Wind power project size and component costs: An Alaska case study

Transcription

1 Wind power project size and component costs: An Alaska case study Jeremy VanderMeer, Marc Mueller-Stoffels, and Erin Whitney Citation: Journal of Renewable and Sustainable Energy 9, (2017); View online: View Table of Contents: Published by the American Institute of Physics Articles you may be interested in Preface: Technology and cost reviews for renewable energy in Alaska: Sharing our experience and know-how Journal of Renewable and Sustainable Energy 9, (2017); / An Alaska case study: Solar photovoltaic technology in remote microgrids Journal of Renewable and Sustainable Energy 9, (2017); / An Alaska case study: Diesel generator technologies Journal of Renewable and Sustainable Energy 9, (2017); / An Alaska case study: Energy storage technologies Journal of Renewable and Sustainable Energy 9, (2017); / An Alaska case study: Cost estimates for integrating renewable technologies Journal of Renewable and Sustainable Energy 9, (2017); / An Alaska case study: Electrical transmission Journal of Renewable and Sustainable Energy 9, (2017); /

2 JOURNAL OF RENEWABLE AND SUSTAINABLE ENERGY 9, (2017) Wind power project size and component costs: An Alaska case study Jeremy VanderMeer, Marc Mueller-Stoffels, and Erin Whitney Alaska Center for Energy and Power, University of Alaska Fairbanks, P.O. Box , Fairbanks, Alaska , USA (Received 6 June 2017; accepted 4 December 2017; published online 21 December 2017) Many communities in Alaska, especially along the coast, have excellent wind resources and serve as ideal laboratories for microgrids around the Arctic and the world. Wind power systems have been installed in a number of locations in Alaska, both in remote areas and along the road system. As more isolated microgrids are developed worldwide, understanding the cost and performance of wind power in these systems is of increased importance. In this review of wind power project sizes and costs in Alaska, the costs per kilowatt for the different components of wind turbine installations generally were found to decrease with increasing sizes. Capacity factors ranged from approximately 10% to 40%, with variability from rated wind speeds, wind turbine heights, and the resolution of the wind power class map. Due to the size of communities in Alaska as well as other factors, turbines installed in Alaska are smaller than the trend in the larger market. Evidence indicates that installing overcapacity wind farms, together with energy storage and significant diversion into thermal loads, would allow communities to achieve diesel-off and least-cost energy when considering displacement of diesel fuel for both electricity and heat. However, the size of wind turbines that can be installed in Alaska is limited by the small size of the state s communities and the challenges of integrating high penetrations of wind power into microgrids. Demonstrations of different high-penetration wind power integration techniques are still needed. Published by AIP Publishing. INTRODUCTION The purpose of this paper is to review the cost and performance of wind power installations in Alaska. Specific metrics in this review include capital costs, operation and maintenance costs, expected life, capacity factors, diesel offsets, costs per kilowatt-hour, conditions for greatest efficiency, cost curves over time, installed costs by major components, transportation costs, technology trends, refurbishment/upgrade markets, and realized cost savings. This is the only paper of its kind specifically for Alaska. The Alaskan wind market is unique from the rest of the United States, due to the large number of isolated microgrids and extreme weather conditions. The isolated microgrids are often not connected to a road system, making it difficult to transport components. As more isolated microgrids are developed worldwide, understanding the cost and performance of wind power in these systems is of increased importance. Many communities in Alaska, especially along the coast, have excellent wind resources. Wind power systems have been installed in several locations in Alaska, both in remote areas and along the road system. Total installed capacity exceeds 60 MW, with installations ranging from 40 kw to 24 MW (McMahon, 2015). The levelized cost of energy (LCOE) of wind power has been decreasing due to improved technology, siting techniques, and maturity of the industry; wind is now one of the most affordable sources of energy based on unsubsidized LCOE in the United States ( falling-wind-energy-costs). In Alaska, the cost of energy is much higher than that in the rest of the nation due to the remoteness, harsh climate, and small size of remote Alaska microgrids /2017/9(6)/061703/19/$ , Published by AIP Publishing.

3 VanderMeer, Mueller-Stoffels, and Whitney J. Renewable Sustainable Energy 9, (2017) (Remote microgrids are not connected to a larger electrical grid, which means that they must generate all electricity consumed within the grid.) Although the cost of wind power is also affected by these conditions, this energy source has helped to reduce the cost of energy in many Alaska communities and more wind power potential remains in locations around the state. Alaska currently has 62 MW of installed wind power capacity ( gov/wind/windexchange/wind_installed_capacity.asp), with installations in 28 remote microgrid communities. Kodiak, Alaska, is an example of a wind power success story. In 2016, this community generated 99.7% of its electricity from renewable energy sources, with 33.8% from wind and 65.9% from hydro. The cost of generating electricity from wind power in Kodiak is less than half the cost of generating electricity from diesel ( generation.html). This review of wind power technology in Alaska is a result of Alaska Senate Bill (SB) 138. In this bill, the Alaska State Legislature created an uncodified section of law entitled: Plan and Recommendations to the Legislature on Infrastructure Needed to Deliver Affordable Energy of the State to Areas That Do Not Have Direct Access to a [proposed] North Slope Natural Gas Pipeline. To support the Alaska Energy Authority (AEA) in its development of an Alaska Affordable Energy Strategy, the Alaska Center for Energy and Power (ACEP) contracted with AEA to document technology development needs specific to Alaska with regard to renewable and sustainable energy technologies. The intention was to determine what targeted, energy technology development solutions could be implemented in Alaska to make energy more affordable in the Alaska Affordable Energy Study area. While the focus was on technology research solutions, other factors such as logistics, labor, and training were also addressed. Drafts of technology reviews were vetted by expert roundtables in late February and early March These reviews are not meant to be exhaustive discussions of energy technologies in Alaska or proper designs for each technology, and they should not be used as guides for the choice and installation of specific systems. As such, not all possible issues with power production and each technology are addressed. Data for each technology were collected from surveys and publically available databases. Only completed projects, or projects with clearly reported data, were included in each technology analysis. These distinctions and descriptions of data sources are included in each technology review. METHODS This analysis of wind power technology in Alaska largely relies on data extracted from applications to the AEA s Renewable Energy Fund (REF), Rounds 1 8. This dataset includes the costs that were budgeted for different wind power projects in Alaska. There are 103 projects spanning the time frame from 2008 to The costs were separated into the categories of Energy Analysis, Conceptual Design, Final Design, Hardware, Transportation, Foundation and Infrastructure, Turbine Installation, Transmission, and Integration. These categories were necessary since individual applications often only included a subset of costs. Linear regressions were used to determine whether the installed capacity, year, or region in Alaska of the installation significantly impacted the different cost categories. The only factor that significantly affected the costs was the installed capacity. This effect corresponds to the well documented effect of economies of scale in wind power projects (Wiser and Bolinger, 2017). Linear regressions were calculated for the capital expenditure (CAPEX) per installed capacity for each individual cost with respect to the installed capacity of the wind power project. The regressions of individual costs were added together to estimate the total cost (in $/ kw) with respect to the installed capacity of the wind power projects. A coefficient was considered significant if its P-value (the probability of there being no relationship between the dependent variable and regressors compared with the given regression) was less than The results were compared with as-built costs from AEA (see Appendix C) and are a good prediction of as-built costs. Costs are in 2015 dollars.

4 VanderMeer, Mueller-Stoffels, and Whitney J. Renewable Sustainable Energy 9, (2017) TABLE I. The coefficients and adjusted R-squared values for the cost categories and the total cost of wind power projects with respect to installed capacity. Equations (1) and (2) show the general form of the equations. Energy analysis Conceptual design Final design Hardware Transportation b 2 b * (0.14) 0.84*** (0.11) 0.71*** (0.17) 0.38*** (0.05) 0.35** (0.11) b 0 6.6*** (0.83) 10.5*** (0.63) 10.5*** (1.0) 10.5*** (0.35) 9.0*** (0.69) R The general form of the equation showing the correlation between the cost Y i;c (in $/kw) with respect to the installed capacity X i (in kw) for cost category c of wind power project i is given by Eq. (1). The general form of the equation showing the total cost Y i (in $/kw) with respect to the installed capacity X i (in kw) for project i is given by Eq. (2). The variable i;c is the error term between the actual cost of a particular cost category of a project and what is predicted by the linear equation with respect to installed capacity. The variable i similarly gives the error for the predicted total cost of a project. Table I shows the values of the coefficients for Eqs. (1) and (2) for the different cost categories and the total cost of the wind power projects considered in this paper logðy i logðy i;c Þ ¼ b 0;i;c þ b 1;i;c logðx i Þþ i;c ; (1) Þ ¼ b 0;i þ b 1;i logðx i Þþb 2;i logðx i Þ 2 þ i : (2) In addition to these cost analyses, we considered data from REF applications and the literature to analyze other cost and performance factors for wind power installation in Alaska. RESULTS AND DISCUSSION Capital costs Tables I and II show the coefficients (b 0, b 1 ; and b 2 ) of the linear regressions for the individual cost categories and total costs of a wind power project with respect to the installed capacity. The general form of the regression equations is given by Eqs. (1) and (2). The asterisks (*) beside the coefficient values indicate the p-value for that coefficient, with * for p < 0.05, ** for p < 0.01, and *** for p < Beneath the value for each coefficient, the standard error is shown in parentheses. The bottom row shows the adjusted R-squared (R 2 ) value of the fit, which is a value between 0 and 1 that indicates how much of the variability in the cost is explained by the linear regression with respect to the installed capacity. The plots showing linear regression for each cost category are shown in Appendix A (Figs. 8 16). Based on data from REF applications, the costs per kilowatt for the different cost categories of wind turbine installations were found to decrease with the increasing size of the installation. Exceptions were the costs for transmission and integration, which did not vary significantly with the capacity of the installed system. Individual costs were tested for correlations TABLE II. The coefficients and adjusted R-squared values for the cost categories and the total cost of wind power projects with respect to installed capacity, continued. Equations (1) and (2) show the general form of the equations. Foundation and infrastructure Turbine installation Transmission lines Integration Total project cost b *** ( ) b *** (0.10) 0.55*** (0.14) 0.77*** ( ) b *** (0.63) 10.6*** (0.91) 6.6*** (0.24) 6.6*** (0.23) 12.9*** ( ) R

5 VanderMeer, Mueller-Stoffels, and Whitney J. Renewable Sustainable Energy 9, (2017) FIG. 1. Cumulative cost of a wind power system. Each line is added to the line below it, with the top line representing the total cost per kilowatt of installed capacity. These costs are based on regressions calculated from REF applications, shown with the data and equations in Appendix A. Very few applications included all cost categories; thus, regressions were found for each category and added to determine an estimate for the total project cost. Costs are in 2015 dollars. with the region and year, but in most cases, the relationship was not significant or did not significantly improve the fit of installed capacity without taking these relationships into consideration. Figure 1 shows the incremental costs of each individual category for different installed capacity systems. Each line represents the vertical summation of the lines below it. Thus, the top line (integration) represents the total cost. Note that the x-axis is plotted on a log scale. Table III shows the cost per installed capacity for several system sizes, which have been calculated based on the regression analysis. The costs are grouped as analysis and design (energy analysis and conceptual and final design), hardware and transport (turbine hardware and transport), and balance of system (foundations and infrastructure, turbine installation, transmission, and integration). Operation and maintenance $/kw Many communities in Alaska have performance-based operation and maintenance (O&M) contracts with the turbine supplier. The supplier performs O&M (often in collaboration with the utility) while guaranteeing a certain level of availability. Operation and maintenance does not include repairs and replacements. Figure 2 shows the predicted maintenance costs based on REF applications, with an average of $0.036/kWh. Predicted maintenance costs tend to be lower than O&M calculations from other sources. The points are plotted against the annual wind generation and sorted by the annual electric consumption of the grid. No clear trends are indicated in the data. Expected life The minimum design lifetime for wind turbines must be 20 years (based on IEC , Wind Turbines-Part 1: Design requirements). Table IV shows average wind turbine lifetimes as TABLE III. Costs per installed capacity taken from the regressions shown in Fig. 1 (costs in 2015 dollars). Size (kw) Analysis and design Hardware and transport Balance of system Total

6 VanderMeer, Mueller-Stoffels, and Whitney J. Renewable Sustainable Energy 9, (2017) FIG. 2. Operation and maintenance costs per kwh of wind energy production. The points are sorted by the annual electricity demand of the grid. These data are from REF applications. TABLE IV. Wind turbine lifetime (source data from NREL, 2016). Wind turbine size (kw) Lifetime (yr) Lifetime standard deviation (yr) < FIG. 3. Capacity factor plotted against the wind power class. The wind power classes are from the Alaska Energy Data Inventory (AEDI) wind power class map ( Capacity factors above 40% were removed since they are possible but unlikely. The capacity factor was calculated as the total wind energy used to supply electrical loads (excluding diversion loads) in 1 year, divided by the energy that would be harvested from the wind if the turbines were outputting their rated (maximum) power the entire year. In general, the capacity factor should depend primarily on the wind power class of the installation site for low and medium penetration hybrid-diesel systems. Wind power class refers to the available energy from the wind as outlined by DOE.

7 VanderMeer, Mueller-Stoffels, and Whitney J. Renewable Sustainable Energy 9, (2017) reported by the National Renewable Energy Laboratory in 2013 (NREL, 2016). For planning purposes, 20 years is typically used. However, considering that many wind turbines installed in Alaska fall in the 10-to-1000-kW range, it might be prudent to revise the values given in Table II or work with the vendor on sufficient warranties. Capacity factor Figure 3 shows predicted capacity factors based on REF applications. Capacity factors above 40% were removed since they are possible but unlikely. The capacity factor was calculated as the total wind energy used to supply electrical loads (excluding diversion loads) in 1 year, divided by the energy that would be harvested from the wind if the turbines were outputting their rated (maximum) power the entire year. In general, the capacity factor should depend primarily on the wind power class of the installation site for low- and mediumpenetration hybrid-diesel systems. Wind power class refers to the available energy from the wind, as outlined by the U.S. Department of Energy (DOE). At high wind power penetrations, it might not be possible to use all the energy from wind for electrical loads. In such cases, thermal loads can be supplied, but this is not reflected in the capacity factor. Thus, the capacity factor might not be as good of a metric as reduction in diesel consumption, and even then, a distinction will have to be made between the displacement of diesel fuel slated for electricity production and heat production. Factors affecting the variance of capacity factors at wind power classes include the rated wind speed and the height of the wind turbines and the resolution of the wind power class map. A turbine rated for low wind speeds will have a higher capacity factor at low wind speeds than a turbine rated for high wind speeds. At high wind speeds, a turbine rated for high wind speeds may have a higher capacity factor due to a higher cutoff wind speed. Taller wind turbines experience higher wind speeds, which usually results in higher capacity factors. The wind power classes in Fig. 3 are from the Alaska Energy Data Inventory (AEDI) wind power class map ( Localized wind speeds can be higher than what is shown by the resolution of the wind power class map, resulting in higher capacity factors used in the REF applications than what is shown in Fig. 3. Diesel offset Figure 4 shows the predicted improvement in grid diesel electric efficiency based on REF applications. This improvement is calculated as the total electric consumption divided by the FIG. 4. System diesel electric efficiency before and after installing wind turbines, plotted against wind energy penetration. Corresponding before-and-after values are connected by a dashed line. Wind energy penetration is the amount of wind energy in kilowatt-hours that can be generated (assuming no diversion), divided by grid electrical consumption.

8 VanderMeer, Mueller-Stoffels, and Whitney J. Renewable Sustainable Energy 9, (2017) FIG. 5. The predicted ratio of diesel offset for thermal loads to electric loads supplied by wind energy. Wind energy penetration is the amount of wind energy in kilowatt-hours that can be generated (assuming no diversion), divided by the grid electrical consumption. diesel consumed. Wind power supplies some of the electric load, reducing diesel consumption. At higher penetrations, not all wind energy can be used for electric loads. This excess generation can be used to supply thermal loads. Figure 4 only accounts for electric loads. Wind energy displaces more diesel by supplying electric loads than thermal loads. Assuming a boiler that has an efficiency of Btu/gal and a diesel generator that has an efficiency of 13 kwh/gal, 33 kwh of wind energy will displace around 1 gal of diesel if it supplies a heating load and 2.5 gal of diesel if it supplies an electrical load. Thus, since diesel is more efficient at supplying thermal loads than electric loads, wind energy displaces more diesel by supplying electric loads. Figure 5 shows the predicted ratio of diesel offset for thermal loads to electric loads supplied by wind energy at different energy penetrations. These ratios simply show what the project plan was for using thermal and electric loads and do not necessarily reflect what is feasible or optimal. For example, a project may plan to install a boiler or several masonry thermoelectric heaters that use some but not all of the excess wind generation. One project below 20% wind energy penetration included a plan to use a large portion of the wind energy to supply thermal loads. Typically, this plan is not economical due to the amount of wind energy required to displace diesel for thermal loads. In general, only projects above around 25% energy penetration included a plan to supply thermal loads with wind energy. Cost per kilowatt-hour Figure 6 shows the resulting levelized cost of electricity (LCOE), assuming yearly O&M costs of $0.036/kWh of wind energy generation that increase with an inflation rate of 2%, an interest rate of 5%, a lifetime of 20 years, and different average capacity factors; LCOE equations are given in Appendix B. These costs do not take into account subsidies; thus, they are higher than the actual cost to the utility and not directly comparable to subsidized diesel generating costs. Table V shows LCOE values for various installed wind capacities and average capacity factors. Conditions for greatest efficiency Consistent, high-speed, non-turbulent winds result in the best wind farm performance. Turbines are rated by IEC design requirements for average wind speed, extreme 50-

9 VanderMeer, Mueller-Stoffels, and Whitney J. Renewable Sustainable Energy 9, (2017) FIG. 6. Levelized cost of energy for different installed capacities and capacity factors, assuming yearly O&M costs of $0.036/kWh of wind energy generation that increase with an inflation rate of 2%, an interest rate of 5%, and a lifetime of 20 years. year gust, and turbulence. The wind power at heights of 10 m and 50 m is classified by DOE using Wind Power Class 1 7, with Class 3 and above usually suitable for utility wind power (Elliot et al., 1986). A grid s ability to accept power from a wind farm at a given moment may result in having to divert or curtail excess generation. Excess generation can be used for either energy storage or controllable loads. Cost curve over time The REF applications did not show any statistically significant change in costs over time. As-built costs from Alaska (outlined in Appendix C) did not show a significant change in cost over time either. In Fig. 7, as-built costs for projects in Alaska are plotted in 2015 dollars against the year the projects were installed. Installed costs by major components Information on installed costs by major components is given in the Capital Costs section and in Appendix A. Transportation Information on transportation costs is given in the Capital Costs section and in Appendix A. TABLE V. LCOE values for various installed wind capacities and average capacity factors. Installed wind capacity (kw) 20% average capacity factor ($/kwh) 30% average capacity factor ($/kwh) 40% average capacity factor ($/kwh)

10 VanderMeer, Mueller-Stoffels, and Whitney J. Renewable Sustainable Energy 9, (2017) FIG. 7. As-built costs for projects in Alaska plotted against the year of installation. Costs are converted to 2015 dollars. Technology trends The wind market is moving toward larger and more powerful wind turbines, often with gearless direct-drive generators (Wiser and Bolinger, 2017). Due to the size of communities in Alaska as well as other factors, turbines installed in Alaska are smaller than the trend in the larger market. Evidence indicates that installing overcapacity wind farms together with energy storage and significant diversion into thermal loads allows communities to achieve diesel-off and least-cost energy when considering the displacement of diesel fuel for both electricity and heat (Simpkins et al., 2015). Robust direct drive turbines are currently popular in Alaska since they require less maintenance. Refurbishment/upgrade market The refurbished wind turbine market is significant. As wind farms upgrade to larger wind turbines, companies such as Windmatic purchase the old wind turbines to refurbish and resell them (Fauy et al., 2010). Upgrades in new and refurbished models that are popular for Alaska include marine-grade paint on the tower and black Teflon paint and heaters on the blades. Modifications to turbines, such as hub extensions, and control and drive upgrades are considered at times. Realized cost savings Cost savings from integrating renewable power are difficult to gauge due to technical and incentive impacts at the entire power system level. At the technical level, for example, the effects of diminished losses of secondary services, such as recovered waste heat and reductions in fuel efficiency, are hard to gauge, as they depend not only on average reductions in load but also on specific operating schemes regarding the minimum allowable load on diesels and spinning reserve kept. CONCLUSIONS Wind power systems have been installed in Alaska, both in remote areas and along the road system. Total installed capacity exceeds 60 MW, with installations ranging from 40 kw to 24 MW. The costs per kilowatt for the different components of wind turbine installations were found to decrease with increasing size of installation. Exceptions were the costs for transmission and integration, which did not vary significantly with the capacity of the installed system.

11 VanderMeer, Mueller-Stoffels, and Whitney J. Renewable Sustainable Energy 9, (2017) Individual costs were tested for correlations with the region and year, but in most cases, the relationship was not significant. Capacity factors range from approximately 10% to 40%. The rated wind speed and height of the wind turbines and the resolution of the wind power class map affect the variance of capacity factors at wind power classes. A turbine rated for low wind speeds will have a higher capacity factor at low wind speeds than a turbine rated for high wind speeds. At high wind speeds, a turbine rated for high wind speeds may have a higher capacity factor due to a higher cutoff wind speed. Taller wind turbines experience higher wind speeds, which usually result in higher capacity factors. Wind power technology is relatively mature internationally, nationally, and in Alaska. Turbines have minimal down time in a year, often for their entire design life and sometimes longer. The remoteness of communities in Alaska leads to higher transportation, infrastructure, and maintenance costs. Significant economies of scale could be gained with larger wind turbines. The size of wind turbines that can be installed in Alaska is limited by the small size of the state s communities and the challenges of integrating high penetrations of wind power into microgrids. Excess wind generation and the need for more complex integration equipment increase with higher penetrations of wind power. Demonstrations of different high-penetration wind power integration techniques are needed, including demand-side management and wind to heat. Improvements and reduction in costs of integration equipment such as energy storage systems will help to achieve higher penetrations of wind power. An integrated approach is needed that includes analysis to understand lowhanging fruit and mechanisms to allow collaboration between government and industry. Funding is a major issue in implementing grid improvements, and state, federal, and private funds could help with this implementation. ACKNOWLEDGMENTS The authors wish to thank the Alaska Energy Authority for its funding to support this project. Many people have contributed information and insight to this report. For their review and comments, we wish to thank David Burlingame of Electric Power Systems, Inc.; Rob Bensin, Energy Efficiency and Renewable Energy Division Manager at Bering Straits Development Company; Ian Baring-Gould of the National Renewable Energy Laboratory; Steve Gilbert of the Alaska Village Electric Cooperative, Inc.; and Dan Smith, Josh Craft, and Neil McMahon of the Alaska Energy Authority. APPENDIX A: LINEAR REGRESSION PLOTS FOR EACH COST CATEGORY FIG. 8. Graph of data and best-fit equation for energy analysis costs of installing wind power.

12 VanderMeer, Mueller-Stoffels, and Whitney J. Renewable Sustainable Energy 9, (2017) FIG. 9. Graph of data and best-fit equation for conceptual design costs of installing wind power. FIG. 10. Graph of data and best-fit equation for final design costs of installing wind power. FIG. 11. Graph of data and best-fit equation for turbine component costs of installing wind power.

13 VanderMeer, Mueller-Stoffels, and Whitney J. Renewable Sustainable Energy 9, (2017) FIG. 12. Graph of data and best-fit equation for turbine installation costs of installing wind power. FIG. 13. Graph of data and best-fit equation for transmission costs of installing wind power. FIG. 14. Graph of data and best-fit equation for foundation and supporting infrastructure costs of installing wind power.

14 VanderMeer, Mueller-Stoffels, and Whitney J. Renewable Sustainable Energy 9, (2017) FIG. 15. Graph of data and best-fit equation for the integration cost of installing wind power. FIG. 16. Graph of data and best-fit equation for the turbine transportation cost of installing wind power. APPENDIX B: LEVELIZED COST OF ENERGY EQUATIONS Equations for LCOE: NPV c CRF LCOE ¼ Annual Energy Production ; NPV c ¼ CAPEX þ XN j 1 þ i OM; 1 þ r CRF ¼ j¼1 r 1 ð1 þ rþ N ; where NPV c is the net present value of the annual cost of the system, CAPEX is the capital expenditure, i is the inflation rate, r is the interest rate, and N is the system lifetime in years.

15 VanderMeer, Mueller-Stoffels, and Whitney J. Renewable Sustainable Energy 9, (2017) APPENDIX C: COMPARISON OF REF APPLICATION COST ESTIMATES WITH AS-BUILT COSTS Figure 17 shows the total project costs and installed capacity predicted by REF applications compared to future as-built costs and installed capacity from the same community. The projects can be seen in Table VI. Note that the REF data do not necessarily represent all costs. For example, many projects do not include the engineering and energy assessment. Other projects do not include transport, foundation, installation, integration, and/or transmission costs. It is not always clear whether these costs are included with other costs. Also, it is not known what exactly the asbuilt costs cover. That being said, a trend of lower power and higher CAPEX/Power for as-built projects compared with the associated REF application is apparent. Figure 18 shows a comparison of the best fits for as-built costs and the costs predicted in the REF applications. The fits are relatively similar, indicating that the costs calculated from the REF applications and used in this paper are representative of actual costs. FIG. 17. REF costs and associated as-built costs and power are connected with dashed lines. The projects can be seen in Table VI. The fit to the REF data is shown by the solid line. FIG. 18. As-built costs with the as-built fit and REF fit. The fit for both datasets is similar. This indicates that the costs calculated from the REF applications and used in this paper are representative of actual costs. The as-built costs show a more dramatic curvature, but for absolute values, they are relatively similar.

16 TABLE VI. Comparison of REF application costs and as-built costs for the same location. Not all as-built costs are for the same project as the REF application. The ones with a dark grey background were built before the REF application. Names Unalakleet Wind Farm Bethel Wind Power Project Times 4 Bethel Wind Farm (BNC land) Year Size (kw) Assessment Design conceptual Design final REF applications Hardware Transportation Foundation Installation Transmission Integration Total Size (kw) Date As built costs CAPE- X ($/ kw) Power ratio (kw/kw) November April April 2014 Bethel April 2014 Kongiganak Wind Farm Quinhagak Wind Farm Mekoryuk Wind Farm Toksook Bay Wind Farm Expansion Hooper Bay Wind Farm December 2012 CAPEX ratio ($/$) November February October June July VanderMeer, Mueller-Stoffels, and Whitney J. Renewable Sustainable Energy 9, (2017)

17 TABLE VI. (Continued.) Names Kotzebue Wind Farm Expansion Year Size (kw) Assessment Design conceptual Design final REF applications Hardware Transportation Foundation Installation Transmission Integration Total July May May May May October August May 2014 Kotzebue July May May May May October 2006 Size (kw) Date As built costs CAPE- X ($/ kw) Power ratio (kw/kw) CAPEX ratio ($/$) #N/A 0.06 #N/A #N/A 0.14 #N/A #N/A 0.03 #N/A #N/A 0.04 #N/A #N/A 0.02 #N/A #N/A 0.06 #N/A #N/A 0.01 #N/A #N/A 0.11 #N/A #N/A 0.25 #N/A #N/A 0.06 #N/A #N/A 0.07 #N/A #N/A 0.04 #N/A #N/A 0.11 #N/A VanderMeer, Mueller-Stoffels, and Whitney J. Renewable Sustainable Energy 9, (2017)

18 TABLE VI. (Continued.) Names Buckland Wind Farm Deering Wind Farm Eva Creek Wind Farm Year Size (kw) Assessment Design conceptual Design final REF applications Hardware Transportation Foundation Installation Transmission Integration Total August May May October October 2012 Size (kw) Date As built costs CAPE- X ($/ kw) Power ratio (kw/kw) CAPEX ratio ($/$) #N/A 0.01 #N/A Eva Creek October Delta Junction Wind Farm September June September 2013 Sand Point Wind February 2011 Sand Point February 2011 Tuntutuliak High Penetration Wind Diesel July VanderMeer, Mueller-Stoffels, and Whitney J. Renewable Sustainable Energy 9, (2017)

19 TABLE VI. (Continued.) Names Year Size (kw) Assessment Design conceptual Design final REF applications Hardware Transportation Foundation Installation Transmission Integration Total Size (kw) Date As built costs CAPE- X ($/ kw) Power ratio (kw/kw) Shaktoolik Wind February Emmonak/ Alakanuk Wind & Trans September Selawik turbine upgrade October 2003 Wales July 1998 CAPEX ratio ($/$) #N/A 1.30 #N/A VanderMeer, Mueller-Stoffels, and Whitney J. Renewable Sustainable Energy 9, (2017)

20 VanderMeer, Mueller-Stoffels, and Whitney J. Renewable Sustainable Energy 9, (2017) APPENDIX D: DIESEL OFFSET PER INSTALLED CAPACITY CALCULATION The total electrical energy produced in a year by a wind farm is E e ¼ 8760 h X 3 P, where X 3 is the average capacity factor and P is the nameplate capacity of the windfarm. Assuming that the energy content of diesel is 38 kwh/gal and an average diesel efficiency of g will displace E e =ð38 gþ gallons of diesel. Thus, the offset per installed capacity in gal/kw is OFF ¼ 230 g X 3. Elliot, D., Holladay, C., Barchet, W., Foote, H., and Sandusky, W., Wind Energy Resource Atlas of United States, Report No. DOE/CH (1986). Fauy, G., Schworer, T., and Keith, K., for Alaska Isolated Wind-Diesel Systems: Performance and Economic Analysis, Alaska Energy Authority; accessed June McMahon, N., personal communication (2015). NREL, for Distributed Generation Renewable Energy Estimate of Costs. NREL, Updated August 2013; accessed 28 January Simpkins, T., Cutler, D., Hirsch, B., Olis, D., and Anderson, K., Cost-optimal pathways to 75% fuel reduction in remote Alaskan villages, Report No. NREL/CP-7A ; accessed 28 October Wiser, R. and Bolinger, M., 2016 Wind Technologies Market Report, Report No. DOE-GO/ (2017).