. Besnard et al. (2013) also incorporated preventative maintenance several times a year. As this thesis did not include preventative maintenance, this difference could account for the differences in measured availability. Preventative maintenance could potentially impact the implementation of onsite maintenance.

(36)

32 of 46

assumed to be less than the lifetime cost of leasing a mother vessel and less than the investment cost to install a brand new offshore platform.

In addition to improvements in performance, it was found that any type of offshore maintenance is less sensitive to changes within the OW environment. The offshore environment is known to be volatile and is characterized by uncertainty. Therefore, any maintenance strategy that is able to stay relatively stable amid variations in the environment is beneficial. The sensitivity analysis tested the energy price and failure rate, two of the most volatile components of the OW environment (Shafiee and Sørensen, 2017). As these components can change over the lifetime of a turbine, it is beneficial to mitigate their impact on the performance of an OWF. Based on the results of the sensitivity analysis, it can be determined that the an offshore maintenance site creates stability in the performance measures that is not seen in the harbor scenario. In spite of this, if improving technologies lead to decreased failure rates, the cost benefit may be too minimal to justify the cost of refurbishment and/or relocation. However, the harsh environment and the inevitable aging of installed turbines suggests that this is not a point that will be reached in the near future.

The distance from shore is something that varies from wind farm to wind farm so any proposed solution should take into account variations in distance. This adds to the applicability of the solution. As expected, the farther offshore a wind farm is located, the more beneficial the relocated platform is. This is because the distance from shore has no impact on the distance needed to travel to complete repairs when the platform is located at the OWF. However, the slight variations seen may be due to a resource delay caused by bad weather not allowing a large enough weather window to deliver spare parts and new technicians to the platform. While this is an increased risk with the relocated platform, it has proven to be almost negligible. On the other hand, the closer to shore a wind farm is located, the less beneficial it is to relocate the platform. This is something that should be taken into consideration when considering the implementation of such as plan.

6.2 Implications

(37)

33 of 46

platform), an OWF operator should know the trade-offs of each and how they can be overcome. This research prevents a solution of consideration which benefits the performance of OWFs while requiring presumably less investment than other onsite maintenance methodologies. This, in turn, leads to the possibility for higher profits, increased return on investments and the possibility to better meet the lofty energy targets that have been set forth by government initiatives. Similarly, this research can aid in decision making within the OO&G industry. The OO&G industry is currently in search of methods to avoid decommissioning. This research presents a viable and practical solution to the mass decommissioning dilemma. Overall, this research should add to the tools available to OWF operators and OO&G companies to deal with the ongoing energy transition.

Theoretically, this research serves as an addition to the extensive literature in the improvement of O&M activities for OWFs. Because the improvement of O&M is vital to the continues success and development of OWFs, all research that can present new and viable solutions is beneficial. This research also identifies a solution to a problem in the literature of onsite maintenance for OWFs. Previous literature presented onsite maintenance that was either non-permanent (i.e. mother vessels) or so expensive that they must be shared (i.e. service islands). Permanent onsite maintenance for a singular wind farm was largely believed to be unobtainable due to the high costs outweighing the benefits. This research identifies a possible solution that can be built upon in future investigations of onsite maintenance. Additionally, this research begins to investigate the idea of the OW industry and the OO&G industry working together to alleviate their individual problems. This collaboration has the potential to be hugely beneficial but has only begun to be researched.

6.3 Limitations and Future Research

(38)

34 of 46

It is recommended that future research be conducted to incorporate these failure rates and assess their impact on the KPIs.

Preventative maintenance was not considered in the research. As mentioned in the discussion, this omission can impact the impact of onsite maintenance. Theoretically, if maintenance can be planned in advance, the impact of delay will be minimized. However, it could also be argued that preventative maintenance, specifically condition-based maintenance, might be more feasible when located onsite as the condition of the turbine can be monitored better than when compared to onshore maintenance. Because of this uncertainty, it is recommended that future research be conducted to investigate the impact of the implementation of preventative maintenance.

When testing the distance from shore input, it was assumed that the refurbished platform would be located 60% of the distance between the OWF and the harbor. This is a limitation and not representative of reality. Additional research should be conducted utilizing a case study approach with specific OO&G platforms and specific OWFs. In line with this, additional research could be conducted to test this solution when multiple platforms and OWFs are incorporated. A network of maintenance platforms could prove to be beneficial.

Lastly, a study should be conducted to determine the feasibility and cost of the actual refurbishment and/or relocation of the platform. This represents a large component in the feasibility of the proposed solution but was considered outside of the scope of this thesis.

7 Conclusions

(39)

35 of 46

Assessing the Impact of Onsite Maintenance on Offshore Wind Farms: The Case of the Repurposed Oil Platform