The design of a service contract between
original equipment manufacturer and
wind farm owner
Kirsten Sommerauer (S2207397)
Student MSc Technology and Operations Management
Supervisors
Dr. X. Zhu (First assessor)
Dr. E. Ursavas (Second assessor)
Mr. A. H. Schrotenboer (Academic advisor)
University of Groningen
February 13, 2019
Abstract
Currently, offshore wind energy is growing in popularity as an alternative energy source. Resulting from higher efficiency due to a more consistent source of wind compared to onshore wind. As this market is growing, it is more difficult to obtain specialized equipment and knowledge about maintenance logistics. To take advantage of this shortage, many origi-nal equipment manufacturers are offering service contracts to wind farm owners after they purchased a wind turbine. As this trend only started recently, these service contracts are not being critically reviewed, while the maintenance of offshore wind farms induce a lot of costs. As a result, preventing unnecessary costs by analyzing service contracts will give wind farms more insight in their expenses. Consequently, this thesis will design different service contracts in order to advise the wind farm owner and original equipment manufacturer in its negotiations for the ideal service contract. The contracts will be distinguished between 2- and 5 year con-tracts, combined with the needs of the original equipment manufacturer or wind farm owner. Furthermore, to develop this model, a plant simu-lation is conducted. Based on the output of the simusimu-lation the original equipment manufacturer could be advised to focus on the failure distri-bution, since a variable failure distribution could cause maintenance costs to fluctuate which induces uncertainty to meet the minimum service level. In contrast, the service level is most striking for the wind farm owner as it assures stable revenue.
Acknowledgements
1
Introduction
Input
Lumpsum x Service Level !(OEM)
Revenue Loss > Helicopter cost (WF owner)
Optimize Method
Determine reorder level S1 and S2, wherefore the profits
of the OEM and WF owner is maximized
ProfitOEM = x – costs (!) ProfitWFowner = Revenue – x
Output
Figure 1: Contract structure according to original equipment manufacturer and wind farm owner.
between the two perspectives, the availability will be determined by the service
level (α), holding costs determined by reorder levels of the main port (S1) and
service operations vessel (S2), and the lump sum (x). This is shown in Figure
1. Consequently, the service level is the boundary whenever the availability becomes too low.
Moreover, applying maintenance also includes the transportation of spare parts and technicians to and from the wind farm. To start, this research considers spare parts to be stocked at the main port and the service operations vessel (hereafter SOV). Between the main port and the distribution centre the trans-portation will be handled by a delivery service. Furthermore, within the inven-tory of the main port and SOV a (s,S) policy will be maintained as inveninven-tory policy, this inventory policy applies an order-up-to level S whenever inventory level reaches or falls below s which is suitable for offshore wind turbines because of flexible order quantity and no fixed time frame to order (Berman et al., 2008; Garg et al., 1997). Additionally, offshore transportation is important to con-sider due to the large impact on the cost of maintenance service logistics. This thesis will focus on the SOV and helicopter as means of offshore transportation, as these are applied for relatively smaller O&M purposes (Dewan et al., 2016). Finally, in practice, ports around the North Sea consider a competitive relation, but also a partnership. This is necessary, since storing all spare parts as indi-vidual port could be considered inefficient. Wherefore, they must collaborate to offer the full maintenance support to wind farms. Figure 2 provides an overview of the previously mentioned parties and their relations.
Main port Supplier port OEM storage Wind farm SOV or helicopter CTV Legend Offshore transportation Onshore transportation
Figure 2: The relation between wind farm, main port, supplier port and original equipment manufacturer storage.
rising trend of outsourcing investigating this topic carries societal relevance for both perspectives. The outcome of this thesis should help wind farm owners and OEM in their decision making with respect to the negotiations for the service contract. This research will be constructed by means of plant simulation and aims to show different outcomes of costs between the various decisions of a wind farm owner and OEM, such as contract length and minimal availability of the wind farm. Moreover, the following research question will be answered:
How is a service contract between original equipment manufact-urer and wind farm owner designed?
This thesis is structured into the following sections; Section 2 contains the the-oretical outline which summarizes current literature and shows the scientific contribution of this thesis. Section 3, elaborates on the applied research. Sec-tion 4 presents the results from the simulaSec-tion study. SecSec-tion 5 consists of the discussion including recommendations and limitations and, finally Section 6 represents the conclusion. In addition, consult Appendix A.1 and Appendix A.2 to check the meaning of the in this thesis applied abbreviations.
2
Theoretical Outline
According to Shafiee (2015) and Nillson (2007), who analyzed logistics of off-shore WFs, the OEM offers a 2- or 5-year full-service contract. This includes the responsibility of the OEM to carry the costs for maintenance whenever the wind turbine fails and must apply refurbishments on critical components to decrease the probability of future failures for the upcoming two or five years. Similarly, Kumar et al. (2004), who have investigated the negotiation of service contracts within the business-to-business sector, agree that the duration of service con-tracts varies between two and five years and argue that after each year the price and delivery scope is reviewed and if necessary revised. Additionally, the dura-tion of a service contract must not be too short (< one year) or too long (> five years). In order to achieve the best fitting duration depends on factors like type of product, the associated price and the insecurity of the market. For instance, rapid changes in technology, market and/or prices of the concerned product are reasons for conflicts between the WF owner and OEM, as these changes imply uncertainties. Moreover, too long service contracts can induce a loss of com-mitment between the involved parties and too short service contracts can cause unnecessary high costs due to meetings/negotiations (Kumar et al., 2004). This thesis will check whether there is a significant difference between a 2- and 5 year service contract concerning WFs availability, costs and revenue. Which has an added value, because Shafiee (2015) and Nillson (2007) investigates the maintenance logistics in the offshore wind market more in general terms and do not go into depth about service contracts.
Whether the contract has a duration of 2- or 5 years, the two parties need to collaborate within the contract. Therefore, both parties must be very clear and transparent about what elements will influence the maintenance service logis-tics and what critical information they need to share with the other party. This can be clarified during the negotiation process. Consequently, if service terms are not clear enough or even incorrectly specified during the negotiations, the relationship is likely to provoke conflicts and service failures. Thus, to have both parties satisfied about the relationship every element of the service must be clearly elaborated in the contract, because all future uncertainties cannot be covered within the contract. To avoid future conflicts, both parties must be prepared for renegotiation about the original agreement (Kumar et al., 2004). This thesis will help both parties involved in the service contract by indicating the most relevant topics in their perspectives. For the OEM will minimizing to-tal costs be the incentive and for the WF owner the setting where most revenue will be generated. In contrast to Kumar et al. (2004) this thesis will take into account the stochastic nature of the offshore wind market.
ne-gotiation position of both parties. The following are applicable in the offshore wind market. First, the involved firm must prevent dependency by outsourcing maintenance to different OEMs. Second, the OEM must not put forth their expertise on R&D all at once. Finally, both parties must focus on their rela-tive bargaining power during the negotiations, as the relarela-tive bargaining power affects whether the service contract is profitable from their perspective. Nev-ertheless, not all techniques concluded in the paper of Cho et al. (1994) are applicable in the offshore wind market, since offshore WF are more dependent on OEMs in comparison to other sectors, due to the shortage of work-boats and specialist knowledge (Dalgic et al., 2015a; Dalgic et al., 2015b).
Furthermore, one of the crucial factors during the negotiations are the mainte-nance costs. These costs are mainly high due to expensive transportation and the frequent need of technicians and spare parts on the WF. To reduce these costs it may be more suitable to combine visits to the WF. However, reduc-ing visits to the WF will increase the risk for keepreduc-ing the wind turbines out of operation too long and could reduce the availability until near the promised service level, which will cause a reduction in revenue (Kaldellis et al., 2013; Shafiee, 2015). This dilemma must be balanced out during the negotiations of the service contract to improve cooperation between the OEM and WF owners. Therefore, this will be tested in combination with inventory optimization. Moreover, more sectors are showing an increase in usage of service contracts, as Stremersch et al. (2001), who studied full-service contracts within the indus-trial maintenance market, argues a trend in demand of full service contracts. As a result, the relationship between the firms and OEM is changing from a competitive into a cooperative. Yet, these full service contracts have not been studied often. There has been research conducted about service contracts, how-ever these are all at least 10 years old and do not focus on the offshore wind market. So, it would be a scientific contribution to the current literature to test service contracts within the offshore wind market.
To sum up, this thesis is relevant for the following reasons. First, no research has been conducted specifically on service contracts in combination with the offshore wind market, as previous research does not include the concerned stochastic circumstances an offshore WF has to deal with. Also, the offshore wind market is a relatively new sector, therefore WFs are more dependent on OEMs for their expertise and work-boats compared to other sectors. Second, the concerned papers which analyzed the offshore wind sector dedicate only a minor part on service contracts. Third, most papers concerning service contracts are outdated,
despite the fact that the topic is still highly relevant. In short, this thesis
3
Methodology
In this thesis contracts will be analyzed with respect to duration and perspec-tives through plant simulation. The plant simulation is done with the aim of cost minimization and revenue maximization and is tested within a 2- and 5 year contract. Starting with determining the optimal reorder levels of the main
port (S1) and SOV (S2) based on cost minimization and revenue maximization.
A flowchart of this model is shown in Figure 3. This section will elaborate on the model parameters of the flowchart and explains how these parameters are specified within the model. The parameters are separated as inputs, simula-tions/subsystems and outputs. The inputs include the parameters which are provided by other papers and sources. Simulations are determined by the in-puts. Resulting from these simulations are the outin-puts. This section has the following structure: Section 3.1 discusses the inputs for the simulation. Sec-tion 3.2 describes the simulaSec-tions/subsystems. SecSec-tion 3.3 discusses the likely outputs derived from the simulations.
Offshore Transportation Inputs Wind Farm (WF) Failure/Repair Weather Conditions ‘Preferred’ Availability Onshore Transportation Spare Parts Storage Supplier Port
Spare Part Failure
Actual Availability Offshore Transport Specific Cost Performance Measures
Simulations and Subsystems Outputs
Maintenance
Failure Specific
Figure 3: Conceptual Model - Flowchart
3.1
Inputs
3.1.1 Wind Farm
An offshore WF consists of a varied number of wind turbines. One of the
et al., 2012). Nevertheless, Carroll et al. (2016), who analyze failures of wind turbines, claimed that there is no clear definition for failure within the market of wind energy. To be consistent they applied ‘an unplanned visit to a turbine in which material is consumed’ as definition of failure. Based on this definition they concluded that around 97.5% of all failures consists of repairs and 2.5% major replacements. The failures related to repairs will be elaborated in Sec-tion 3.1.4. Furthermore, the distance between the WF and shore influences the accessibility of the WF. As a result, a greater distance is one of the factors for higher maintenance costs (Bussel, van et al., 2001). Consequently, these factors will influence the determination of inventories policies.
3.1.2 Weather Conditions
As mentioned the Section 1, whenever weather conditions causes hazardous sit-uations to access the WF, a helicopter can complement or replace the SOV. In those situations there is an impulse to transfer spare parts to the WF in order to prevent long term failures of wind turbines (Dewan et al., 2016). Ac-cording to Scholz-Reiter et al. (2010), which analyzed the planning and control of offshore logistics, the supply chain of offshore wind farms must be able to respond quickly on the varying weather conditions. Since, it is more difficult for WFs to react on failures during bad weather conditions. As a result, WFs must transfer sufficient supplies in case of good weather conditions. To pre-vent material shortage, all components in the chain must be well matched to maximize the efficiency of the transportation. Furthermore, to include weather in the planning, it is essential to determine a certain approach to forecast the weather. One example of forecasts Scholz-Reiter et al. (2010) mentions is the seasonal weather forecast. These forecasts uses historical data to determine the forecast, therefore it is applicable long-term and suitable for longer periods, such as months. The different months are distinguished by the probabilities of good or bad weather. For example November has a higher probability of bad weather
than May. Similarly, L¨utjen et al. (2012) who developed an inventory
con-trol system between ports and offshore installations, applied a seasonal weather forecast. The seasonal weather forecast is presented in Appendix A.4, where Figure 9 illustrates how the probabilities are distinguished between months, Ta-ble 13 shows what the weather conditions imply in terms of consequences for transport and maintenance, and Table 14 presents these probabilities expressed
in percentages (L¨utjen et al., 2012).
3.1.3 ‘Preferred’ Availability
MTTF MTTF Weather conditions allow maintenance? Transport time to reach wind turbine Service level α intent to be insufficient?*
*Original Equipment Manufacturer (OEM) perspective: Minimize maintenance cost **Wind Farm (WF) perspective: Maximize revenue
Failure occurs? Yes? Sent helicopter Repair time MTTR Requisite Spare Part is not in stock S2? Yes? Revenue Loss > helicopter cost?** No?
Figure 4: Mean time to repair consisting of decisions and unavoidable lead times.
SOV is offshore. Moreover, the first four steps are explained as follows. The first question verifies whether a failure occurs, otherwise there is no need for action. The second step questions whether the weather conditions allow failure repair. The third step checks the availability of the failed spare part in the inventory
of the SOV (S2). If the spare part is in stock then the next three steps must be
skipped. The fourth step from the OEM perspective measures the threat of the service level becoming too low. In contrast, the WF owner checks whether the lost revenue until resupply will exceed helicopter costs.
3.1.4 Failure/Repair
minor failures and 95% major failures. Similarly, the offshore wind turbines are originated from onshore wind turbines, so the pattern of failure is alike. Yet, as mentioned before, the accessibility of an offshore WF is limited. Consequently, due to limited accessibility the downtime caused by minor failures will increase and therefore is expected that total downtime will increase as well (Faulstich et al., 2011). Nevertheless, during the lifetime of 20 years the failure rate of a wind turbine is covered by the early life phase and useful life phase. After 20 years the wear-out period, the last phase, will enter. First, the early life phase with at the start a very high failure rate which falls during the phase. This phase has an duration of around one to two years. Second, the useful life phase, consisting of a constant failure rate with ‘intrinsic or random failure’. The duration of this phase is around 18 years until the life time has reached. Third, the wear-out period, with an increasing failure rate due to the operational age of the wind turbine. This pattern of first decreasing, constant and increasing failure rate is also called the bathtub curve (Faulstich et al., 2011; Tavner et al., 2007). This bathtub curve is shown in Figure 5.
In cr eas ed F ail ur e Ra te Time Early life Falling failure rate Useful life
Constant failure rate Wear-out periodRising failure rate The Bathtub Curve
Hypothetical Failure Rate versus Time
Figure 5: The hypothetical failure rate over time of wind turbines (Faulstich et al., 2011; Tavner et al., 2007).
3.2
Simulations and Subsystems
3.2.1 Spare Part Storage
In ve nt or y Le ve l Time s S Lead Time Inventory Position
Figure 6: (s,S) Policy; orders (S-s) at point s and is restocked after the lead time has passed.
not easily accomplished, as forecasts are mostly based on historical data of de-manded, which is not always available.
This thesis will attain the minimum inventory level by optimizing the inventory policy, concerning the (s,S) policy illustrated in Figure 6. This policy restocks inventory level to order-up-to level S if inventory level falls below or reaches re-order point s (Berman et al., 2008). The inventory position is defined as the inventory level in case lead times are not considered. Furthermore, for the (s,S) policy the optimal parameters s and S must be determined by the performance measures of the diverse spare parts (Garg et al. 1997). Moreover, this policy will be evaluated over the period the service contract is applicable and will be measured by the average failure rate and repair time of the considered subsys-tem, average back orders transported by helicopter, mean service level provided by the contract and average holding costs per subsystem.
In addition, this thesis will consider an order-up-to level S equal to re-order point s, so the re-order point will be neglected. In other words, whenever inventory level drops below order-up-to level S it will be restocked immediately (after reorder time) until the order-up-to level is maintained again. This inventory
policy is considered for the main port (S1) and SOV (S2). Both order-up-to
levels are differentiated by the holding costs and capacity limitation, which are elaborated in Section 3.2.2.
3.2.2 Transportation
halve of the O&M costs. Moreover, according to a practical example; ‘Gronin-gen Seaports’, the SOV is regularly deployed based on a cycle of fourteen days to perform preventive and corrective maintenance. In addition, a helicopter could be applied whenever the SOV is unable to travel or repair due to weather conditions.
SOV
Table 11 and Table 12 shows the specifications of the SOV. In this table it is assumed that the transportation is shared, as the OEM is often a big player, like Siemens, within the wind energy market and will serve multiple WFs at the same time. Additionally, the SOV is deployed in a fourteen days cycle, as mentioned in this section. Between the fourteen days cycle the SOV returns to the main port and will resupply, what is assumed to have a duration of one day. Only weather conditions could influence the fourteen days cycle whether the SOV is able to transport or maintain. Whereas, the SOV can only resist
weather conditions up to level 3 (Dewan et al. 2016; L¨utjen et al., 2004). In this
research the SOV consists of an inventory level which is set to S2. The inventory
of the SOV is relatively expensive, since the SOV has limited inventory capacity and so contains higher holding costs in comparison to the storage at the main
port (S1) (Dewan et al. 2016). Besides, it is assumed that the storage of the
main port is shared as well, what will keep the costs low. Helicopter
In addition, the specifications of the helicopter are shown in Table 11 with the input parameters in Table 12. In contrast to the SOV the helicopter is resistant to all weather conditions and could be applied every day. Nevertheless, the costs of the helicopter are significantly higher in comparison to the SOV (Dewan et al., 2016). Therefore, the helicopter is not regularly used and is even deployed in different context depending on the perspective. Following the OEM perspective, the helicopter is deployed depending on the service level (α). Consequently, if the service level becomes too low the OEM is forced to deploy the helicopter otherwise he/she will receive a significant large fine. In contrast, the WF owner will compare the revenue loss generated until resupply to the helicopter costs. Other Transport
This section concerns the onshore transportation and the supplier port. Start-ing with the onshore transportation which represents the transportation between the distribution and the storage at the main port (Eemshaven). Consequently, this is executed by a transport service, like DHL. The transportation is
con-cerned before resupplying the main port (S1). Secondly, the supplier port which
in Section 4.4.
3.2.3 Maintenance
Each failure per subsystem per day is recorded in a list where a first-in, first-out policy is applied. The reason to maintain this policy is in relation to the service level. Since, the service level is determined by the availability of the WF and the OEM has as goal to minimize total costs. Also, each failure contains their corresponding technicians and required hours to maintain. In other words, the OEM prefers the cheapest circumstances which means failures with the shortest duration are preferred to repair and maintain the service level. As a result, the failures with the longest duration will never be resolved, while the service level will remain sufficient. Nevertheless, the maintenance is to some extent still influenced by both perspectives, due to the application of the helicopter. Besides, the daily maintenance is restricted as the sum of the required hours of all daily failures must not exceed the total hours from the twenty daily deployed technicians, where the daily shift contains twelve hours. If this does not hold then the repair will be postponed to the next day.
3.2.4 Spare Part Failure
Failu re r at e sp ar e p ar ts [1/y e ar] Failu re r at e wi n d tu rb in e [1/y e ar]
Figure 7: Failure rate developed over time (Faulstich et al., 2011).
In addition, the corresponding failure rate is computed with a turbine specified as G ≥ 1M W . Following that the failure rates per components are added up to a final failure rate representing the overall failure rate of the subsystem (Reder et al., 2016). In addition, Table 15 illustrates the minimal weather condition the
subsystems can be repaired as well (L¨utjen et al., 2012). Whenever the weather
conditions are worse the repair must be postponed to the next day. Besides, subsystem ‘Nacelle’ is ordered from ‘Cuxhaven’ and contains a make-to-order
policy instead of using the inventory of the main port (S1).
3.2.5 Performance Measures
The performance measures are determined by all other simulations within Sec-tion 3.2. Whereas, the outputs will be based on the performance measures, such as service level, total costs and revenue loss.
Performance
To construct a bathtub shaped curve for the total failure rate a Weibull distri-bution is applied with an α = 0,5 and a time parameter β equal to 6 hours. Additionally, the failures are distinguished per subsystem, as mentioned in Sec-tion 3.2.4 and shown in Table 15. Consequently, the performance is measured by the number of failures waiting for to be repaired which is expressed in the mean daily queue length. This performance measure checks daily the number of failures that needs to be maintained. Preferred is to minimize the daily queue length, since a longer queue length will intent less revenue and a lower service level. Besides, the relative availability of the WF is measured by the mean service level, expressed in Equation 1.
M eanServiceLevel = Na
(Na+ Nu)
, (1)
where Na represents the availability of the WF and Nu the unavailability of
Tuptime = M SL · Hy, (2)
Tdowntime = (1 − M SL) · Hy, (3)
where M SL represents the mean service level and Hy the number of hours for
the total duration of the service contract. For example a 2 year contract the number of hours is 490 days times 24 hours equals 11.760 hours and for a 5 year contract will it be 29.400 hours. The reason for 245 days per year is explained in Section 4.2.
Costs
The cost outputs are determined by the following costs: offshore transportation costs, the costs per technician per day, and holding cost per spare part. First, the offshore transportation costs which consists of constant yearly SOV cost, the travel cost of a SOV per trip and the transport costs of the helicopter per day. These costs are specified in Table 11 in Appendix A.3. Furthermore, the costs per technician per day are determined by the input data of Dinwoodie et al. (2015), which analyzed simulations related to offshore WFs. They set the technician costs on 80.000 pounds per year, which equals 246,51 euro per day (based on the interest rate of 3-12-2018). Finally, the holding costs which contains the reorder costs and the holding costs itself. The difference between the two considered holding costs is shown in Table 15. Moreover, the sum of holding costs will vary within experiments.
Revenue loss
Subsequently, the lost revenue per day is derived by the capacity of the wind turbines. This research applies the Siemens SWT - 2.3 93 as wind turbine, since this turbine is one of the most applied wind turbines in the UK and is suitable for offshore installment (Staffell, 2012). This wind turbine reaches a capacity of 2.3 MW given wind speeds of 14 up to 25 m/s, see Figure 8. To implement the power curve within the five weather conditions the average of the power outputs is taken within the captured wind speeds, which is shown in Table 1. Nevertheless, the capacity of 2.3 MW will never be reached in this simulation, due to the average results. In addition, these captured speeds are determined by the max wind speeds from the weather conditions. To express this in ‘lost rev-enue’ the electricity price of 7 cents per kWh (Wind Energy Solutions (WES), 2016) is used. Lost revenue occurs whenever a wind turbine fails and could not be maintained immediately, as the SOV is not able to repair at that moment.
One of the reasons is the spare part shortage in the SOV (S2), due to bounded
capacity and the associated holding costs. The latter limits the cost efficiency, since keeping infinite spare parts is likely to be expensive.
3.3
Outputs
Multiple simulations are performed to compute each output for both perspec-tives. As mentioned in Section 3 there are four main cases: a 2- and 5 year contract from the OEM perspective and a 2- and 5 year contract from the WF
owner perspective. Within these cases different experiments are conducted.
Table 1: Turbine Specific Inputs
Weather Condition Max Speed Captured Speed Power Output
[m/s] [m/s] [KW ] Very good <5 1-5 61,6 Good <6,5 6 376,0 Medium <11 7-10 1.154,5 Bad <12 11 2.164,0 Very bad >12 12-30 1.693,8
Figure 8: Power Curve of a Siemens SWT - 2.3 93 (Staffell, 2012).
changing inventory levels S1 and S2 within the experiments. Thereafter, with
the ideal inventory level a sensitivity analysis is conducted by adjusting the time parameter of the failure distribution, changing the required work hours per fail-ure and, depending on perspective, change service level or helicopter costs to check performance measurements. Finally, the costs for the subsystem Nacelle will be tested in a sensitivity analysis. The associated outputs are listed below:
• Actual availability: Mean service level;
• Offshore transport specific: Implementation of helicopter and SOV with their associated costs;
• Failure specific: Application of different failure distributions and the mean queue length;
• Cost: Holding costs of inventory levels, transport costs and technician costs.
4
Results
within each scenario a confidence interval is applied. This confidence interval is calculated by Equation 4.
CI = X ± 2, 576 ·√σ
n, (4)
where CI stands for confidence interval distinguished between upper- and lower confidence interval, X represents the mean, 2,576 is the critical value or the determined z-value for 99% confidence level, σ is the standard deviation and
the n is the number of observations. Furthermore, in Section 4.2 the base
scenario is determined by finding the optimal inventory levels for the main
port (S1) and SOV (S2). Starting with experiments based on the inventory
combinations shown in Table 2, where CC stands for the subsystem Control and Communication, DT the abbreviation is of the subsystem Drive Train and
PM signifies the subsystem Power Module which are stored at the port (S1)
and SOV (S2). The inventory combinations are based on the time parameter
of the failure distribution and the failure rate. From this the inventory levels are decreased and increased. Similarly, these inventory combinations will be applied in Table 3, Table 6, Table 7, Appendix C and Appendix D. All experiments do not contain a warm-up period, as it is assumed that contracts start at a beginning of a cycle, during resupply, when all inventories are full. The corresponding assumptions can be found in Appendix B.
Table 2: Inventory Levels per experiment, where CC means Control & Commu-nication, DT stands for Drive Train and PM is Power Module.
Experiment CCS1 DT S1 P M S1 CCS2 DT S2 P M S2 1 15 15 15 13 8 10 2 10 10 10 10 8 10 3 9 9 9 9 8 9 4 8 8 8 8 8 8 5 7 7 7 7 7 7 6 16 16 16 14 9 10 7 17 17 17 15 10 11
4.1
Model design
if the following holds: the SOV is at the WF for maintenance, the weather conditions are sufficient, the requisite spare part is available at the SOV and the sum of repair times of all maintained spare parts from that day does not exceed the total daily available time for maintenance (240 hours). Nevertheless, it could occur that the SOV is short in the requisite spare part to repair a certain failure. Depending on the perspective, the helicopter will be deployed for different reasons. Considering the OEM perspective, a helicopter will only be deployed when the service level falls below a specified threshold. However, if we consider the WF perspective, a helicopter will only be deployed if the revenue loss will exceed the helicopter costs. Whenever a failure is repaired the corresponding spare part will exit the model. Besides, the resupply day only considers subsystems Control and Communication, Drive Train and Power Module. Since, the spare parts of subsystem Nacelle concerns a make-to-order policy and do not pass the inventory at the main port. From this design of the simulation experiments the performance measures, costs and revenue are
generated. Following, these outcomes will determine which topics are most
important to negotiation during the design of the service contract.
4.2
Base scenario
The model considers a WF consisting of 100 wind turbines. To put this size into perspective, this WF will be one of the largest WFs affiliated to the Eemshaven. Whenever a wind turbine fails and is maintained, a spare part from the SOV
inventory (S2) is selected. To refill inventory S2 the storage at the main port
(S1) must be replenished first by the distributor. In this scenario the inventory
at the main port will be refilled right after inventory level becomes lower than
S1. In contrast, the SOV can only be restocked one day in a cycle of 15 days.
Since, the cycle consists of one day SOV replenishment, one day traveling to the WF, twelve days of maintenance and one day traveling back. Consequently,
with one day replenishment of the SOV (S2) it can occur that inventory level
becomes equal to zero before the day of replenishment. Before applying a he-licopter to send the missing spare part, the OEM will check the service level and the WF owner will compare the revenue loss generated until replenishment day with the helicopter cost. Appendix A.3 shows the SOV and Helicopter specifications.
4.2.1 Original Equipment Manufacturer
This subsection will elaborate on the perspective of the OEM within a contract of 2- and 5 years. Typical for the OEM perspective is the dependency on the service level which is set to 85%. Important is that the service level will not drop below 85%, so the turbine level is set to a minimum of 90 turbines. Since, haz-ardous weather conditions could delay maintenance for minimal one day while the number of failures continues to increase. Furthermore, Figure 12 shows the total holding costs and helicopter costs for 2- and 5 years in the perspective of the original equipment manufacturer. As can be seen, the helicopter costs never exceeds the total holding costs. Similarly, this has a logical explanation, as Control and Communication contains relatively high holding costs. Striking is that all total costs of a 5 year contract are around 2,5 times higher than the costs of a 2 year contract. Since, the failure distribution is shaped with parame-ter 0,5 which has to be decreasing over time. More experiments are done to test whether a higher or smaller shape parameter will distinguish the 2 year contract from the 5 year contract. The results from these experiments are shown in Ap-pendix G. From the associated figures can be seen that with shape parameters 0,1 and 0,2 a difference in ratio between a 2- and 5 year contract is measured, however shape parameters smaller than 0,5 distribute failures that decreases too fast over time. Consequently, on average seems to be no significant difference to be determined by the performance measurements. Nevertheless, shape pa-rameters larger than 0,5 do not create a difference in the ratio between 2- and 5 year contracts.
Table 3 illustrates the total O&M costs and revenue, and shows in bold the best situation and in italics the worst. To calculate the total O&M costs the following costs are added up: helicopter costs, holding costs, yearly SOV costs, technician costs and SOV travel costs. Consequently, the different O&M costs are mainly distinguished by helicopter costs and holding costs, shown in Ap-pendix C. This is also noticeable in Table 3, since the optimal total costs also represent the lowest inventory level and vice versa. There can be noticed that the total costs and revenue are minimized at experiment 5 (containing the low-est holding costs and highlow-est helicopter costs compared to other experiments, shown in Figure 12). The OEM prefers this experiment, because their profit concerns the lump sum minus the total costs. To check whether costs could be reduced more, lower inventory levels are tested as well. Lower inventory levels could be considered, as the current mean service levels are all minimal 90%, shown in Figure 10 from Appendix C.
Table 3: Total O&M costs and revenue, OEM scenario for 2 years and 5 years,
respectively (mile).
Experiments 1 2 3 4 5 6 7
Total Costs 36,93 33,63 32,81 32,08 31,40 37,95 39,15
Total Revenue 86,17 85,24 84,60 83,98 83,25 86,52 87,05
Total Costs 92,53 84,20 82,10 80,27 78,52 95,12 98,20
Table 4: Inventory Levels per experiment, where CC means Control & Commu-nication, DT stands for Drive Train and PM is Power Module.
Experiment CCS1 DT S1 P M S1 CCS2 DT S2 P M S2 1 6 6 6 6 6 6 2 5 5 5 5 5 5 3 4 4 4 4 4 4 4 3 3 3 3 3 3 5 2 2 2 2 2 2
Table 5: Total O&M costs and revenue, OEM scenario for 2 years (mile).
Experiments 1 2 3 4 5
Total Costs 30,86 30,44 30,13 29,92 29,77
Total Revenue 82,56 81,93 81,36 80,89 80,56
Mean service level 89,63% 89,02% 88,47% 88,03% 87,73%
The lower inventory levels within the experiments are set to the parameters of Table 4. From these input parameters Table 5 is created for a 2 year contract. Only a 2 year contract is tested, as a 2- and 5 year contract react in the same manner on the inventory levels. Also with these experiments, the total costs are minimized with the lowest inventory level. Probably, due to the large holding
costs in comparison to the helicopter costs. Similarly, the experiment with
the lowest cost also has the lowest revenue and the highest costs consist of the highest revenue, which will cause a conflict of interest between the OEM and WF owner. To design the ideal contract, compromises must be made on the lowest (as possible) costs and the highest (as possible) revenue. Moreover, all percentages of mean service levels are acceptable within these experiments. Nevertheless, 87,73% is a risky mean service level in connection to the risk of bad weather conditions.
4.2.2 Wind Farm Owner
The WF owner checks whether the helicopter cost exceeds the total revenue of the upcoming days until resupply. In other words, whenever the helicopter costs are smaller than the total revenue loss then the helicopter will be deployed to pick up the considered spare part. In Figure 13 illustrated in Appendix C are the helicopter cost versus holding costs presented. Noticeable is the comparison between the helicopter cost and holding costs of OEM and WF, specified in Appendix D. Resulting that, the helicopter costs in perspective of OEM are slightly lower than the helicopter costs in perspective of WF owner. Taking that into account, the WF owner would deploy the helicopter more in comparison to the OEM. Nevertheless, the difference between the perspectives is relatively small. Although, it seems more logical that all holding costs must be smaller for the WF owner, as the WF owner generates more helicopter costs. However,
the cost of the inventory on the SOV (S2) is larger for the WF owner
the spare parts take when transported by the helicopter, namely from the port to SOV. So, all (extra) spare parts transported by helicopter must go through
inventory S2, which will create additional holding cost. Moreover, on average
of all experiments the OEM perspective contains during the contract of 2 years e278,13 thousand less costs in comparison to the WF owner and in a 5 year
contract e634,26 thousand less costs. As a result, the OEM perspective has
2,28 times more costs in a 5 year contract in comparison to a 2 year contract. Meaning that the difference between the costs of WF perspective and OEM perspective converges over time.
Furthermore, Table 6 illustrates the total O&M costs, revenues and difference between both values. Also, the best and worst case are illustrated in bold and italicized, respectively. Likewise, the total costs and revenue are maximized with the highest inventory levels, so the WF owner would prefer experiment 7. Nevertheless, concerning the negotiations, whenever the total costs are at its highest the OEM will demand a higher lump sum to cover total costs. Therefore, the difference between revenues and total costs must also be taken into account. Noticeable, concerning Table 6, the most optimal ‘difference’ is experiment 5 with the smallest inventory, probably also due to the relative high holding costs. To check whether this is the case, the experiments are repeated with half of the initial holding costs, resulting in Table 7. However, these experiments also show the lowest inventory as most optimal concerning the difference between revenue and total costs. Providing that, Table 8 is based on the experiment settings of Table 4. In this setting there is an optimal solution greater than the lowest inventory level, namely experiment 1 regarding inventory levels equal to 6. Table 6: Total O&M costs, revenue and the difference, Wind Farm owner
sce-nario for 2 years and 5 years, respectively (mile).
1 2 3 4 5 6 7 Total Revenue 87,78 87,61 87,50 87,42 87,32 87,88 88,02 Total Costs 37,19 33,95 33,13 32,38 31,66 38,20 39,39 Difference 50,60 53,65 54,38 55,03 55,66 49,68 48,63 Total Revenue 217,81 217,36 217,11 216,88 216,63 218,03 218,39 Total Costs 93,12 84,96 82,86 80,96 79,11 95,67 98,69 Difference 124,69 132,40 134,25 135,92 137,52 122,36 119,70
Table 7: Total O&M costs, revenue and the difference, Wind Farm owner
sce-nario for 2 years and half of holding costs (mile).
Experiments 1 2 3 4 5 6 7
Total Revenue 87,78 87,61 87,50 87,42 87,32 87,88 88,02
Total Costs 30,83 29,47 29,17 28,94 28,76 31,26 31,77
Difference 56,95 58,14 58,33 58,48 58,56 56,61 56,25
Table 8: Total O&M costs, revenue and the difference, Wind Farm owner
sce-nario for 2 years and half of holding costs (mile).
Experiments 1 2 3 4 5
Total Revenue 87,23 87,17 78,12 87,08 87,06
Total Costs 28,66 28,65 28,72 28,87 29,08
Difference 58,57 58,52 58,40 58,22 57,98
Mean service level 94,41% 94,32% 94,26% 94,21% 94,18%
4.3
Sensitivity Analysis
To analyze more determinants of service contracts the following input variables are tested in a sensitivity analysis: failure distribution and repair times in both perspectives, as these variables could vary in practice. Additionally, the heli-copter costs and service level will be tested in the WF perspective and OEM perspective, respectively. Furthermore, the base scenario consists of 85%
ser-vice level, 6 hours as time parameter, repair times shown in Table 15 ande6000
helicopter costs per day. Moreover, the different sensitivity analysis are distin-guished between both perspectives, presented in Appendix E.
The sensitivity analysis of the OEM perspective consists of service level, failure distribution and repair times. Resulting from Section 4.2.1 the optimal inven-tory level is determined based on the lowest total costs, which will also be applied in this section. Providing that the smallest inventory levels induce the lowest total costs, still the mean service level must not become below 85%. To check whether this is the case the minimum availability and maximum unavailability are combined to calculate the mean service level. Starting with experiment 5 from Table 4 during a 5 year contract which creates an insufficient mean service level of 84,64%. Subsequently, experiment 4 simulates also an insufficient mean service level of 84,89%. Finally, experiment 3 provides a combination which cor-responds to the service level of 85% and so will be applied as best combination in a 5 year contract for the OEM perspective with a mean service level of 85,25%. However, for a 2-years contract all experiments from Table 4 do not hold to 85% service level. Therefore, experiment 5 of Table 2 is tested and seems that this experiment does hold with 85,13%. As a result, the base scenario is for the OEM set to inventory levels of seven in a 2 year contract and four in a 5 year contract. Whereas, the WF perspective includes helicopter cost, failure distribution and repair times in the sensitivity analysis. The sensitivity analysis will be executed with the base scenario determined with half holding costs, which is at all inventory levels set to six.
4.4
Nacelle
fail-ure has already occurred. Since, holding spare parts from this subsystem is very expensive, a nearby other port holds these spare parts. Providing that, the costs for the subsystem Nacelle consists of travel costs and holding costs
of the inventory on the SOV (S2). It is assumed the supplier port applies a
Crew Transfer Vessel (hereafter CTV) for the transportation to the main port, which is a smaller, but faster boat in comparison to the SOV. Also the transfer between the main port and the WF is by CTV. Consequently, the travel costs
are set, based on Dewan et al. (2016), to e500 per trip, which equals e1000
per spare part.
A sensitivity analysis is performed to test the generated costs of the Nacelle, shown in Table 9 and Table 10. First, noticeable within the OEM perspective is the large gap between base scenario and the adjustment of failure distribution. Which could be expected, as in case of the time parameter of 3 hours the failure distribution is twice as high. Equally, with a time parameter of 9 hours the fail-ure distribution is only two-thirds of initial failfail-ure distribution. Furthermore, changes in the repair times and service level do not adjust the costs significantly. Which seems logical, as in any case the spare part will be sent due to the make-to-order policy. Only when the failure distribution is changed, the spare part will be applied more or less compared to the base scenario. In addition, com-paring both contracts in the OEM perspective consisting of different inventory levels, determined in Section 4.2, the contract with a higher inventory level gen-erates more costs with regard to the Nacelle. Only if the capacity is reached, seen at failure distribution equals 3:00:00 or 95% service level, inventory levels make no difference to the costs anymore. Second, the WF perspective has the largest adjustments in the change of the time parameter of failure distribution. Whereas, the rest of the adjustments do not show any significant effect of more
thane500. The 5 year contract even shows less absolute effect than the 2 year
contract.
Table 9: Sensitivity Analysis: Total O&M costs for Nacelle for OEM scenario.
(a) 2 years
OEM 2-years Nacelle costs
Service Level 75% e154.492,00
Base e155.740,00
95% e156.806,00
Failure Distribution 3:00:00 e310.596,00
Base e155.740,00
9:00:00 e104.416,00
Repair Times -20% e155.792,00
Base e155.740,00
+20% e155.636,00
(b) 5 years
OEM 5-years Nacelle costs
Service Level 75% e385.476,00
Base e386,854,00
95% e387.790,00
Failure Distribution 3:00:00 e781.196,00
Base e386.854,00
9:00:00 e260.208,00
Repair Times -20% e386.906,00
Base e386.854,00
Table 10: Sensitivity Analysis: Total O&M costs for Nacelle for WF scenario.
(a) 2 years
WF 2-years Nacelle costs
Helicopter Cost -20% e156.728,00
Base e156.702,00
+20% e156.702,00
Failure Distribution 3:00:00 e309.010,00
Base e156.702,00
9:00:00 e104.910,00
Repair Times -20% e156.884,00
Base e156.702,00
+20% e156.416,00
(b) 5 years
WF 5-years Nacelle costs
Helicopter Cost -20% e387.244,00
Base e387.218,00
+20% e387.218,00
Failure Distribution 3:00:00 e778.778,00
Base e387.218,00
9:00:00 e261.014,00
Repair Times -20% e387.244,00
Base e387.218,00
+20% e387.166,00
5
Discussion
The WF owner perspective includes half of the holding costs. As, higher holding costs has a bigger impact on the WF perspective due to more regular deployment
of the helicopter which creates extra holding costs on the SOV (S2). Following
This thesis contributes to current scientific literature for the following reasons. First, scientific papers that fully focus on service contracts combined with off-shore wind energy do not exist currently. Furthermore, this thesis also covers its impact on the design of the logistics. Second, within this logistic model the inventory management is observed as well, involving another port as third party. Third, many papers only include one perspective which is often the WF perspective. In this research both perspectives are taken into consideration to create coherency between the two parties with their design of the service con-tract. Also, the 2- and 5 year contract are included to check whether there is a significant difference between these contracts.
However, this thesis also contains several limitations. Firstly, the developed sim-ulation model includes mainly stylized input variables. All input variables are determined by papers or other sources except for the failure distribution. Nev-ertheless, the weather conditions are measured by the probabilities per month, so stochastic in nature. However, these are determined in advance using ex-cel. Secondly, while designing a service contract it is only to a certain extent reliable to retrieve weather data from the past. Because trends in weather con-ditions and ocean flows change over time and could deviate from the suggested mean weather probabilities. For example, this summer the temperatures were exceptionally high for an extended period (World Meteorological Organization (WMO), 2018). Thirdly, this thesis ignores four months of weather data per year, which makes output less realistic. Finally, the simulated inventory could be more realistic and more detailed. Since, the simulated inventory is simplified compared to specialized research on inventory management due to data and time constraints.
Considering the earlier mentioned points several recommendations can be for-mulated for alternative approaches. Elaborating on the first limitation, more or adjusted input variables could be added. First, more detailed costs variables, such as reorder cost and costs associated with resupply could be added. Sec-ond, more variables that influence offshore transport, such as more detail in the weather conditions and transportation time. Third, focusing on more types of maintenance, as this research only focuses on corrective maintenance. Fourth, implementing the bathtub shaped failure rate differently, since the Weibull dis-tribution did not have the ideal outcome. Even if the shape parameter was adjusted. This implies as stated before that more stochastic factors could be created in the simulation model. Besides, another period of the wind turbine’s lifetime could be highlighted or different states for each wind turbine could be implemented. Finally, regarding the fourth limitation more OEMs could poten-tially share their data to promote more reliable research.
6
Conclusion
available literature is limited in scope and quantity. The need of more research in service contracts is supported by the increasing demand for offshore WFs. The overall results of this thesis shows that even though OEM and WF owner interests can be conflicted the reasoning and costs associated with these interests could also bring the two parties together. Since, from the OEM perspective it is striking that the total costs do not change significantly whenever the service level will be adjusted. However, the revenue and the mean daily queue length will increase. Nevertheless, to prevent large cost increases, it is necessary to indicate the failure distribution during the negotiations. To conclude, based on the sensitivity analysis the trade off within the negotiations would be the service level for the WF owner and a well determined failure distribution for the OEM. In addition, the OEM will prefer the 5 year contract, whereas the WF owner will prefer the 2 year contract. Depending on the outcome of the negotiations the lump sum could be determined.
References
Bilgili, M., Yasar, A., & Simsek, E. (2011). Offshore wind power develop-ment in Europe and its comparison with onshore counterpart. Renewable and Sustainable Energy Reviews, 15(2), 905-915. doi:10.1016/j.rser.2010.11.006
Bussel, van, G. J. W., & Henderson, A. (2001). State of the Art and Tech-nology Trends for Offshore Wind Energy: Operation and Maintenance Issues. Section Wind Energy, Faculty Civil Engineering and Geosciences Delft Univer-sity of Technology, (1), 1-4. Retrieved October 28, 2018, from
http : //www.of f shorewindenergy.org/CA − OW EE/downloads/Brussels01 O&M.pdf
Carroll, J., McDonald, A., & McMillan, D. (2016). Failure rate, repair time and unschedules O&M cost analysis of offshore wind turbines. Wind Energy, 19(6), 1107-1119. http://doi.org/10.1002/we.1887
Cho, D., & Chu, W. (1994). Determinants of bargaining power in OEM nego-tiations. Industrial Marketing Management, 23(4), 343-355. doi:10.1016/0019-8501(94)90050-7
Dalgic, Y., Lazakis, I., & Turan, O. (2013). Vessel charter rate estimation for offshore wind O&M activities. Developments in Maritime Transportation and Exploitation of Sea Resources, 899-907. doi:10.1201/b15813-113
Dalgic, Y., Lazakis, I., & Turan, O. (2015a). Investigation of Optimum Crew Transfer Vessel Fleet for Offshore Wind Farm Maintenance Operations. Wind Engineering, 39(1), 31-52. doi:10.1260/0309-524x.39.1.31
Dekker, R., Pin¸ce, C¸ , Zuidwijk, R., & Jalil, M. N. (2013). On the use of installed base information for spare parts logistics: A review of ideas and industry practice. International Journal of Production Economics, 143(2), 536-545. doi:10.1016/j.ijpe.2011.11.025
Dewan, A., & Asgarpour, M. (2016). Reference O&M Concepts for Near and Far Offshore Wind Farms (ECN-E—16-055, pp. 1-49, Rep. No. 1). Pet-ten: ECN. Retrieved October 19, 2018, from
https : //www.ecn.nl/publications/P df F etch.aspx?nr = ECN −E −−16−055 Dinwoodie, I., Endrerud, O. V., Hofmann, M., Martin, R., & Sperstad, I. B. (2015). Reference Cases for Verification of Operation and Maintenance Simulation Models for Offshore Wind Farms. Wind Engineering, 39(1), 1-14. doi:10.1260/0309-524x.39.1.1
Esteban, M. D., Diez, J. J., L´opez, J. S., & Negro, V. (2011). Why offshore
wind energy? Renewable Energy, 36(2), 444-450. doi:10.1016/j.renene.2010.07.009 Faulstich, S., Hahn, B., & Tavner, P. J. (2011). Wind turbine downtime
and its importance for offshore deployment. Wind Energy, 14(3), 327-337.
doi:10.1002/we.421
Garg, A., & Kalagnanam, R. (1997). Method for fast and accurate eval-uation of periodic review inventory policy. International Business Machine Cor-poration.
Ghandi, A., & Lin, C. C. (2014). Oil and gas service contracts around the world: A review. Energy Strategy Reviews, 3, 63-71. doi:10.1016/j.esr.2014.03.001
Jackson, C., & Pascual, R. (2008). Optimal maintenance service contract negotiation with aging equipment. European Journal of Operational Research, 189(2), 387-398. doi:10.1016/j.ejor.2007.05.039
Jin, T., Ding, Y., & Guo, H. (2012). Managing wind turbine reliability and maintenance via performance-based contract. 2012 IEEE Power and En-ergy Society General Meeting. doi:10.1109/pesgm.2012.6344739
Kaldellis, J., & Kapsali, M. (2013). Shifting towards offshore wind
en-ergy—Recent acivity and future development. Energy Policy, 53, 136-148.
doi:10.1016/j.enpol.2012.10.032
Kumar, R., Markeset, T., & Kumar, U. (2004). Maintenance of machin-ery. International Journal of Service Industry Management, 15(4), 400-413. doi:10.1108/09564230410552077
L¨utjen, M., & Karimi, H. R. (2012). Approach of a Port Inventory
Con-trol System for the Offshore Installation of Wind Turbines, The Twenty-second International Offshore and Polar Engineering Conference (ISOPE). Rhodes, Greek, 2012, pp. 502-508. Retrieved October 19, 2018, from
259930920 Approach of a P ort Inventory Control System f or the Of f shore Installation of W ind T urbines/links/0a85e52e9391f 2daa 7000000.pdf
Nilsson, J., & Bertling, L. (2007). Maintenance Management of Wind Power Systems Using Condition Monitoring Systems—Life Cycle Cost Analysis for Two Case Studies. IEEE Transactions on Energy Conversion, 22(1), 223-229. doi:10.1109/tec.2006.889623
Reder, M. D., Gonzalez, E., & Melero, J. J. (2016). Wind Turbine Fail-ures - Tackling current Problems in Failure Data Analysis. Journal of Physics: Conference Series, 753, 072027. doi:10.1088/1742-6596/753/7/072027
Scholz-Reiter, B., L¨utjen, M., Heger, J., & Schweizer, A. (2010).
Plan-ning and control of logistics for offshore wind farms. BIBA - Bremer
In-stitut f¨ur Production und Logistik GmbH, Retrieved October 19, 2018, from
http : //www.biba.uni − bremen.de.
Shafiee, M. (2015). Maintenance logistics organization for offshore wind
energy: Current progress and future perspectives. Renewable Energy, 77, 182-193. doi:10.1016/j.renene.2014.11.045
Shafiee, M., & Sørensen, J. D. (2017). Maintenance optimization and inspec-tion planning of wind energy assets: Models, methods and strategies. Reliability Engineering and System Safety. https://doi.org/10.1016/j.ress.2017.10.025
SIEMENS. (2013). Plant Simulation. Plant Simulation: Simulation and Optimization of Production Systems and Processes, 87. Retrieved from: http:// community.plm.automation.siemens.com/siemensplm/attachments/siemensplm/Plant-Simulation-Tecnomatix/181/1/Plant Simulation Fact Sheet Book HQ.pdf
Staffel, I. (2012). Wind Turbine Power Curves. W1P5, Retrieved November 29, 2018, from https : //www.academia.edu/1489838/W ind T urbine P ower Curves
Stremersch, S., Wuyts, S., & Frambach, R. T. (2001). The Purchasing of
Full-Service Contracts:. Industrial Marketing Management, 30(1), 1-12. doi:10.1016/s0019-8501(99)00090-5
Tavner, P. J., Xiang, J., & Spinato, F. (2007). Reliability Analysis for
Wind Turbines. Wiley Interscience, 10, 1-18. doi:10.1002/we.204 Wind
Appendix
Appendix A - Model Components
A.1 Definitions frequently used abbreviations.Definitions Meaning
O&M Operations and Maintenance
OEM Original Equipment Manufacturer
WF Wind Farm
SOV Service Operations Vessel
CTV Crew Transfer Vessel
S1 Inventory level at main port
S2 Inventory level at SOV
CC Control and Communication
DT Drive Train
PM Power Module
A.2 Definitions derived from Faulstich et al. (2011).
Definitions Descriptions
Reliability Probability an item performs adequately
Availability Ratio production time to total time
Mean time to failure (M T T F ) Time the machine is working
Mean time to repair/Downtime (M T T R) The duration to correct the failure
Failure rate (λ) The frequency a failure occurs
A.3 Transportation Details
Table 11: SOV and Helicopter parameters
Service Operation Helicopter
Transport input Vessel
Cycle 14 days 1 day
Max wave height 3.0 m 99 m
Max wind speed 17 m/s 20 m/s
Cost e12M/yr + e6000/day
e100/trip
Speed 10 knots 120 km/hr
Lead time resupply 24 hours 2 hours
Table 12: Input parameters offshore transport
Input parameters Description
Max wave height Limitation due to offshore weather conditions
Max wind speed Limitation due to offshore weather conditions
Cost Equipment costs
Speed Average speed of transportation
Lead time resupply The time it takes to resupply
Max technicians Number of technicians per transportation
Cycle Duration offshore transport
A.4 Weather Forecast
60% 50% 40% 30% 20% 10% 0% Sh ar e of mo nt h J F M A M J J A S O N D Medium Good Very bad Very good Bad Month
Figure 9: The probability of offshore weather conditions in terms of months based on data from the last 50 years until 2007.
Table 13: Feasible offshore operations in terms of weather conditions.
Operation max. wind speed max. wave min. weather
[m/s] height [m] condition
No operation possible >12 >4.8 very bad
Sea-fastening <12 <4.8 bad
Installation of substructures <11 <3.5 medium
Installation of nacelles <6.5 <2.5 good
Table 14: Probabilities of weather conditions divided per month.
very good (%) good (%) medium (%) bad (%) very bad (%)
Jan 3.0 10.0 33.0 9.0 45.0 Feb 3.0 13.0 39.0 10.0 35.0 Mar 5.0 14.0 42.0 9.0 30.0 Apr 7.0 22.0 47.0 7.0 17.0 May 7.0 23.0 53.0 7.0 10.0 Jun 7.0 27.0 51.0 6.0 9.0 Jul 8.0 27.0 50.0 6.0 9.0 Aug 7.5 27.0 48.0 6.5 11.0 Sep 6.5 20.0 45.0 7.5 21.0 Oct 5.0 13.0 40.0 7.0 35.0 Nov 2.5 9.0 35.5 9.0 44.0 Dec 4.0 10.5 33.5 10.0 42.0
A.5 Subsystems specified
Table 15: Subsystems input parameters
Subsystem Control & Drive train Power Nacelle
Input Communication Modules
Failure rate 26% 15% 20% 4%
Holding cost a day S1 510 euro 120 euro 150 euro make-to-order
Holding cost a day S2 1275 euro 300 euro 375 euro 1600 euro
Specified Minor Medium Medium Major
Repair time 7,5 h 22 h 22 h 26 h
Number of technicians 2 3 3 4
Min. weather condition Medium Medium Medium Medium
Table 16: Components per Subsystem
Subsystem Components
Control & Communication Sensors, Controller, Communication System,
Emergency Control & Communication Series, Data Aquisition System
Drive Train Gearbox, Main Bearing, Bearings, Mechanical Brake,
High speed Shaft, Silent Blocks, Low Speed (Main) Shaft
Power Module Frequency Converter, Generator, Switch Gear,
MV/LV Transformer, Power Feeder Cables, Power Cabinet, Power Module Other, Power Protection Unit
Appendix B - Assumptions
1. Main port and SOV stocks spare parts for subsystems Control & Commu-nication, Drive Train and Power Module;
2. Supplier port stocks spare part for subsystem Nacelle;
3. Spare parts from the subsystem Nacelle are always available to supply to main port;
4. The four spare parts have different failure rates and repair times; 5. The entire wind turbine breaks down if one spare part fails;
6. The Storage of the port and SOV maintain a (s,S) policy, except the Nacelle subsystem, which contains a make-to-order policy;
7. Staff is always available and in optimal state; 8. The SOV and helicopter has a constant speed; 9. The SOV or helicopter do not fail;
10. The SOV and helicopters are shared by OEM with other WFs, which has consequence on transport costs;
11. SOV and helicopter are always available;
12. The seasonal weather forecast of L¨utjen et al., (2012) is assumed to
fore-cast the weather and sea conditions;
13. All wind turbines have the same specifications; 14. All wind turbines start at year 0;
Appendix C - Revenues, costs and mean service level
86,0% 87,0% 88,0% 89,0% 90,0% 91,0% 92,0% 93,0% 94,0% 95,0% 96,0% €-€5,00 €10,00 €15,00 €20,00 €25,00 €30,00 €35,00 €40,00 €45,00 1 2 3 4 5 6 7 Mil lionsTotal Cost Mean Service Level
(a) 2 year contract
86,0% 87,0% 88,0% 89,0% 90,0% 91,0% 92,0% 93,0% 94,0% 95,0% 96,0% €-€10,00 €20,00 €30,00 €40,00 €50,00 €60,00 €70,00 €80,00 €90,00 €100,00 1 2 3 4 5 6 7 Mil lions
Total costs Mean Service Level
(b) 5 year contract
Figure 10: The total costs and mean service level for the perspective of original equipment manufacturer. 86,0% 87,0% 88,0% 89,0% 90,0% 91,0% 92,0% 93,0% 94,0% 95,0% 96,0% €80,00 €81,00 €82,00 €83,00 €84,00 €85,00 €86,00 €87,00 €88,00 €89,00 1 2 3 4 5 6 7 Mil lions
Revenue Mean Service Level
(a) 2 year contract
86,0% 87,0% 88,0% 89,0% 90,0% 91,0% 92,0% 93,0% 94,0% 95,0% 96,0% €200,00 €202,00 €204,00 €206,00 €208,00 €210,00 €212,00 €214,00 €216,00 €218,00 €220,00 1 2 3 4 5 6 7 Mil lions
Revenue Mean Service Level
(b) 5 year contract
Figure 11: The revenue and mean service level for the perspective of Wind Farm owner. €-€1,00 €2,00 €3,00 €4,00 €5,00 €6,00 Helicopter Cost CCS1 DTS1 PMS1 CCS2 DTS2 PMS2 Mil lions 1 2 3 4 5 6 7
(a) 2 year contract
€-€2,00 €4,00 €6,00 €8,00 €10,00 €12,00 €14,00 Helicopter Cost CCS1 DTS1 PMS1 CCS2 DTS2 PMS2 Mil lions 1 2 3 4 5 6 7 (b) 5 year contract
€-€1,00 €2,00 €3,00 €4,00 €5,00 €6,00 Helicopter Cost CCS1 DTS1 PMS1 CCS2 DTS2 PMS2 Mil lions 1 2 3 4 5 6 7
(a) 2 year contract
€-€2,00 €4,00 €6,00 €8,00 €10,00 €12,00 €14,00 Helicopter Cost CCS1 DTS1 PMS1 CCS2 DTS2 PMS2 Mil lions 1 2 3 4 5 6 7 (b) 5 year contract
Figure 13: Helicopter costs versus holding costs: Wind Farm owner perspective.
Appendix D: Difference Original Equipment Manufacturer
and Wind Farm owner in Base scenario
To exemplify the difference within the initial experiments between OEM- and WF owner perspective with respect to holding costs and helicopter costs, Table 17 and Table 18 are created.
Table 17: The cost difference from OEM perspective relative to the WF
per-spective of 2 year contracts (Thousande).
Experiments Helicopter costs CCS1 DT S1 P M S1 CCS2 DT S2 P M S2
1 -58,92 -0,52 0,03 -0,13 -140,47 -16,67 -38,24 2 -61,02 0,61 0,07 -0,05 -221,26 -12,49 -31,36 3 -58,08 1,19 0,12 0,06 -225,09 -6,88 -32,67 4 -51,96 2,22 0,17 0,19 -220,58 -1,73 -31,51 5 -41,04 3,19 0,27 0,45 -198,21 3,16 -25,91 6 -47,64 5,13 0,89 -0,14 -149,94 -19,47 -38,75 7 -34,62 5,32 0,88 1,42 -148,05 -22,64 -36,27
Table 18: The cost difference from OEM perspective relative to the WF
per-spective of 5 year contracts (Thousande).
Experiments Helicopter costs CCS1 DT S1 P M S1 CCS2 DT S2 P M S2
Appendix E: Sensitivity Analysis
For each perspective and duration a sensitivity analysis is created. The results are diverge from large impact to no significant impact. To begin with the OEM perspective. In this perspective both contracts are based on different base sce-narios. For both contracts the total costs do not change significantly whenever the service level is adapted. Also the adjustment of the failure distribution has a large impact on the overall costs and revenue, especially the helicopter costs, the
total holding costs for the inventory of the SOV (S2) and technician costs. This
Appendix F: Confidence interval base scenario
A confidence interval is determined for the base scenario of both perspectives and both contract durations. The confidence intervals are determined by the average of all 100 observations within the simulation. A 99% confidence level is applied for the confidence interval, as the simulation model only consists of one stochastic variable. What can be observed is that within these 100 observations all confidence intervals are relatively narrow. Which suggests that the amount of observations was sufficient.
Table 23: Results at 99% confidence level: OEM perspective
OEM year: M easure M ean StandardDeviation CIlow CIup
2 Helicopter cost e1.481.940,00 e273.262,44 e1.411.547,60 e1.552.332,40
CCS1 e1.650.778,20 e3.388,68 e1.651.905,28 e1.651.651,38
DTS1 e394.317,60 e1.542,89 e393.920,15 e394.715,05
PMS1 e488.397,00 e1.425,42 e488.029,81 e488.764,19
CCS2 e1.544.241,75 e152.895,54 e1.504.855,86 e1.583.627,64
DTS2 e597.936,00 e39.709,66 e587.706,79 e608.165,21
PMS2 e590.073,75 e48.253,93 e577.643,54 e602.503,96
Technician cost e646.653,08 e46.983,09 e634.550,24 e658.755,92
Revenue e83.245.5195,64 e598.545,35 e83.091.010,36 e83.399.380,92
Revenue Loss e7.368.610,42 e598.532,76 e7.214.428,38 e7.522.792,46
Mean Service Level 90,361% 0,671% 90,189% 90,534%
Mean Daily Queue Length 8,06 0,64 7,90 8,23
5 Helicopter cost e6.811.500,00 e546.077,83 e6.670.830,35 e6.952.169,65
CCS1 e2.375.325,00 e2.306,56 e2.374.730,83 e2.375.919,17
DTS1 e562.948,80 e1.045,32 e562.679,53 e563.218,07
PMS1 e700.410,00 e1.049,03 e700.139,77 e700.680,23
CCS2 e1.839.774,00 e102.652,94 e1.813.330,60 e1.866.217,40
DTS2 e687.870,00 e36.627,85 e678.434,67 e697.305,33
PMS2 e675.708,75 e39.856,56 e665.441,70 e685.975,80
Technician cost e1.591.604,75 e75.212,49 e1.572.230,01 e1.610.979,49
Revenue e201.825.242,17 e673.553,47 e201.651.734,80 e201.998.749,54
Revenue Loss e22.642.749,94 e673.547,85 e22.469.244,01 e22.816.255,87
Mean Service Level 88,532% 0,299% 88,446% 88,600%
Mean Daily Queue Length 10,07 0,29 9,99 10,14
Table 24: Results at 99% confidence level: WF perspective
WF year: M easure M ean StandardDeviation CIlow CIup
2 Helicopter cost e1.890.600,00 e293.832,98 e1.814.908,63 e1.966.291,37
CCS1 e567.630,65 e1.530,27 e567.236,45 e568.024,85
DTS1 e168.478,80 e683,15 e168.302,82 e168.654,78
PMS1 e208.884,75 e673,41 e208.711,28 e209.058,22
CCS2 e690.717,94 e64.250,79 e674.166,94 e707.268,94
DTS2 e235.725,00 e18.626,26 e230.926,88 e240.523,12
PMS2 e243.561,52 e20.619,57 e238.249,92 e248.873,12
Technician cost e646.653,08 e46.983,09 e634.550,24 e658.755,92
Revenue e87.230.725,11 e431.640,27 e87.119.534,58 e87.341.915,64
Revenue Loss e3.383.234,77 e431.631,87 e3.272.046,40 e3.494.423,14
Mean Service Level 94,407% 0,667% 94,235% 94,579%
Mean Daily Queue Length 3,72 0,46 3,60 3,84
5 Helicopter cost e4.583.880,00 e468.188,76 e4.463.274,58 e4.704.485,42
CCS1 e1.421.605,30 e2.655,86 e1.420.921,15 e1.422.289,45
DTS1 e422.205,00 e1.031,16 e421.939,37 e422.470,63
PMS1 e523.231,50 e1.060,88 e522.958,22 e523.504,78
CCS2 e1.753.370,74 e111.298,48 e1.724.700,25 e1.782.041,23
DTS2 e597.336,00 e29.208,74 e589.811,83 e604.860,17
PMS2 e615.654,88 e35.329,81 e606.533,92 e624.755,84
Technician cost e1.591.604,75 e75.212,49 e1.572.230,01 e1.610.979,49
Revenue e216.413.864,61 e678.920,25 e216.238.974,75 e216.588.754,47
Revenue Loss e8.054.635,33 e678.903,11 e7.879.749,89 e8.229.520,77
Mean Service Level 94,471% 0,414% 94,364% 94,577%