Risk assessment of marine energy projects

(1)

Master of Science Thesis

KTH School of Industrial Engineering and Management Energy Technology EGI-2016-005MSC EKV1123

Division of Heat & Power SE-100 44 STOCKHOLM

Risk assessment of marine energy projects

Steven GUEGUEN

(2)

Master of Science Thesis EGI-2016-005MSC EKV1123

Risk assessment of marine energy projects

Steven Gueguen

Approved

2016-02-19

Examiner

Miroslav Petrov- KTH/ITM/EGI

Supervisor at KTH

Miroslav Petrov

Commissioner

EDF R&D, France

Contact person

Jean-Marie Fourmigué

Abstract

The energy marine technologies (Wave Energy Converters, Tidal Energy Converters and floating wind turbines) are still at an early stage of development since no commercial array exists today. Their development has to face three main issues: the lack of feedback from the field, the great diversity of the existing technologies and the harsh marine environment. In order to attract investors, these technologies need to prove both their reliability and their economic viability. Thus this thesis proposes a methodology and a support tool to assess the financial risks linked to a marine energy project based on the reliability of systems.

The first part of the methodology is a reliability and maintainability assessment which is directly based on the classic FMECA (Failure Modes, Effects and Criticality Analysis) methodology. Then an Excel prediction tool was developed so as to assess the financial risks linked to a marine energy projects. This support tool is based on a Monte-Carlo method and relies on reliability data provided by the FMECA.

The whole methodology developed is simple since no accurate data exists. Moreover, the support tool is designed so as to be adapted for a large variety of technologies and maintenance strategies. Finally, all the outputs of this support tool are quantitative. The main outputof the methodology is the LCOE (Levelized Cost Of Energy) since this value is a function of the CAPEX (CAPital EXpenditures), OPEX (OPerational EXpenditures) and the availability of the array.

To prove the validity of the support tool, this one was tested on several existing WECs thanks to the input parameters provided by the developers. So as to keep the confidentiality of these data, it was decided to develop in this thesis one fictive example including two different WECs. In depth analysis and comparison of technologies are carried out: sensibility analysis on several parameters, optimization of the design for one technology (dry-mate connector Vs. wet-mate connector), optimization of the maintenance strategy for one technology.

The strength of the methodology developed lies in the ability to really calculate both the OPEX and the availability of the array based on reliability data and without performing an exhaustive analysis of each system. However, the support tool is limited to 13 components per technology. So, if more components need to be taken into account, a functional analysis has to be done in order to gather components in functions. Moreover, because of the lack of data, the FMECA performed is simplified: failure rates are linked to components instead of failure modes.

(3)

-i-

List of Figures

Figure 1: Wave energy conversion methods in electricity (JRC, 2014)... 2

Figure 2: Picture of a wave. ... 2

Figure 3: Picture of corrosion ... 3

Figure 4: Picture of bio-fouling... 3

Figure 5: General overview of the link between the marine energy context and the expected tool characteristics ... 4

Figure 6: General overview of the risk assessment in marine energy projects ... 6

Figure 7: Overview of one output expected with the methodology, LCOE Vs. MTTF ... 7

Figure 8: Overview of the general methodology ... 7

Figure 9: Illustration of the functional analysis ... 9

Figure 10: View of the SADT basis element ... 10

Figure 11: View of the FAST diagram ... 10

Figure 12: View of the array modelling ... 11

Figure 13: Overview of the improvement process ... 15

Figure 14: The general work flow of the prediction tool ... 17

Figure 15: Example of typical results ... 18

Figure 16: Several systems compared together according to their CAPEX, OPEX and LCOE. ... 18

Figure 17: One example of a sensibility analysis carried out for one system ... 19

Figure 18: Drawing of the Bathtub Curve (source: (Thies, 2012)) ... 20

Figure 19: Overview of the failure modeling strategy ... 22

Figure 20: View of the failures management scenarios ... 23

Figure 21: Screen print of the main parameters of the general maintenance strategy ... 24

Figure 22: Diagram of the maintenance strategies available ... 25

Figure 23: Screen print of the vessels input data ... 26

Figure 24: Picture of the two kinds of maintenance vessels used ... 26

Figure 25: The scattered diagram (Hmo/Te) of SEM-REV. ... 27

Figure 26: The scattered diagram (Hmo/Te) of the Yeu Island. ... 27

Figure 27: Screen print of the 6 sheets available in the prediction tool. ... 29

Figure 28: Overview of the framework of the prediction tool ... 29

Figure 29: View of the array selected in the example ... 33

Figure 30: SADT A0 ... 34

Figure 31: SADT main view ... 34

Figure 34: SADT main view ... 37

Figure 35: The radar profile of the two technologies ... 41

Figure 36: Plot of the LCOE in function of the time between two preventive maintenance operations when there is no optimum ... 44

Figure 37: Plot of the LCOE in function of the time between two preventive maintenance operations when there is one optimum ... 45

Figure 38: Cost of each funcion in the OPEX for the WEC_heaving device ... 46

Figure 39: Cost of each funcion in the OPEX for the WEC_surging device ... 46

Figure 40: Plot of the real production over maximum production expected for the WEC_heaving device ... 47

Figure 41: Plot of the real production over maximum production expected for the WEC_surging device ... 47

Figure 42: Distribution of CAPEX and OPEX in the total cost for tehWEC_heaving device ... 48

Figure 43: Distribution of CAPEX and OPEX in the total cost for tehWEC_surging device ... 48

(6)

-iv-

Figure 44: Plot of the variation of the LCOE with Capacity factor for the two devices ... 50

Figure 45: Plot of the variation of the LCOE with CAPEX for the two devices ... 50

Figure 46: LCOE variations when Capacity factor and CAPEX are 20% lower or higher ... 51

Figure 47: Diagram showing the variation of LCOE when all the MTTFs are 50% lower ... 52

Figure 48: Diagram showing the variation of LCOE when all the MTTRs are 50% higher ... 52

Figure 49: LCOE predictions for wave arrays, (JRC, 2014) ... 53

(7)

-v-

List of Tables

Table1: Example of dysfunctional analysis carried out on the buoy system ... 12

Table 2: Examples of failure mechanisms and failure modes ... 12

Table3: Example of assessment of criticality ... 14

Table4: Comparative table of MTBF for established wind turbine technology ... 21

Table5: Input parameters of the 3.3.2 section example ... 23

Table6: Summary of the WEC_heaving and WEC_surging characteristics ... 32

Table7: Summary of all the functions for the WEC_heaving device ... 36

Table8: Summary of all the functions for the WEC_surging device ... 38

Table9: Summary of the reliability and costs assumptions for the WEC_heaving device ... 39

Table10:Summary of the reliability and costs assumptions for the WEC_surging device ... 40

Tableau 11: First results for the two technologies ... 41

Table12: Comparison of the dry-mate and wet-mate assumptions... 42

Table13: Summary of the main results for the exampledry-mateVs. wet-mate ... 42

Table14: Summary of the main results for the example SWAP Vs. without SWAP ... 43

Table15: Summary of all the results of the example ... 49

Table16: Results for 9 runs of the WEC_surgingdry-mate sheet ... 49

(8)

-vi-

Nomenclature

α

: Shape parameter of the Weibull distribution 1/k: Scale parameter of the Weibull distribution

λ

: Failure rate Cp: Capacity factor

Hmo: Significant wave height Te: Average energy period

(9)

-vii-

Abbreviations

LCOE: Levelized Cost Of Energy CAPEX: CAPitalEXpenditures OPEX: OPerationalEXpenditures TRL: Technology Readiness Level

FMECA: Failure Modes, Effects and Criticality Analysis WEC: Wave Energy Converter

TEC: Tidal Energy Converter NPV: Net Present Value PV: Present Value

DEVEX: DEVelopment Expenditure MTTF: Mean Time To Failure

MTBF: Mean Time Between Failure TTF: Time To Failure

IEA: International Energy Agency

DNV: Det Norske Veritas - international standardization and certification committee for the offshore sector

SADT: Structured Analysis and Design Technique FAST: Functional Analysis System Technique

(10)

-viii-

ACKNOWLEDGMENTS

Firstly I would like to thank Nicolas Relun, the head of the marine energy project within EDF R&D for welcoming me as a trainee.

My thanks also go to Jean-Marie Fourmigué, research engineer in EDF R&D, who supervised and mentored my work, for his attention, and for all the time he awarded to me.

I would like to thank Elisabeth Mallet, research Engineer in EDF R&D, who was always ready to help me and who gave me thousands of useful advices.

I also thank Antoine Despujols, research Engineer in EDF R&D, for sharing with me a piece of his knowledge about risks assessment in industrial projects.

I would like to thank every person I have worked with within the EPI department in Les Renardières for their cheerfulness.

Finally, I would like to thank Dr.Miroslav Petrov, my KTH supervisor, who helped me realizing this work and who gave me all his attention.

(11)

-1-

1 INTRODUCTION

1.1 Background

1.1.1 Renewable Marine Energies

1.1.1.1 Wave energy all over the world

According to IEA, the world wave energy resource represents 20000TWh/year.

Moreover, in Europe, among all the marine energies available, wave energy is believed to become one of the top contributors to the European energy system by year 2025-2030 (JRC, 2014).

Therefore, today the wave energy is becoming more and more attractive as it offers sustainable electricity generation from a comparatively predictable renewable resource. However, according to the same report (JRC, 2014), the growth of this sector is slower than expected and developers have still to prove the viability of their technologies. That is why, reliability issues remain a key challenge towards developing economically viable wave energy conversion devices and systems (Thies et al., 2009).

1.1.1.2 A great diversity of technologies

So far more than 45 different kinds of technologies have been developed and have reached an open sea deployment (JRC, 2014). The first commercial systems are maturing and suggest the possibility for energy companies to invest in these solutions.

All these technologies are based on different kinds of power take-off systems as one can see in Figure 1.

(12)

-2-

Figure 1: Wave energy conversion methods in electricity (JRC, 2014)

1.1.1.3 A binding marine environment

Despite its dynamism, the wave energy field has to face a harsh marine environment (Figure 2).

Wave energy converters are submitted to extreme conditions firstly due to the wave overload.

For instance, a wave of 10 m high represents a weight of 12 tons per square meter.

Figure 2: Picture of a wave.

Moreover, offshore devices have to face corrosion (figure 3) and bio fouling (figure 4).

(13)

-3-

Figure 3: Picture of corrosion

Figure 4: Picture of bio-fouling

Also, the electrical connection at sea continues to be a largely unresolved issue for many offshore projects. Today, two kinds of grid connectors exist: dry-mate connectors and wet-mate connectors, however, none of them are easy to implement.

1.1.1.4 A lack of maturity

According to (Thies et al., 2009), the lack of reliability data in the offshore field leads to rough values when one tries performing a reliability assessment. According to the same author, this is due to:

• Low feedback from the field

• The confidentiality in some projects.

Moreover, when one tries finding some figures regarding the OPEX linked to a specific WEC project, it is hard to find a value based on true calculations. Indeed, it is generally assumed that the annual OPEX is a certain percentage of CAPEX.

1.1.1.5 The need

The marine energy field is a very constraining environment that required specific tools (figure 5).

Firstly, the methodology developed has to be simple and based on a limited number of parameters since data from the marine renewable energy field are limited and not so accurate.

Moreover, in order to be suitable for a wide range of technology design options and

(14)

-4-

implementation pathways, this methodology should be flexible and open, i.e. a large panel of input parameters has to be accessible. This is further explained in Figure 5.

Figure 5: General overview of the link between the marine energy context and the expected tool characteristics

1.1.2 Risk assessment

Generally, the three main risks involved in marine offshore projects are (figure 6):

• The financial risk

• The security

• The environmental risk

Today, the security linked to design, manufacturing and operation of marine energy technologies is the first concern of developers. To support them and provide confidence between them and the finance or insurance authorities, both standards and certification by third parties have been implemented. Developers can use the DNV Offshore Standards during the design stage. For instance, the Offshore Standard (OS) DNV-OS-C401 gives recommended practices regarding fabrication and testing of offshore structures. In May 2005, a guide gathering engineering standards and recommended practices for Wave Energy Converters was commissioned by the Carbon Trust and carried out by DNV (DNV, 2005). To sum-up, the security risk includes the safety of the device and also the safety for the surrounding environment (maintenance staff, vessels...).

The environmental risk has to be controlled by performing environmental studies. For these one, a few criteria need to be evaluated including noise, electromagnetic field, pollution, antifouling, modification of the submarine relief, reversibility of the foundation installation, and the possibility for creating an artificial reef. Another aspect can be the social acceptability of the installation of an array including the visual aspect, the seabed occupation, the electrical installation needed onshore or the possibility of creating jobs near the array.

(15)

-5-

Finally, the financial risk is well described by the LCOE. As one can see below, the LCOE is a function including the CAPEX, the OPEX and the production.

𝐿𝐿𝐿𝐿 = 𝐿𝐶𝐶𝐿𝐶 + 𝐶𝑃(𝐿𝐶𝐿𝐶)

𝐶𝑃(𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒 𝑝𝑒𝑝𝑝𝑝𝑒𝑒𝑒𝑝𝑝)

In short, the LCOE is the price to produce electricity (€/MWh). By 2020, for instance, the cheapest way to produce electricity is expected to be the advanced natural gas-fired combined cycle (72.6 $/MWh) (EIA, 2015).So one can understand that if the LCOE is higher, the financial risk is higher.

These considerations are summarized in Figure 6.

1.2 Objectives

As one can notice, the marine energy field is a very dynamic one but both the reliability and the profitability of the proposed energy conversion technologies have to be proven in order to attract investors. Moreover, as seen above, security and environmental risks can be well managed by following international standards and by performing environmental studies.

Thus the goal of this thesis will be to propose both a general methodology and a support tool aiming at assessing in an objective manner the financial risk linked to the installation and the operation of an array of WECs (figure 6). The methodology proposed should respect 3 important points.

Firstly, reliability analysis has already been developed in many mature industries, but for the wave energy field the goals of such analysis cannot be the same. The lack of reliability data makes useless the utilization of high level statistical tools. So the purpose of this thesis is to find a reliability assessment methodology which operates at acceptable level in terms of data accuracy.

Then, the methodology and mainly the support tool have to be very flexible in order to be suitable for the large range of technology design options and types of applications.

Finally, this methodology has to be quantitative and based on true calculations in order to preserve its objectiveness and representability for the involved sectors.

(16)

-6-

Figure 6: General overview of the risk assessment in marine energy projects

1.3 Methodology

In order to lead an in-depth study of a wave energy converter, three main steps have been identified:

Firstly, both reliability and maintainability analyses of the technology have to be performed in order to identify the main functions, their failure modes and the economic consequences related to those.

Secondly, based on the previous analyses, an economic assessment of the technology is carried out. In the framework of the master thesis, the author has developed a software tool based on the Monte-Carlo method, to calculate the OPEX and the LCOE linked to an array of a specific devices. Unfortunately, because of the lack of accuracy in input data, the output data are only rough values.

Finally, since the output data are not accurate, for each device sensitivity analyses with critical parameters have to be done. As a result of these sensitivity analyses, the LCOE is no longer a value but a range (figure 7) which is much more meaningful. This sensitivity analysis is also a good way to identify the main bottlenecks linked to a device.

(17)

-7-

Figure 7: Overview of one output expected with the methodology, LCOE Vs. MTTF

Figure 8: Overview of the general methodology

The end of this thesis will be dedicated to a relevant case study which is the best way to illustrate the main three steps of the methodology proposed (figure 8). In this example, the reader is going to see how it is possible to use the methodology and the support tool in order to optimize firstly the design and then the maintenance strategy of one technology. For instance, is it better to use a dry-mate or a wet-mate connector?

The author would like to underline the fact that the whole methodology implies to work with a limited number of functions and components. This choice was made to allow for simplifications with the goal to conclude the study in a reasonable amount of time.

200 250 300 350 400 450

System 1 System 2 System 3 System 4

LC OE (e ur os /M W h)

WECs

LCOE Vs. MTTF

Low MTTF

High MTTF

Low MTTF

High MTTF Low MTTF High MTTF

Low MTTF High MTTF

(18)

-8-

2 RELIABILITY AND MAINTAINABILITY

The FMECA (Failure Mode Effects and Criticality Analysis) aims to assess the reliability and maintainability of a wave energy converter and its main components. Both these criteria will then be combined to obtain the criticality of the system and its components. To sum up a FMECA is an analysis which aims to assess causes and consequences of failure modes.

A methodology to achieve a FMECA is proposed below.

2.1 Functional analysis

2.1.1 Presentation

In this first part, one identifies the functions, sub-functions and components of the system that one plans to analyse:

The system considered is an array (or network)of identical WEC's

• Sub-systems at level 1 are the whole WEC including all its major functions,

• Sub-systems at level 2 describe the main functions of the WEC’s components, such as wave energy capture, mechanical power transfer to an intermediate stage, electric power generation, electric power conversion and power output to shore, anchoring and link to the sea bed, etc.

• Sub-systems at level 3 and 4 can either describe a sub-function or a component of the systems that fulfil a given function.

The more sub-systems and levels, the longer and complex will be the analysis work. Thus, it is more relevant to carry out the analysis step by step by increasing the number of sub system or components at each step.

- STEP 1: This first step is limited to max 10 items: either sub-systems of level 2 or components.

- STEP 2: The analysis is extended to 10 new items: either sub-systems or components depending on the need for deeper or wider analysis.

- STEP 3: Analysis of ten more items.

In the example below, we have considered a fictive WEC consisting of a buoy that pumps a hydraulic fluid and then produces electricity from a hydraulic motor coupled to a generator and frequency converter. All components except the output power cable are inside the hull. This is illustrated in Figure 9.

(19)

-9- Figure 9: Illustration of the functional analysis

1- Fixation to seabed 2- Wave Energy Capture 3- Control

4- Energy conversion 5- Electric energy export

6- Piston

7- High pressure hydraulic circuit 8- Manifold

9- Motor 10- Generator

Moreover, all the components of one function must be either in parallel or in series. But a mix of this two configurations has to be avoided (explanation in 3.3.1).

The SADT analysis is firstly performed since it is a very exhaustive method not only to identify the functions carried out by the system but also the links between all these. Moreover, the flows of energy and motion can be easily represented by this method. Such a representation is going to be very useful for the dysfunctional analysis. For instance, if one wants to analyse the failure of a function, it is important to know exactly both the link between this function and the other functions (consequences of the failure) and which components carry out this function (causes and maintenance).

Next, the FAST analysis is achieved by choosing the main functions and the components of the SADT analysis. The FAST analysis is more simplistic but it gives a clear overview of the system one wants to deeper analyse. It includes only the functions and the components that are going to be considered in the following parts.

Finally it is relevant to give a representation of an array including the system under the spotlight.

2.1.2 SADT

Structured Analysis and Design Technique (SADT) is a systems engineering methodology for describing systems as a hierarchy of functions. It was developed in the late 1960s by Douglas T.

Ross, and further formalized and published as the standard IDEF0 in 1981.

Structured Analysis and Design Technique (SADT) is a diagrammatic notation designed specifically to help people describe and understand systems. It offers building blocks to represent entities and activities, and a variety of arrows to link boxes between them. These boxes and arrows have an associated informal semantics. SADT can be used as a functional analysis tool of a given process, using successive levels of details.

Below one can find the SADT basis element depicted in Figure 10.

(20)

-10-

Figure 10: View of the SADT basis element

2.1.3 FAST

Functional Analysis System Technique (FAST) is a diagrammatic notation designed specifically to help people to describe and understand systems.

A FAST diagram is firstly a translation of service functions into technical functions and then physically in technical solutions. FAST diagram is built from left to right in the logic of “why” to

“how”.

Figure 11 gives a typical example of a FAST diagram:

Figure 11: View of the FAST diagram 2.1.4 The array

All the technologies are compared when they are operating in an array with the same nominal power. It is more convenient to compare two different technologies if the power of the array is the same. In an array several parameters have to be defined at the beginning of the project: the number of systems operating, the number of substations offshore, the distance from the shore, etc. Figure 12 presents a generalized view of an array model.

(21)

-11-

Figure 12: View of the array modelling

Moreover, a unit system includes: one system, one umbilical cable, one offshore substation and one grid export cable. In the next section, the failure modes analysis is performed on this unit system.

2.2 Dysfunctional analysis

For each function and each component, one searches for the possible causes of failure limited to circumstances that result in failure during operation and maintenance (specification, design and manufacture are not treated here).

The best is to use as a starting point a list of possible failure mechanisms, such as corrosion or fatigue. Based on that list, one analyses for each function or each component the occurrence of such mechanism, and one details what are the related failure modes; or, in other words, the effect of this mechanism on the component or sub-function analysed. Finally, the consequences of the failure mode on the whole system should be described (e.g. 10% loss of production).

(22)

-12-

Table1: Example of dysfunctional analysis carried out on the buoy system

# Failure

mechanism Components Function Failure modes

Consequence of the failure

mode

Mean Time to Failure

6 Mechanical

fatigue Piston

Move the hydraulic fluid into the high pressure circuit

Unwanted stop of the fluid

compression (Mechanical breakdown of the main piston)

100% loss of component function

100% loss of system function

5 years

7 Corrosion

High Pressure hydraulic circuit

Storage and transport of the hydraulic fluid at high pressure

Leakage in the circuit and loss of hydraulic fluid

Up to 30% loss of the function

30% loss of motor efficiency

2 years

1 Mechanical

rupture Mooring Link the system

to the seabed

Rupture of all moorings lines

Loss of system

Uncontrolled system floating

10 years

Table 2: Examples of failure mechanisms and failure modes List of failure mechanisms List of failure modes

Corrosion Leakage

Erosion Cessation of fluid filtration

Bio fouling Rupture

Mechanical fatigue Loss of electrical transmission

Mechanical rupture Loss of heat protection

If the FMECA (Failure Modes, Effects and Criticality Analysis) is based on the analysis of functions without considering the components, it only exists in five failure modes:

• No working of the function

• Partial working of the function

• No start of the function

• No stop of the function

• Unwanted working of the function

(23)

-13-

2.3 Maintainability

The maintainability is the probability that a given active maintenance action can be carried out within a stated time interval. It can be expressed in terms of Mean Time To Restore (MTTR), which include the reparation time and the delay time before the maintenance procedures commence.

There are two main aims to this study:

• First the costs associated to each failure mode must be understood. To do so the maintenance strategy must be analysed and the costs of maintenance tasks determined.

• Then a reflexion must be carried out on safety and simplicity of maintenance tasks. A list of questions should be answered to assess the ability to perform the tasks.

For instance:

• Are divers needed?

• How can the system be connected/disconnected to/from the grid?

• Does the system need to be dragged back to the shore for maintenance?

2.4 Assessment of a criticality of a failure

In the dysfunctional analysis we examined the consequences of the loss of function. Here the quantitative effect of these losses is assessed.

2.4.1 Assessment of the severity

The first step of the analysis is to qualify the severity of a failure mode. The potential or actual detrimental consequences of a failure or a fault is assessed.

To do so the following steps are used:

• First one assesses qualitatively the effects, such as: Severe for Safety (SS), Severe for Environment (SE), and Severe for Production (SP).

• If the fault is SP, one tries to assess its quantitative impact:

i. Through direct costs: cost of repair.

ii. On indirect costs: loss of production when system is down, contract penalties, etc.

2.4.2 Assessment of the criticality

The severity does not account for the probability of occurrence of a failure but merely describes the impacts of this failure. The criticality accounts for both: the costs induced by a failure mode are multiplied by the probability of occurrence of this failure mode; therefore giving an estimate of the potential costs induced by the failure mode over the lifetime.

(24)

-14-

An example is provided in Table2, in which the failure mode #6 happens in average once every 5 years and costs 65k€. Therefore this failure mode costs in average 13k€/year and is less critical than the failure mode #7 that costs 47k€/year. It is to be noted that a failure can be critical because of its severity but also because of its occurrence.

Table3: Example of assessment of criticality

# Sub system Component or

Failure Mode

Consequence of the failure

mode

Strategy Direct costs Indirect costs Severity MTTF Criticality (k€/year)

6 Hydraulic pump

Unwanted stop of the fluid compression

Loss of 100%

of component function Loss of 100%

of system function

Replacement when broken

Small boat: 3 days Labor: 3 days Part: 5 k€

Total: 15 k€

8 weeks with 0% production Total : 50 k€

SP

65k€ 5 years 13 k€

7 High Pressure Hydraulic circuit

Leakage in the circuit and loss of hydraulic fluid

Loss of up to 30% of the function 30% loss of motor efficiency

Repair when loss>30%

function

Small boat: 3 days Labor: 3 days Part: 5 k€

Total: 15 k€

16 weeks with 30%

production loss 8 weeks at 0%

production Total : 80 k€

SP 95k€

SE?

2 years 47k€

1 Mooring

Rupture of all moorings lines

Loss of system Uncontrolled system floating

Need a new installation (and a change of design)

Loss of 100%

CAPEX Small boat: 3 days Labor: 5 days Part: 1000 k€

Total: 1010 k€

Loss of 100%

production 6 months with 0% production Total : 150 k€

SS Safety for other ships

10 years

Max 1160k€

2.5 Analysis and conclusion

This part is devoted to an analysis of the strengths and weaknesses of the system. Based on the previous analysis, the weakest parts can be identified by sorting them in terms of criticality.

Additionally, parts that represent a threat for safety (SS) or for environment (SE) may be regarded as crucial, even they are not critical for the production and do not represent a high potential cost.

(25)

-15-

After this analysis, strengths of weaknesses of the system have been studied. Further actions can be taken:

• If needed, the decision can be taken to start again a new analysis by increasing the deepness and/or the wideness of its functional analysis. Sensible parts can be broken down further in more sub-components (e.g.: detail the moorings to understand better which specific part is crucial in preventing the rupture).

• Otherwise, improvements can be planned based on the conclusions, as proposed in 2.6.

2.6 Optimization through design and maintenance strategy The idea in this part is to optimize functions or components that appeared as weak points in the sections before. By changing the design or plan the maintenance procedures differently, one can expect improvements of the reliability. The improvement process can lead to restarting a new Failure Mode Analysis and providing newer possibilities for further alterations in a multiple loop optimization process, as shown in Figure 13 below.

Figure 13: Overview of the improvement process

(26)

-16-

3 ECONOMIC ASSESSMENT

3.1 Development of a support tool

3.1.1 Context

A calculation Excel tool has been developed within an EDF R&D project of Wave Energy Converters (WEC) reliability assessment (July-December 2015). A reliability assessment starts with both a functional analysis to identify the main functions within the system and a dysfunctional analysis to identify the main failure modes. Regarding the main failure modes, severity and criticality have to be found. At this point, one has to estimate the costs of each failure mode. Resulting from several discussions with developers of WEC, it is obvious that OPEX keeps being a blurred data when one tries to estimate the LCOE of an array of WECs.

Generally, OPEX is seen as an assumed fraction of the CAPEX.

Moreover, some software tools able to calculate the CAPEX, OPEX and LCOE for offshore wind turbines have already been developed. However, still resulting from several discussions with WECs developers, it is clear that maintenance strategies are not the same for WECs and offshore wind turbines. For instance, an onshore maintenance is required for a lot of WECs which doesn’t exist for wind turbines. So, regarding the maintenance strategies, it is not recommendable to use the wind-targeted software tools for modelling arrays of wave energy converters.

To sum-up, the hereby proposed prediction tool in Excel is needed to bring more accurate figures regarding OPEX and LCOE for deployed WECs.

3.1.2 General presentation of the tool

The calculation Excel tool has been developed to estimate several costs (CAPEX, OPEX and LCOE) and the availability of the WEC array or the loss of production. To reach this goal, the prediction tool relies on repeated random sampling. Each sampling is the probable life of the array in which:

• Failures occur as random events depending on the reliability model chosen for each function in the input parameters.

• Failed system are fixed or replaced depending on the maintenance strategy chosen in the input parameters. Several strategies exist and depend on several parameters: When is the maintenance operation going to be carried out? Before the failure, directly after or some months after? Where is the maintenance operation going to be carried out…onshore or offshore? What kind of logistic means are going to be used? How many days does it take to repair the system?

• Regarding the interventions, the weather windows depend on the weather data of one specific site (The SEM-REV test site is used by default).

• The average production per year of the array depends also on these weather data and the power matrix of the system.

• CAPEX and inspection costs are calculated in a deterministic manner.

Figure 14 below illustrates the general work flow of the prediction tool.

(27)

-17-

Figure 14: The general work flow of the prediction tool

During one simulated life of the array, MTTF values are generated and depend on two parameters: the failure rate and the wearing parameter of each function under interest. Regarding the mathematical approach, the well-known Monte-Carlo method can serve for this purpose. To get a mean result, more than 100 lives of the array are simulated.

Furthermore, the management of failure rates follows the “As Good As New” method: when a corrective or preventive operation is carried out on a component, the component is assumed to be brought back to its initial state attaining the failure rate of a new component.

Finally, in order to simplify the problem, it is assumed that failure rates are linked to functions or components instead of failure modes.

The main results constituting the prediction tool output parameters (figure 15) are:

• CAPEX

• OPEX (per year)

• LCOE

• Availability and electrical production (per year)

• The impact of each function on the OPEX

(28)

-18-

Figure 15: Example of typical results

3.1.3 Goal of the prediction tool

The main goal of the tool is either to compare several marine energy systems together or to compare several maintenance strategies together for one system or to compare several designs of one system among each other. As illustrated in figure 16, systems or maintenance strategies or designs can be compared using CAPEX, OPEX, LCOE results.

Moreover, as shown in figure 17, sensibility analyses can be performed for one system by varying the reliability of components.

Figure 16: Several systems compared together according to their CAPEX, OPEX and LCOE.

(29)

-19-

Figure 17: One example of a sensibility analysis carried out for one system

3.1.4 Limitations

The prediction tool is designed to simulate the life of a maximum of 50 systems. The economic lifetime of the array is 25 years. This figure is a fixed input parameter and cannot be modified.

Finally, the number of functions per system is limited to 13.

Otherwise, this tool cannot be used without a preliminary study which consists in carrying out a functional analysis and a dysfunctional analysis of the system. It is really important to identify the right functions with the right components inside. For instance, it is impossible to put 20 hydraulic cylinders and one HP-circuit in the same function since the consequences of their failures are totally different. One has to keep in mind that there are specific functions for components in parallel. Moreover, for each functions one has to select the main failure mode since only one failure mode can be represented for each function. Generally it is recommended to select the worst failure mode among the five failure modes existing for a function (section 2.2), i.e. “no working of the function”.

3.2 Modeling of the reliability

3.2.1 The methodology of reliability modeling

The reliability factor can be defined as the probability for a function to be performed for

“specified conditions and time” (Thies, 2012).

(30)

-20-

The failure rate is the frequency with which an engineered system or component fails, expressed, for example, in number of failures per hour. Therefore failure rates can be used to compute the Mean Time To Failure (MTTF). The “bathtub curve” (figure 18) is usually used in reliability engineering to represent the time function of the failure rate and related parameters. It describes a particular form of the failure rate behaviour along the life span of a system, which includes three main parts:

• The first part is a decreasing failure rate, known as early failures whose severity is usually small and whose frequency drops considerably with the running-in of the system.

• The second part, relatively constant, consists of the normal life of the system or device during which it is in its best mechanical condition.

• The last part is an increasing failure rate, known as wear-out failures, which is related to a growing frequency of major faults and break-downs.

.

Figure 18: Drawing of the Bathtub Curve (source: (Thies, 2012))

Moreover, one has to keep in mind that it is really hard to find an exact MTTF or MTBF of each component. A literature review was conducted on failure rates of existing wind turbines. Table 4 below shows how difficult it is to assess the failure rate of one component. For instance, regarding the brake system of wind turbines, the MTBF can be between 4 years and 83 years.

(31)

-21-

Table 4: Comparative table of MTBF for established wind turbine technology

3.2.2 The failure distribution

According to P.R.Thies, the Weibull probability distribution is “one of the most widely used distribution functions in reliability applications as it is very flexible and the choice of different distribution parameters allows to model a multitude of failure behaviours” (Thies, 2012). So, in the next steps of the thesis, the failure distribution will be a Weibull function in which α is called the ageing parameter or shape parameter and (1/k) is the scale parameter. The scale parameter determines the statistical dispersion of the distribution.

The Weibull distribution is:

𝐹(𝑘, 𝛼, 𝑒) = 1 − 𝑒^{−(𝑘∗𝑡)}^𝛼 The Weibull failure rate is:

𝜆(𝑒) = 𝑘 ∗ 𝛼 ∗ (𝑘 ∗ 𝑒)^𝛼−1 The Weibull mean life or MTTF is ( Г() is the Gamma function):

𝑀𝑀𝑀𝐹 =1

𝑘 ∗ 𝛤(1 + 1 𝛼)

In case of constant failure rate, the exponential distribution can be used (Thies, 2012). The ageing is not taken into account and so the item is only subjected to a random number of failures.

Regarding the aging parameter, four values are considered in the prediction tool. The value 1 refers to the lack of aging. Then, the aging parameter increases with the importance of the aging.

(32)

-22-

Figure 19: Overview of the failure modeling strategy

Finally, from the Weibull distribution and the Excel function RAND(), the real life time (TTF) of a component is calculated with the following formula:

𝑀𝑀𝐹 =1

𝑘 ∗ exp (�

1

𝛼� ∗ ln(− ln(1 − 𝑅𝐶𝑅𝑅()))) 3.2.3 The management of failure

The prediction tool includes three different states regarding the evolution of the failure for 1 component (see Figure 20 below):

• The initial state

• The partial failure

• The total failure

When the partial failure is reached it means that the component is still working but a preventive maintenance is needed. Then if the preventive maintenance is not carried out, the total failure state is reached.

If no preventive maintenance operation is scheduled, the component passes directly from the initial state to the total failure.

The time spent by a component in each state is an input parameter.

(33)

-23-

Figure 20: View of the failures management scenarios

3.3 From the components to the functions

When working with functions, the big issue is to find both the failure rate and the severity cost linked to one function from the data of components included in this function.

3.3.1 The failure rate

In the prediction tool developed, when two components operate in parallel in the same function, the tool considers them as two different functions. That is why it is really important to have either all the components in parallel in one function or all the components in series.

If all the components are in series, the failure rate of the function is the sum of all the components failure rates (Thies, 2012).

3.3.2 The severity cost

Let’s take an example to illustrate how to calculate the severity cost of one function with several components. In this example, the function has two components and their features are shown in Table 4 below:

Table5: Input parameters of the 3.3.2 section example

The severity cost of the function is written Cf and can be calculated as the average cost of components pondered by the failure rates:

𝐿𝐶 = 𝜆1

(𝜆1 + 𝜆2) . 𝐿1 + 𝜆2

(𝜆1 + 𝜆2) . 𝐿2 Failure rate

Severity cost of the component

Component 1 λ1 C1 Component 2 λ2 C2 Function

(34)

-24-

3.4 Maintenance strategy The maintenance strategy includes two levels:

• The general maintenance strategy, illustrated in Figure 21.

• A maintenance strategy per function.

If the onshore maintenance strategy is chosen for the general maintenance procedures:

• If there is a failure on a function with an onshore strategy, a new system is connected;

• If there is a failure on a function with an offshore, the offshore strategy is implemented.

If the offshore maintenance strategy is chosen for the general maintenance procedures:

• If there is a failure on a function with an offshore strategy, the offshore strategy is implemented.

• If there is a failure on a function with an onshore strategy, only the components concerned are taken to the shore but not the whole system.

There is a second important parameter in the general maintenance strategy. Once the WEC is onshore, one can choose to check the whole WEC and replace old components even if there is no failure. So the user can set a life time limit before replacement.

Regarding the onshore maintenance, another parameter has to be taken into account: the time required to disconnect the system. It is obvious that this time depends on the connection technology used (dry-mate or wet-mate).

Otherwise, the maintenance of the grid export cable is taken into account in the general maintenance strategy. In the prediction tool, it is assumed that no failure can happen on the subsea cable. So the user has just to decide the number of maintenance operations foreseen for this component. Moreover, by default, a jack up vessel is used for this kind of maintenance.

Figure 21: Screen print of the main parameters of the general maintenance strategy Regarding the maintenance strategy per function, more options can be chosen.

In a run-to-failure strategy, there are two types of situations:

1. As soon as a failure occurs, a vessel is sent toward the WEC in order to fix the problem.

2. No direct intervention is carried out when a failure occurs but the user can schedule inspections of the whole array every 3 months, 6 months… During the inspection, all broken components are fixed.

(35)

-25-

In a preventive maintenance strategy, there are several operations available. Firstly, as soon as there is a failure, a vessel goes toward the system in order to fix the problem. Moreover, the user can schedule maintenance operation every 3 months, 6 months… On top of that, if a monitoring system exists, the user can decide to replace an old component before the failure occurs.

Generally, old components are less efficient and this data can be monitored.

One can notice that onshore maintenance can be carried out. For instance, in order to improve the bio-fouling protection of the system, one can bring it to the shore.

The scheme of the maintenance strategy per function is summarized in figure 22 below:

Figure 22: Diagram of the maintenance strategies available

The prediction tool takes into account two kinds of vessels:

• A workboat

• A jack-up vessel

The features of these boats are summarized in figures 23 and 24:

Onshore

Corrective

maintenance Run to failure

Preventive Maintenance

Scheduled Conditioned

Offshore

Corrective

maintenance Run to failure

Preventive maintenance

Scheduled

Conditioned

(36)

-26-

Figure 23: Screen print of the vessels input data

These features are already filled but they can be modified. These vessels are linked to the two operations available: the workboat is used for minor operations and the jack-up vessel is used for heavy operations.

Figure 24: Picture of the two kinds of maintenance vessels used

Regarding the human resources, in this prediction tool only one kind of technician is available.

The cost of such technician is set to 110 euros per hour and a day of work is 10 hours. All these values can be modified by the user.

(37)

-27-

3.5 Modelling of the surrounding environment

3.5.1 Scattered diagrams

In the “sea states” sheet two scattered diagrams are already filled (SEM-REV and YEU Island) and one is free.

The prediction tool needs these data to:

• Calculate the capacity factor (Cp) of one device on one specific site

• Calculate the weather window before intervention of the maintenance crew on one specific site.

The scattered diagram (Significant wave height (Hmo)/Average energy period (Te)) has to be filled with data in hours per year.

Regarding the SEM-REV (figure 25) and the Yeu Island (figure 26) data has been freely downloaded from the CANDHIS data base (Ministre de l'Ecologie, 2015).

Figure 25: The scattered diagram (Hmo/Te) of SEM-REV.

Figure 26: The scattered diagram (Hmo/Te) of the Yeu Island.

From the scattered diagram, it is possible to get the percentage of wave with a size under either 1m or 1,5m or 2m.

From these data, the prediction tool can calculate the mean number of days before getting wave with a size under a specified value, using a geometric law. It is assumed that the probability to expect waves under a specific size follows a geometric law.

(38)

-28- 3.5.2 Power matrix

The power matrix of the system has to be filled in the sheet “Power matrix”. Then, the prediction tool would multiply it with the scattered diagram selected so as to get the capacity factor of the system. For some systems or in certain locations it is very important to think about the scattered diagram before the power matrix since the power absorbed by the system depends on the sea states.

The power matrix (Hmo/Te) has to be filled with data in kW.

3.6 Monte Carlo Method

According to Michael Heath (Heath, 2002), Monte Carlo methods involve the use of random numbers as a tool to compute something that is not random. Therefore, Monte Carlo methods are not deterministic methods.

According to the same author, Monte Carlo methods are used when the problem is multi- dimensional and approximations that factor the problem into products of lower dimensional problems are inaccurate.

Monte Carlo methods are mainly used in three distinct problem classes: generating draws from a probability distribution, optimization and numerical integration.

Monte Carlo methods are widely used in reliability engineering to generate MTTF and MTTR for components. In such problems, MTTF and MTTR are linked to probability distributions.

(39)

-29-

3.7 Working principle of the developed tool

Regarding the framework of the prediction tool, it includes 6 different sheets (figure 27): 1 sheet with the results and 5 sheets including input parameters.

Figure 27: Screen print of the 6 sheets available in the prediction tool.

Regarding the calculation, there are one sheet per function, one sheet to synchronise all the functions and one sheet to iterate for several systems and on several life cycles. This is presented in Figure 28.

Figure 28: Overview of the framework of the prediction tool

On each function sheet, a probable life of the function is simulated taking into account both the probable life of the other functions and the global maintenance strategy.

For each function, in order to simulate time to failure, the Excel function RAND() is used to simulate random numbers. Then these numbers are pondered by the Weibull distribution of the function so as to get the right repartition of the times to failure.

In the Synchronisation sheet, two main variables are monitored:

1. The total time with at least one failure existing in the array multiplied with the right output power of each WEC.

2. In an onshore maintenance strategy, if one function has failed, this event can be read on this sheet. So, every function sheet knows that the WEC is going to be hauled back to the shore for maintenance.

3.8 The output parameters

In order to assess the reliability of one technology, several outputs can be relevant even though the reliability is indirectly included in its value: the CAPEX, the OPEX, the LCOE and the availability of the array.

The assumptions done are summarized below.

(40)

-30- 3.8.1 CAPEX

The CAPEX is split into five main costs:

• The development and consent (permission and certification) cost

• The installation and commissioning cost

• The purchase of devices

• The cost of spare part

• The decommissioning cost

In order to assess the installation costs, several assumptions are made.

It is assumed that a jack-up vessel is used to transport foundations and to drill if necessary. So the user has to provide input parameters for the jack-up vessel location time.

Next, in order to install or repair parts of the system, it is assumed that workboats are used. So the user has just to insert input parameters for the time needed to connect a system.

This scheme has been chosen since a lot of WEC technologies are buoys: the foundation and mooring are heavy to install but not the floating parts.

However, this scheme can be adapted to different cases: one can use only jack up vessel for all installations or just workboats when the foundations are light.

3.8.2 OPEX

To calculate the OPEX, three costs are taken into account:

• The costs of installation

• The costs of inspection

• The logistic costs

3.8.2.1 The cost of inspection

Within the inspection cost, three constituting parts are taken into account:

• The costs of manpower

• The hourly or daily rate cost of vessels times the time required to inspect the whole array

• The extra costs for the vessel to travel to the array and return to harbour.

3.8.2.2 The cost of installation

Within the installation cist, two constituting parts are taken into account:

• The costs of manpower

• The costs of components 3.8.2.3 The logistics cost

In the logistics costs, the costs for inspection are not taken into account. So this value includes:

• Daily rate of vessels

• Mobilization cost

• Costs to go and return to the array

(41)

-31-

3.8.3 Production and availability of the array

There are several factors impacting the availability of the array. These factors depend mainly on the maintenance strategy chosen.

In a run-to-failure strategy, if a failure occurs the WEC doesn’t produce power until the corrective operation is carried out. The idling time can be split into 4 categories:

• The spare part supply time

• The travelling time

• The mobilization time

• The average time to wait a suitable weather window (generally if the height of waves is above 2 meters, the vessels cannot leave the harbour).

In a preventive maintenance strategy, the WEC doesn’t produce power only during the replacement stage. In an onshore strategy, the WEC doesn’t produce power only during the disconnection/connection procedures.

Other losses are due to the power matrix of the WEC. Generally the capacity factor of such a system is expected to be between 20% and 40%.

3.8.4 LCOE

The LCOE is defined as:

𝐿𝐿𝐿𝐿 = 𝐿𝐶𝐶𝐿𝐶 + 𝐶𝑃(𝐿𝐶𝐿𝐶)

𝐶𝑃(𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒 𝑝𝑒𝑝𝑝𝑝𝑒𝑒𝑒𝑝𝑝)

If the decommissioning cost is not taken into account, only the OPEX and the electrical production are discounted in the calculation of the LCOE.

The LCOE is a very interesting value since it takes into consideration several parameters that can describe one technology:

• The discount rate represents the financial risk linked to one project

• The OPEX represents the reliability and the maintenance strategy

• The electrical production includes not only the performance of one technology but also the reliability and the weather conditions for one specific site

The only financial data existing in the tool is the discount rate used to calculate the net present values.

Regarding WEC projects, it is recommended to take a value of the discount rate between 8% and 15%. This value represents the risks taken by investors.

(42)

-32-

4 Case study

4.1 Objectives

The goal of this section is to illustrate the potential of the methodology and the tool developed.

Through this section, the three main steps of the methodology will be implemented:

1. The FMECA

2. The in depth study of one technology which includes a design optimization and a maintenance strategy optimization.

3. The comparison of several technologies based on sensitivity analysis.

4.2 Methodology

So as to illustrate the overall methodology, two different technologies will be under the spotlight (table 5). The first one will be called WEC_heaving and the second one will be called WEC_surging.

The WEC_heaving is a point absorber and has all the features of this category of WEC. This system captures the wave energy at sea surface where the energy available is maximum. It means that a 1MW system at the sea surface can produce a power close to 1MW more often that a 1MW system at the bottom of the sea.

However, there are two main issues with the point absorber system. At sea surface the energy available is higher but the fatigue due to wave loading is higher too. So this kind of system has to be designed for harsh constraints. On top of that, the energy is higher but the resonance has to be reached to get all the energy. So point absorbers need to have a very sophisticated motion control system in order to have a large spectrum for the resonance. As a result of these two issues, point absorbers are generally more expensive since they are over-designed and very complex.

On the opposite, the WEC_surging is a surge converter which lies at the bottom of the sea where the energy available is lower. This kind of system produces less but generally they are less expensive due to two factors:

• At the bottom of the sea the environment is less constraining

• They don’t need a very complex motion control system.

Table6: Summary of the WEC_heaving and WEC_surging characteristics

WEC_heaving WEC_surging

Produces more energy Produces less energy

Needs a sophisticated motion control system Doesn’t need a motion control system

Has a high CAPEX Has a lower CAPEX

The hydrodynamics explanation of WEC systems can be found in (Folley, et al., 2015).

Risk assessment of marine energy projects

Risk assessment of marine energy projects

Steven GUEGUEN

Abstract

Contents

List of Figures

List of Tables

Nomenclature

α

λ

Abbreviations

ACKNOWLEDGMENTS

1 INTRODUCTION

System 1 System 2 System 3 System 4

LC OE (e ur os /M W h)

WECs

LCOE Vs. MTTF

2 RELIABILITY AND MAINTAINABILITY

3 ECONOMIC ASSESSMENT

Onshore

Corrective

maintenance Run to failure

Preventive Maintenance

Scheduled Conditioned

Offshore

Corrective

maintenance Run to failure

Preventive maintenance

Scheduled

Conditioned

4 Case study