FACULTY MECHANICAL, MARITIME AND MATERIALS ENGINEERING Delft University of Technology Department Marine and Transport Technology

Mekelweg 2 2628 CD Delft the Netherlands Phone +31 (0)15-2782889 Fax +31 (0)15-2781397 www.mtt.tudelft.nl

Specialization: Transport Engineering and Logistics

Report number: 2017.TEL.8167

Title: Optimization of O&M of offshore wind farms

Author: J.P.R. Triepels

Title (in Dutch) Optimalisatie van O&M van offshore windparken

Assignment: Literature project

Confidential: no

Supervisor: Dr. ir. X. Jiang

Date: August 1st, 2017

This report consists of 42 pages and 4 appendices. It may only be reproduced literally and as a whole. For commercial purposes only with written authorization of Delft University of Technology. Requests for consult are only taken into consideration under the condition that the applicant denies all legal rights on liabilities concerning the contents of the advice.

FACULTY OF MECHANICAL, MARITIME AND MATERIALS ENGINEERING Delft University of Technology Department of Marine and Transport Technology

Mekelweg 2 2628 CD Delft the Netherlands Phone +31 (0)15-2782889 Fax +31 (0)15-2781397 www.mtt.tudelft.nl

Student: Assignment type: Review Supervisor: Dr.ir. Jiang Report number: 2017.TEL.xxxx Specialization: TEL Confidential: Creditpoints (EC): 10

Subject: Deployment of big data technology to reduce O&M cost of offshore wind

Offshore wind is a relatively new industry and in general offshore wind is more expensive to generate than many alternative renewable sources. Operation & Maintenance (O&M) makes up a significant part of the overall cost of running offshore wind turbines. The complication of O&M lies in that its responsibility has been split between turbine manufacturers, operators and the offshore transmission owners. This has resulted in procedures and systems that have evolved to provide short term solutions. Thus this has inevitably led to areas of inefficiency, duplication of effort and lack of information.

Big data technology is one great technologies that will drive future growth. Big data would offer three huge benefits that can transform an industry in terms of visualization of real time data; development of decision support tools based on disparate data sources; and data mining to aid planning and find performance inefficiencies to improve real time operations. It should be noted that even with the data in place, there are several challenges facing converting data into valuable information, such as posing the right question, developing the right software using the right technique to answer the questions, etc.

This assignment aims is to investigate the possible impact of big data on O&M of offshore wind turbine in order to answer such questions as whether and how a better collection, management and presentation of data (coupled with an integrated software solution) would provide significant cost reductions in O&M. The following aspects are required to be illustrated in the report:

• The definition of big data in context of offshore wind technology

• The state of the art – Application of big data in offshore wind industry addressing on - Wind turbine monitoring - Supply chain management - Marine operations - Other aspects, if applicable Including available database, analysis models and methods, software and techniques, etc.

• Exploration of future work (remaining issues and possible ways for improvement / solutions)

This report should be arranged in such a way that all data is structurally presented in graphs, tables, and lists with belonging descriptions and explanations in text.

The report should comply with the guidelines of the section. Details can be found on the website. If you would like to know more about the assignment, you may contact with Dr. X Jiang through [email protected].

X Jiang

Abstract

Europe is working on a transition to a sustainable, reliable and affordable energy supply for everyone. As de- termined in the Innovation Outlook Offshore Wind (2016) of IRENA (International Renewable Energy Agency) [1] will have to become the leading power generation technology by 2030 to ensure a decar- bonization of the global economy. In order to achieve such growth, the levelized cost of energy (LCOE) needs to be reduced. One of the important factors of the LCOE is operations and maintenance (O&M). Due to the offshore location and the uncertain weather conditions, wind farms at sea have different requirements, chal- lenges and costs compared to onshore wind farms. O&M responsibility is currently split between wind farm owners, turbine manufacturers, offshore transmission owners and sometimes independent service contrac- tors. This makes the already difficult process of operations even more complicated. There is a demand for an approach to make the O&M more effective. Furthermore, the availability of cheap, reliable sensors and the general recognition of the value of data has lead to an explosion in interest in the area of “big data”.

Operations and maintenance covers all activities from completion of installation works to the start of decom- missioning. Different stakeholders can have different interests making it sometimes difficult to determine the best strategy. There are three main actors involved: project owners, wind turbine original equipment manufacturers (OEM) and offshore transmission owners (OFTO).

Operations and maintenance of offshore wind farms can be categorized into three fields: supply chain man- agement, wind turbine monitoring and marine operations. Supply chain management is the management of all the flows of materials, parts, equipment and storage thereof. There is limited literature available on the topic of supply chain management specifically related to offshore wind farms. Wind turbine monitoring has to do with gathering data on the wind turbine itself for both performance control and condition based maintenance. A lot of research is going in this field at the moment. Marine operations include all operations that are necessary to get workers and equipment to and from the wind farm. Although not much researched, marine operations is also an interesting field for data collection and analysis.

The most likely developments in operation and maintenance are in the following areas: improvements in weather forecasting and analysis, introduction of turbine condition-based maintenance strategies, improve- ments in OMS strategy for far-offshore wind farms, improvements in personnel transfer and access, intro- duction of remote and automated maintenance, and introduction of wind farm-wide control strategies. In particular, a lot is expected from the implementation of condition-based maintenance in combination with wind farm-wide optimizing control strategies and the improvements and innovations of the PTVs and SOVs.

There have been a few attempts to optimize the maintenance control strategy with the help of an inte- grated system. ECN is an important Dutch organization working on this principle. Other recent projects are DAISY4Offshore and the in start-up phase project ZEPHYROS. ZEPHYROS is a project by World Class Main- tenance trying to bring together relevant actors in the field with researchers, students and local authorities. Projects like these are important in the process of innovation. By working together on improvements, Europe will only have the leading role, but also keep this leading role.

v

Contents

1 Introduction 1 2 Offshore wind in The Netherlands3 2.1 Current situation...... 3 2.2 Roadmap to 2023...... 3 3 Offshore wind operations and maintenance5 3.1 Stakeholders...... 5 3.1.1 Contracting types...... 5 3.2 Activities...... 6 3.2.1 Operations...... 6 3.2.2 Maintenance...... 6 4 Role of big data in offshore wind9 4.1 Supply chain management...... 9 4.1.1 Supply chain in offshore wind industry...... 9 4.1.2 Purchasing and supply management practices...... 10 4.1.3 Supply chain assessment The Netherlands...... 10 4.1.4 Big data in supply chain management...... 11 4.2 Wind turbine monitoring...... 12 4.2.1 Wind Turbine Maintenance...... 12 4.2.2 Monitoring system: CMS...... 12 4.2.3 Monitoring system: SCADA...... 13 4.2.4 Opportunities...... 13 4.2.5 Big data in wind turbine monitoring...... 14 4.3 Marine operations...... 15 4.3.1 Offshore logistics...... 15 4.3.2 Case study by UEA...... 15 4.3.3 Operations at Gemini...... 16 4.3.4 Opportunities...... 17 4.3.5 Big data in marine operations...... 18 5 Projects on optimizing control strategies 19 5.1 ECN...... 19 5.2 Decision support tool by H. Koopstra...... 19 5.3 World Class Maintenance...... 20 5.3.1 DAISY and DAISY4Offshore...... 21 5.3.2 ZEPHYROS...... 21 6 Conclusions & future work 23 6.1 Conclusions...... 23 6.2 Future work...... 24 Appendix A: Wind farms in The Netherlands 25 Appendix B: What is big data? 27 Appendix C: TKI Offshore Supply Chain Assessment 29 Appendix D: Phone interview with Bart Hoefakker, Gemini 32 Bibliography 33

vii

1 Introduction

Europe is working on a transition to a sustainable, reliable and affordable energy supply for everyone. In 2016 Europe raised a total of AC43bn for the construction of new wind farms, refinancing operations, project ac- quisitions, and public market fund raising. The countries investing the most are UK, Germany, Belgium and Norway together taking 80% of the new capacity financed. [2] Also The Netherlands is working on more sus- tainable energy sources. In 2015 over forty organizations including ministries, energy organizations, employ- ers organizations, unions, NGO’s and others agreed on ’The National Energy Agreement’. In this agreement a goal of 16% sustainable energy in 2023 has been set. Part of this agreement states that, in order to reach this goal, the offshore wind capacity has to increase from 1000 MW to 4500 MW in 2023. [3]

As determined in the Innovation Outlook Offshore Wind (2016) of IRENA (International Renewable Energy Agency) [1] wind power will have to become the leading power generation technology by 2030 to ensure a decarbonization of the global economy. The amount offshore wind capacity built each year is rising, see also Figure 1.2 IRENA’s analysis is that offshore wind capacity can reach 100 gigawatts (GW) by 2030 when innovation continues as fast as it does currently. In the past 15 years offshore wind has already grown from a few megawatts capacity to a current worldwide total capacity of more than 12 gigawatts, most of which is generated off the coasts of Europe as can be seen in Figure 1.1.

Figure 1.1: Offshore wind deployment at the end of 2015 (source: IRENA, 2016)

In order to achieve such growth, the levelized cost of energy (LCOE) needs to be reduced. Currently the LCOE is still higher than alternative renewable energy sources such as solar energy, or onshore wind. The aim to lower the costs of offshore wind energy requires design optimization and efficiency gains in many different aspects within the offshore wind sector. Examples of areas of interest are wind turbines, foundations, electri- cal installations and operation and maintenance. The remote location and harsh conditions of offshore wind

1 2 1. Introduction

Figure 1.2: Global annually installed offshore wind capacity (source: IRENA, 2016) farms increases the complexity and requirements of most of the components. Rough sea conditions cause complex loads to be exerted on the structures, requiring special designed platforms and turbines. Wind is usually stronger at sea, changing the design parameters of the turbines. Also construction and maintenance of an offshore wind farm is relatively complex and expensive compared to onshore wind farms. Special- purpose built ships are required to get the parts out on sea and in place. With an average lifetime of 20-25 years [4], operations and maintenance (O&M) becomes an important factor in the total LCOE. Offshore op- erations are costly compared to onshore due to the remoteness of these farms. One of the challenges here is to minimize unnecessary expensive journeys to and from the wind farm. Weather conditions play an impor- tant role in the possibilities, cost, and planning of the O&M. Combined with failures that require last-minute changes in activities, it is necessary to have a day-to-day planning that is taking into account many different factors. Good strategies can minimize production losses and O&M costs.

O&M responsibility is currently split between wind farm owners, turbine manufacturers, offshore transmis- sion owners and sometimes independent service contractors. This makes the already difficult process of operations even more complicated. There is demand for an approach to make the O&M more effective. Fur- thermore, the availability of cheap, reliable sensors and the general recognition of the value of data has lead to an explosion in interest in the area of “big data”.

This paper will give an overview on the latest situation in O&M of offshore wind farms and provide sugges- tions for future improvements. In Chapter2 the current situation in The Netherlands will be described and an overview of the future plans of offshore wind until 2023 will be given. Chapter3 will explain which stake- holders and what activities belong to the operations and maintenance aspect of offshore wind farms. The concept of "big data" and how it relates to offshore wind farms will be described in Chapter 6.2. Operations and maintenance will then be further analyzed in Chapter4 on the basis of the following aspects: supply chain management, wind turbine monitoring and marine operations. In Chapter5 some examples will be given of past and on-going projects on O&M optimization. in the field of operations and maintenance. Lastly, some opportunities for future developments will be discussed and the research will be concluded in Chapter6. 2 Offshore wind in The Netherlands

The Netherlands is working on a more sustainable energy production. In 2013 an energy agreement has been set for an expansion of our current four offshore wind farms with an additional five wind farm zones of 700 MW each. This chapter will describe the current situation and the future plans up to 2023 in The Netherlands .

2.1. Current situation Currently there are four wind farms operational in The Netherlands together generating almost 1000 MW. The four wind farms are mentioned below, together with the corresponding manufacturer of the wind turbines and the total power output, see Table 2.1[5] A map of the current and future wind farms in The Netherlands can be found in Appendix A.

Table 2.1: Overview of wind farms in The Netherlands [6][7][8][9]

Turbine #Turbines x Windfarm Owner(s) Total capacity manufacturer capacity Offshore Windfarm Egmond aan Zee Nuon(Vattenval) and Shell Vestas 36 x 3 MW 108 MW (2006) Princess Amalia Wind Farm Eneco Vestas 60 x 2 MW 120 MW (2008) Wind farm Luchterduinen Eneco and Mitsubishi Vestas 43 x 3 MW 129 MW (2015) Corporation Gemini wind farms Northland Power, Siemens, Siemens 150 x 4 MW 600MW (2017) Van Oord and HVC

The Gemini wind farm recently opened (May 2017) and is one of the biggest offshore wind farms in Europe with a total capacity of 600MW. [9] The five new wind farms that are going to be constructed in the coming years will each be 700MW.

2.2. Roadmap to 2023 In 2013 a big step towards sustainable energy supply has been made. The government and the Social and Economic Council of the Netherlands (SER) have set up The Energy Agreement for Sustainable Growth to- gether with over 40 employers, trade unions, environmental organizations and others, containing provisions on energy conservation, boosting energy from renewable sources and job creation. [10] Part of this agreement was the goal of 16% sustainable energy by 2023 (in 2013 this was 4.4%). To reach this goal, the following plans related to wind energy are made:

1. Scale to 4450 Mw by 2023 through yearly phased procurement procedures.

• Assuming cost of offshore wind will reduce 40% in the years ahead

3 4 2. Offshore wind in The Netherlands

Figure 2.1: Annual tendering plan (source: SER, 2013 [10])

• Assuming wind farms are operational within 4 years after decision of funding • Assuming state-of-the art technology

2. Scale onshore wind power to 6000 Mw by 2020

An offshore network will be constructed in case this is more efficient than connecting wind farms directly to the national high-voltage network. Responsibility for this will be allocated to TenneT. Progress towards achieving the 14% target for 2020 and the 16% target for 2023 will be assessed according to a set of clearly defined criteria. [11]

The process is as followed: 1. Annual tendering of 700 MW 2015-2019 Starting in 2015, every year a wind farm site of 700 MW will be made available for companies through a tendering procedure (see Figure 2.1). The government has set a maximum price per MWh they are willing to subsidize, the lowest bidder will be granted the relevant site. [3] 2. Government provides preliminary work The government will provide all necessary information for the interested parties to be able to do a bid on a wind farm site. They investigate the physical environment of the site on aspects such as soil- wind- and water conditions. This information includes (a) geological, morphodynamical and geomorphological data, (b) archaeological and unexploded ordnance analysis, (c) metocean data, (d) wind resource assessment, and (e) geophysical and geotechnical data (based on surveys). [3] 3. TSO TenneT realises grid connection The national electricity Transmission System Operator TenneT will construct five standardised platforms with a capacity of 700 MW each within the wind farm zones. TenneT will also take care of the grid connections to the national grid. [5] 4. Government assigns area to lowest bidder Government gives a certain time window in which parties can make an offer for which price per MWh they are willing to supply the energy. The lowest bidder will be assigned to the new site. [3]

Recently the first area Borssele has been assigned to different bidders: Borssele I en II Danish company Dong Energy offered to deliver at a subsidy price of 7,27 cent per KWh. According to our minster of Economic Affairs H.G.J. Kamp this means the condition of 40% price decrease of offshore wind has already been met. Dong Energy is a well-known party with currently 3.000 MW of offshore wind operational and 3.300 MW in process of building. End of 2015 Dong Energy was the biggest owner of offshore wind farms in Europe with a share of 15,6%. The Danish state holds a majority interest in the company. Borssele III en IV Borssele IV and V: assigned to Blauwwind II c.v. (consisting of Eneco, diamond Generation Europe (Mitsibishi), Shell and Van Oord. Borssele V: Innovation site Government wants to provide an oppertunity for Dutch businesses to test and demonstrate state-of- the-art technologies. In this way contribute to cost price reduction of wind energy, Dutch economy, knowledge building in The Netherlands. 3 Offshore wind operations and maintenance

Operations and maintenance covers all activities after completion of installation work up to the start of de- commissioning. Many actors are involved in the construction and operations and maintenance of an offshore wind park. During the construction phase most of these actors each have their own specific task (e.g. wind turbines, platforms, electricity grid). However, during the operations phase there is much more overlap in these tasks. Actors often have to work together on the same task, either simultaneously or sequential. Dif- ferent actors can have different interests making it sometimes difficult to determine the best strategy. This chapter will give an overview of the different stakeholders and the relations between them. Furthermore, the O&M activities will be explained.

3.1. Stakeholders The key actors involved will be described, based on the work of GL Garrad Hassan (2013). [12] There are three main actors involved:

• Project owners

• Wind turbine original equipment manufacturers (OEM)

• Offshore transmission owners (OFTO)

Project owner The project owner is in charge of all operational services associated with the offshore wind farm up to the point of interface with the OFTO. In most cases, as in The Netherlands, the substation is part of OFTO’s re- sponsibility. The owners decide on operational strategy and maintenance contracts. The owners are usually large organizations (and collaborations thereof) such as Vattenval, Eneco, Shell and DONG Energy. Turbine manufacturer The turbine manufacturer are often not only supplying the turbines, but also responsible for the mainte- nance of the turbines during the warranty period (usually 5 years). Depending on the contracting strategy of the project owner, either the manufacturer stays responsible for the maintenance after this period or other outside independent companies can be contracted. In The Netherlands the leading turbine manufacturers are Vestas and Siemens. Offshore transmission owners The offshore infrastructure and wind turbine platforms are owned by the OFTO. These can in turn contract others for the maintenance, sometimes this is the project owner itself. In The Netherlands TenneT is the only company in charge of all offshore infrastructure. [3]

3.1.1. Contracting types Energy production The wind turbine manufacturer usually gives an availability guarantee per year. For example, a guaranteed

5 6 3. Offshore wind operations and maintenance availability of 95% on average per year over the total wind farm. When it turns out to be higher this is rewarded with a bonus, in case it turns out to be lower a compensation penalty needs to be paid. [13] Maintenance As already mentioned, when the main warranty on the wind turbines expires, the project owner can decide on its strategy. There are three possible options (or a combination of thereof): renew the maintenance contract with the OEM manufacturer, take all O&M in-house or contract the O&M to an independent service provider (ISP). "Hands-on" or "hands-off" approach The approach of a project owner on their operations and maintenance strategy can be classified as a hands- off or hands-on approach. A hands-off approach is when the project owner is relying on several independent contractors to look after the project. Usually a contract is made with the turbine manufacturer to be respon- sible for all day-to-day operation management. With a hands-on approach the project owner is taking own responsibility for most or all of the operations, and performing all operations and activities in-house working in partnership with the turbine manufacturer, vessel operators and others. This can minimize the operation expenses but comes with more risk. An example is Dong Energy, owning and operating marine assets such as work boats and providing technicians and other personnel. Maintenance is performed by own personnel, but in cooperation with the turbine manufacturers. The approach can also be somewhere in between fully hands-on or fully hands-off, where a contract is placed with the turbine manufacturer to do only the mainte- nance on turbines. Other specialized independent contractors are contracted for different parts of the wind farm. In this case the wind farm owner provides some or all of the operations management. [12][1] According to the ’Innovation Outlook Offshore Wind’ published by IRENA (2006) debt providers often insist on a more hands-off approach in order to minimize risk, but for turbine maintenance the hands-on approach is now becoming the state-of-the-art approach for many wind farms which have utilities as majority shareholders. [1]

3.2. Activities O&M activities will be explained in the following two paragraphs.

3.2.1. Operations Operations has to do with the management of all the tasks involved in keeping the wind farm running, and management of use of personnel and equipment to perform these tasks. This includes remote monitoring of the wind turbines, environmental monitoring (weather conditions, wave height), electricity sales, marketing, administration and other back office tasks. [14] In order to do so, the operations department has access to real-time and historical data for wind turbines, substations, crew and vessels. With this information a planning of scheduled and unscheduled activities can be made on day-to-day basis, taking into account the costs of operations, spare parts availability, available crew and weather conditions. [12]

3.2.2. Maintenance The majority of O&M expenses, effort and risk goes to keeping the wind farm and its systems working and repairing it when something brakes down or fails. Hence, good and efficient maintenance is very important. Maintenance can be classified into two types: corrective maintenance and preventive maintenance. Some examples of maintenance tasks can be find in Table 3.1.

Corrective maintenance is the repair or replacement of failed or damaged components after the failure has occurred. This results in the wind turbine having downtime in which no energy is being generated. As down- time is very expensive, this method is avoided as much as possible.

Preventive maintenance tasks are performed prior to failure. This type of maintenance can be divided into time-based maintenance and condition based maintenance (also called scheduled maintenance). Time- based maintenance is a scheduled activity and does not require active monitoring of the equipment. How- ever, in order to avoid failure a certain safety margin needs to be kept. This results in parts being changed before they actually reach the end of their lifetime. Therefore, this maintenance strategy is unnecessarily costly. [15]. Condition based maintenance solves this by monitoring the actual state of the equipment. Mon- itoring the wind turbine requires expensive equipment and analysis tools, but in the end condition based maintenance might be the most economical option. It can save on costs for parts and maintenance oper- 3.2. Activities 7 ation expenses which are relatively high due to the more extreme conditions and remote location of these offshore farms.

Maintenance Tasks

• Wind turbine maintenance

– Scheduled inspection of blades, lubrication, hydraulics, replacement of parts – Unscheduled maintenance – Provision of parts – Training of technicians

• Transfer cable maintenance (partially by OFTO, partially by farm operator)

– Detection of cable exposure, inspection of depth of cable burials

• Offshore substation (OFTO)

– Inspection of transformers, switchgear, foundation and topside structural inspection

• Onshore substation (OFTO)

– Inspection of tranformers, switchgear etc

Table 3.1: Examples of maintenance tasks (source: UEA, 2015 [14])

Operations and maintenance of offshore wind farms can be categorized into three fields:

• Supply chain management

• Wind turbine monitoring

• Marine operations

Supply chain management is the management of all the flows of materials, parts, equipment and storage thereof. The aim is to determine the most optimal strategy of these flows. There is limited literature available on the topic of supply chain management specifically related to offshore wind farms. Big data is not being mentioned in relation to supply chain management.

Wind turbine monitoring has to do with gathering data on the wind turbine itself for both performance con- trol and condition based maintenance. A lot of research is going in this field at the moment. Big data plays an important role in this field.

Marine operations include all operations that are necessary to get workers and equipment to and from the wind farm. Onshore planning of maintenance activities, bringing workers and equipment to the offshore wind farms and vessel management are all part of it. Although much less researched, marine operations is also an interesting field for data collection and analysis. [14]

These three categories will be explored in detail in Chapter 4.1, 4.2 and 4.3 respectively. First some theoretical background on ’big data’ and the use of data in offshore wind will be given in Chapter 6.2.

4 Role of big data in offshore wind

There is a lot of on-going research related to offshore wind operations and maintenance and how to reduce the costs of it. One of the recurring themes is the use of data as an analysis tool to create a better insight in O&M activities and decision making processes. The data coming from these wind farms, however, are of such gigantic quantities that this data cannot be handled with standard processing solutions. This is where the term ’big data’ is all about. According to Michael Borges from Vestas, "’Big data’ is not just a buzz word. It is something that companies are massively investing in right now, and there is no question that this is a pure gold mine for companies seeking to achieve greater efficiency. Leading companies no longer make decisions based on assumptions. They make decisions based on facts." [16] For a more scientific definition of ’big data’, see Appendix B. The role of data in the offshore wind sector will be explored on the basis of the three themes as mentioned in previous chapter: supply chain management, wind turbine monitoring and marine operations.

4.1. Supply chain management Supply chain in the offshore wind sector refers to "the management of the flows of goods and services in- cluding the movement and storage of components with the objective of efficiently delivering a completed project according to plan." [17] In order to do so, it requires sufficient planning early in the project develop- ment to choose the most optimal strategies, such as which ports and vessels to use and which methods of transportation and installation to use.

4.1.1. Supply chain in offshore wind industry Federico D’Amico et al. (2017) provide a simplified introduction to the complex supply chain that goes with the installation and operation of an offshore wind farm, which can involve over 40 firms. As example the is used, a 175 turbines farm with a total capacity of 630 MW. The offshore wind supply chain consists of three principal phases: supply, construction and management. These phases should not be seen as sequential phases, as they overlap during the lifetime of the wind farm. In the first phase consultants offer environmental, technical and financial advice. Suppliers are responsi- ble for the supply of offshore wind farm components, such as turbines, foundations, cables and substations. E.g. turbine suppliers produce the turbine components, cable suppliers produce array and export cables and substation suppliers provide substations. The second phase includes assembly, installation and commission- ing of the offshore wind farm. This phase mainly involves port operators, installers and vessel suppliers. The wind farm operators are responsible for the overall management of the farm and the day-to-day manage- ment. This usually is the wind farm owner, although part of the day-to-day operations like maintenance can also be the responsibility of the turbine supplier, as for example the case in the new Gemini wind park in The Netherlands. [13] As the supply and construction period can take quite some time (4 years in case of the London Array), high investments are needed over a longer period before the wind farm is operational and can return on investments. This problem can be solved by time compression strategies where some clusters of turbines are commissioned and become already operational while other clusters are still being installed. The main resource availability problems for the offshore wind supply chain are vessel availability, capacity of staging areas and suppliers’ capacity. The new trend of offshore wind farms getting bigger and bigger makes

9 10 4. Role of big data in offshore wind the supply chain even more complex. [18]

4.1.2. Purchasing and supply management practices Literature defines three purchasing and supply management practices that address supply chain complexi- ties [18]:

• Make-or-buy decisions

• Contract forms

• Local-to-global sourcing decisions

F. D’Amico et al. (2017) identified "three particularly innovative practices that can address the increased complexity of supply chains resulting from growing wind farms." [18] One in each of the three supply man- agement practices mentioned above. An explanation of each of these three practises with the corresponding innovative practise as advised by F.D’Amico et al will be provided below.

Make-or-buy decisions Make-or-buy decisions deal with whether company performs a process in-house or outsources it to an ex- ternal contractor. Reasons to take processes in-house can be the avoidance of double marginalization and to increase their control over manufacturing processes. Reasons to outsource certain processes can be because external transaction costs are lower than the internal ones or capabilities are inferior to those of potential suppliers. The scale of offshore wind projects is growing fast, and F. D’Amico et al. (2017) argue that in terms of make-or-buy decisions this trend should be associated with more outsourcing. There is a high degree of vertical integration in the offshore wind industry. This has to do with the fact that until recently there were only a few companies with expertise in this industry. Furthermore, the suppliers have the tendency to build everything in-house rather than outsource certain components in order to protect their intellectual property. Turbine supplier Senvion is a good example of a company outsourcing almost all of the turbine components and possessing mostly the intellectual properties. Contract forms Contract forms refer to the conditions of delivery of products or services. By signing contracts, buyers and suppliers agree on the price, quality and time of delivery of products or services. They also determine how to share risks if anything does not go according to plan. In the earlier stages of new product development there are still many uncertainties making it difficult to write contracts with exact specifications. By sharing the risks through mechanisms such as buyback, revenue-sharing and quantity-flexibility contracts, suppliers are getting increasing responsibility for supply chain processes. Different contract forms can be applicable depending on the collaboration between multiple companies. The owner of a wind farm can take control in the developing and operating tasks in own hands, or contract outside parties to do most of it. In that case responsibilities and operational risks of each of the involved companies should be clearly identified, in order to share the risks between them. Local-to-global sourcing decisions Local-to-global sourcing decisions have to do with identifying, selecting, evaluating and engaging with sup- pliers. Working together with companies within the same country can be defined as local sourcing. This is not always possible, forcing buyers to sign contracts with global suppliers. In case of uncertain demand, high customer service, high costs of expediting shipments or more complex product manufacturing this is not the preferred option.[18] For the local-to-global sourcing decisions D’amico et al. recommends the use of a more local sourcing strategy. Sourcing from well-established suppliers from all over Europe is currently a popular way to go. However, this strategy increases logistical challenges.

4.1.3. Supply chain assessment The Netherlands In 2014 the Dutch organization TKI published a supply chain assessment performed by GL Garrad Hassan to see whether problems can be expected in the future with the planned growth of the offshore wind sector. A prediction of the growth can be found in Figure 4.1.[19] An assessment is made of 19 elements of the offshore wind supply chain, categorized into 5 categories. An overview of the results of the assessments can be found in Figure5 of Appendix C. 4.1. Supply chain management 11

Figure 4.1: Annual European offshore wind installation rate - MW (source: GL Garrad Hassan, 2014 [19])

From the assessment it can be concluded that the following items are identified as possible bottlenecks:

• Turbine assembly

• Export cables

• DC offshore substations

• Jacket foundations

• Foundations installation vessels

• Operations and maintenance technicians

4.1.4. Big data in supply chain management The use of big data in supply chain management is not being mentioned in related research. Big data analysis has no function here as there is no data being collected in extremely high quantities and the data that might be available can be processed with simple methods. 12 4. Role of big data in offshore wind

4.2. Wind turbine monitoring Monitoring of wind turbines has become an important aspect of O&M in the shift towards doing mostly con- dition based maintenance instead of time-based or corrective maintenance. Condition based maintenance requires the monitoring of information on the actual state of health of the wind turbine, also called condi- tion monitoring (CM). There are monitoring systems which are installed by the OEM manufacturer (SCADA) and systems that can be added later as add-ons (CMS). All large utility scale wind turbines are standard al- ready equipped with a supervisory control and data acquisition (SCADA) system, fitted by the manufacturer to monitor the wind turbine performance. These monitoring systems were initially not meant for monitoring the health state of the wind turbine. Independent companies introduced the CM systems with the primary goal to monitor health parameters such as drive train vibrations, oil quality and temperatures in some of the main assemblies. [20] The downfall is that these additions can be quite expensive (11.000 Euros per turbine). Current research is working on a solution to use these SCADA data not only for performance measurements, but also health monitoring. This will reduce the required add-on equipment. As wind turbine monitoring is not only interesting for offshore but also onshore wind farms, there is al- ready a lot of on-going research on this topic. [21][20][22][23][24][25] The details on these systems are beyond the scope of this research. However, a short description on the state-of-art on both CMS and SCADA as CM will be given.

4.2.1. Wind Turbine Maintenance Usually wind farm owners agree to have the turbine maintenance performed by the turbine manufacturers for the initial warranty period of 5 years with a possible extension to 15 years. A wind turbine needs to be visited once every 6 or 12 months for scheduled maintenance. When the SCADA or an additional CM system sets off an alarm, extra unscheduled visits are performed. These condition monitoring systems are more and more used not only for real-time measurements but also for prediction of equipment failure. The better the prognosis of failure, the sooner necessary parts and equipment can be mobilized. Owners can choose to do this prognosis with in-house tools or pay specialist contractors to provide this service. The techniques are likely to become more accurate at prognosis, minimizing lost generation. According to the ’Innovation Outlook Offshore Wind’ published by IRENA (2006) this is likely to become business as usual for offshore wind farms by the early-to-mid 2020. [1] Monitoring tools are being developed in-house by turbine manufacturers such as Vestas, but also by third-party service such as Romax (UK). Examples of operators that have developed their own in-house models using SCADA and other additional CM systems to prioritize the maintenance tasks are E.ON (UK/Germany), Scottish Power (UK) and Vattenfall (Sweden). [1]

4.2.2. Monitoring system: CMS Research by Ahmed and Kamaruddin (2012) shows that 99% of equipment failures are preceded by certain signs, conditions, or indications that such a failure was going to occur. [26] This makes the use of condition based maintenance very valuable compared to traditional time-based maintenance. Early fault detection makes it possible to do maintenance in-time before actual failure is going to occur. In order to be able to use these indications a three step process is performed as shown in Figure 4.2. First step is data acquisition. This involves the use of sensors to measure certain parameters like vibrations and oil viscosity. This data then needs to be processed and diagnosed to detect if a fault is developing and where it is located. The last step is the prognosis of the remaining useful life. The challenge in the field of offshore wind is that due to strongly varying loads and more complex forces, it is difficult to make a reliable prognosis.

Figure 4.2: Three-step process of CMS-data utilisation for condition-based maintenance (source: Coronade & Fischer , 2015 [21])

The most used condition monitoring technique in the additional CMS is vibration-based monitoring of the 4.2. Wind turbine monitoring 13 drive train components. They are recommended as standard equipment for most offshore wind turbines. Other often used CMS techniques are complementary particle-counting system to monitor the quality of the gearbox oil an structural-health monitoring of rotor blades and support structures. [21]

4.2.3. Monitoring system: SCADA As additional CMS is still expensive, the use of already available SCADA data from the OEM installed sensors is being considered to help in the monitoring of the actual state of health. However, according to a state-of- the-art review in 2015 condition monitoring based on 10min-averaged SCADA data is found unsuitable as a standalone solution. [21] Nonetheless, when combined with purpose-designed CMS it can be a valuable component to achieve better fault-detection. Parameters that are usually recorded for SCADA are listed in Table 4.1. These parameters are typically recorded with a frequency of one sample per second and then averaged over 10 seconds. This information is stored, together with a maximum, minimum and standard deviation. As can also be seen in the table, vibration sensors are usually not part of the SCADA system and need to be added as additional CMS. [20]

Environmental Electrical characteristics Part temperatures Control variables wind speed active power output gearbox bearing pitch angle wind direction power factor gearbox lubricant oil yaw angle ambient temperature reactive power generator winding rotor shaft speed nacelle temperature generator voltages generator bearing generator speed generator phase current main bearing fan speed / status voltage frequency rotor shaft cooling pump status generator shaft number of yaw movements generator slip ring set pitch angle / deviation inverter phase number of starts / stops converter cooling water operations status code transformer phase hub controller top controller converter controller grid busbar

Table 4.1: Basic SCADA parameters (source: Tauts-Weinert and Watson, 2017 [20])

Analysis methods According to Tauts-Weinert and Watson (2016) different analysis methods have been developed. These meth- ods include (i) ‘trending’, (ii) ‘clustering’, (iii) ‘normal behaviour modelling’ (iv) ‘damage modelling’ and (v) ‘assessment of alarms and expert systems’. Trending methods considers data stored over a longer period and looks at ratios of SCADA parameters and how they change over time. The downside is that is requires visual interpretation which can become problematic when monitoring a large fleet of wind turbines. [20] The clustering method is using complicated algorithms to compare values of new input data to previous stored ’training’ data. Interpretation has proven to be difficult so the analysis shifted to the normal behaviour modelling method. With the NBM method, the measured data values are compared to a target value. The difference between these values should normally be around 0, with a certain tolerance. Exceeding this tolerance means changed condition or failure. Multiple studies have proven that NBM method can be used to detect failures, but it depends a lot on the used target values. These values depend on training data and manually set thresholds, which can result in undetected failures or frequent false alarm. Next step is the damage modelling approach, which is making use of physical models to interpret the measured data instead of using empirical data. Assessment of alarms and expert sys- tems are focusing on reducing the number of unnecessary alarms and interpreting the necessary alarms on how urgent they are.

4.2.4. Opportunities Today, most maintenance activities are performed on-site. However, as automated and remote maintenance is gaining in other sectors, it is being adapted for offshore use. An example is the use of drones for aerial in- spection of wind turbine blades. Innovations in this area can reduce costs of energy through lower personnel 14 4. Role of big data in offshore wind and transportation costs. [1]

4.2.5. Big data in wind turbine monitoring Data analyses is the basis of conditional monitoring. The role of big data in wind turbine monitoring is there- fore of huge importance. Especially in the field of failure prognosis still a lot of research is required in order to make this it more reliable. 4.3. Marine operations 15

4.3. Marine operations Marine operations include all operations that are necessary to get workers and equipment to and from the wind farm. This includes onshore planning of maintenance activities, bringing workers and equipment to the offshore wind farms and vessel management. Marine operations is also an interesting field for data collection and analysis. Marine operations constitute 10-15% of the total O&M costs, and according to Bagnall et al. (2015) it has high potential for immediate efficiency gains. [14] Their report is one of the few literature with a focus on how big data can help in the marine operations instead of wind turbine monitoring. First, a short introduction to the offshore logistics.

4.3.1. Offshore logistics For short distances to wind farms, transport of technicians, equipment and spare parts from the onshore operations base to the offshore wind farm is usually provided by several PTVs. These custom designed vessels carry up to 24 passengers and 20 tonnes of spares and equipment. Most vessels have strong aluminum hulls, however, the use of cheaper and lighter fibreglass vessels is becoming more popular. [1] Often independent service providers are contracted for the supply and operation of these vessels. Depending on the distance of the wind farm to the shore, wind farm owners can choose to have their operations base onshore or offshore. The transition point is around 50 kilometers. In case of far-offshore wind farms, the addition of helicopters is an often used strategy for quick access to the wind turbines. Another option, introduced by Siemens, is the use of a larger Service Operations Vessels (SOVs). These vessels offer accommodation for technicians and are fitted with hydraulic bridges to access the wind turbines under a greater range of sea conditions. [1] These vessels can stay out on sea for extended periods (e.g. 2 weeks in case of the Siemens vessel at Gemini Wind Farm [9]), providing a fast reaction time even with far-offshore wind farms. The greatest impact on reducing the LCOE in O&M since 2001 were from improvements in personnel transfer and access systems. [1]

4.3.2. Case study by UEA The University of East Anglia (UEA) has published an independent report commissioned by James Fisher and Sons plc, a marine service company in the UK. The most important findings in this report will be summarized below. Vessel monitoring Vessels are a necessary tool in offshore wind farm maintenance. When a wind turbine needs maintenance, the workers and equipment are brought to the wind turbine platform with vessels. Usually standard crew transfer vessels are sufficient, but in some occasions of major repairs or part replacements specialized jack- up vessels are needed. Wind farm operators that prefer a ’hands-on’ method owns their own vessels and maintenance crew. However, vessels can also be leased on a long term basis, or sometimes even chartered daily. Vessel cost makes up around 15% of the total marine operations costs. Fuel costs would be around a third of this. The vessel journey can be split into two components: the trip to and from the wind farm and push-on events at the wind turbine platforms. These push-on events are during the transfer of the crew and equipment. During this transfer, the vessel is "pushing on" to the turbine at high throttle. When multiple turbines are being visited on a trip, this involves large amounts of fuel use. The skipper is responsible for most of the decisions on the vessel, including route, speed, and method for the push-on events. Most decisions are based on the experience of the skipper. The vessel route, speed, fuel usage and all other related data is often not monitored. This results in the operator having no insight in what is happening with the vessels, except for a fuel bill at the end of the day. If data would be available on vessel locations, speed and fuel usage, these processes could be optimized. For example, the fuel usage depends a lot on the speed of the vessels. Analyses of this data can help in finding an optimum in speed vs fuel usage. With additional data like vessel motion, weather data and wave height, this optimum can even be conditional to the current conditions. There are already new ship designs where these push-on events are avoided with the use of motion compensated gateways. [27][28][29] Data collection and analysis could lead to proactive management of sailing strategy and saving on unnec- essary fuel costs, instead of having no clue on the exact costs of vessel operations. Integrated system for data storage and visualization Another problem reported by the UEA was that currently information systems that are already available (for weather information and information on the vessels, staff and equipment) are all provided through its own system. This makes it complicated to use, especially in decision making processes. The desire from the operators is to have one integrated system combining all information sources. This can help in providing a 16 4. Role of big data in offshore wind better oversight. One of the difficult but important decisions operator have to make is whether to call a "weather day" or not. A "weather day" is when a vessel cannot go out to the wind farm due to the heavy wind and high waves, making it to dangerous to transfer the crew and to work on the wind turbine. However, in case of corrective maintenance, not going out to the wind turbine means that the wind turbine is not generating energy for a whole day. According to the report by UEA this can cost around 4000 British pounds each day the wind turbine is not working. The maximum safe wave height is particularly relevant to the transfer of the crew. If the waves are too high, it is not safe for the crew to get to the wind turbine platform. These decisions are not based on only one variable, f.e. wave height. They are much more complex and decision makers have to account for parameters such as current weather, weather forecast, sailing time, operation time, different wave heights at different location within the farm, availability of the staff and vessels. This decision is even more complex when there is not a good understanding of the exact costs. Right now, operators have to weigh operation costs against the cost of lost generation mostly based on experience.

Two weaknesses have been identified by the UEA in the current method of operation:

• The decision is not explicitly quantified in terms of cost.

• The huge amount of weather data is not quantitatively incorporated into the decision.

Decision tool An example provided by UEA is a simple overview of resulting costs depending on the choices made. Imagine a case where there is a wind turbine that needs to be fixed in order for it to work again. The weather is in such a condition, that is not sure whether the maintenance activities can be performed. The operator now has to decide whether to go to the farm or not. If the vessel is send out and the turbine can be fixed, the only costs are the operation costs (in this example £4000). If it turns out to be a wrong decision, the vessel has to return and total costs are operation costs (£3000) plus lost generation of one day (£4000). If the operator decides not to send the vessel but it turned out that in fact it was possible to perform the required maintenance, it is an unnecessary lost day of generation (£4000). If it was the right decision not to send the vessel, it is an unavoidable extra day’s lost generation (£8000). See figure 4.3.

Figure 4.3: Example of decision table (source: UEA, 2015 [14])

When this decision table is extended with probability factors, a better insight in the estimated costs can be achieved. For example, when considering all variables the estimated chance to get safely to the farm and fix the turbine is 50:50. The decision table will then look as followed, see Figure 4.4. In order to calculate the right probability factor, the data of all sources like weather forecast, current wave height, vessel information can be combined and analyzed. Even more reliable estimates can be obtained when also using historical data from these sources and learning from the decisions made in the past. This example is of course an oversimplification of the reality, but it shows how decision tools can be help- ful. In order to make a decision tool, complex data analysis and a large amount of factors need to be imple- mented. This is definitely a ’big data’ problem.

4.3.3. Operations at Gemini A personal interview with Bart Hoefakker, the Operations Manager of the Dutch wind farm Gemini, has been conducted to get more insight in the current work method of recently constructed wind farms. A summary of the findings will be presented here, the full interview can be found in Appendix D. 4.3. Marine operations 17

Figure 4.4: Example of decision table with probability. (source: UEA, 2015 [14])

Gemini is a so-called special purpose company. A one-time company with a unique shareholders structure. There are 4 shareholders: Northland Power (60%), Siemens (20%), Van oord (10%) and HVC (10%). Siemens is responsible for the maintenance of the wind turbines, with a contract of 15 years. They are maintaining the park, resolve malfunctions, and decide on the use of their logistic resources. Gemini just keeping an ’oper- ational oversight’. In the early start of this project, Siemens has made some assessments with the help from ECN on best strategy for accessibility. Outcome was their current maintenance fleet of one big mother vessel and a helicopter. Every day it is determined what activities are happening and where they take place in the park. Currently, this is a manual job and there is no decision model helping with advice on the best options. The challenge is how much unscheduled maintenance there will be today, tomorrow, and the days after. How many teams are needed to keep available? Based on historical data Siemens can do a reasonably estimate but they cannot fully change the planning per day as the crew is on-board of the big vessel for 2 weeks straight. So for Siemens, who is working with daily decisions based on stock, logistics, weather, people, malfunctions, the help of a decision tool might be a very interesting area.

Research on the development of such a decision tool has been done by a master student from the Delft University of Technology (H. Koopstra, 2015) [30]. Another project, DAISY4Offshore is working on the op- timization of O&M with the use of sensors and providing an integrated management system to analyze, control and optimize the maintenance and inspections of wind turbines. More on Koopstra’s thesis and the DAISY4Offshore project can be found in the next chapter.

4.3.4. Opportunities Many of the opportunities as pointed out in the innovation outlook by IRENA [1] are related to the marine operations section. This also indicates that especially in this area there is still a lot of room left for improve- ments. The following opportunities currently have the most potential: Improvements in weather forecasting and analysis One of the opportunities as was improvements in weather forecasting and analysis. Currently, weather fore- casting can be used as a reliable source of information only for five days ahead. However, for optimal use of resources reasonable accuracy needs to be extended to 21 days. In a more ready-to-use format, the forecasts should provide weather windows instead of indicators such as wind speed and wave height. A better forecast can reduce the costs of energy due to less down-time, better planning of maintenance schedules and more efficient use of vessels and personnel. Improvements in OMS strategy for far-offshore wind farms As the sites for short distance offshore wind farms are getting used up, wind farms are moving further away from the coast. Strategies have to adapt in order to keep up with this new trend. The use of SOVs will be optimized further and can become interesting for sites currently still using an onshore maintenance base. At a certain point, fixed offshore bases may become cost effective for some projects as well. Improvements in personnel transfer and access Improvements of PTVs will result in larger passenger capacity, more comfort and greater payload capacity. Innovations for better access such as the heave compensated walkways by Ampellman [27] can increase ac- cessibility from about 70% to 95% and are likely to become standard on larger SOVs and some PTVs. Improve- ments in this sector will lower the cost of energy by lowering the cost of both planned and unplanned OPEX and reducing availability losses. Introduction of wind farm-wide control strategies The development of more integrated control strategies can improve the control of wind farms and help in 18 4. Role of big data in offshore wind decision making in order to minimize the cost of energy. Collaboration is needed between turbine manufac- tures, operators, external consultancies, research organizations and technology organizations (such as ECN).

In particular, a lot is expected from the implementation of condition-based maintenance in combination with wind farm-wide optimizing control strategies. Also the improvements and innovations of the PTVs and SOVs are seen as important developments. [1] Some projects on wind farm-wide optimizing control strategies will be discussed in the next chapter.

4.3.5. Big data in marine operations In the field of marine operations, big data does not yet play an important role. However, there is an important link between big data analyses related to the wind turbine monitoring and marine operations. As the outcome of the first, will result in actions of the latter. It is important to combine these two field in order to improve the total offshore wind operations. Furthermore, specific data related to marine operations such as vessel data could also help in making these processes more efficient. 5 Projects on optimizing control strategies

Optimization of operations and maintenance is a rising demand in the offshore wind sector. There have been several projects attempting to develop a tool that gives insight in all incoming data and helps in the decision process. Some of these attempt incorporates the help of an integrated system to optimize the maintenance control strategy. The most recent projects will be described in this section.

5.1. ECN ECN has developed the O&M Cost Estimator (OMCE) to estimate the the O&M effort in the operational phase. The tool is programmed in MATLAB. However, this tool is only useful for what-if analyses and is not able to do any optimization. The tool has no planning aspect integrated and condition based maintenance is not included either. [31] The latest news on this tool dates from 2014 and no updates have been released after that.[32]

5.2. Decision support tool by H. Koopstra In 2015 a student from the Delft University of Technology (H. Koopstra) has made a start on an integrated decision support tool for the offshore operations industry for his master thesis. The outline of his thesis will be discussed in this paragraph. Koopstra states that previous research shows a lack of an approach that is:

• integrated: including all the stakeholders’ requirements that are necessary to assess the effectiveness of the O&M, and

• generic: applicable for all wind farms.

Therefore, the goal of Koopstra’s thesis was to develop an integrated and generic approach that enable users to find the optimal strategy in the production, costs and planning trade-off triangle. [31] The integrated decision support tool is specified as Discrete Event System Specification (DEVS) tool. It has the following specifications [31]:

• use stochastic parameters or state variables for e.g. weather conditions (wave and wind), failure rates, repair time and transfer times,

• to include the resource (man and material) utilization to test their (un)availability and optimize their capacity,

• include queues which make their able to test the (waiting) time within the entire process and identify bottlenecks,

• to combine simulation and P&S,

• to test different O&M strategies, and

• provide animation in order to get insight in the process.

19 20 5. Projects on optimizing control strategies

Figure 5.1: Requirements of Koopstra’s decision tool (source: Koopstra, 2015 [31])

Koopstra identified 7 model requirements through interviews with Director Operation UK at Vattenfall and Service Manager Offshore at MHI Vestas Offshore Wind, as shown in Figure 5.1. Each requirement is divided into sub requirement forming a total list of 35 sub requirements. From multiple examined studies and existing tools, Koopstra identified 9 of these sub requirements currently to be not or only partially fulfilled. Existing models are lacking the distance between the turbines as a parameter and do not have a preventive maintenance threshold. The other seven are related to the planning aspect, to the insight in the O&M process (e.g. visualization or process animation) or to the optimization capability. [31] Besides the requirements, Koopstra identified four integration challenges: different weather components, these components together with energy based availability, the different O&M strategies together with the planning aspect and the different maintenance types (preventive and corrective) are considered as integra- tion challenges based on examined studies. [31] The software used for the implementation of Koopstra’s tool is Simio. The models made with this software package are object-orientated and shows these graphically. Risk-based Planning and scheduling(RPS) is in- cluded as a feature, which can be used to implement risk and uncertainty into the models.

The developed O&M tool features four main capabilities:

1. The resource planning of scheduled maintenance

2. The robustness of the preventive maintenance plan

3. Comparison and optimization of different strategies

4. Insight in the O&M process

The developed tool is validated by a comparison with the validated ECN tool and by a face validation. In order to improve the tool, it needs more realistic additions, overall extension of the model and improvements in the decision making process. [31]

5.3. World Class Maintenance World Class Maintenance (WCM) is a foundation which is established to boost the Dutch maintenance sector with innovations. WCM encourages ’smart maintenance’ (maintenance based on data) for Dutch industry. This is done by developing, distributing and applying maintenance knowledge. Some of their focus points: condition based maintenance, use of simulation at maintenance, remote and non-destructive inspections, data mining and maintenance and logistics. However, smart maintenance not only about data. For example, it also has to do with design of the product, the management of the maintenance and the method to perform the maintenance. The intention of WCM is to work on the whole cycle from start to end instead of only looking at the use of sensors and data analytics. WCM does this by setting up certain Field Labs. Each Field Lab has its own focus. A Field Lab consists of multiple Living Labs. This is a physical location where organizations can experiment and showcase their progress. Contributors to a field lab are: 5.3. World Class Maintenance 21

• representatives of companies that are a member of the WCM group,

• representatives of educational or research institutions,

• representatives of government or branch organizations, and

• representatives of WCM.

One of these Field Labs, called ZEPHYROS, is in the field of offshore wind. This Field Lab is currently in the start-up phase. Another similar project that WCM is involved in is DAISY4Offshore. DAISY4Offshore is also aimed at maintenance of offshore wind turbines. Both DAISY4Offshore and ZEPHYROS will be discussed below. [33]

5.3.1. DAISY and DAISY4Offshore For the project DAISY, a team has developed a fully integrated package of condition monitoring techniques to monitor any type of brand wind turbine. The system, equipment and processing software is a fully digi- tized asset management system. For now it is an automated intelligence that takes short-term maintenance decisions. In the future the system should also be able to optimize the operation and maintenance process. DAISY4Offshore is a new initiative which focuses on the offshore wind parks. This project is in cooperation with different companies such as Delta, IMS International / KEC, ECN and the TU/e. [34][35] It is working together with the Eindhoven University of Technology to tackle the big data problems. [36]

Figure 5.2: DAISY Asset & Maintenance Management System (source: Daisy Solutions [34])

5.3.2. ZEPHYROS Lately, cost of offshore wind energy has already dropped to a level where it can compete with other technolo- gies. A significant part of this cost reduction is achieved by developing smarter engineering and construc- tion. Further optimization therefore also has to come from reducing costs in other fields such as Operations & Maintenance. The focus of the ZEPHYROS field lab is exactly this. Optimizing operations, management and maintenance of offshore wind farms. As already mentioned, this Field Lab is just in start-up phase and WCM is currently looking for and assembling interested contributors. The current planning is to start with the Living Labs around first or second quarter of 2018.

6 Conclusions & future work

6.1. Conclusions Offshore wind development The offshore wind sector is gaining interest and popularity. Europe is taking a leading role in the develop- ment of sustainable power generation technologies, including (offshore) wind. With this leading role Europe is at the front of innovation and first to work on improvements. Improvements which are also necessary in order to lower the costs of offshore wind and making it a feasible primary source of energy. Therefore, all stakeholders (government, companies, research institutes, educational institutes, local authorities) need to work together to keep pushing the technologies to the next level. On every aspect improvements can be made. Wind turbines are becoming bigger and more efficient, grid connections are becoming smarter, gov- ernment legislation is evolving, offshore installations will be simplified and wind farms are becoming bigger and bigger. And especially the latter is creating a new challenge in operations and maintenance, which at the moment is usually done manually. The increasing size of offshore wind farms results in more complex operations and maintenance. Therefore, improvements in this field are vital in the goal of bringing down the levelized cost of energy.

Big data usage for OM O&M responsibility is currently split between wind farm owners, turbine manufacturers, offshore transmis- sion owners and sometimes independent service contractors. The increasing size and complexity of wind farms requires a more integral approach combined with the use of data intelligence in O&M as this can no longer efficiently be done manually. Recent developments have helped the maintenance operation by providing feedback from SCADA and additional conditional monitoring systems. However, in order to use this data to its full potential we still have a long way to go. This ’big data’ is a relative new phenomenon but already gaining popularity fast. At the moment, data is already an important indicator used in wind turbine performance measurements and conditional monitoring. But there is more potential. In the near future, big data cannot only provide current feedback, but by processing and analyzing historical data it will also be used for prediction and optimization. And by learning from itself, it will become more and more accurate.

Recent academic research Academic research tends to concentrate mainly on the conditional monitoring of wind turbines and data analysis necessary to optimize the maintenance. [20–25] In order to reduce costs of O&M, improvements are needed in O&M strategy. By combining the data from wind turbines with data from other areas such as weather, replacements parts, equipment, personnel transfer and other marine operations, a more considered and efficient strategy can be determined. For this to work, more research is required in areas such as weather forecasting and analysis, big data analysis, automated maintenance, personnel transfer and access and inte- grated system development. Some of these areas are currently not well understood leading to unnecessary inefficiencies. [14] There is no comprehensive work on the application of big data technology in the offshore wind energy maintenance. However, parts are being addressed in different research papers. As big data is emerging in

23 24 6. Conclusions & future work many fields, it also does in the world of offshore wind. The desire to cut down costs makes it necessary to gather data on a wide range of aspects.

New projects in O&M DAISY4Offshore and ZEPHYROS are examples of projects working on developing a system that integrates all these different sources of information into one integrated management asset system. ZEPHYROS, a project by World Class Maintenance is trying to bring together relevant actors in the field with researchers, students and local authorities. Projects like these are important in the process of innovation. By working together on improvements, Europe will not only have the leading role in offshore wind technology, but also keep this leading role in the future.

6.2. Future work In this paper three different fields in the offshore wind sector have been mentioned: supply chain manage- ment, wind turbine monitoring and marine operations. Looking at these three field, currently marine oper- ations seems to be lagging the most extensive research. The other two field, supply chain management and wind turbine monitoring are already widely researched. The applications are not always specifically in the offshore wind sector, but the knowledge is present from applications in other sectors. Marine operations is therefore recommended to be considered firstly in future research. Substantial O&M expense reductions can be achieved by combining marine operations research with condition monitoring techniques. Based on gaps in existing research, some interesting topics for future research in the field of marine oper- ations are:

• Introducing/improving vessel monitoring

• Improvements in wind farm specific weather forecasting and analysis

• Improvements in O&M strategy for far-offshore wind farms (e.g. the use of larger service operating vessels versus smaller personnel transfer vessels)

• Introducing decision tools for O&M strategy

Another interesting development is the introduction of wind farm-wide control strategies. This is closely related to the improvements of marine operations. The idea of this is to develop a universal applicable pack- age of turbine monitoring equipment and maintenance/marine operations optimization software. This will function as a full-solution package that can be used by a wind farm owner with only some minor tweaks to adapt it to the wind farm specific parameters. World Class Maintenance already started working on this idea and is currently orientating to find partners to help them develop and test this new concept. It is an extensive project that requires knowledge and research in many different fields. In addition to the topics listed above, the following topics are relevant for this project:

• Design of an O&M management software package

• Data management (management, security, storage)

• Data analysis (wind turbine failure prognosis)

• Automated and remote maintenance

• Supply chain integration (stock management, equipment management)

• Scheduling optimization

Good coordination of all different research fields will be vital to the success of the project. Point of atten- tion will be combining all different research fields and integrating each individual product (such as weather forecasting, condition monitoring, vessel management etc.) into one final product. Appendix A: Wind farms in The Netherlands

SCHIERMONNIKOOG

AMELAND

TERSCHELLING

VLIELAND

HARLINGEN

TEXEL

DEN HELDER

2019 "

IJMUIDEN

AMSTERDAM 2017 2018 " "

SCHEVENINGEN

HOEK VAN HOLLAND

ROTTERDAM

GOEREE

2016 OVERFLAKKEE " SCHOUWEN

2015 " Overzicht windenergiegebieden Noordzee Aangewezen gebieden

WALCHEREN Indicatief kabeltrace " Indicatief hubplatform VLISSINGEN Overige informatie Andere aangewezen gebieden Bestaande parken

ZEEUWS VLAANDEREN Nautische 12 mijl (22,2 km) ZEEBRUGGE

ANTWERPEN 0 2.25 4.5 9 13.5 nm

0 3 6 12 18 24 ´ km

Figure 1: Overview of current and future wind farms in The Netherlands (source: Noordzeeloket [5])

25

Appendix B: What is big data?

Definition of big data Many literature have tried to define the definition of ’big data’. And everyone has his or her own view on it. According to Davenport (2012) big data is more than just smarter, more insightful data analysis. "Companies taking advantage of big data will use real-time information from sensors, rfid identification and other identi- fying devices to understand their business environments at a more granular level." [37] He describes big data as "data that is too big to be processed on one server, too fast-moving to be sequestered in a data warehouse, or too unstructured to fit into a conventional database [38] Goes (2014) defines big data by means of 4 V’s: volume, velocity, variety and veracity. With volume being the amount of data, velocity the frequency of data, variety the different types of data and veracity the valida- tion of data. He mentions that many science disciplines focus only on volume, and perhaps also velocity, but the 4 V’s together is the best quantification method. Integration of several data sources, different data types, and making sure the data is valid are all important aspect when working with big data. [39] Since their is a lack of a formal definition Mauro et al. (2015) did a review on different definitions in lit- erature and combined it to propose a new formal definition. The definition proposed by Mauro et al. is as followed: “Big Data represents the Information assets characterized by such a High Volume, Velocity and Vari- ety to require specific Technology and Analytical Methods for its transformation into Value.” [40]

Big data analysis Big data analytics is a part of big data technology. Often the terms diagnosis and prognosis are used related to data analysis. There is a subtle but important distinction between those two terms. The term diagnosis is used for monitoring of the incoming data and the detection of abnormalities, whereas the term prognosis means the understanding of a problem before it occurs.

27

Appendix C: TKI Offshore Supply Chain Assessment

Figure 2: Number of operational turbines in European waters - projection (source: GL Garrad Hassan, 2014 [19])

Figure 3: netherlands offshore wind projections (source: GL Garrad Hassan, 2014 [19])

The result of the assessment has been classified with the use of a ’traffic light’ grading. Green meaning good supply - unlikely to constrain deployment, orange meaning tight supply - may constrain deployment with- out timely investment and red meaning very tight supply - likely to constrain deployment and significant investment required.

29 30 6. Conclusions & future work

Figure 4: Water depth of Dutch projects (source: GL Garrad Hassan, 2014 [19])

Figure 5: Summary of supply chain assessment (source: GL Garrad Hassan, 2014 [19]) Appendix D: Interview with Bart Hoefakker, Gemini

How should I see Gemini? Gemini is a so-called special purpose company. One-time company with a unique shareholders structure. There will not be developed any second or third wind park by Gemini. There are 4 shareholders each with a different share, and below that there is the group Gemini. A company of about 40 people taking care of the wind park operations and finances. But this is for the whole duration of the park, right? Yes, we just finished the construction phase. During this phase we were working with around 40 people man- aging the ‘contractors’ and currently we are in the operations phase working with around 15-20 people for long-term operations of the park. And who is responsible of the maintenance of the park? For the wind turbines this is Siemens (the supplier), we have a maintenance contract of 15 years with them. For all the other maintenance we wrote out a tender and made contracts with different independent service providers. Van Oord, who performed the construction of the site, are still responsible of the warranty period, but they do not perform the maintenance themselves. For example the electrical infrastructure, this is EWE from Germany, a spin-off from “Utility?” They also own some German offshore wind farms themselves where they perform the maintenance of the substations as well. Since we are a one-time project, you cannot put up a whole organization because there is a need for many different functions and competences, but not all full-time. Therefore, we made the choice to gather competent parties around us who can do the maintenance, and also a part of the operations. So Siemens is doing the maintenance of the wind turbines. I saw they also have their own maintenance ship. Are they then also the ones monitoring the wind turbines? Yes, Siemens is responsible for the maintenance and the performance of the turbines. The give an availability guarantee per year. So they guarantee over the total farm an availability of 95%. When it’s higher they get a bonus, when it is lower the pay a compensation penalty. And normally everything is in the hands of Siemens to maintain the park, resolve malfunctions, the use of logistic resources and decide on the use of these resources. So this is actually outside the operations task of Gemini? Partly, in principal this is outside our scope but as owners but we must ensure that they perform the work well and properly, and whether it meets the contractual conditions. So we actually keep ’operational oversight’. Is Siemens using condition monitoring? Yes, they use them and that information is also visible to us. This is important for us so we can keep an eye on what’s going on. For Siemens, there may be other interests than for us. For example, they could try to postpone certain repairs, while for us it is best to do them in the summer period and not in winter. Turbines sitting still in winter costs more than in summer. And the compensation we receive from Siemens does not really compensate for the actual missed revenue. So it’s a kind of game with each other in which you try to influence and control each other so that you can find a common ground within these different interests. And for this we also get the information from the condition monitor systems, and all performance data from Siemens. We monitor and analyse this information to see if we see a trend of things that are not yet officially known and to go back to Siemens with a kind of “early warning”. They do that themselves as well, but they communicate to us especially what is important in their interests. So you must ensure that you monitor your own installation as well. So you’re trying to monitor your own park and decide and calculate what’s the best strategy and then go to Siemens with that strategy to say "we’d like you to do it this way"? Partly, but you have to be careful with that. At the moment you’re giving instructions, you are also responsible. You do not want to do that. So you have to influence a bit. So you get data from those CM systems, how is this data analysed? Do you make use of software packages like those of ECN? ECN had developed a kind of cost estimator to determine a kind of OPEX cost outline for parks and determine what you needed resources to achieve a particular accessibility. In the early start of this project, Siemens have made some assessments using the ECN tool and ECN also gave advice to make certain decisions on questions like ‘if we are only using small crew transfer vessels, how much percent accessibility do we have then, and what if we sail with a big mother vessel. And it was then realized that with that big service vessel with a crew bridge

31 32 6. Conclusions & future work

(kind of Ampellmann but different) plus a helicopter that we had the biggest range. It was not always the cheapest, but in that way we had the highest accessibility. Theoretically at least, in 2010, of course, relatively little was known about this park. This access tool by ECN is used to determine what the best strategy is beforehand. However, this is not a tool used in day-to-day operations taking into account the current status of the CMS in conjunction with current planning and weather impacts. Such systems are not used yet? No, it is indeed more on feeling. It’s almost like let’s look outside, of course there’s a little bit more to it than that. Every day we look at the planned maintenance, which is then grouped as much as possible so that the ship does not have to go criss-cross through the park. The unplanned maintenance we also do with this ship or with a small maintenance ship. So every day it is determined what activities are happening and where they take place in the park, but there is no decision model where you enter some numbers and parameters and it gives you advice on the best options. The decisions are often very complicated. I know a student from TU Delft has started the design of an inte- grated simulations tool taking into account all the different sources of information. It then calculated on different strategy methods what the total costs would be. Are these problems that you are running into? Is there a desire to get more overview in this with such a tool? We do not use this right now. But that would be a question for Siemens for the day-to-day management. There is a planning for the scheduled maintenance. The challenge is how much unscheduled maintenance there will be tomorrow, and the days after. How many teams do I need to keep available? Based on historical data we can do a reasonably estimate. We cannot fully change the planning per day. We are for 2 weeks on the big vessel, and with this crew and materials we’ll have to deal with the problems. We have a couple of scheduled maintenance teams and a couple of unscheduled maintenance teams. Each with their limitations. And when a turbine has failed and stops generating electricity, we have to option: do I go now with a helicopter and solve the problem as fast as possible? Or do I wait until tomorrow and maybe save money because I can go with a vessel? This is now decided on based on experience and ‘feeling’. We do not have a model that calculates the different options for us. Siemens might be interested to discuss this with you. Are there desires to develop certain tools or to get more insight in certain strategy options? I do think that for Siemens, who is working with daily decisions based on stock, logistics, weather, people, mal- functions, that this is a very interesting area to dig into. Also when you compared it to for example the electrical infrastructure, which is much more planned work. What are the costs of a wind turbine that is not working for a day? We have wind turbines of 4MW with an average capacity factor of 50%. So 2MW per hour, times 24 hours makes on average 48 MW per WT per day. The sales price with subsidy is 168 euros per MW. But this subsidy has a cap in it. So over a certain amount of MW’s we only get the regular price per MW which is much lower. But with sub- sidy the lost generations come down to 168 x 48 MW which is around 8000 euro per day. This is on an average day, on a windy day this can be doubled. The thing is when Siemens has reached their guaranteed availability of 95%, for them there is no hurry to fix the problem as soon as possible. While for us, every day the turbine is not operational we lose possible income. Are there particular things that you are working on or want to work on to improve? Yes, 6 years ago an assessment has been made related to the accessibility of the wind farm. Best would be to do an evaluation to see whether the estimates that are made at that moment are still accurate. The mother vessel, for example, currently costs Siemens around 30.000 – 40.000 euro per day. In summer, it is fully occupied because of all the scheduled maintenance that is planned mostly in summer. But in winter, we try to keep the park operational with as few people as possible. And to have this ship sailing around with only four teams is very expensive. It might be interesting to see if it is smarter to do all winter maintenance by helicopter. At the moment, all this is fixed in contracts. But if we go to Siemens with possible cost reductions and it is a win-win situation, they are of course interested as well. They can then maybe use the ship for a second park. Today this is a rising discussion, to combine things as much as possible instead of working only with dedicated ships. Is data being recorded related to the use of this vessel and helicopter? To answer questions such as ‘How much does it actually cost to send a ship?’ or ‘How many times do we use them?’, ‘Is it smarter to change to another strategy?’ No, currently we are not doing this. For us this is not our priority as we have an all-in service agreement with Siemens. But we do know from every turbine the failures that have occurred, how many times it has been vis- ited, when and how which activities have been performed. You could try to modulate this and see if there would have been a more optimal strategy to solve. Of course, we are only operational for 2 months now, so this would be something to do in a later stadium, after a year or so. Bibliography

[1] IRENA. Innovation Outlook: Off shore Wind. International Renewable Energy Agency, Abu Dhabi, 2016.

[2] WindEurope asbl/vzw (windeurope.org). Financing and investments trends. https://windeurope.org/wp-content/uploads/files/about-wind/reports/ Financing-and-Investment-Trends-2016.pdf, 2017. [Online; 31-05-2017].

[3] Netherlands Enterprise Agency (RVO.nl). Offshore wind energy in The Netherlands. https: //www.rvo.nl/sites/default/files/2015/03/Offshore%20wind%20energy%20in%20the% 20Netherlands.pdf, 2015. [Online; 23-05-2017].

[4] MEGAVIND. Strategy for Extending the Useful Lifetime of a Wind Turbine. 2016.

[5] Noordzeeloket. https://www.noordzeeloket.nl/functies-en-gebruik/windenergie/ bestaande_windparken/. [Online; 23-05-2017].

[6] Noordzeewind. http://www.noordzeewind.nl/project/. [Online; 23-05-2017].

[7] Wind farm luchterduinen. http://projecten.eneco.nl/prinses-amaliawindpark. [Online; 23- 05-2017].

[8] Prinses amalia wind farm. https://www.eneco.nl/over-ons/projecten/ windpark-luchterduinen/. [Online; 23-05-2017].

[9] Wind farms gemini. http://geminiwindpark.nl/. [Online; 23-05-2017].

[10] Sociaal-Economische Raad (SER). Energieakkoord voor duurzame groei. 2013.

[11] Social and Economic Council (SER). Energy agreement for sustainable growth - summary. https:// www.ser.nl/, 2013. [Online; 23-05-2017].

[12] GL Garrad Hassan. A Guide to UK Offshore Wind Operations and Maintenance. Scottish Enterprise and The Crown Estate, 2013.

[13] B. Hoefakker. Personal interview. 2017, June 6.

[14] A. Bagnall, A. Hadjer, I. Weeks, and M. Blo. Big data: can it reduce the cost of wind turbine operations and maintenance? UEA Consulting Ltd., 2015.

[15] E. G. Nabati and K. D. Thoben. Big Data Analytics in the Maintenance of Off-Shore Wind Turbines: A Study on Data Characteristics. Dynamics in Logistics. Springer International Publishing, 2015.

[16] Vestas Wind Systems. Vestas wins deloitte’s big data award. https://www.vestas.com/en/media/~/ media/3617e16e4a454859827fdaec74ade44a.ashx. [Online; 24-05-2017].

[17] M. drunsic, D. Ekici, and M. White. Logistics and Supply-Chain management in Offshore Wind Farm OWF Applications. IET Renewable Power Generation, 2016.

[18] Federico D’Amico, Riccardo Mogre, Steve Clarke, Adam Lindgreen, and Martin Hingley. How purchasing and supply management practices affect key success factors: the case of the offshore-wind supply chain. Journal of Business Industrial Marketing, 32(2), pages 218–226, 2017.

[19] O.F.Roy, P.Reynolds, and J. Clayton. Offshore Wind Supply Chain Assessment. GL Garrad Hassan Ned- erland B.V., 2014.

[20] J. Tautz-Weinert and S.J. Watson. Using SCADA data for wind turbine condition monitoring – a review. IET Renewable Power Generation, 2016.

33 34 Bibliography

[21] D. Coronado and K. Fisher. Condition Monitoring of Wind Turbines: State of the Art, User Experience and Recommendations. 2015.

[22] P.Tchakoua, R. Wamkeu, M. Ouhrouche, F.Slaoui-Hasnaoui, T.A. Tameghe, and G. Ekemb. Wind Turbine Condition Monitoring: State-of-the-Art Review, New Trends, and Future Challenges. Energies, 2014.

[23] F.P.G. Márquez, A.M. Tobias, J.M.P. Pérez, and M. Papaelias. Condition monitoring of wind turbines: Techniques and methods. Elsevier, Renewable Energy 46, pages 169–178, 2012.

[24] A. Kusiak and W. Li. The prediction and diagnosis of wind turbine faults. Elsevier, Renewable Energy 36, pages 16–23, 2011.

[25] M. Schlechtingen, I.F.Santos, and S. Achiche. Wind turbine condition monitoring based on SCADA data using normal behavior models. Part 1: System description. Elsevier, Applied Soft Computing 13, pages 259–270, 2013.

[26] R. Ahmad and S. Kamaruddin. An overview of time-based and condition-based maintenance in indus- trial application. Elsevier, Computers Industrial Engineering 63, pages 135–149, 2012.

[27] Ampelmann Operations B.V. Safe offshore access - ampelmann - motion compensated gangways. http: //www.ampelmann.nl/offshore-access. [Online; 29-05-2017].

[28] Kongsberg Maritime. State-of-the-art kongsberg motion compensated, integrated ’walk-to-work’ sys- tem to be installed on olympic orion mpsv. https://www.km.kongsberg.com/ks/web/nokbg0238. nsf/AllWeb/9046E209B5FFD448C12580C70029EDD0. [Online; 29-05-2017].

[29] Royal Wagenborg. Innovative walk to walk vessel, our new type of off- shore maintenance support vessel. https://www.wagenborg.com/our-stories/ the-new-standard-in-offshore-support-and-maintenance. [Online; 29-05-2017].

[30] H. Koopstra. An Integrated and Generic Approach for Effective Offshore Wind Farm Operations Main- tenance. Master thesis, 2015.

[31] H. Koopstra. An Integrated and Generic Approach for Effective Operations Maintenance of Offshore Wind Farms. 2015.

[32] Energy research Centre of the Netherlands (ECN). New release of the ECN OM Calculator v2.2. https: //www.ecn.nl/news/item/ecn-om-calculator/. [Online; 20-06-2017].

[33] Royal Wagenborg. World Class Maintenance. http://www.worldclassmaintenance.com/nl/. [On- line; 15-06-2017].

[34] Royal Wagenborg. Daisy. http://www.daisy-solutions.com/. [Online; 15-06-2017].

[35] Royal Wagenborg. Innovative walk to walk vessel, our new type of offshore main- tenance support vessel. http://www.worldclassmaintenance.com/nl/projecten/ onderhoud-aan-offshore-windparken-daisy4offshore. [Online; 29-05-2017].

[36] F.Visser. Presentation at Windday Conference. 2017, June 14.

[37] T. H. Davenport, P. Barth, and R. Bean. How “Big Data” Is Different. MIT Sloan Management Review, 54(1), pages 43–46, 2012.

[38] N. Smith. Book Interview: Thomas Davenport. Engineering Technology, 9(4), pages 92–93, 2014.

[39] P.Goes. Big Data and IS Research. MIS Quarterly, 38(3), pages 3–8, 2014.

[40] A. Mauro, Greco, M., and M. Grimaldi. What is Big Data? A Consensual Definition and a Review of Key Research Topics. AIP Conference Proceedings, 1644(1), page 97–104, 2015.