Risk Analysis, Vol. 35, No. 4, 2015
DOI: 10.1111/risa.12324
Perspective
Risk Analysis for U.S. Offshore Wind Farms: The Need for an Integrated Approach Andrea Staid∗ and Seth D. Guikema
Wind power is becoming an increasingly important part of the global energy portfolio, and there is growing interest in developing offshore wind farms in the United States to better utilize this resource. Wind farms have certain environmental benefits, notably near-zero emissions of greenhouse gases, particulates, and other contaminants of concern. However, there are significant challenges ahead in achieving large-scale integration of wind power in the United States, particularly offshore wind. Environmental impacts from wind farms are a concern, and these are subject to a number of on-going studies focused on risks to the environment. However, once a wind farm is built, the farm itself will face a number of risks from a variety of hazards, and managing these risks is critical to the ultimate achievement of longterm reductions in pollutant emissions from clean energy sources such as wind. No integrated framework currently exists for assessing risks to offshore wind farms in the United States, which poses a challenge for wind farm risk management. In this “Perspective”, we provide an overview of the risks faced by an offshore wind farm, argue that an integrated framework is needed, and give a preliminary starting point for such a framework to illustrate what it might look like. This is not a final framework; substantial work remains. Our intention here is to highlight the research need in this area in the hope of spurring additional research about the risks to wind farms to complement the substantial amount of on-going research on the risks from wind farms. KEY WORDS: Bayesian network; offshore wind; risk analysis; risk framework
1. INTRODUCTION
in recent years but the growth has been focused on onshore farms, many of which have been built in the middle of the country. However, one of the hurdles facing renewable energy implementation is the disparity between where the energy resources and energy demand are located. For example, the strongest onshore wind resources in the United States are located in the center of the country in the Great Plains, far from many of the major urban areas.(2) Many of the most densely populated areas of the United States are located along the coasts, far from the strongest onshore wind resources. Without sizable investments in new transmission lines, this will continue to be a barrier to wind. Because of this, there is a growing interest in the development of offshore wind in the United States.
Wind energy is currently the fastest growing form of electricity generation in the world.(1) With concerns about climate change, carbon dioxide emissions, and the impacts of a growing population, renewable energy will likely continue to be a topic of substantial interest in terms of technologies, implementation, policies, and incentives. Wind energy in the United States has been growing rapidly Department of Geography and Environmental Engineering, Johns Hopkins University, Baltimore, MD, USA. ∗ Address correspondence to Andrea Staid, Department of Geography and Environmental Engineering, Johns Hopkins University, Baltimore, MD 21218, USA; tel: 857–928–8316; [email protected].
587
C 2015 Society for Risk Analysis 0272-4332/15/0100-0587$22.00/1
588 The offshore wind resource potential is estimated to be more than 4,000 GW, or almost four times greater than the current U.S. energy generation capacity.(3,4) While not all of this potential can be captured, due to various siting and feasibility considerations, there is a large untapped resource available that would help considerably in moving toward a cleaner energy portfolio. Offshore locations typically have higher and more consistent wind speeds when compared to onshore development, with offshore farms reaching capacity factors of 50%.(5) However, offshore farms are more expensive to install and maintain. The United States is lagging behind many European countries in offshore development. Denmark, the United Kingdom, the Netherlands, Germany, Sweden, and China are just some of the countries that are successfully operating large offshore wind farms. The oldest of these farms has been operational for over a decade, and there is much to be learned from these early adopters. The United States currently has no offshore wind farms, but several proposed sites are actively being planned. Not surprisingly, these new projects will face a number of hazards during their operational life. Bad weather can bring strong winds, high waves, ice, and lightning, and these factors, along with the normal wear and tear of a long operational life, can result in significant damage and loss of productivity to wind turbines and wind farms. As with any new project, risk analysis is a critical component, and proper risk management for the early offshore wind farms in the United States is crucial for the longerterm adoption of this technology. Economic and technical success with the early farms would greatly increase the willingness of others to develop offshore wind farms in the United States. The methods that led to the success of European offshore farms serve as a starting point for developing farms in the United States, but there are some factors that will differentiate farms in the United States from farms in Europe. U.S. farms, particularly those along the Gulf and Atlantic coasts will be operating in different geographical and meteorological conditions. For example, the presence of hurricanes poses a significant challenge to U.S. farms, and European farms have little experience with storms of this strength. Icing events and winter storms, as well as ship traffic, will also impact U.S. farms differently from those in Europe. Thus, a risk management framework should contain the hazards present in a specific location. The methodology of analyzing risks to offshore wind farms is general in nature, but we will focus
Staid and Guikema our discussion on locations in the United States. Although we present a preliminary risk framework here, our focus is on arguing that more research is needed in this area, not on providing a final risk framework. There are also many political and public perception issues that could impact the success of new wind farms, and these should be brought into future wind farm risk models. For example, public acceptance of wind has been shown to be a strong indicator of project success.(6) Public opposition and political pressure can severely delay and impede the progress of new offshore projects, as evidenced in the case of the Cape Wind farm in Massachusetts.(7) These social and political issues strongly affect the early stages of planning and development. Here, we focus on the technical risk issues faced by a farm in the operational phase while acknowledging the importance of public perception and political issues that should be considered earlier in the process. In this article we argue that an integrated, coherent risk framework at the wind-farm level is both lacking and needed for U.S. offshore wind farms, and we give an initial starting point for such a framework. We hope that this will spur the further research needed to fully develop a risk management framework for U.S. offshore wind farms. 2. BACKGROUND The topics of risk and wind turbines often appear together in the literature. However, much of the risk that is being studied and assessed regarding wind turbines is related to the risks that the turbine poses to the natural environment. This includes studies on the impacts to avian populations, migration patterns, bat habitats, marine life, and possible climate change in the form of altered sea-surface roughness.(8–12) There have also been proposals for the use of an integrated framework to capture all of these environmental risks together.(13) These environmental risk studies have garnered significant media attention and have often formed the basis of arguments put forth by opponents of wind projects.(6) While the environmental concerns are certainly important, there is also a need to assess the risks that will be faced by the turbines themselves. These risks to the turbines are critical for assessing the long-term financial viability of a farm, and they need to be considered in tandem with the environmental risks when making decisions about investments, operations, and locations of wind farms. Some of these risks and hazards affecting turbines have been
Perspective looked at, but they have been primarily studied on an individual basis. For example, Rose et al.(14) have conducted an in-depth study of the impact of hurricanes on offshore wind farms and Tavner et al.(15) have studied turbine reliability and failure rates. ˜ In addition, Duenas-Osorio and Basu have studied dangerous accelerations from wind.(16) Ice throws from wind turbines are yet another hazard that has been studied.(17) Although much of the work in the area of risk to wind farms has been highly individual, whether to a single turbine or from a specific failure mode, there has been some work done on an overall risk framework for offshore wind, most notably by Vijayakumar.(18) It focused on component-level failures and the conditions necessary to cause a malfunction of the various components of a wind turbine. Work such as this provides a starting point but does not yet provide the type of integrated framework that we suggest is needed. Assessing individual risks ignores the interactions between multiple events or hazards, as well as the possibility of joint events. Integrating multiple risk hazards into a single framework, on the other hand, allows these complex interactions to be captured. It also provides a much more integrated platform to support decision making in which limited resources must be allocated to address multiple hazards. Single-hazard models cannot offer this level of decision support, making an integrated framework critical. There does not yet exist a coherent, integrated framework for risk assessment and management for U.S. offshore wind farms. Such a framework would need to be multihazard, including both humaninduced and natural hazards, and it would need to explicitly include maintenance and operations. It should be extensible to the scale of large, multiturbine wind farms. If such a framework were developed, it could have substantial benefit for wind farm developers and operators. In the next section, we present some preliminary steps toward the development of such a framework as a way of better illustrating our suggestion. 3. A PRELIMINARY RISK FRAMEWORK To illustrate the proposed idea that an integrated framework is needed, we develop a preliminary and admittedly incomplete framework for risk assessment and management for an offshore wind farm. We base this framework on a Bayesian belief network (BBN).(19) The framework is best understood through the use of an example, so for
589 clarity, we present it in this manner. Fig. 1 shows the structure of the proposed framework, the components of which will be described in detail below. Such a framework can be used to estimate energy output, costs, and revenues for a given wind farm, helping the developer and operator maximize the success of the project while accounting for those uncertainties included in the model. In addition, the flexible nature of the framework allows it to be used for situational evaluation, such as the impact of a certain, damaging event, or to study the interactions of multiple situations. Because the outcome of each node is probabilistically dependent on the preceding linked nodes, changing the outcome of one node propagates through the framework and updates the probabilities based on the conditions selected. Again, this is a preliminary framework, the primary purpose of which is to show the potential benefits of an integrated framework in a simplified, easily accessible manner. More research is needed to develop a more complete integrated risk framework for offshore wind farms in the United States. To demonstrate the use of such a risk framework, we gathered data for a candidate project, Cape Wind in Massachusetts, a site that is in the planning phases for a major offshore wind farm. Because this wind farm does not yet exist, there were significant gaps in the data needed, and our example was conducted with a simplified version of the framework. It should be noted that this demonstration serves only to show the use of the framework itself, not to make specific recommendations about the Cape Wind site. Fig. 1 shows the populated framework for Cape Wind. Data were taken from site-specific, open sources when possible.(20,21) When this was not possible, expert judgment (by the authors) or data from European wind farms were used for the sake of demonstration.(22,23) The thought process of creating and populating such a framework for one’s own analysis can be almost as useful as the framework itself, as has shown to be the case in many areas of risk analysis.(24) The steps for using the framework are similar to those for any risk assessment model, such as that proposed by Vose.(25) For a simplified view of the risks to a wind farm, we first consider many of the major factors that could impact the energy output of a wind farm. It is then necessary to characterize the interactions among the various hazards and risk factors. In our example, we discretize the events for simplicity; the levels for each node should be chosen and defined. We present a model for a pared-down version
590
Staid and Guikema
Fig. 1. Proposed framework shown with Cape Wind sample data.
of Cape Wind that only contains three turbines for a clearer demonstration. This simplification does not account for all possible interactions and outcomes, but it allows for a clear view of the thought process behind building up such a framework. A full-scale framework would need to account for substantially more turbines, but our simplified example serves to illustrate the intention behind the risk framework. For locations along the Atlantic coast of the United States—the area that we chose to focus on in our example—many of the hazards are weather related. Hurricanes, ice storms, lightning, strong waves (not presented in our preliminary framework), and normal wind variability all have the potential to greatly impact power generation and can force individual turbines, or an entire farm, to shut down temporarily. Other hazards can be unrelated to weather, such as ship collisions. As presented, the framework focuses on factors influencing energy output, but it is directly extensible to addressing economic outcomes as well. Power prices, cost of maintenance, and capital investments can all be factored into the framework. It can be used to analyze or compare farms on an annual or lifetime basis for profit maximization. By including both probabilities and related consequences of relevant hazards, as well as the interactions among these hazards, the user can build up an understanding of the biggest risk factors and which steps may be appropriate to take for risk mitigation. The risk factors present in the framework were chosen based on the known issues that cause damage to or affect the power generation of wind
turbines. Hurricanes have the potential to cause significant damage to offshore wind turbines in many locations in the United States.(14) Wind farms located in colder regions often have issues caused by icing on wind turbines. Ice storms and icing events during the winter months can result in decreased power generation and can also damage the turbines themselves.(26,27) Another factor of concern is lightning striking a wind turbine. There have been a number of instances of damage attributed to lightning strikes, and these cases have led to studies on the best methods to protect against lightning strikes.(22,28) Potential ship collisions have also been identified as a risk factor to offshore wind turbines. It is thought to be a rare occurrence, but it may become more likely if wind farms are built in areas with high shipping traffic. Since the consequences of a large ship colliding with a wind turbine would likely be substantial, some studies have been carried out to try and reduce the risks.(23) Once the framework hazards have been identified, the hazard interactions need to be modeled as well. The integration of the hazards identified allows the user to assess the collective effect that a concurrence of certain events may have on the energy production of a wind farm. While this example does not include every possible hazard, it does include major factors that have been shown to cause damage to wind turbines and downtime to turbines or a farm. Other factors have not been included here, but could be added in if deemed necessary, especially for
Perspective applications to locations other than the U.S. Atlantic coast, which is demonstrated here. The resulting consequences are just as important as the hazards. The consequences in our preliminary framework all relate to one major factor: whether or not the wind farm is producing power. This could be broadened to directly model economic effects in future frameworks. Damage to the turbines caused by one or more of the hazards identified would likely result in turbine downtime, or a period without power production. The consequence nodes, such as damage, downtime, and the resulting energy production value, are probabilistically dependent on the upstream nodes. Thus, if a turbine is exposed to a hazard, there is a conditional probability distribution associated with the consequence of damage. In some cases, the damage will be minimal and the turbine can still operate, whereas in other scenarios severe damage will result in significant downtime. All damage requires repair, but the time required for repair can also vary, depending on the level of damage. Thus, severe damage affects the ultimate turbine power production in two ways: lost production time due to the damage itself and also to allow for the appropriate maintenance to be carried out. Mild damage can often wait to be repaired, and it can be done during periods of previously planned maintenance so as to minimize additional downtime. Of course, this is just an example of what the framework could look like; the specifics would be determined by the user to address the areas of most concern for the purposes of his or her analysis. The wind capacity factor, while not a hazard per se, is of great concern to developers of wind farms, and it has a strong impact on the energy production. In this demonstration, we divide the capacity factor into three discrete levels based on measured wind data from the area around Cape Wind. The expected value and standard deviation for capacity factor (0.364 ± 0.065) is shown at the bottom of the node box, which is based on the associated probabilities of the three levels. The capacity factor and turbine downtime determine the energy output of the turbines for the year, based on the rated turbine capacity of 3.6 MW each.(29) These values then lead into the farm output, which is also discretized into levels based on the resulting energy output of each turbine. The expected value of 33,600 MWh is shown at the bottom of the node. Each combination of levels (e.g., two turbines operating at high power and one turbine not producing) from the three turbines results in a level of output for the farm, in this
591 simplified and discretized case. For example, if all three turbines are operating for one year at a highcapacity factor, that results in the highest power level of 9. The reductions in levels come when either the capacity factor decreases or one or more of the turbines experiences downtime during the year. The next level down, level 8, corresponds to the case of all three turbines operating at medium power. Due to the close proximity of turbines, we assume that the wind conditions will be the same at each turbine at a given time, so the turbines will either be operating or not, and if any are operating, those will be at the same level (i.e., high, medium, or low). We therefore do not consider cases with one turbine operating at the highest level while another turbine operates at the lowest. In order to apply an integrated framework, information should be collected for the chosen location. If possible, real and measured data can be used, but expert knowledge elicited in the form of probability distributions can also be used.(30) As of the writing of this article, there are currently no offshore wind farms operating in the United States. Early applications of this framework will depend on analysis, expert opinion, or studies in other countries. BBNs are a natural modeling approach for this framework because of their ability to incorporate a variety of data types. A framework such as this can be used to assist in decision making and to gain insight into the impacts and interactions of different hazards at a given site. Since many hazards depend greatly on location, decisions regarding wind farm locations that minimize risk would be an obvious application. For hazard analysis specifically, a few illustrative examples comparing different hazards for the Cape Wind example are shown in Table I, where the downtime and loss of production if selected hazards occur is shown. For the simplified three-turbine wind farm located at the site of Cape Wind, the average annual energy production is 33,600 MWh. The first event evaluated is a category 4 hurricane. Nominally, the probability of this occurring off the coast of Massachusetts is fairly low at 1.5%. However, Table I shows that if such a hurricane does occur, there would be a large increase in the turbine downtime and a significant decrease in energy production for that year. Similarly, the Cape Wind framework was also analyzed for the events of an ice storm and a year with below average wind speeds. In this case, ice storms do not appear to have a considerable impact on downtime or energy production. A year of low overall wind speeds,
592
Staid and Guikema Table I. Results of Event Findings for Cape Wind Cape Wind Hazards
Event Nominal Category 4 hurricane Ice storm Low annual capacity factor
Expected single turbine downtime
Expected annual energy production
Change from nominal
2.24% 12.20% 3.42% 2.24%
33,600 MWh ± 6,900 30,200 MWh ± 8,700 33,200 MWh ± 7,200 22,900 MWh ± 2,200
– −10.1% −1.2% −31.8%
however, can seriously reduce the amount of energy produced, which would result in a reduction in profits. 4. EXPANDING THE MODEL As it currently stands, the model presents a simplified view of the risks faced by an offshore wind farm, and many factors were intentionally left out of the first iteration. It serves more to illustrate the potential for integrated risk models for offshore wind farms than as a full, ready to deploy model. We chose to initially focus on the parameters that were relatively straightforward to model, were important to wind farm risk based on our knowledge and the literature in the field, and had clear, well-understood connections in the framework. As research on integrated frameworks progresses, there are many other aspects that could be incorporated. The grid connection is modeled as simply as possible in this initial framework, but the realities of the grid-side interactions are quite complicated and potentially significant in determining the useful energy output of a wind farm. Wind forecasting could also be considered in this framework, as inaccurate forecasts can greatly affect the overall profits of the wind farm operators. 5. CONCLUSION This article argues that an integrated framework is needed for wind farm risk assessment and management, and it presents a preliminary example of such a framework to illustrate the benefit of integrated models. While not complete, the example framework demonstrates the functionality of an integrated, systems-level approach to risk assessment that can be modified and applied as needed to wind farms or potential wind farms. It is necessary to look at overall risk to a wind farm, building from detailed studies of individual failure modes, hazards, or
vulnerabilities. The hazards likely to be faced by an offshore wind farm need to be identified and characterized, and an assessment of the hazard interactions and consequences should be done to fully utilize the framework. These parameters will likely change with location, so individual farms will need unique assessments. Data limitations may be a challenge for implementation in many U.S. locations, but expert elicitation and a thorough understanding of results from international studies, if used correctly, can result in a framework with high value to the users. As stated earlier, there are other, nontechnical hazards faced by offshore wind, primarily in the form of public and political opposition. These aspects of risk to project success are critical, and they can be incorporated into such a framework as the one proposed here, but we have not included them here. These opposition concerns, as well as other hazards that may be of concern in a given location, can easily be added into our framework. A systematic approach to developing an integrated risk model for offshore wind farms will yield comprehensive results, which can be used to better inform decisionmakers of the risks faced by offshore wind in the United States. One of the strengths of such an approach is the resulting insight into the relative influence and sensitivity of each parameter, especially when the interactions of multiple hazards play a large role in the results. This information can be used to manage and plan for these known risks by making better decisions early on in the design and development process. The success of the initial offshore wind projects in the United States will set the stage for further investment in wind power as a source of clean, renewable energy. ACKNOWLEDGMENTS This research was conducted with support of the Lee & Albert H. Halff Doctoral Student Award and the Gordon Croft Fellowship as part of
Perspective the Environment, Energy, Sustainability & Health Institute of Johns Hopkins University in addition to partial support from the National Science Foundation (Grant CMMI 0968711).
593
14. Rose S, Jaramillo P, Small MJ, Apt J. Quantifying the hurricane catastrophe risk to offshore wind power. Risk Analysis, 2013; 33(12):2126–2141. 15. Tavner P, Xiang J, Spinato F. Reliability analysis for wind turbines. Wind Energy, 2007; 10(1):1–18. ˜ 16. Duenas-Osorio L, Basu B. Unavailability of wind turbines due to wind-induced accelerations. Engineering Structures, 2008; 30(4):885–893. ¨ 17. Seifert H, Westerhellweg A, Kroning J. Risk analysis of ice REFERENCES throw from wind turbines. Boreas, 2003; 6(9). 18. Vijayakumar RRN. Risk analysis of offshore wind turbine 1. U.S. Energy Information Administration Wind Data. Availfarm in normal and extreme conditions. Master’s Thesis, able at: www.eia.gov/cneaf/solar.renewables/page/wind/wind. 2007. Available at: http://www.cee.elektro.dtu.dk/education/ html, Accessed August 10, 2012. student˙projects/07/rrnv, Accessed July 25, 2012. 2. National Renewable Energy Laboratory Wind Maps. Avail19. Pearl J. Probabilistic Reasoning in Intelligent Systems: Netable at: www.nrel.gov/gis/wind.html, Accessed September 4, works of Plausible Inference. Morgan Kaufmann: Elsevier, 2013. 1988. 3. Musial W, Ram B. Large-scale offshore wind power in the 20. National Oceanographic and Atmospheric Administration. United States: Assessment of opportunities and barriers. Historical Hurricane Tracks, 2013. Available at: http://www. NREL/TP-500-40745. National Renewable Energy Laboratocsc.noaa.gov/hurricanes, Accessed August 1, 2013. ries, 2010. 21. Levitan & Associates, Inc. Technical Assessment of Onshore 4. International Energy Statistics. 2011. Available at: http:// www.eia.gov/cfapps/ipdbproject/IEDIndex3.cfm?tid=2&pid=2&aid=7, and Offshore Wind Generation Potential in New England. Boston, MA:ISO Incorporated, 2007. Accessed August 10, 2012. 22. Sorensen T, Jensen F, Raben N, Lykkegaard J, Saxov J. 5. Capacity factors at Danish offshore wind farms, 2014. AvailLightning protection for offshore wind turbines. Conference able at: http://energynumbers.info/capacity-factors-at-danishPaper, DOI: 10.1049/cp:20010840 Conference: Electricity Disoffshore-wind-farms, Accessed August 4, 2013. tribution, 2001. Part 1: Contributions, 2001. 6. Firestone J, Kempton W, Krueger A. Public acceptance of off23. den Boon H, Just H, Hansen P, Ravn E, Frouws K, Otto S, shore wind power projects in the USA. Wind Energy, 2009; Dalhoff P, Stein J, van der Tak C, van Rooij J. Reduction 12(2):183–202. of ship collision risks for offshore wind Farms—SAFESHIP. 7. Layzer JA. The Environmental Case: Translating Values into EWEC London, 2004. Policy. Washington, DC: CQ Press, 2002. 24. Apostolakis GE. How useful is quantitative risk assessment? 8. Desholm M, Kahlert J. Avian collision risk at an offshore wind Risk Analysis, 2004; 24(3):515–520. farm. Biology Letters, 2005; 1(3):296–298. 25. Vose D. Risk Analysis: A Quantitative Guide. New York: Wi9. Madsen PT, Wahlberg M, Tougaard J, Lucke K, Tyack PL. ley, 2008. Wind turbine underwater noise and marine mammals: Impli26. Jasinski WJ, Noe SC, Selig MS, Bragg MB. Wind turbine cations of current knowledge and data needs. Marine Ecology performance under icing conditions. Transactions—American Progress Series, 2006; 309:279–295. Society of Mechanical Engineers Journal of Solar Energy 10. Kirk-Davidoff DB, Keith DW. On the climate impact of surEngineering, 1998; 120:60–65. face roughness anomalies. Journal of the Atmospheric Sci27. Lacroix A, Manwell JF. Wind Energy: Cold Weather Issues. ences 2008; 65(7):2215–2234. University of Massachusetts at Amherst, Renewable Energy 11. Hueppop O, Dierschke J, Exo K, Fredrich E, Hill R. Bird miLaboratory, 2000. gration studies and potential collision risk with offshore wind 28. McNiff B. Wind Turbine Lightning Protection Project. turbines. Ibis, 2006; 148(s1):90–109. NREL/SR-500-31115. Golden, CO: National Renewable En12. Arnett EB, Brown W, Erickson WP, Fiedler JK, Hamilton BL, ergy Laboratory, 2002. Henry TH, Jain A, Johnson GD, Kerns J, Koford RR. Pat29. Cape Wind. 2012. Available at: http://www.capewind.org/, Acterns of bat fatalities at wind energy facilities in North Amercessed December 10, 2012. ica. Journal of Wildlife Management, 2008; 72(1):61–78. 30. Cooke NJ. Varieties of knowledge elicitation techniques. 13. Ram B. Assessing integrated risks of offshore wind projects: International Journal of Human-Computer Studies, 1994; Moving towards gigawatt-scale deployments. Wind Engineer41(6):801–849. ing, 2011; 35(3):247–266.
DOI: 10.1111/risa.12324
Perspective
Risk Analysis for U.S. Offshore Wind Farms: The Need for an Integrated Approach Andrea Staid∗ and Seth D. Guikema
Wind power is becoming an increasingly important part of the global energy portfolio, and there is growing interest in developing offshore wind farms in the United States to better utilize this resource. Wind farms have certain environmental benefits, notably near-zero emissions of greenhouse gases, particulates, and other contaminants of concern. However, there are significant challenges ahead in achieving large-scale integration of wind power in the United States, particularly offshore wind. Environmental impacts from wind farms are a concern, and these are subject to a number of on-going studies focused on risks to the environment. However, once a wind farm is built, the farm itself will face a number of risks from a variety of hazards, and managing these risks is critical to the ultimate achievement of longterm reductions in pollutant emissions from clean energy sources such as wind. No integrated framework currently exists for assessing risks to offshore wind farms in the United States, which poses a challenge for wind farm risk management. In this “Perspective”, we provide an overview of the risks faced by an offshore wind farm, argue that an integrated framework is needed, and give a preliminary starting point for such a framework to illustrate what it might look like. This is not a final framework; substantial work remains. Our intention here is to highlight the research need in this area in the hope of spurring additional research about the risks to wind farms to complement the substantial amount of on-going research on the risks from wind farms. KEY WORDS: Bayesian network; offshore wind; risk analysis; risk framework
1. INTRODUCTION
in recent years but the growth has been focused on onshore farms, many of which have been built in the middle of the country. However, one of the hurdles facing renewable energy implementation is the disparity between where the energy resources and energy demand are located. For example, the strongest onshore wind resources in the United States are located in the center of the country in the Great Plains, far from many of the major urban areas.(2) Many of the most densely populated areas of the United States are located along the coasts, far from the strongest onshore wind resources. Without sizable investments in new transmission lines, this will continue to be a barrier to wind. Because of this, there is a growing interest in the development of offshore wind in the United States.
Wind energy is currently the fastest growing form of electricity generation in the world.(1) With concerns about climate change, carbon dioxide emissions, and the impacts of a growing population, renewable energy will likely continue to be a topic of substantial interest in terms of technologies, implementation, policies, and incentives. Wind energy in the United States has been growing rapidly Department of Geography and Environmental Engineering, Johns Hopkins University, Baltimore, MD, USA. ∗ Address correspondence to Andrea Staid, Department of Geography and Environmental Engineering, Johns Hopkins University, Baltimore, MD 21218, USA; tel: 857–928–8316; [email protected].
587
C 2015 Society for Risk Analysis 0272-4332/15/0100-0587$22.00/1
588 The offshore wind resource potential is estimated to be more than 4,000 GW, or almost four times greater than the current U.S. energy generation capacity.(3,4) While not all of this potential can be captured, due to various siting and feasibility considerations, there is a large untapped resource available that would help considerably in moving toward a cleaner energy portfolio. Offshore locations typically have higher and more consistent wind speeds when compared to onshore development, with offshore farms reaching capacity factors of 50%.(5) However, offshore farms are more expensive to install and maintain. The United States is lagging behind many European countries in offshore development. Denmark, the United Kingdom, the Netherlands, Germany, Sweden, and China are just some of the countries that are successfully operating large offshore wind farms. The oldest of these farms has been operational for over a decade, and there is much to be learned from these early adopters. The United States currently has no offshore wind farms, but several proposed sites are actively being planned. Not surprisingly, these new projects will face a number of hazards during their operational life. Bad weather can bring strong winds, high waves, ice, and lightning, and these factors, along with the normal wear and tear of a long operational life, can result in significant damage and loss of productivity to wind turbines and wind farms. As with any new project, risk analysis is a critical component, and proper risk management for the early offshore wind farms in the United States is crucial for the longerterm adoption of this technology. Economic and technical success with the early farms would greatly increase the willingness of others to develop offshore wind farms in the United States. The methods that led to the success of European offshore farms serve as a starting point for developing farms in the United States, but there are some factors that will differentiate farms in the United States from farms in Europe. U.S. farms, particularly those along the Gulf and Atlantic coasts will be operating in different geographical and meteorological conditions. For example, the presence of hurricanes poses a significant challenge to U.S. farms, and European farms have little experience with storms of this strength. Icing events and winter storms, as well as ship traffic, will also impact U.S. farms differently from those in Europe. Thus, a risk management framework should contain the hazards present in a specific location. The methodology of analyzing risks to offshore wind farms is general in nature, but we will focus
Staid and Guikema our discussion on locations in the United States. Although we present a preliminary risk framework here, our focus is on arguing that more research is needed in this area, not on providing a final risk framework. There are also many political and public perception issues that could impact the success of new wind farms, and these should be brought into future wind farm risk models. For example, public acceptance of wind has been shown to be a strong indicator of project success.(6) Public opposition and political pressure can severely delay and impede the progress of new offshore projects, as evidenced in the case of the Cape Wind farm in Massachusetts.(7) These social and political issues strongly affect the early stages of planning and development. Here, we focus on the technical risk issues faced by a farm in the operational phase while acknowledging the importance of public perception and political issues that should be considered earlier in the process. In this article we argue that an integrated, coherent risk framework at the wind-farm level is both lacking and needed for U.S. offshore wind farms, and we give an initial starting point for such a framework. We hope that this will spur the further research needed to fully develop a risk management framework for U.S. offshore wind farms. 2. BACKGROUND The topics of risk and wind turbines often appear together in the literature. However, much of the risk that is being studied and assessed regarding wind turbines is related to the risks that the turbine poses to the natural environment. This includes studies on the impacts to avian populations, migration patterns, bat habitats, marine life, and possible climate change in the form of altered sea-surface roughness.(8–12) There have also been proposals for the use of an integrated framework to capture all of these environmental risks together.(13) These environmental risk studies have garnered significant media attention and have often formed the basis of arguments put forth by opponents of wind projects.(6) While the environmental concerns are certainly important, there is also a need to assess the risks that will be faced by the turbines themselves. These risks to the turbines are critical for assessing the long-term financial viability of a farm, and they need to be considered in tandem with the environmental risks when making decisions about investments, operations, and locations of wind farms. Some of these risks and hazards affecting turbines have been
Perspective looked at, but they have been primarily studied on an individual basis. For example, Rose et al.(14) have conducted an in-depth study of the impact of hurricanes on offshore wind farms and Tavner et al.(15) have studied turbine reliability and failure rates. ˜ In addition, Duenas-Osorio and Basu have studied dangerous accelerations from wind.(16) Ice throws from wind turbines are yet another hazard that has been studied.(17) Although much of the work in the area of risk to wind farms has been highly individual, whether to a single turbine or from a specific failure mode, there has been some work done on an overall risk framework for offshore wind, most notably by Vijayakumar.(18) It focused on component-level failures and the conditions necessary to cause a malfunction of the various components of a wind turbine. Work such as this provides a starting point but does not yet provide the type of integrated framework that we suggest is needed. Assessing individual risks ignores the interactions between multiple events or hazards, as well as the possibility of joint events. Integrating multiple risk hazards into a single framework, on the other hand, allows these complex interactions to be captured. It also provides a much more integrated platform to support decision making in which limited resources must be allocated to address multiple hazards. Single-hazard models cannot offer this level of decision support, making an integrated framework critical. There does not yet exist a coherent, integrated framework for risk assessment and management for U.S. offshore wind farms. Such a framework would need to be multihazard, including both humaninduced and natural hazards, and it would need to explicitly include maintenance and operations. It should be extensible to the scale of large, multiturbine wind farms. If such a framework were developed, it could have substantial benefit for wind farm developers and operators. In the next section, we present some preliminary steps toward the development of such a framework as a way of better illustrating our suggestion. 3. A PRELIMINARY RISK FRAMEWORK To illustrate the proposed idea that an integrated framework is needed, we develop a preliminary and admittedly incomplete framework for risk assessment and management for an offshore wind farm. We base this framework on a Bayesian belief network (BBN).(19) The framework is best understood through the use of an example, so for
589 clarity, we present it in this manner. Fig. 1 shows the structure of the proposed framework, the components of which will be described in detail below. Such a framework can be used to estimate energy output, costs, and revenues for a given wind farm, helping the developer and operator maximize the success of the project while accounting for those uncertainties included in the model. In addition, the flexible nature of the framework allows it to be used for situational evaluation, such as the impact of a certain, damaging event, or to study the interactions of multiple situations. Because the outcome of each node is probabilistically dependent on the preceding linked nodes, changing the outcome of one node propagates through the framework and updates the probabilities based on the conditions selected. Again, this is a preliminary framework, the primary purpose of which is to show the potential benefits of an integrated framework in a simplified, easily accessible manner. More research is needed to develop a more complete integrated risk framework for offshore wind farms in the United States. To demonstrate the use of such a risk framework, we gathered data for a candidate project, Cape Wind in Massachusetts, a site that is in the planning phases for a major offshore wind farm. Because this wind farm does not yet exist, there were significant gaps in the data needed, and our example was conducted with a simplified version of the framework. It should be noted that this demonstration serves only to show the use of the framework itself, not to make specific recommendations about the Cape Wind site. Fig. 1 shows the populated framework for Cape Wind. Data were taken from site-specific, open sources when possible.(20,21) When this was not possible, expert judgment (by the authors) or data from European wind farms were used for the sake of demonstration.(22,23) The thought process of creating and populating such a framework for one’s own analysis can be almost as useful as the framework itself, as has shown to be the case in many areas of risk analysis.(24) The steps for using the framework are similar to those for any risk assessment model, such as that proposed by Vose.(25) For a simplified view of the risks to a wind farm, we first consider many of the major factors that could impact the energy output of a wind farm. It is then necessary to characterize the interactions among the various hazards and risk factors. In our example, we discretize the events for simplicity; the levels for each node should be chosen and defined. We present a model for a pared-down version
590
Staid and Guikema
Fig. 1. Proposed framework shown with Cape Wind sample data.
of Cape Wind that only contains three turbines for a clearer demonstration. This simplification does not account for all possible interactions and outcomes, but it allows for a clear view of the thought process behind building up such a framework. A full-scale framework would need to account for substantially more turbines, but our simplified example serves to illustrate the intention behind the risk framework. For locations along the Atlantic coast of the United States—the area that we chose to focus on in our example—many of the hazards are weather related. Hurricanes, ice storms, lightning, strong waves (not presented in our preliminary framework), and normal wind variability all have the potential to greatly impact power generation and can force individual turbines, or an entire farm, to shut down temporarily. Other hazards can be unrelated to weather, such as ship collisions. As presented, the framework focuses on factors influencing energy output, but it is directly extensible to addressing economic outcomes as well. Power prices, cost of maintenance, and capital investments can all be factored into the framework. It can be used to analyze or compare farms on an annual or lifetime basis for profit maximization. By including both probabilities and related consequences of relevant hazards, as well as the interactions among these hazards, the user can build up an understanding of the biggest risk factors and which steps may be appropriate to take for risk mitigation. The risk factors present in the framework were chosen based on the known issues that cause damage to or affect the power generation of wind
turbines. Hurricanes have the potential to cause significant damage to offshore wind turbines in many locations in the United States.(14) Wind farms located in colder regions often have issues caused by icing on wind turbines. Ice storms and icing events during the winter months can result in decreased power generation and can also damage the turbines themselves.(26,27) Another factor of concern is lightning striking a wind turbine. There have been a number of instances of damage attributed to lightning strikes, and these cases have led to studies on the best methods to protect against lightning strikes.(22,28) Potential ship collisions have also been identified as a risk factor to offshore wind turbines. It is thought to be a rare occurrence, but it may become more likely if wind farms are built in areas with high shipping traffic. Since the consequences of a large ship colliding with a wind turbine would likely be substantial, some studies have been carried out to try and reduce the risks.(23) Once the framework hazards have been identified, the hazard interactions need to be modeled as well. The integration of the hazards identified allows the user to assess the collective effect that a concurrence of certain events may have on the energy production of a wind farm. While this example does not include every possible hazard, it does include major factors that have been shown to cause damage to wind turbines and downtime to turbines or a farm. Other factors have not been included here, but could be added in if deemed necessary, especially for
Perspective applications to locations other than the U.S. Atlantic coast, which is demonstrated here. The resulting consequences are just as important as the hazards. The consequences in our preliminary framework all relate to one major factor: whether or not the wind farm is producing power. This could be broadened to directly model economic effects in future frameworks. Damage to the turbines caused by one or more of the hazards identified would likely result in turbine downtime, or a period without power production. The consequence nodes, such as damage, downtime, and the resulting energy production value, are probabilistically dependent on the upstream nodes. Thus, if a turbine is exposed to a hazard, there is a conditional probability distribution associated with the consequence of damage. In some cases, the damage will be minimal and the turbine can still operate, whereas in other scenarios severe damage will result in significant downtime. All damage requires repair, but the time required for repair can also vary, depending on the level of damage. Thus, severe damage affects the ultimate turbine power production in two ways: lost production time due to the damage itself and also to allow for the appropriate maintenance to be carried out. Mild damage can often wait to be repaired, and it can be done during periods of previously planned maintenance so as to minimize additional downtime. Of course, this is just an example of what the framework could look like; the specifics would be determined by the user to address the areas of most concern for the purposes of his or her analysis. The wind capacity factor, while not a hazard per se, is of great concern to developers of wind farms, and it has a strong impact on the energy production. In this demonstration, we divide the capacity factor into three discrete levels based on measured wind data from the area around Cape Wind. The expected value and standard deviation for capacity factor (0.364 ± 0.065) is shown at the bottom of the node box, which is based on the associated probabilities of the three levels. The capacity factor and turbine downtime determine the energy output of the turbines for the year, based on the rated turbine capacity of 3.6 MW each.(29) These values then lead into the farm output, which is also discretized into levels based on the resulting energy output of each turbine. The expected value of 33,600 MWh is shown at the bottom of the node. Each combination of levels (e.g., two turbines operating at high power and one turbine not producing) from the three turbines results in a level of output for the farm, in this
591 simplified and discretized case. For example, if all three turbines are operating for one year at a highcapacity factor, that results in the highest power level of 9. The reductions in levels come when either the capacity factor decreases or one or more of the turbines experiences downtime during the year. The next level down, level 8, corresponds to the case of all three turbines operating at medium power. Due to the close proximity of turbines, we assume that the wind conditions will be the same at each turbine at a given time, so the turbines will either be operating or not, and if any are operating, those will be at the same level (i.e., high, medium, or low). We therefore do not consider cases with one turbine operating at the highest level while another turbine operates at the lowest. In order to apply an integrated framework, information should be collected for the chosen location. If possible, real and measured data can be used, but expert knowledge elicited in the form of probability distributions can also be used.(30) As of the writing of this article, there are currently no offshore wind farms operating in the United States. Early applications of this framework will depend on analysis, expert opinion, or studies in other countries. BBNs are a natural modeling approach for this framework because of their ability to incorporate a variety of data types. A framework such as this can be used to assist in decision making and to gain insight into the impacts and interactions of different hazards at a given site. Since many hazards depend greatly on location, decisions regarding wind farm locations that minimize risk would be an obvious application. For hazard analysis specifically, a few illustrative examples comparing different hazards for the Cape Wind example are shown in Table I, where the downtime and loss of production if selected hazards occur is shown. For the simplified three-turbine wind farm located at the site of Cape Wind, the average annual energy production is 33,600 MWh. The first event evaluated is a category 4 hurricane. Nominally, the probability of this occurring off the coast of Massachusetts is fairly low at 1.5%. However, Table I shows that if such a hurricane does occur, there would be a large increase in the turbine downtime and a significant decrease in energy production for that year. Similarly, the Cape Wind framework was also analyzed for the events of an ice storm and a year with below average wind speeds. In this case, ice storms do not appear to have a considerable impact on downtime or energy production. A year of low overall wind speeds,
592
Staid and Guikema Table I. Results of Event Findings for Cape Wind Cape Wind Hazards
Event Nominal Category 4 hurricane Ice storm Low annual capacity factor
Expected single turbine downtime
Expected annual energy production
Change from nominal
2.24% 12.20% 3.42% 2.24%
33,600 MWh ± 6,900 30,200 MWh ± 8,700 33,200 MWh ± 7,200 22,900 MWh ± 2,200
– −10.1% −1.2% −31.8%
however, can seriously reduce the amount of energy produced, which would result in a reduction in profits. 4. EXPANDING THE MODEL As it currently stands, the model presents a simplified view of the risks faced by an offshore wind farm, and many factors were intentionally left out of the first iteration. It serves more to illustrate the potential for integrated risk models for offshore wind farms than as a full, ready to deploy model. We chose to initially focus on the parameters that were relatively straightforward to model, were important to wind farm risk based on our knowledge and the literature in the field, and had clear, well-understood connections in the framework. As research on integrated frameworks progresses, there are many other aspects that could be incorporated. The grid connection is modeled as simply as possible in this initial framework, but the realities of the grid-side interactions are quite complicated and potentially significant in determining the useful energy output of a wind farm. Wind forecasting could also be considered in this framework, as inaccurate forecasts can greatly affect the overall profits of the wind farm operators. 5. CONCLUSION This article argues that an integrated framework is needed for wind farm risk assessment and management, and it presents a preliminary example of such a framework to illustrate the benefit of integrated models. While not complete, the example framework demonstrates the functionality of an integrated, systems-level approach to risk assessment that can be modified and applied as needed to wind farms or potential wind farms. It is necessary to look at overall risk to a wind farm, building from detailed studies of individual failure modes, hazards, or
vulnerabilities. The hazards likely to be faced by an offshore wind farm need to be identified and characterized, and an assessment of the hazard interactions and consequences should be done to fully utilize the framework. These parameters will likely change with location, so individual farms will need unique assessments. Data limitations may be a challenge for implementation in many U.S. locations, but expert elicitation and a thorough understanding of results from international studies, if used correctly, can result in a framework with high value to the users. As stated earlier, there are other, nontechnical hazards faced by offshore wind, primarily in the form of public and political opposition. These aspects of risk to project success are critical, and they can be incorporated into such a framework as the one proposed here, but we have not included them here. These opposition concerns, as well as other hazards that may be of concern in a given location, can easily be added into our framework. A systematic approach to developing an integrated risk model for offshore wind farms will yield comprehensive results, which can be used to better inform decisionmakers of the risks faced by offshore wind in the United States. One of the strengths of such an approach is the resulting insight into the relative influence and sensitivity of each parameter, especially when the interactions of multiple hazards play a large role in the results. This information can be used to manage and plan for these known risks by making better decisions early on in the design and development process. The success of the initial offshore wind projects in the United States will set the stage for further investment in wind power as a source of clean, renewable energy. ACKNOWLEDGMENTS This research was conducted with support of the Lee & Albert H. Halff Doctoral Student Award and the Gordon Croft Fellowship as part of
Perspective the Environment, Energy, Sustainability & Health Institute of Johns Hopkins University in addition to partial support from the National Science Foundation (Grant CMMI 0968711).
593
14. Rose S, Jaramillo P, Small MJ, Apt J. Quantifying the hurricane catastrophe risk to offshore wind power. Risk Analysis, 2013; 33(12):2126–2141. 15. Tavner P, Xiang J, Spinato F. Reliability analysis for wind turbines. Wind Energy, 2007; 10(1):1–18. ˜ 16. Duenas-Osorio L, Basu B. Unavailability of wind turbines due to wind-induced accelerations. Engineering Structures, 2008; 30(4):885–893. ¨ 17. Seifert H, Westerhellweg A, Kroning J. Risk analysis of ice REFERENCES throw from wind turbines. Boreas, 2003; 6(9). 18. Vijayakumar RRN. Risk analysis of offshore wind turbine 1. U.S. Energy Information Administration Wind Data. Availfarm in normal and extreme conditions. Master’s Thesis, able at: www.eia.gov/cneaf/solar.renewables/page/wind/wind. 2007. Available at: http://www.cee.elektro.dtu.dk/education/ html, Accessed August 10, 2012. student˙projects/07/rrnv, Accessed July 25, 2012. 2. National Renewable Energy Laboratory Wind Maps. Avail19. Pearl J. Probabilistic Reasoning in Intelligent Systems: Netable at: www.nrel.gov/gis/wind.html, Accessed September 4, works of Plausible Inference. Morgan Kaufmann: Elsevier, 2013. 1988. 3. Musial W, Ram B. Large-scale offshore wind power in the 20. National Oceanographic and Atmospheric Administration. United States: Assessment of opportunities and barriers. Historical Hurricane Tracks, 2013. Available at: http://www. NREL/TP-500-40745. National Renewable Energy Laboratocsc.noaa.gov/hurricanes, Accessed August 1, 2013. ries, 2010. 21. Levitan & Associates, Inc. Technical Assessment of Onshore 4. International Energy Statistics. 2011. Available at: http:// www.eia.gov/cfapps/ipdbproject/IEDIndex3.cfm?tid=2&pid=2&aid=7, and Offshore Wind Generation Potential in New England. Boston, MA:ISO Incorporated, 2007. Accessed August 10, 2012. 22. Sorensen T, Jensen F, Raben N, Lykkegaard J, Saxov J. 5. Capacity factors at Danish offshore wind farms, 2014. AvailLightning protection for offshore wind turbines. Conference able at: http://energynumbers.info/capacity-factors-at-danishPaper, DOI: 10.1049/cp:20010840 Conference: Electricity Disoffshore-wind-farms, Accessed August 4, 2013. tribution, 2001. Part 1: Contributions, 2001. 6. Firestone J, Kempton W, Krueger A. Public acceptance of off23. den Boon H, Just H, Hansen P, Ravn E, Frouws K, Otto S, shore wind power projects in the USA. Wind Energy, 2009; Dalhoff P, Stein J, van der Tak C, van Rooij J. Reduction 12(2):183–202. of ship collision risks for offshore wind Farms—SAFESHIP. 7. Layzer JA. The Environmental Case: Translating Values into EWEC London, 2004. Policy. Washington, DC: CQ Press, 2002. 24. Apostolakis GE. How useful is quantitative risk assessment? 8. Desholm M, Kahlert J. Avian collision risk at an offshore wind Risk Analysis, 2004; 24(3):515–520. farm. Biology Letters, 2005; 1(3):296–298. 25. Vose D. Risk Analysis: A Quantitative Guide. New York: Wi9. Madsen PT, Wahlberg M, Tougaard J, Lucke K, Tyack PL. ley, 2008. Wind turbine underwater noise and marine mammals: Impli26. Jasinski WJ, Noe SC, Selig MS, Bragg MB. Wind turbine cations of current knowledge and data needs. Marine Ecology performance under icing conditions. Transactions—American Progress Series, 2006; 309:279–295. Society of Mechanical Engineers Journal of Solar Energy 10. Kirk-Davidoff DB, Keith DW. On the climate impact of surEngineering, 1998; 120:60–65. face roughness anomalies. Journal of the Atmospheric Sci27. Lacroix A, Manwell JF. Wind Energy: Cold Weather Issues. ences 2008; 65(7):2215–2234. University of Massachusetts at Amherst, Renewable Energy 11. Hueppop O, Dierschke J, Exo K, Fredrich E, Hill R. Bird miLaboratory, 2000. gration studies and potential collision risk with offshore wind 28. McNiff B. Wind Turbine Lightning Protection Project. turbines. Ibis, 2006; 148(s1):90–109. NREL/SR-500-31115. Golden, CO: National Renewable En12. Arnett EB, Brown W, Erickson WP, Fiedler JK, Hamilton BL, ergy Laboratory, 2002. Henry TH, Jain A, Johnson GD, Kerns J, Koford RR. Pat29. Cape Wind. 2012. Available at: http://www.capewind.org/, Acterns of bat fatalities at wind energy facilities in North Amercessed December 10, 2012. ica. Journal of Wildlife Management, 2008; 72(1):61–78. 30. Cooke NJ. Varieties of knowledge elicitation techniques. 13. Ram B. Assessing integrated risks of offshore wind projects: International Journal of Human-Computer Studies, 1994; Moving towards gigawatt-scale deployments. Wind Engineer41(6):801–849. ing, 2011; 35(3):247–266.
