The Wave Energy Converter Design Process: Methods Applied in Industry and Shortcomings of Current Practices
Abstract
:1. Introduction
2. Generalized Definition of a WEC Design Project
2.1. Identify the Need and Clarify the Problem
2.2. Define Requirements and Functions
2.3. Metrics of Success
3. Stages of a WEC Design Process
3.1. Project and Product Definition
- Systems Engineering is a more traditional engineering practice than those mentioned above. Systems Engineering and standard product design methods guide designers through similar initial design steps of determining a mission, identifying stakeholders and stakeholder needs, identifying the functional requirements to satisfy those needs, and using the relationship between the stakeholder needs and the functional requirements to set targets for the functional requirements. Bull et al. applied the Systems Engineering approach to a wave energy farm to propose taxonomies for WEC capabilities (sometimes called stakeholder needs or customer requirements) and WEC functional requirements [32]. Systems Engineering encourages designers to consider the whole life cycle of the WEC and to decompose the WEC into rational subsystems.
- Set-Based Design is a design methodology which encourages designers to develop multiple concepts concurrently. Instead of choosing the best concept variant with the limited knowledge intrinsic of the conceptual design stages, Set-Based Design focuses on eliminating inferior concepts while iteratively defining and developing the other concepts in order to avoid choosing a concept based on imprecise data [55]. The methodology encourages designers to update their problem statement as they learn more about the problem. Set-Based Design has been acknowledged as particularly useful for design problems with high degrees of uncertainty [56], such as WEC design [49]. Delaying commitment to a single concept has shown to decrease the time and money spent throughout the design process [57]. Set-Based Design could help WEC designers follow the TPL-TRL curve suggested by Weber as a pathway toward successful commercialization [47] by integrating performance assessment prior to concept selection and refocusing WEC design toward performance rather than exclusively readiness.
- Axiomatic Design is a design theory for general systems, including non-physical systems, which uses a rigorous decision-making framework to guide designers toward rational designs of reduced complexity. Axiomatic Design theory is based on the theorem that the best design is the one in which all functional requirements are independent (Axiom I) and the information content is minimized (Axiom II). System architecture is defined using matrices and flow diagrams [48]. Proponents of Axiomatic Design claim that it reduces technical and business risk compared to heuristic design methods. Axiomatic Design was integrated into an early design stage of marine energy design by Ruiz-Minguela et al. and determined to help determine risk factors, focus designers on key properties of the system, and compare concept alternative [33]. The theoretical definition of a successful system in Axiomatic Design could help WEC designers assess their WECs with less uncertainty and its rigorous process for decision making could guide WEC designers toward better designs before they conduct detailed hydrodynamic modeling and testing campaigns. Non-physical requirements of WECs, such as community and government acceptability, may be more challenging, but by no means impossible to integrate into Axiomatic Design.
- Principles of Ecological Engineering provide guidance for the design of systems which are integrated into the natural environment. Typically, Ecological Engineering processes are applied to projects at the junction of ecology and engineering such as wetland restoration or or sustainable timber harvest [58], but Bergen et al. assert that the principles of Ecological Engineering can be applied to any engineered system which extracts natural resources [59]. This would include WECs harvesting energy from ocean waves. The principles of Ecological Engineering include the two design axioms from Axiomatic Design as well as “design consistent with ecological principles”, “design for site-specific context”, and “acknowledge the values and purposes that motivate design” [59]. The principles of Ecological Engineering could help WEC designers improve the resilience of WEC systems, better account for upstream and downstream effects, utilize natural ecological functions, and integrate their primary purpose throughout the design process. Ecological engineering practice may also lead to unique WEC concepts through its emphasis on functional diversity, site-specific solutions, and human values.
3.2. Conceptual Design
3.2.1. Concept Generation
3.2.2. Concept Evaluation
3.3. Embodiment Design
3.3.1. Numerical Modeling
3.3.2. Prototyping and Testing
3.4. Detail Design
4. Design For WEC Requirements
4.1. Power Production
4.2. Capital Cost
4.3. Operational Cost
4.4. Availability
4.5. Reliability and Survivability
4.6. Manufacturability and Materials Selection
4.7. Installation and Maintenance
4.8. Grid Integration
4.9. Environmental Impacts and Safety
4.10. Acceptability
4.11. Global Deployability
5. WEC Designer and Developer Methods
5.1. Survey Overview
5.2. Survey Results
5.2.1. Design Philosophy and Conceptual Design
5.2.2. Common Design Methods
5.2.3. Common Metrics and Evaluation Methods
5.2.4. Dynamic Modeling
5.2.5. Timeline of Requirement Consideration and Subsystem Design
5.2.6. Deployment Site-Agnostic Design
5.2.7. Designer Satisfaction
5.2.8. Under-Utilized Methods and Choosing a Method
- Set-Based Design.
- Ecological Engineering.
- Axiomatic Design.
- Ethnographic Need-Finding.
- Participatory Design.
- Quality Function Deployment.
- Conceptual Design Methods.
- Installation Storyboarding.
- Redundancy of Critical Function.
- Subsystem Co-design.
6. Conclusions
- Relating design decisions to customer requirements It will be the role of researchers to clarify how different design decisions impact a WEC’s ability to meet each design requirement and to create the tools that can help designers understand, visualize, and quantify those impacts. An example of such a tool would be one that relates design parameters to deployment site criteria in order to characterize how individual design decisions impact the wave resource available globally to a WEC.
- Early assessment of all design requirements Although usable power production is the primary goal of WECs, and improving power production continues to be the main focus of much of the academic research, wave energy development is at a point where many of the methods of energy absorption and conversion are well understood. For that reason, designers will need to begin to consider requirements other than power production and hydrodynamics earlier and more often. This will require assessment techniques geared toward WEC concepts with high uncertainty.
- Addressing grid integration and end use Grid integration is a requirement that consistently stood out among others. There were no common design or evaluation methods for grid integration, it was the requirement considered least often when making design decisions, the fewest respondents considered it prior to concept selection, and it was one of the requirements for which designers were least satisfied with the tools they had available.The widespread use of LCOE as a performance metric may contribute to the challenges designers face in designing for grid integration. The metric does not value any ancillary benefits that WECs could provide to the grid, which could become more important as more renewable energy sources come online. WEC designers need better tools for considering grid integration which are less computationally expensive than wave-to-wire models and do not require a fully-defined WEC concept.
- Conceptual design processes As has been emphasized in previous WEC design research, engaging in structured conceptual design processes stands to save WEC designers time and money. With so many WEC concepts being proposed, conceptual design methods can help designers begin with a clean sheet. Concept evaluation methods can offer designers opportunities to evaluate concepts before creating detailed models.
- Exploring new design philosophies As we have seen throughout this paper, systems engineering approaches tend to dominate the WEC design process although other design philosophies such as Ecological Engineering, Set-Based Design, and User-Center/Participatory design for emerging market WECs have the potential to guide WEC design in new directions. Further research is needed to determine whether any of these other design philosophies will lead to improvements in WEC design.
- The impacts of model surrogates As discussed in Section 3.3.1, WEC designers may use surrogate representations of subsystems in early numerical models of WECs. How they do so depends on the prioritization of subsystems, which we analyzed for the survey respondents in Section 5.2.4. No research exists which explores the impacts of using these surrogates on the eventual performance of a WEC device. Such research could better inform design approaches (such as the extent to which co-design should be implemented), as well as the way that designers decompose WEC subsystems.
- Materials selection at various design stages Prototype testing and the deployment of scaled WECs will be essential to gaining the experience necessary to drive down costs, reduce risk, and gain acceptance in the public eye. Gaining a better understanding of what components can be tested and what investigations can be performed at various scales of prototyping and how results scale to the full-sized WEC can help researchers and developers determine ways to cut material and manufacturing costs of prototyping.
- Need-finding and site-specific design Given the opportunities for WECs which include grid-scale development and emerging market off-grid development as well as the driver of WEC development—climate change—there is more than one potential path for wave energy. Although we summarize stakeholder and functional WEC requirements in this paper, a particular project or site will have its own set of unique requirements. Developers should not forgo the need-finding design practices that allow them to determine those unique requirements. Just as the device requirements are site-specific, researchers have shown that the economic viability of a WEC is also site-specific. These facts challenge us to more closely evaluate the meaning and value of technology convergence and global deployability to determine the best pathway for WEC development. The pathway chosen will, as discussed, impact which design methodologies which are most appropriate.
Funding
Acknowledgments
Conflicts of Interest
Abbreviations
WEC | Wave Energy Converter |
PTO | Power Take-Off |
LCOE | Levelized Cost of Energy |
CWR | Capture Width Ratio |
MAEP | Mean Average Energy Production |
NPV | Net Present Value |
IRR | Inertial Rate of Returns |
PBP | Payback Period |
HPQ | Hydrodynamic Performance Quality |
ACE | Average climate capture width divided by characteristic capital expenditure |
TRL | Technology readiness Level |
TPL | Technology Performance Level |
QFD | Quality Function Deployment |
HoQ | House of Quality |
AUV | Autonomous Underwater Vehicle |
TRIZ | Theory of Inventive Problem Solving |
DSM | Design Structure Matrix |
SWOT | Strengths Weakness Opportunities and Threats |
BEM | Boundary Elements Methods |
CFD | Computational Fluid Dynamics |
SPH | Smoothed Particle Hydrodynamics |
OWC | Oscillating Water Column |
CAD | Computer Aided Design |
CAPEX | Capital Expenditure |
OPEX | Operational Expenditure |
MTBF | Mean Time Between Failure |
FMEA | Failure Modes and Effects Analysis |
VMEA | Variations Modes and Effects Analysis |
EMEC | European Marine Energy Centre |
IEC | International Electrotechnical Commission |
NREL | National Renewable Energy Laboratory |
MRL | Manufacturing Readiness Level |
DFM | Design for Manufacturing |
DfE | Design for the Environment |
LCA | Lifetime Cost Analysis |
GIS | Geographic Information Systems |
WSTAT | Whole System Trades Analysis |
References
- Edenhofer, O.; Madruga, R.P.; Sokona, Y.; Seyboth, K.; Matschoss, P.; Kadner, S.; Zwickel, T.; Eickemeier, P.; Hansen, G.; Schlömer, S.; et al. Renewable Energy Sources and Climate Change Mitigation—IPCC; Technical Report; Cambridge Univerrsity Press: New York, NY, USA, 2012. [Google Scholar]
- Reikard, G.; Robertson, B.; Bidlot, J.R. Combining wave energy with wind and solar: Short-term forecasting. Renew. Energy 2015, 81, 442–456. [Google Scholar] [CrossRef]
- EERE. Powering the Blue Economy: Exploring Opportunities for Marine Renewable Energy in Maritime Markets; Technical Report; U.S. Department of Energy, Office of Energy Efficiency and Renewable Energy: Washington, DC, USA, 2019.
- Clément, A.; McCullen, P.; Falcão, A.; Fiorentino, A.; Gardner, F.; Hammarlund, K.; Lemonis, G.; Lewis, T.; Nielsen, K.; Petroncini, S.; et al. Wave energy in Europe: Current status and perspectives. Renew. Sustain. Energy Rev. 2002, 6, 405–431. [Google Scholar] [CrossRef]
- Robertson, B.; Hiles, C.; Luczko, E.; Buckham, B. Quantifying wave power and wave energy converter array production potential. Int. J. Mar. Energy 2016, 14, 143–160. [Google Scholar] [CrossRef]
- Langhamer, O.; Haikonen, K.; Sundberg, J. Wave power-Sustainable energy or environmentally costly? A review with special emphasis on linear wave energy converters. Renew. Sustain. Energy Rev. 2010, 14, 1329–1335. [Google Scholar] [CrossRef]
- Babarit, A. A database of capture width ratio of wave energy converters. Renew. Energy 2015, 80, 610–628. [Google Scholar] [CrossRef]
- Ruehl, K.; Bull, D. Wave Energy Development Roadmap: Design to commercialization. In Proceedings of the OCEANS 2012 MTS/IEEE: Harnessing the Power of the Ocean, Hampton Roads, VA, USA, 14–19 October 2012. [Google Scholar] [CrossRef]
- DTOceanPlus—Design Tools for Ocean Energy Systems; DTOcean2 Consortium: Edinburgh, UK, 2020.
- MaRINET2. Objectives—MaRINET2; Marine Renewables Infrastructure Network; European Union: Brussels, Belgium, 2020. [Google Scholar]
- Weber, J.; Roberts, J. Wave-SPARC Systematic Process and Analysis for Reaching Commercialization; Technical Report; U.S Department of Energy Office of Energy Efficiency and Renewable Energy: Washington, DC, USA, 2019.
- Aderinto, T.; Li, H. Ocean Wave energy converters: Status and challenges. Energies 2018, 11, 1250. [Google Scholar] [CrossRef] [Green Version]
- Falcão, A.F.O. Wave energy utilization: A review of the technologies. Renew. Sustain. Energy Rev. 2010, 14, 899–918. [Google Scholar] [CrossRef]
- Ullman, D. The Mechanical Design Process, 6th ed.; McGraw Hill: New York, NY, USA, 2017. [Google Scholar]
- Falnes, J. A review of wave-energy extraction. Mar. Struct. 2007, 20, 185–201. [Google Scholar] [CrossRef]
- Sims, R.E.; Schock, R.N.; Adegbululgbe, A.; Fenhann, J.; Konstantinaviciute, I.; Moomaw, W.; Nimir, H.B.; Schlamadinger, B.; Torres-Martínez, J.; Turner, C.; et al. Energy Supply. In Climate Change 2007: Mitigation of Climate Change. Contribution of Working Group III to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change; Metz, B., Davidson, O., Bosch, P., Dave, R., Meyer, L., Eds.; Cmabridge University Press: Cmabridge, UK, 2007; pp. 251–322. [Google Scholar]
- Krewitt, W.; Nienhaus, K.; Kleßmann, C.; Capone, C.; Stricker, E.; Graus, W.; Hoogwijk, M.; Supersberger, N.; von Uta, W.; Sascha, S. Role and Potential of Renewable Energy and Energy Efficiency for Global Energy Supply; Technical Report; Federal Environment Agency: Dessau-Roßlau, Germany, 2009.
- Mørk, G.; Barstow, S.; Kabuth, A.; Pontes, M.T. Assessing the global wave energy potential. In Proceedings of the International Conference on Offshore Mechanics and Arctic Engineering—OMAE, Shanghai, China, 6–11 June 2010; Volume 3, pp. 447–454. [Google Scholar] [CrossRef]
- Gunn, K.; Stock-Williams, C. Quantifying the global wave power resource. Renew. Energy 2012, 44, 296–304. [Google Scholar] [CrossRef]
- Hughes, M.G.; Heap, A.D. National-scale wave energy resource assessment for Australia. Renew. Energy 2010, 35, 1783–1791. [Google Scholar] [CrossRef]
- Liberti, L.; Carillo, A.; Sannino, G. Wave energy resource assessment in the Mediterranean, the Italian perspective. Renew. Energy 2013, 50, 938–949. [Google Scholar] [CrossRef]
- Guedes Soares, C.; Bento, A.R.; Gonçalves, M.; Silva, D.; Martinho, P. Numerical evaluation of the wave energy resource along the Atlantic European coast. Comput. Geosci. 2014, 71, 37–49. [Google Scholar] [CrossRef]
- Neary, V.; Copping, A.; Rieks, J.; Lawson, M.; Hallett, K.C.; Murray, D.; Previsic, M.; LaBonte, A. Methodology for design and economic analysis of marine energy conversion technologies. In Proceedings of the 2nd Marine Energy Technology Symposium, Seattle, WA, USA, 15–18 April 2014. [Google Scholar]
- De Andrés, A.; Macgillivray, A.; Guanche, R.; Jeffrey, H. Factors affecting LCOE of Ocean energy technologies: A study of technology and deployment attractiveness. In Proceedings of the International Conference on Ocean Energy, Halifax, NS, Canada, 4–6 November 2014. [Google Scholar]
- Chang, G.; Jones, C.A.; Roberts, J.D.; Neary, V.S. A comprehensive evaluation of factors affecting the levelized cost of wave energy conversion projects. Renew. Energy 2018, 127, 344–354. [Google Scholar] [CrossRef]
- O’Hagan, A.M.; Huertas, C.; O’Callaghan, J.; Greaves, D. Wave energy in Europe: Views on experiences and progress to date. Int. J. Mar. Energy 2016, 14, 180–197. [Google Scholar] [CrossRef] [Green Version]
- Cantarero, M.V.; Domene, G.A.; Noble, D.R.; Pennock, S.; Jeffrey, H.; Ruiz, P.M.; Tunga, I.; Morrison, N.; Apolonia, M.; Luxcey, N.; et al. D8.1 Potential Markets for Ocean Energy; Technical Report; DTOceanPlus: Edinburgh, UK, 2020. [Google Scholar]
- Caio, A.; Davey, T.; Mcnatt, C. Tackling the Wave Energy Paradox—Stepping towards Commercial Deployment. In Proceedings of the Twenty-Ninth (2019) International Ocean and Polar Engineering Conference, Honolulu, HI, USA, 16–21 June 2019. [Google Scholar]
- Robertson, B.; Bekker, J.; Buckham, B. Renewable integration for remote communities: Comparative allowable cost analyses for hydro, solar and wave energy. Appl. Energy 2020, 264, 114677. [Google Scholar] [CrossRef]
- Clark, C.E.; Miller, A.; DuPont, B. An analytical cost model for co-located floating wind-wave energy arrays. Renew. Energy 2019, 132, 885–897. [Google Scholar] [CrossRef]
- Babarit, A.; Bull, D.; Dykes, K.; Malins, R.; Nielsen, K.; Costello, R.; Roberts, J.; Bittencourt Ferreira, C.; Kennedy, B.; Weber, J. Stakeholder requirements for commercially successful wave energy converter farms. Renew. Energy 2017, 113, 742–755. [Google Scholar] [CrossRef]
- Bull, D.; Roberts, J.; Malins, R.; Babarit, A.; Weber, J.; Dykes, K.; Costello, R.; Kennedy, B.; Neilson, K. Systems Engineering Applied to the Development of a Wave Energy Farm; Technical Report; Sandia National Laboratories: Albuquerque, NM, USA, 2017.
- Ruiz-minguela, P.; Blanco, J.M.; Nava, V. Novel methodology for holistic assessment of wave energy design options. In Proceedings of the 13th European Wave and Tidal Energy Conference, Naples, Italy, 1–6 September 2019. [Google Scholar]
- De Andres, A.; Medina-Lopez, E.; Crooks, D.; Roberts, O.; Jeffrey, H. On the reversed LCOE calculation: Design constraints for wave energy commercialization. Int. J. Mar. Energy 2017, 18, 88–108. [Google Scholar] [CrossRef]
- De Andres, A.; MacGillivray, A.; Roberts, O.; Guanche, R.; Jeffrey, H. Beyond LCOE: A study of ocean energy technology development and deployment attractiveness. Sustain. Energy Technol. Assess. 2017, 19, 1–16. [Google Scholar] [CrossRef]
- Frost, C.; Findlay, D.; Macpherson, E.; Sayer, P.; Johanning, L. A model to map levelised cost of energy for wave energy projects. Ocean Eng. 2018, 149, 438–451. [Google Scholar] [CrossRef]
- Dallman, A.; Jenne, D.S.; Neary, V.; Driscoll, F.; Thresher, R.; Gunawan, B. Evaluation of performance metrics for the Wave Energy Prize converters tested at 1/20th scale. Renew. Sustain. Energy Rev. 2018, 98, 79–91. [Google Scholar] [CrossRef]
- Babarit, A.; Hals, J.; Muliawan, M.J.; Kurniawan, A.; Moan, T.; Krokstad, J. Numerical benchmarking study of a selection of wave energy converters. Renew. Energy 2012, 41, 44–63. [Google Scholar] [CrossRef]
- Hiles, C.E.; Beatty, S.J.; de Andres, A. Wave energy converter annual energy production uncertainty using simulations. J. Mar. Sci. Eng. 2016, 4, 53. [Google Scholar] [CrossRef] [Green Version]
- Guanche, R.; de Andrés, A.D.; Simal, P.D.; Vidal, C.; Losada, I.J. Uncertainty analysis of wave energy farms financial indicators. Renew. Energy 2014, 68, 570–580. [Google Scholar] [CrossRef]
- Henriques, J.C.; Portillo, J.C.; Gato, L.M.; Gomes, R.P.; Ferreira, D.N.; Falcão, A.F. Design of oscillating-water-column wave energy converters with an application to self-powered sensor buoys. Energy 2016, 112, 852–867. [Google Scholar] [CrossRef] [Green Version]
- Clarke, F.; Lorenzoni, A. Applied Cost Engineering; CRC Press: Boca Raton, FL, USA, 1979. [Google Scholar]
- Rush, C.; Roy, R. Analysis of cost estimating processes used within a concurrent engineering environment throughout a product life cycle. In Proceedings of the CE2000 Conference, Lyon, France, 17–20 July 2000. [Google Scholar]
- Pahl, G.; Beitz, W.; Feldhuen, J.; Grote, K.H. Engineering Design: A Systematic Approach, 3rd ed.; Springer International Publishing: Berlin/Heidelberg, Germany, 2007. [Google Scholar]
- Portillo, J.C.; Collins, K.M.; Gomes, R.P.; Henriques, J.C.; Gato, L.M.; Howey, B.D.; Hann, M.R.; Greaves, D.M.; Falcão, A.F. Wave energy converter physical model design and testing: The case of floating oscillating-water-columns. Appl. Energy 2020, 278, 115638. [Google Scholar] [CrossRef]
- Falcão, A.F.; Sarmento, A.J.; Gato, L.M.; Brito-Melo, A. The Pico OWC wave power plant: Its lifetime from conception to closure 1986–2018. Appl. Ocean Res. 2020, 98, 102104. [Google Scholar] [CrossRef]
- Weber, J. WEC Technology Readiness and Performance Matrix—Finding the best research technology development trajectory. In Proceedings of the International Conference on Ocean Energy, ICOE, Dublin, Ireland, 17 October 2012; pp. 1–10. [Google Scholar]
- Suh, N.P. Axiomatic Design Theory for Systems. Res. Eng. Des. 1998, 10, 189–209. [Google Scholar] [CrossRef]
- Trueworthy, A.M.; DuPont, B.L.; Maurer, B.R.; Cavagnaro, R.J. A Set-Based Design Approach for the Design of High—Performance Wave Energy Converters. In Proceedings of the European Wave and Tidal Energy Conference, Naples, Italy, 1–6 September 2019; Volume 98105. [Google Scholar]
- Tunga, I.; Bradley, S.; Eraut, N.; Bowick, L.; Noble, D.; Henderson, J. D3.1 Technical Requirements for the Implementation of Structured Innovation in Ocean Energy Systems; Technical Report 1.0, Report by DTOceanPlus. Report for DTOceanPlus; DTOceanPlus: Edinburgh, UK, 2019. [Google Scholar]
- Sahsa Costanza-Chock. Design Justice: Community-Led Practices to Build the Worlds We Need; MIT Press: Cambridge, MA, USA, 2020; pp. 77–101. [Google Scholar]
- Smith, R.C.; Bossen, C.; Kanstrup, A.M. Participatory design in an era of participation. CoDesign 2017, 13, 65–69. [Google Scholar] [CrossRef] [Green Version]
- Kurfman, M.A.; Stone, R.B.; Rajan, J.R.; Wood, K.L. Functional modeling experimental studies. In Proceedings of the ASME Design Engineering Technical Conference, Pittsburgh, PA, USA, 9–12 September 2001; Volume 4, pp. 267–279. [Google Scholar]
- Zurita, N.F.; Stone, R.B.; Demirel, O.; Tumer, I.Y. The Function-Human Error Design Method (FHEDM). In Proceedings of the ASME Design Engineering Technical Conference, Quebec City, QC, Canada, 26–29 August 2018; Volume 7. [Google Scholar] [CrossRef]
- Singer, D.J.; Doerry, N.; Buckley, M.E. What is set-based design? Nav. Eng. J. 2009, 121, 31–43. [Google Scholar] [CrossRef]
- Finch, W.W.; Ward, A.C. A Set-Based System for Eliminating Infeasible Designs in Engineering Problems Dominated by Uncertainty. In Proceedings of the ASME International Design Engineering Technical Conferences, Sacramento, CA, USA, 14–17 September 1997; pp. 1–12. [Google Scholar]
- Ward, A.; Liker, J.K.; Cristiano, J.J.; Sobek, D.K. The second Toyota paradox: How delaying decisions can make better cars faster. In MIT Sloan Managment Review; Spring: Berlin/Heidelberg, Germany, 1995. [Google Scholar] [CrossRef]
- Mitsch, W.J. What is ecological engineering? Ecol. Eng. 2012, 45, 5–12. [Google Scholar] [CrossRef]
- Bergen, S.D.; Bolton, S.M.; Fridley, J.L. Design principles for ecological engineering. Ecol. Eng. 2001, 18, 201–210. [Google Scholar] [CrossRef]
- Ruellan, M.; Ahmed, H.B.; Multon, B.; Josset, C.; Babarit, A.; Clément, A.; Benahmed, H.; Babarit, A.; Clement, A. Design Methodology for a SEAREV Wave Energy Converter. IEEE Trans. Energy Convers. 2010, 25, 760–767. [Google Scholar] [CrossRef]
- Boscaino, V.; Cipriani, G.; Di Dio, V.; Franzitta, V.; Trapanense, M. Experimental test and simulations on a linear generator-based prototype of a wave energy conversion system designed with a reliability-oriented approach. Sustainability 2017, 9, 98. [Google Scholar] [CrossRef] [Green Version]
- Bracco, G.; Giorcelli, E.; Mattiazzo, G. ISWEC: Design of a prototype model for wave tank testing. In Proceedings of the ASME 10th Biennial Conference on Engineering Systems Design and Analysis, Istanbul, Turkey, 12–14 July 2010. [Google Scholar]
- Cordonnier, J.; Gorintin, F.; De Cagny, A.; Clément, A.; Babarit, A. SEAREV: Case study of the development of a wave energy converter. Renew. Energy 2015, 80, 40–52. [Google Scholar] [CrossRef]
- Costello, R.; Nielsen, K.; Weber, J.; Tom, N.; Roberts, J. WaveSPARC: Evaluation of Innovation Techniques for Wave Energy. In Proceedings of the 13th Euopean Wave and Tidal Energy Conference, Naples, Italy, 1–6 September 2019. [Google Scholar]
- PAO and Olivier Goguel. TRIZ Matrix/40 Principles/TRIZ Contradictions Table; SolidCreativity: Bordeaux, France, 2014. [Google Scholar]
- Zhang, H.; Aggidis, G.A. Nature rules hidden in the biomimetic wave energy converters. Renew. Sustain. Energy Rev. 2018, 97, 28–37. [Google Scholar] [CrossRef] [Green Version]
- Wang, N.; Zou, J.; Yang, Y.; Li, X.; Guo, Y.; Jiang, C.; Jia, X.; Cao, X. Kelp-inspired biomimetic triboelectric nanogenerator boosts wave energy harvesting. Nano Energy 2019, 55, 541–547. [Google Scholar] [CrossRef]
- So, R.; Michelen, C.; Bosma, B.; Lenee-Bluhm, P.; Brekken, T.K. Statistical Analysis of a 1:7 Scale Field Test Wave Energy Converter Using WEC-Sim. IEEE Trans. Sustain. Energy 2017, 8, 1118–1126. [Google Scholar] [CrossRef]
- Strawbridge, Z.; McAdams, D.A.; Stone, R.B. A computational approach to conceptual design. In Proceedings of the ASME Design Engineering Technical Conference, Montreal, QC, Canada, 29 September–2 October 2002; Volume 4, pp. 15–25. [Google Scholar] [CrossRef]
- Sobek, D.; Liker, J.K. Toyota’s Principles of Set-Based Concurrent Engineering. MIT Sloan Manag. Rev. 1999, 40, 67. [Google Scholar]
- Truong, K.N.; Hayes, G.R.; Abowd, G.D. Storyboarding: An empirical determination of best practices and effective guidelines. In Proceedings of the Conference on Designing Interactive Systems: Processes, Practices, Methods, and Techniques, DIS, University Park, PA, USA, 26–28 June 2006; Volume 2006, pp. 12–21. [Google Scholar]
- Alfaris, A.; Siddiqi, A.; Rizk, C.; De Weck, O.; Svetinovic, D. Hierarchical decomposition and multidomain formulation for the design of complex sustainable systems. J. Mech. Des. Trans. ASME 2010, 132, 091003. [Google Scholar] [CrossRef] [Green Version]
- Guindon, R. Designing the Design Process: Exploiting Opportunistic thoughts Guindon Contents. Hum.-Comput. Interact. 1990, 5, 305–344. [Google Scholar] [CrossRef]
- Liu, T.; Azarm, S.; Chopra, N. Decentralized Multisubsystem Co-Design Optimization Using Direct Collocation and Decomposition-Based Methods. J. Mech. Des. 2020, 142. [Google Scholar] [CrossRef]
- OSullivan, D.; Murray, D.; Hayes, J.; Egan, M.G.; Lewis, A.W. The Benefits of Device Level Short Term Energy Storage in Ocean Wave Energy Converters. In Energy Storage in the Emerging Era of Smart Grids; IntechOpen: Rijeka, Croatia, 2011. [Google Scholar] [CrossRef] [Green Version]
- Ringwood, J.V.; Bacelli, G. Control of Wave-Energy Converters. IEEE Control Syst. Mag. 2014, 34, 30–55. [Google Scholar]
- Garcia-Rosa, P.B.; Bacelli, G.; Ringwood, J.V. Control-informed geometric optimization of wave energy converters: The impact of device motion and force constraints. Energies 2015, 8, 2386. [Google Scholar] [CrossRef]
- Gilloteaux, J.; Ringwood, J. Control-informed geometric optimisation of wave energy converters. IFAC Proc. Vol. 2010, 43, 399–404. [Google Scholar] [CrossRef] [Green Version]
- Garcia-Rosa, P.B.; Bacelli, G.; Ringwood, J.V. Control-Informed Optimal Array Layout for Wave Farms. IEEE Trans. Sustain. Energy 2015, 6, 575–582. [Google Scholar] [CrossRef] [Green Version]
- Eppinger, S.D. The Design Structure Matrix—DSM; MIT Press: Cambridge, MA, USA, 2005. [Google Scholar]
- Malak, R.J.; Aughenbaugh, J.M.; Paredis, C.J. Multi-attribute utility analysis in set-based conceptual design. Comput.-Aided Des. 2009, 41, 214–227. [Google Scholar] [CrossRef]
- Upadhye, A.A.; Qi, W.; Huber, G.W. Conceptual process design: A systematic method to evaluate and develop renewable energy technologies. AIChE J. 2011, 57, 2292–2301. [Google Scholar] [CrossRef]
- Wisthoff, A.; DuPont, B. A method for understanding sustainable design trade-offs during the early design phase. In Smart Innovation, Systems and Technologies; Springer Science and Business Media Deutschland GmbH: Berlin/Heidelberg, Germany, 2016; Volume 52, pp. 271–280. [Google Scholar]
- Bubbar, K.; Buckham, B.; Wild, P. A method for comparing wave energy converter conceptual designs based on potential power capture. Renew. Energy 2018, 115, 797–807. [Google Scholar] [CrossRef]
- Folley, M.; Babarit, A.; Child, B.; Forehand, D.; O’Boyle, L.; Silverthorne, K.; Spinneken, J.; Stratigaki, V.; Troch, P. A review of numerical modelling of wave energy converter arrays. In Proceedings of the International Conference on Offshore Mechanics and Arctic Engineering—OMAE, Rio de Janeiro, Brazil, 1–6 July 2012; American Society of Mechanical Engineers Digital Collection: New York, NY, USA, 2012; Volume 7, pp. 535–546. [Google Scholar] [CrossRef] [Green Version]
- Penalba, M.; Kelly, T.; Ringwood, J.V. Using NEMOH for Modelling Wave Energy Converters: A Comparative Study with WAMIT. In Proceedings of the 12th European Wave and Tidal Energy Conference, Cork, Ireland, 27 August–1 September 2017; p. 10. [Google Scholar]
- Wendt, F.; Nielsen, K.; Yu, Y.H.; Bingham, H.; Eskilsson, C.; Kramer, M.; Babarit, A.; Bunnik, T.; Costello, R.; Wendt, F. Ocean Energy Systems Wave Energy Modelling Task: Modelling, Verification and Validation of Wave Energy Converters. J. Mar. Sci. Eng. 2019, 7, 379. [Google Scholar] [CrossRef] [Green Version]
- Ancellin, M.; Dias, F. Capytaine: A Python-based linear potential flow solver. J. Open Source Softw. 2019, 4, 1341. [Google Scholar] [CrossRef]
- Sheng, W.; Alcorn, R.; Lewis, T. Physical modelling of wave energy converters. Ocean Eng. 2014, 84, 29–36. [Google Scholar] [CrossRef]
- Gaebele, D.T.; Magaña, M.E.; Brekken, T.K.; Sawodny, O. State space model of an array of oscillating water column wave energy converters with inter-body hydrodynamic coupling. Ocean Eng. 2020, 195, 106668. [Google Scholar] [CrossRef]
- Yu, Y.H.; Lawson, M.; Ruehl, K.; Michelen, C. Development and Demonstration of the WEC-Sim Wave Energy Converter Simulation Tool. In Proceedings of the 2nd Marine Energy Technology Symposium, Seattle, WA, USA, 15–18 April 2014. [Google Scholar]
- Sirigu, S.A.; Foglietta, L.; Giorgi, G.; Bonfanti, M.; Cervelli, G.; Bracco, G.; Mattiazzo, G. Techno-Economic Optimisation for a Wave Energy Converter via Genetic Algorithm. J. Mar. Sci. Eng. 2020, 8, 482. [Google Scholar] [CrossRef]
- Li, Y.; Yu, Y.H. A synthesis of numerical methods for modeling wave energy converter-point absorbers. Renew. Sustain. Energy Rev. 2012, 16, 4352–4364. [Google Scholar] [CrossRef]
- Davidson, J.; Costello, R. Efficient nonlinear hydrodynamic models for wave energy converter design—A scoping study. J. Mar. Sci. Eng. 2020, 8, 35. [Google Scholar] [CrossRef] [Green Version]
- Davidson, J.; Giorgi, S.; Ringwood, J.V. Linear parametric hydrodynamic models for ocean wave energy converters identified from numerical wave tank experiments. Ocean Eng. 2015, 103, 31–39. [Google Scholar] [CrossRef] [Green Version]
- Paduano, B.; Giorgi, G.; Gomes, R.P.; Pasta, E.; Henriques, J.C.; Gato, L.M.; Mattiazzo, G. Experimental validation and comparison of numerical models for the mooring system of a floating wave energy converter. J. Mar. Sci. Eng. 2020, 8, 565. [Google Scholar] [CrossRef]
- Yu, Y.H.; Jenne, D.S.; Thresher, R.; Copping, A.; Geerlofs, S.; Hanna, L.A. Reference Model 5 (RM5): Oscillating Surge Wave Energy Converter; Technical Report; National Renewable Energy Laboratory: Golden, CO, USA, 2013.
- Bull, D.; Smith, C.; Jenne, D.S.; Jacob, P.; Copping, A.; Willits, S.; Fontaine, A.; Brefort, D.; Copeland, G.; Gordon, M.; et al. Reference Model 6 (RM6): Oscillating Wave Energy Converter; Technical Report; Sandia National Laboratories: Albuquerque, NM, USA, 2014.
- Davidson, J.; Ringwood, J.V. Mathematical modelling of mooring systems for wave energy converters—A review. Energies 2017, 10, 666. [Google Scholar] [CrossRef] [Green Version]
- O’Connor, M.; Lewis, T.; Dalton, G. Techno-economic performance of the Pelamis P1 and Wavestar at different ratings and various locations in Europe. Renew. Energy 2013, 50, 889–900. [Google Scholar] [CrossRef]
- Topper, M.B.; Olson, S.S.; Roberts, J.D. Techno-economic modelling of tidal energy converter arrays in the tacoma narrows. J. Mar. Sci. Eng. 2020, 8, 646. [Google Scholar] [CrossRef]
- Têtu, A.; Frigaard, P.; Kofoed, J.P.; Lopes, M.; Iyer, A.; Bourdier, S.; Marina, D.; Stripling, S.; Johanning, L. WP2: Marine Energy System Testing-Standardisation and Best Practice Deliverable 2.5 EC Report on Instrumentation Best Practice; Technical Report; MARINET2; European Union: Brussels, Belgium, 2013. [Google Scholar]
- Falcão, A.F.; Henriques, J.C. Model-prototype similarity of oscillating-water-column wave energy converters. Int. J. Mar. Energy 2014, 6, 18–34. [Google Scholar] [CrossRef]
- Sheng, W.; Alcorn, R.; Lewis, A. Assessment of primary energy conversions of oscillating water columns. I. Hydrodynamic analysis. J. Renew. Sustain. Energy 2014, 6. [Google Scholar] [CrossRef] [Green Version]
- Schmitt, P.; Elsäßer, B. The application of Froude scaling to model tests of Oscillating Wave Surge Converters. Ocean Eng. 2017, 141, 108–115. [Google Scholar] [CrossRef] [Green Version]
- Thiébaut, F.; Sutton, G.; Johnstone, C. WP2: Marine Energy System Testing—Standardisation and Best Practice D2.28Model Construction Methods; Technical Report; MARINET2; European Union: Brussels, Belgium, 2015. [Google Scholar]
- Herrmann, J.W.; Cooper, J.; Gupta, S.K.; Hayes, C.C.; Ishii, K.; Kazmer, D.; Sandborn, P.A.; Wood, W.H. New directions in design for manufacturing. In Proceedings of the ASME Design Engineering Technical Conference, Salt Lake City, UT, USA, 28 September–2 October 2004; Volume 3, pp. 853–861. [Google Scholar]
- Arora, J.S. Introduction to Optimum Design, 4th ed.; Academic Press: Cambridge, MA, USA, 2016. [Google Scholar]
- Thomsen, J.B.; Ferri, F.; Kofoed, J.P.; Black, K. Cost optimization of mooring solutions for large floating wave energy converters. Energies 2018, 11, 159. [Google Scholar] [CrossRef] [Green Version]
- Cretel, J.A.; Lightbody, G.; Thomas, G.P.; Lewis, A.W. Maximisation of energy capture by a wave-energy point absorber using model predictive control. IFAC Proc. Vol. 2011, 44, 3714–3721. [Google Scholar] [CrossRef] [Green Version]
- Shadman, M.; Estefen, S.F.; Rodriguez, C.A.; Nogueira, I.C. A geometrical optimization method applied to a heaving point absorber wave energy converter. Renew. Energy 2018, 115, 533–546. [Google Scholar] [CrossRef]
- Abdelkhalik, O.; Coe, R.G.; Bacelli, G.; Wilson, D.G. WEC geometry optimization with advanced control. In Proceedings of the International Conference on Offshore Mechanics and Arctic Engineering—OMAE, Trondheim, Norway, 25–30 June 2017; American Society of Mechanical Engineers (ASME): New York, NY, USA, 2017; 10. [Google Scholar] [CrossRef]
- Clark, C.E.; Garcia-Teruel, A.; DuPont, B.; Forehand, D. Towards Reliability-Based Geometry Optimization of a Point-Absorber with PTO Reliability Objectives. In Proceedings of the European Wave and Tidal Energy Conference, Naples, Italy, 1–6 September 2019. [Google Scholar]
- Alamian, R.; Shafaghat, R.; Safaei, M.R. Multi-objective optimization of a pitch point absorber wave energy converter. Water 2019, 11, 969. [Google Scholar] [CrossRef] [Green Version]
- Koh, H.J.; Ruy, W.S.; Cho, I.H.; Kweon, H.M. Multi-objective optimum design of a buoy for the resonant-type wave energy converter. J. Mar. Sci. Technol. 2015, 20, 53–63. [Google Scholar] [CrossRef]
- Gomes, R.P.; Henriques, J.C.; Gato, L.M.; Falcão, A.F. Hydrodynamic optimization of an axisymmetric floating oscillating water column for wave energy conversion. Renew. Energy 2012, 44, 328–339. [Google Scholar] [CrossRef]
- Giassi, M.; Göteman, M. Layout design of wave energy parks by a genetic algorithm. Ocean Eng. 2018, 154, 252–261. [Google Scholar] [CrossRef]
- Sharp, C.; DuPont, B. Wave energy converter array optimization: A genetic algorithm approach and minimum separation distance study. Ocean Eng. 2018, 163, 148–156. [Google Scholar] [CrossRef]
- Pichard, A.; Wale, C.; Rafiee, A. Techno-economical tools for WEC scale optimisation. In Proceedings of the 13th Euopean Wave and Tidal Energy Conference, Naples, Italy, 1–6 September 2019; pp. 1–8. [Google Scholar]
- Bull, D.; Costello, R.; Babarit, A.; Nielsen, K.; Ferreira, C.B.; Kennedy, B.; Malins, R.; Dykes, K.; Roberts, J.; Pi, W. Technology Performance Level Assessment Methodology; Technical Report; Sandia National Laboratory: Albuquerque, NM, USA, April 2017. [Google Scholar]
- Hudson, B.; Henderson, J.; Hodges, J.; Holland, M.; Noble, D.; Tunga, I.; Fonseca, F.; Ruiz-Minguela, P. D4.2 Stage Gate Tool—Alpha Version; Technical Report; DTOceanPlus: Edinburgh, UK, 2020. [Google Scholar]
- Nava, V.; Gonzalaz, I.T.; Mendia, L.L.; Noble, D.R.; Tunga, I.; Fonseca, F.; Henderson, J.; Hudson, B.; Ferri, F.; Pons, F.; et al. D6.2 Performance and Energy Yield Tools—Alpha Version; Technical Report; DTOceanPlus: Edinburgh, UK, 2018. [Google Scholar]
- Babarit, A.; Clément, A.H. Optimal latching control of a wave energy device in regular and irregular waves. Appl. Ocean Res. 2006, 28, 77–91. [Google Scholar] [CrossRef]
- Farrell, N.; Donoghue, C.O.; Morrissey, K. Quantifying the uncertainty of wave energy conversion device cost for policy appraisal: An Irish case study. Energy Policy 2015, 78, 62–77. [Google Scholar] [CrossRef]
- Costello, R.; Teillant, B.; Weber, J.; Ringwood, J.V. Techno-Economic Optimisation for Wave Energy Converters. In Proceedings of the International Conference on Ocean Energy, Dublin, Ireland, 17–19 October 2012. [Google Scholar]
- Dalton, G.J.; Alcorn, R.; Lewis, T. Case study feasibility analysis of the Pelamis wave energy convertor in Ireland, Portugal and North America. Renew. Energy 2010, 35, 443–455. [Google Scholar] [CrossRef]
- O’Connor, M.; Lewis, T.; Dalton, G. Operational expenditure costs for wave energy projects and impacts on financial returns. Renew. Energy 2013, 50, 1119–1131. [Google Scholar] [CrossRef] [Green Version]
- Teillant, B.; Costello, R.; Weber, J.; Ringwood, J. Productivity and economic assessment of wave energy projects through operational simulations. Renew. Energy 2012, 48, 220–230. [Google Scholar] [CrossRef] [Green Version]
- Zawawi, N.A.A.; Liew, M.S.; Na, K.L. Decommissioning of offshore platform: A sustainable framework. In Proceedings of the CHUSER—IEEE Colloquium on Humanities, Science and Engineering Research, Kota Kinabalu, Malaysia, 3–4 December 2012; pp. 26–31. [Google Scholar] [CrossRef]
- Sundin, E. Product and Process Design for Successful Remanufacturing. Ph.D. Thesis, Linköping University, Linköping, Switzerland, 2004. [Google Scholar]
- Wilson, C.; Grubler, A.; Bento, N.; Healey, S.; De Stercke, S.; Zimm, C. Granular technologies to accelerate decarbonization. Science 2020, 368, 36–39. [Google Scholar] [CrossRef] [Green Version]
- Abdulla, K.; Skelton, J.; Doherty, K.; O’Kane, P.; Doherty, R.; Bryans, G. Statistical availability analysis of Wave Energy Converters. In Proceedings of the International Offshore and Polar Engineering Conference, Maui, HI, USA, 19–24 June 2011; pp. 572–577. [Google Scholar]
- Angelis-Dimakis, A.; Biberacher, M.; Dominguez, J.; Fiorese, G.; Gadocha, S.; Gnansounou, E.; Guariso, G.; Kartalidis, A.; Panichelli, L.; Pinedo, I.; et al. Methods and tools to evaluate the availability of renewable energy sources. Renew. Sustain. Energy Rev. 2011, 15, 1182–1200. [Google Scholar] [CrossRef]
- Janakiraman, G.; Santos, J.R.; Turner, Y. Automated system design for availability. In Proceedings of the International Conference on Dependable Systems and Networks, Florence, Italy, 28 June–1 July 2004; pp. 411–420. [Google Scholar] [CrossRef]
- Clark, C.E.; DuPont, B. Reliability-based design optimization in offshore renewable energy systems. Renew. Sustain. Energy Rev. 2018, 97, 390–400. [Google Scholar] [CrossRef]
- ISO. ISO 2394:2015 General Principles on Reliability for Structures; ISO: Geneva, Switzerland, 2015. [Google Scholar]
- Johannesson, P.; Svensson, T. Reliability Evaluation using Variation Mode and Effect Analysis: Application to CorPower’s mooring pre-tension cylinder. In Proceedings of the 13th European Wave and Tidal Energy Conference, Naples, Italy, 1–6 September 2019. [Google Scholar]
- Atcheson, M.; Cruz, J.; Martins, T.; Johannesson, P.; Svensson, T. Quantification of load uncertainties in the design process of a WEC. In Proceedings of the 13th European Wave and Tidal Energy Conference, Naples, Italy, 1–6 September 2019; pp. 1–9. [Google Scholar]
- British Standards Institution. BS 5760-0: Reliability of Constructed or Manufactured Products, Systems, Equipment and Components—A Guide to Reliability; Technical Report; British Standards Institution: London, UK, 2014. [Google Scholar]
- EMEC. Guidelines for Reliability, Maintainability and Survivability of Marine Energy Conversion Systems; European Marine Energy Centre Limited: Orkney, UK, 2009. [Google Scholar]
- International Electrotechnical Commission. IEC TC 114: Marine Energy—Wave, Tidal and Other Water Current Converters; Technical Report; International Electrotechnical Commission: Geneva, Switzerland, 2017. [Google Scholar]
- Margheritini, L.; Vicinanza, D.; Frigaard, P. Sea Slot Cone Generator Overtopping Performance in 3D Conditions. In Proceedings of the International Offshore and Polar Engineering Conference, Vancouver, BC, Canada, 6–11 July 2008. [Google Scholar]
- Coe, R.G.; Michelen, C.; Eckert-Gallup, A.; Yu, Y.H.; van Rij, J. WDRT: A Toolbox for Design-Response Analysis of Wave Energy Converters; Technical Report; Sandia National Laboratories: Albuquerque, NM, USA; National Renewable Energy Laboratory: Golden, CO, USA, 2016.
- Mundon, T.; Rosenberg, B. Development of a Survival Configuration for the Triton Wave Energy Converter. In Proceedings of the Marine Energy Technology Symposium, Washington, DC, USA, 30 April–2 May 2018. [Google Scholar]
- Stillinger, C.J.; Brekken, T.K.; Von Jouanne, A. Furthering the study of real-time life extending control for ocean energy conversion. In Proceedings of the IEEE Power and Energy Society General Meeting, San Diego, CA, USA, 22–26 July 2012. [Google Scholar] [CrossRef]
- Tiron, R.; Mallon, F.; Dias, F.; Reynaud, E.G. The challenging life of wave energy devices at sea: A few points to consider. Renew. Sustain. Energy Rev. 2015, 43, 1263–1272. [Google Scholar] [CrossRef]
- OSD Manufacturing Technology Program and The Joint Service/Industry MRL Working Group. Manufacturing Readiness Level (MRL) Deskbook Version 2018; Technical Report; U.S. Department of Defense Manufacturing Technology Program: Arlington, VA, USA, 2020.
- Hudson, J.A.; Phillips, D.C.; Wilkins, N.J. Materials aspects of wave energy converters. J. Mater. Sci. 1980, 15, 1337–1363. [Google Scholar] [CrossRef]
- PCCI Inc. Wave and Current Energy Generating Devices Criteria and Standards; Technical Report; Mineral Management Service: Alexandria, VA, USA, 2009.
- Malça, C.M.S.P.; Pedro, J.B.F.N.; Felismina, R.P. Influence of material selection on the structural behavior of a wave energy converter. AIMS Energy 2014, 2, 359–372. [Google Scholar] [CrossRef]
- Le, H.R.; Collins, K.M.; Greaves, D.M.; Bellamy, N.W. Mechanics and materials in the design of a buckling diaphragm wave energy converter. Mater. Des. 2015, 79, 86–93. [Google Scholar] [CrossRef] [Green Version]
- Das, S.; Kanchanapiboon, A. A multi-criteria model for evaluating design for manufacturability. Int. J. Prod. Res. 2011, 49, 1197–1217. [Google Scholar] [CrossRef]
- MatWeb Material Property Data, MatWeb LLC. Copyright 1996–2020. Available online: http://www.matweb.com/ (accessed on 10 July 2020).
- Dalton, G.J.; Alcorn, R.; Lewis, T. A 10 year installation program for wave energy in Ireland: A case study sensitivity analysis on financial returns. Renew. Energy 2012, 40, 80–89. [Google Scholar] [CrossRef]
- Correia da Fonseca, F.; Amaral, L.; Rentschler, M.; Arede, F.; Chainho, P.; Yang, Y.; Noble, D.R.; Petrov, A.; Nava, V.; Germain, N.; et al. D5.7 Logistics and Marine Operations Tools—Alpha Version; Technical Report; DTOceanPlus: Edinburgh, UK, 2020. [Google Scholar]
- Rémouit, F.; Chatzigiannakou, M.A.; Bender, A.; Temiz, I.; Sundberg, J.; Engström, J. Deployment and maintenance ofwave energy converters at the Lysekil Research Site: A comparative study on the use of divers and remotely-operated vehicles. J. Mar. Sci. Eng. 2018, 6, 39. [Google Scholar] [CrossRef] [Green Version]
- Mustapa, M.A.; Yaakob, O.B.; Ahmed, Y.M.; Rheem, C.K.; Koh, K.K.; Adnan, F.A. Wave energy device and breakwater integration: A review. Renew. Sustain. Energy Rev. 2017, 77, 43–58. [Google Scholar] [CrossRef]
- Zanuttigh, B.; Angelelli, E.; Kofoed, J.P. Effects of mooring systems on the performance of a wave activated body energy converter. Renew. Energy 2013, 57, 422–431. [Google Scholar] [CrossRef]
- Johanning, L.; Smith, G.H.; Wolfram, J. Mooring design approach for wave energy converters. Proc. Inst. Mech. Eng. Part M J. Eng. Marit. Environ. 2006, 220, 159–174. [Google Scholar] [CrossRef]
- Rinaldi, G.; Crossley, G.; Parkinson, R.; Johanning, L. The O&M driven design of a multi-row platform tidal project. In Proceedings of the 13th Euopean Wave and Tidal Energy Conference, Naples, Italy, 1–6 September 2019; pp. 1–8. [Google Scholar]
- Lawso, J. A perfect match?: Oil and gas companies have learned to overcome many offshore difficulties, so what, if anything, can they bring to the renewable energy table? Renew. Energy Focus 2011, 12, 38–40. [Google Scholar] [CrossRef]
- Reikard, G. Integrating wave energy into the power grid: Simulation and forecasting. Ocean Eng. 2013, 73, 168–178. [Google Scholar] [CrossRef]
- Parwal, A.; Fregelius, M.; Temiz, I.; Göteman, M.; Oliveira, J.G.; Boström, C.; Leijon, M. Energy management for a grid-connected wave energy park through a hybrid energy storage system. Appl. Energy 2018, 231, 399–411. [Google Scholar] [CrossRef]
- Henriques, J.C.; Gato, L.M.; Lemos, J.M.; Gomes, R.P.; Falcão, A.F. Peak-power control of a grid-integrated oscillating water column wave energy converter. Energy 2016, 109, 378–390. [Google Scholar] [CrossRef] [Green Version]
- Molinas, M.; Skjervheim, O.; Sørby, B.; Andreasen, P.; Lundberg, S.; Undeland, T. Power Smoothing by Aggregation of Wave Energy Converters for Minimizing Electrical Energy Storage Requirements. In Proceedings of the 7th European Wave and Tidal Energy Conference, Porto, Portugal, 11–13 September 2007; pp. 3–8. [Google Scholar] [CrossRef]
- Engström, J.; Eriksson, M.; Göteman, M.; Isberg, J.; Leijon, M. Performance of large arrays of point absorbing direct-driven wave energy converters. J. Appl. Phys. 2013, 114, 204502. [Google Scholar] [CrossRef] [Green Version]
- Tissandier, J.; Babarit, A.; Clément, A.H. Study of the smoothing effect on the power production in an array of SEAREV wave energy converters. In Proceedings of the International Offshore and Polar Engineering Conference, Vancouver, BC, Canada, 6–11 July 2008; International Society of Offshore and Polar Engineers: Cupertino, CA, USA, 2008. [Google Scholar]
- Kracht, P.; Giebhardt, J.; Dick, C.; Salcedo, F. Deliverable 4.3 Report on Grid Integration and Power Quality Testing; Technical Report; Marinet2: Marine Renewables Infrastructure Network Transnational Project; European Union: Brussels, Belgium, 2014. [Google Scholar]
- Torres-Olguin, R.E.; Tedeschi, E.; Endegnanew, A.G. D4.14 Demand Side Grid Compatibility; Technical Report; MaRINET2: Marine Renewables Infrastructure Network Transnational Project; European Union: Brussels, Belgium, 2014. [Google Scholar]
- Kovaltchouk, T.; Armstrong, S.; Blavette, A.; Ben Ahmed, H.; Multon, B. Wave farm flicker severity: Comparative analysis and solutions. Renew. Energy 2016, 91, 32–39. [Google Scholar] [CrossRef] [Green Version]
- Blavette, A.; O’Sullivan, D.L.; Alcorn, R.; Lewis, T.W.; Egan, M.G. Impact of a Medium-Size Wave Farm on Grids of Different Strength Levels. IEEE Trans. Power Syst. 2014, 29, 917–923. [Google Scholar] [CrossRef]
- Falcao, A. Modelling and control of oscillating-body wave energy converters with hydraulic power take-off and gas accumulator. Ocean Eng. 2007, 34, 2021–2032. [Google Scholar] [CrossRef]
- Costello, R.; Ringwood, J.V.; Weber, J. Comparison of Two Alternative Hydraulic PTO Concepts for Wave Energy Conversion. In Proceedings of the Ninth European Wave and Tidal Energy Conference, Southhampton, UK, 5–9 September 2011. [Google Scholar]
- Fusco, F.; Nolan, G.; Ringwood, J.V. Variability reduction through optimal combination of wind/wave resources—An Irish case study. Energy 2010, 35, 314–325. [Google Scholar] [CrossRef] [Green Version]
- Wang, L.; Isberg, J.; Tedeschi, E. Review of control strategies for wave energy conversion systems and their validation: The wave-to-wire approach. Renew. Sustain. Energy Rev. 2018, 81, 366–379. [Google Scholar] [CrossRef]
- Penalba, M.; Ringwood, J.V. A high-fidelity wave-to-wire model for wave energy converters. Renew. Energy 2019, 134, 367–378. [Google Scholar] [CrossRef]
- Parkinson, S.C.; Dragoon, K.; Reikard, G.; García-Medina, G.; Özkan-Haller, H.T.; Brekken, T.K. Integrating ocean wave energy at large-scales: A study of the US Pacific Northwest. Renew. Energy 2015, 76, 551–559. [Google Scholar] [CrossRef]
- Wright, G. Marine governance in an industrialised ocean: A case study of the emerging marine renewable energy industry. Mar. Policy 2015, 52, 77–84. [Google Scholar] [CrossRef]
- Apolonia, M.; Silva, A.; Simas, T. Developing an Environmental Impact Assessment model for nearshore wave energy devices. In Proceedings of the 13th Euopean Wave and Tidal Energy Conference, Naples, Italy, 1–6 September 2019; pp. 1–7. [Google Scholar]
- Willsteed, E.; Gill, A.B.; Birchenough, S.N.; Jude, S. Assessing the cumulative environmental effects of marine renewable energy developments: Establishing common ground. Sci. Total Environ. 2017, 577, 19–32. [Google Scholar] [CrossRef] [Green Version]
- Henkel, S.K.; Conway, F.D.; Boehlert, G.W. Environmental and human dimensions of ocean renewable energy development. Proc. IEEE 2013, 101, 991–998. [Google Scholar] [CrossRef]
- Joslin, J.; Rush, B.; Stewart, A.; Polagye, B. Development of an Adaptable Monitoring Package for Marine Renewable Energy Projects Part I: Conceptual Design and Operation. In Proceedings of the 2013 OCEANS, San Diego, CA, USA, 23–27 September 2013; pp. 1–10. [Google Scholar] [CrossRef]
- Telenko, C.; Seepersad, C.C.; Webber, M.E. A Compilation of Design for Environment Principles and Guidelines. In Proceedings of the ASME 2008 International Design Engineering Technical Conferences & Computers and Information in Engineering Conference, Brooklyn, NY, USA, 3–6 August 2008. [Google Scholar] [CrossRef] [Green Version]
- De Pauw, I.C.; Karana, E.; Kandachar, P.; Poppelaars, F. Comparing Biomimicry and Cradle to Cradle with Ecodesign: A case study of student design projects. J. Clean. Prod. 2014, 78, 174–183. [Google Scholar] [CrossRef]
- Gattuso, J.P.; Magnan, A.K.; Bopp, L.; Cheung, W.W.; Duarte, C.M.; Hinkel, J.; Mcleod, E.; Micheli, F.; Oschlies, A.; Williamson, P.; et al. Ocean solutions to address climate change and its effects on marine ecosystems. Front. Mar. Sci. 2018, 5, 337. [Google Scholar] [CrossRef] [Green Version]
- Jacobson, A.; Milman, A.D.; Kammen, D.M. Letting the (energy) Gini out of the bottle: Lorenz curves of cumulative electricity consumption and Gini coefficients as metrics of energy distribution and equity. Energy Policy 2005, 33, 1825–1832. [Google Scholar] [CrossRef]
- McLachlan, C. Technologies in Place: Symbolic Interpretations of Renewable Energy. Sociol. Rev. 2009, 57, 181–199. [Google Scholar] [CrossRef]
- Conway, F.; Stevenson, J.; Hunter, D.; Stefanovich, M.; Campbell, H.; Covell, Z.; Yin, Y. Ocean space, ocean place the human dimensions of wave energy in oregon. Oceanography 2010, 23, 82–91. [Google Scholar] [CrossRef]
- Lange, M.; Page, G.; Cummins, V. Governance challenges of marine renewable energy developments in the U.S.—Creating the enabling conditions for successful project development. Mar. Policy 2018, 90, 37–46. [Google Scholar] [CrossRef]
- Iversen, O.S.; Halskov, K.; Leong, T.W. Values-led participatory design. CoDesign 2012, 8, 87–103. [Google Scholar] [CrossRef]
- Ball, L.J.; Ormerod, T.C. Applying ethnography in the analysis and support of expertise in engineering design. Des. Stud. 2000, 21, 403–421. [Google Scholar] [CrossRef] [Green Version]
- Cradden, L.; Kalogeri, C.; Barrios, I.M.; Galanis, G.; Ingram, D.; Kallos, G. Multi-criteria site selection for offshore renewable energy platforms. Renew. Energy 2016, 87, 791–806. [Google Scholar] [CrossRef] [Green Version]
- Nobre, A.; Pacheco, M.; Jorge, R.; Lopes, M.F.; Gato, L.M. Geo-spatial multi-criteria analysis for wave energy conversion system deployment. Renew. Energy 2009, 34, 97–111. [Google Scholar] [CrossRef]
- Vasileiou, M.; Loukogeorgaki, E.; Vagiona, D.G. GIS-based multi-criteria decision analysis for site selection of hybrid offshore wind and wave energy systems in Greece. Renew. Sustain. Energy Rev. 2017, 73, 745–757. [Google Scholar] [CrossRef]
- Ghosh, S.; Chakraborty, T.; Saha, S.; Majumder, M.; Pal, M. Development of the location suitability index for wave energy production by ANN and MCDM techniques. Renew. Sustain. Energy Rev. 2016, 59, 1017–1028. [Google Scholar] [CrossRef]
- Vance, J.; Giambalvo, J.; Hoffenson, S. Navigating the common approaches to product development. Proc. Int. Conf. Eng. Des. ICED 2017, 9, 169–178. [Google Scholar]
- Schmidt, J.B.; Calantone, R.J. Escalation of Commitment During New Product Development. J. Acad. Mark. Sci. 2002, 30, 103–118. [Google Scholar] [CrossRef]
- Ulrich, K.; Eppinger, S. Product Design and Development, 3rd ed.; McGraw Hill: New York, NY, USA, 2003. [Google Scholar]
- Nielsen, K.; Kennedy, B.; Bull, D.; Costello, R.; Roberts, J.; Weber, J. Technical Submission Form Technical Specification of a Wave Energy Farm; Technical Report; Sandia National Laboratories: Albuquerque, NM, USA, 2017.
- Otto, K.; Wood, K. Product Design Techniques in Reverse Engineering and New Product Development; Pearson: London, UK, 2001. [Google Scholar]
- Giambalvo, J.W.; Vance, J.K.; Hoffenson, S. Toward a decision support tool for selecting engineering design methodologies. In Proceedings of the ASEE Annual Conference and Exposition, Columbus, OH, USA, 25–28 June 2017. [Google Scholar]
- Brown, M.A.; Chandler, J.; Lapsa, M.V.; Sovacool, B.K. Carbon Lock-In: Barriers To Deploying Climate Change Mitigation Technologies; Technical Report; Oak Ridge National Laboratory: Oak Ridge, TN, USA, 2007.
Method | Requirement | Percentage of Users |
---|---|---|
Iterative design of WEC and PTO Subsystems (model, simulate, change parameters, model…) [45] | Power Production | 68 |
Controls optimization [60,110] | Power Production | 48 |
Iterative design by approximating the cost of all components and redesigning the most expensive [23] | Capital Cost | 56 |
Selection of components based on lifetime maintenance schedule [132] | Availability | 64 |
Design for a 50-year wave [144] | Survivability and Reliability | 56 |
Prototyping and prototype testing [106] | Survivability and Reliability | 64 |
Stakeholder meetings with manufacturers [63] | Manufacturing and Materials Selection | 48 |
Design for Manufacturing [107] | Manufacturing and Materials Selection | 48 |
Installation and maintenance storyboarding [50] | Installability and Maintainability | 52 |
Application of conceptual design methodologies to installation and maintenance planning [121] | Installability and Maintainability | 52 |
Stakeholder meetings with installation and maintenance personnel [31] | Installability and Maintainability | 52 |
Eliminating or minimizing entanglement hazards [146] | Environmental Impacts and Safety | 56 |
Eliminating hazardous fluids [121,199] | Environmental Impacts and Safety | 60 |
Minimizing human–device interaction [199] | Environmental Impacts and Safety | 60 |
Reducing visibility [121] | Acceptability | 52 |
Reducing ecosystem impact [146,199] | Acceptability | 52 |
Design for local manufacturing [121] | Acceptability | 52 |
Community engagement [188] | Acceptability | 48 |
Design for flexibility of wave conditions [199] | Global Deployability | 60 |
Wave resource assessment [19] | Global Deployability | 72 |
Design for modularity [152] | Global Deployability | 56 |
Standardization of manufacturing, construction, assembly, and installation needs [106,152] | Global Deployability | 48 |
Method | Requirement | Percentage of Users |
---|---|---|
Multi-objective optimization | Power Production | 24 |
Controls optimization | Power Production | 48 |
Optimization with power production as objective function | Power Production | 32 |
Hydrodynamic optimization to determine PTO characteristics | Power Production | 28 |
Hydrodynamic optimization to determine WEC shape/size | Power Production | 40 |
Optimization with genetic algorithms | Power Production | 20 |
Array optimization | Power Production | 16 |
Optimization algorithms which represent cost as mass of weight | Capital Cost | 20 |
Optimization algorithms which represent cost as volume | Capital Cost | 8 |
Optimization algorithms which estimate and minimize capital cost | Capital Cost | 24 |
Supply chain optimization | Capital Cost | 32 |
Optimization using operational cost as an objective | Operational Cost | 16 |
Optimization using operational cost as a constraint | Operational Cost | 4 |
Optimization using availability as an objective | Availability | 16 |
Optimization using availability as a constraint | Availability | 8 |
Reliability-based optimization | Reliability | 24 |
Optimization using installability or maintainability as an objective | Installability and Maintainability | 4 |
Optimization using availability as a constraint | Installability and Maintainability | 0 |
Optimization to minimize variability | Grid Integration | 12 |
Optimization using grid characteristics of variability as constraints | Grid Integration | 8 |
Optimization using environmental impacts of safety as an objective | Environmental Impacts and Safety | 20 |
Optimization using environmental impacts of safety as a constraint | Environmental Impacts and Safety | 20 |
Requirement | Percent Using Multiple Design Methods | Percent Using Multiple Evaluation Methods |
---|---|---|
Power Production | 84 | 75 |
Capital Cost | 84 | 75 |
Operational Cost | 57 | 76 |
Availability | 76 | 50 |
Survivability and Reliability | 76 | 68 |
Manufacturing and Materials Selection | 83 | 88 |
Installability and Maintainability | 78 | 61 |
Grid Integration | 60 | 67 |
Environmental Impacts and Safety | 95 | 76 |
Acceptability | 100 | 76 |
Global Deployability | 89 | 78 |
Method | Requirement | Percentage of Users |
---|---|---|
Cost Estimates by subcontractors [63] | Capital Cost | 48 |
In-house capital cost estimates based on research and stakeholder engagement [34] | Capital Cost | 52 |
LCOE | Capital Cost | 60 |
In-house operational cost estimates based on research and stakeholder engagement [34] | Operational Cost | 56 |
LCOE | Operational Cost | 48 |
Failure Modes and Effects Analysis [50] | Availability | 48 |
Extreme sea state numerical simulations [144] | Survivability and Reliability | 60 |
Manufacturing cost estimates and timelines provided by subcontractors | Manufacturing and Materials Selection | 52 |
Installation and Maintenance timelines and estimates provided by subcontractors [121] | Installability and Maintainability | 48 |
Estimate of a minimum feasible wave resource for an attractive LCOE [120] | Global Deployability | 56 |
Depth and geophysical requirements [120] | Global Deployability | 52 |
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. |
© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
Share and Cite
Trueworthy, A.; DuPont, B. The Wave Energy Converter Design Process: Methods Applied in Industry and Shortcomings of Current Practices. J. Mar. Sci. Eng. 2020, 8, 932. https://doi.org/10.3390/jmse8110932
Trueworthy A, DuPont B. The Wave Energy Converter Design Process: Methods Applied in Industry and Shortcomings of Current Practices. Journal of Marine Science and Engineering. 2020; 8(11):932. https://doi.org/10.3390/jmse8110932
Chicago/Turabian StyleTrueworthy, Ali, and Bryony DuPont. 2020. "The Wave Energy Converter Design Process: Methods Applied in Industry and Shortcomings of Current Practices" Journal of Marine Science and Engineering 8, no. 11: 932. https://doi.org/10.3390/jmse8110932
APA StyleTrueworthy, A., & DuPont, B. (2020). The Wave Energy Converter Design Process: Methods Applied in Industry and Shortcomings of Current Practices. Journal of Marine Science and Engineering, 8(11), 932. https://doi.org/10.3390/jmse8110932