Long-Term Freezing Temperatures Frequency Change Effect on Wind Energy Gain (Eurasia and North America, 1950 & 2019)

Citation: Aizpurua-Etxezarreta, M.; Carreno-Madinabeitia, S.; Ulazia, A.; Sáenz, J.; Saenz-Aguirre, A. Long-Term Freezing Temperatures Frequency Change Effect on Wind Energy Gain (Eurasia and North America, 1950–2019). Sustainability 2022,14, 5630. https://doi.org/ 10.3390/su14095630 Academic Editor: Domenico Curto Received: 5 April 2022 Accepted: 1 May 2022 Published: 7 May 2022 Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. Copyright: © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/). sustainability Article Long-Term Freezing Temperatures Frequency Change Effect on Wind Energy Gain (Eurasia and North America, 1950–2019) Maddi Aizpurua-Etxezarreta 1, Sheila Carreno-Madinabeitia 2,* , Alain Ulazia 1, Jon Sáenz 3,4 and Aitor Saenz-Aguirre 1 1Energy Engineering Department, University of the Basque Country (UPV/EHU), E-20600 Eibar, Spain; [email protected] (M.A.-E.); [email protected] (A.U.); aitor[email protected] (A.S.-A.) 2Department of Mathematics, University of the Basque Country (UPV/EHU), E-01006 Vitoria-Gasteiz, Spain 3Department of Physics, University of the Basque Country (UPV/EHU), E-48080 Leioa, Spain; [email protected] 4Plentziako Itsas Estazioa (BEGIK), University of Basque Country (UPV/EHU), E-48620 Plentzia, Spain *Correspondence: sheila.carr[email protected] Abstract: The persistent freezing conditions in cold regions are the cause of ice accretion on mechanical and instrumental elements of wind turbines. Consequently, remarkable Annual Energy Production (AEP) losses are prone to occur in those wind farms. Following global expansion of wind energy, these areas have had increased study interest in recent years. The goal of these studies is an improved characterisation of the site for the installation of turbines, which could prevent unexpected high AEP losses due to ice accretion on them. In this context, this paper provides an estimation of the freezing temperatures frequency (FTF) at 100 m over latitudes and evaluates the changes during the last 70 years. To that end, hourly surface temperature data (2 m above surface) from the ERA5 reanalysis is used in the [50 ◦ N, 75 ◦ N] latitudinal belt for the period 1950–2019. The obtained results show an average reduction of FTF hours of 72.5 h/decade for all the domain, reaching a maximum decrease of 621 h/decade on the southeast coast of Greenland and a 60% annual reduction at a specific location in Scandinavia. In terms of AEP a maximum gain of more than 26% would be projected, as categorised by the the International Energy Agency. Keywords: wind energy potential; global warming; ice accretion; annual energy production; ERA5; temperature; applied mathematics 1. Introduction Wind energy is one of the fastest-growing global renewable electricity generation technologies. An increasing number of wind farms are erected, breaking records in new wind power installations since 2015. Global installed wind power capacity reached 744 GW in 2020, which is capable of generating 7% of the world’s electricity demand, according to data provided by the World Wind Energy Association [ 1 ]. Many of the favourable areas for wind farm construction are in cold regions of high latitudes, such as Eastern and Northern Europe, North America, and Asia [ 2 ]. According to statistics from the International Energy Agency (IEA), by the end of 2015, approximately 30% of all wind turbine installations were located in cold regions [3]. These areas present great potential for wind energy exploitation for several reasons. On the one hand, from the point of view of global atmospheric circulation, polar areas are characterised by high pressure due to surface cooling; as a result of the subsidence in them, polar regions tend to be quite stable. However, in regions closer to the Equator that are thereby affected by the Ferrel cell (around 60 ◦ N in our case), there exist many low-pressure areas due to baroclinic instability. These extratropical cyclones produce wind regimes of remarkable value for the exploitation of wind energy (e.g., the Baltic Sea and the UK). It is in these latitudes where the synoptic surface westerly regime manifests itself with maximal Sustainability 2022,14, 5630. https://doi.org/10.3390/su14095630 https://www.mdpi.com/journal/sustainability Sustainability 2022,14, 5630 2 of 15 intensity in accordance with general atmospheric circulation patterns [ 4 ]. On the other hand, in cold climate areas, wind energy potential is 10% higher than that in other regions due to increased air density at low temperatures within seasonal changes, mainly from summer to winter [ 5 – 9 ]. Numerous locations in cold climate areas, such as the Swiss Alps, the northern Scandinavian coastline, many parts of China, northern USA, and Canada offer high wind-energy potential, especially during the winter months [10]. However, more than 50% of wind turbines located in cold regions are affected by various icing events in winter [ 11 ]. Temperatures below the freezing point of water and wet environments potentially lead to ice formation and icing persistence on structures exposed to wind, resulting in considerable power losses. Processes of ice formation, type, duration, and severity depend on a subset of conditions [ 2 ]. Temperature, wind speed, liquid water content, and droplet size distribution are the most important meteorological conditions to consider when studying ice formation. Nevertheless, there is a substantial lack of information, especially on the latter two variables, which can be used to assess and predict the severity and frequency of the phenomenon [ 12 ]. In addition, identical weather conditions do not always cause the same icing effects on the same turbine or even different turbines. At different operating speeds, depending on the rotor, angle of attack, rotational speed, and other factors, icing effects are different [2]. This is why it is so complex to carry out a detailed study on ice formation and production losses. Even so, several studies investigated the ice accretion process [12–17] on wind turbine blades, and its detrimental effect on the performance and power production of wind turbines [ 2 , 10 , 18 , 19 ]. Mapping all potential risks and site-assessment studies are keys to success in a wind farm investment [ 20 ]. The main drawback associated with ice accretion on wind turbines from which wind turbine manufacturers suffer is the considerable Annual Energy Production (AEP) loss experienced by wind turbines located in cold climate areas. The two reasons behind these losses are the reduced aerodynamic performance of iced wind-turbine blades and the unexpected stoppage periods of wind turbines due to high aerodynamic degradation. However, according to the Intergovernmental Panel on Climate Change (IPCC) reports, annual warming [21] in the northern hemisphere is expected with global warming. Every decade since 1880, the temperature of Earth has warmed by 0.14 ◦ C [ 22 ]. Moreover, as shown in Collins et al. [23] , the number of frost days (number of days below 0 ◦ C) is significantly reduced globally. According to Yan et al. [24], frost days, cold days, and cold nights are showing a decreasing trend in China. This suggests that temperatures below the freezing point of water for atmospheric ice formation will occur less frequently in the coming years. In this context, the estimation of AEP losses before the installation of wind turbines is necessary for wind turbine manufacturers, especially in cold climate areas. The most suitable location for the installation of wind turbines can be selected on the basis of AEP estimations, and posterior unexpected losses that might compromise the contracts can be minimised. Our analysis examines the temporal evolution of freezing temperatures as a necessary condition for the formation of meteorological ice, and subsequent AEP losses from 1950 to 2019 to identify clear patterns in its evolution. As a result, a foundation for the site assessment, operation, and maintenance of wind turbines in northern latitudes was developed on the basis of their expected AEP losses due to icing. The paper is organised as follows: Section 2presents datasets and the method used in this study. Section 3describes the results, Section 4discusses the results, and Section 5 concludes this study. 2. Data and Methodology 2.1. Data In this study, temperatures estimated by ERA5 [ 25 ] reanalysis at 2 m over the surface in high latitudes of the northern hemisphere were used. ERA5 is the most recent reanalysis [ 26 ] developed at the European Centre for Medium-Range Weather Forecasts Sustainability 2022,14, 5630 3 of 15 (ECMWF) and it is freely available through the Copernicus Climate Data Store. It is the fifthgeneration reanalysis by the ECMWF. The available ERA5 data period was from 1950 to the present; this reanalysis archives data at an hourly output resolution and a recommended 0.25 ◦× 0.25 ◦ spatial resolution. Different researchers successfully validated ERA5 in wind potential assessments for both wind speed and density, and consequently temperature, even suggesting that it can lead to wind energy models [7,27,28]. According to ERA5 data observation documentation [ 29 ], temperature data from buoys are not assimilated. For this reason, in this study, two buoys were validated. One corresponded to Station 44078, OOI Irminger Sea Surface Mooring managed by the National Oceanic and Atmospheric Administration (NOAA) in Greenland (68.47 ◦ N, 9.26 ◦ W), which was used to perform validation during the period of 10 September 2014–31 December 2019. The second buoy is located on the Iceland coast (59.95 ◦ N, 39.57 ◦ W) and is available from the Pangaea Project (PANG) [ 30 ]. These data were used for validation during the period of 23 November 2007–21 August 2009. Validation was carried out by means of a Taylor diagram [ 31 ] because it shows the correlation, centred root mean standard error, and standard deviation (represented by the horizontal axis, radial distance, and horizontal axis, respectively) in the same plot. The Taylor diagram is a successful tool for the evaluation of model simulations against observed data. It is possible to visually compare three different quality indicators simultaneously. ERA5 reanalysis data are represented by a grey point in the x axis in the diagram (see Figure 1), and visual inspection allows for easily determining how close the model data (green dots) were to the buoy data (i.e., present lower error). It also randomly created 1000 new series using the bootstrap technique to ascertain the sensitivity of the point position in the Taylor diagram to the available sample. 2.2. Annual Freezing Frequency Temperature at 100 m In order to obtain preliminary information on the presence of favourable conditions for icing in wind energy zones, a new variable was defined, namely, annual freezing temperature frequency (FTF), for which the annual hours of temperatures below 0 ◦ C at each grid point at the hub height of the turbines were counted. Furthermore, before obtaining the annual FTF, it is necessary to calculate temperature at hub height, which was 100 m here. According to the American Meteorological Society’s Glossary of Meteorology [ 32 ], temperature inversion corresponds to a layer of the atmosphere at which temperature increases with height. This is not consistent with the most frequent behaviour, which is represented by a reduction in temperature with height. Although temperature inversions are typical at high latitudes due to surface cooling processes, this study approximated a standard atmosphere for the whole domain where temperature decreased through a constant vertical temperature gradient (Γ) in K/m (Equations (1) and (2)). Γ=−0.0065 K m−1(1) t100m =t2m +Γ(100 m −2 m)(2) 2.3. Annual FTF Mean Data and Decadal Trends With the annual FTF data described in the previous section, the mean annual value at each grid point was calculated for the period of 1950–2019, and results are shown in Section 3.2 as a percentage of annual hours. In order to obtain a general decadal trend for the whole domain, annual FTF results were grouped for each decade, and linear regression was calculated with the median of each decade. Lastly, in order to analyse the temporal evolution of this variable in more detail, linear regression with annual data was calculated at each grid point. The objective was to identify areas with the steepest slope where the most important changes occurred. For this purpose, the Theil–Sen estimator was used [ 33 , 34 ], which robustly fits simple linear regression by choosing the median of slopes of all lines crossing each pair of points [ 35 ]. Coefficients of linear regressions were calculated, obtaining the 95% confidence interval. Sustainability 2022,14, 5630 4 of 15 In this work, the mblm() function of the R-cran [ 36 ] library of the same name was used. This function from the mblm package [ 37 ] is used to fit linear models on the basis of the Theil–Sen method. 2.4. AEP Loss Approximation The calculation of the AEP losses of a wind turbine is based on the methodology proposed by the IEA (see Table 1). This table relates the duration of meteorological icing events with an ice class and associated potential AEP losses on the basis of expected aerodynamic performance reduction and stoppage periods due to the ice accretion that the meteorological icing would cause. Ice classes are based on measurements, and an estimate of possible best and worst cases under given icing conditions [38]. Table 1. Ice classification according to IEA. IEA Ice Class Meteorological Icing AEP Loss Due to Icing [−]% of Year % 1 0–0.5 0–0.5 2 0.5–3 0.5–5 3 3–5 3–12 4 5–10 10–25 5>10 >20 This IEA ice classification provides a first idea of the transcendence of freezing and its consequences on the AEP for a given location. This ice classification refers to long-term conditions; for individual years or winters, results may lie outside these classes. If the initial categorisation of ice is the second or higher, the regulation recommends an ice measurement campaign and detailed study of related losses [3]. The approximation of FTF to the meteorological icing duration is appropriate given that air temperature below 0 ◦ C is a necessary condition for icing [ 18 , 39 ]. Therefore, provided maps and associated trends with this variable constitute a rough estimate of potential icing conditions, and should not be used as the only indicator when it comes to studying real icing conditions for the design of wind farms in cold climates. For that case, direct measurements of icing parameters (liquid water content, medium volume droplet, and wind speed) are vital and can provide more accurate information than ice maps can. In this paper, on the basis of annual FTF approximation data, the yearly duration of meteorological icing conditions was estimated, and ice categorisation and AEP loss associated with that meteorological icing can thereby be computed, following Table 3-1 of wind energy projects in cold climates [3]. 2.5. AEP Gain The general reduction in freezing days due to global warming produces changes in AEP during the entire period ( ∆AEP ) that can be quantified using values in Table 1. The values of the table were linearly interpolated to establish a continuous relationship between meteorological icing duration and AEP loss intervals, and averages over decades were then calculated. ∆AEP is AEP loss in percentage in the entire period (Equation (3)). ∆AEP =AEP1950−1959 −AEP2010−2019, (3) given that it is expected that AEP2010−2019 <AEP1950−1959. 3. Results 3.1. ERA5 t2mValidation on Two Buoys The nearest grid point of t2m from ERA5 reanalysis and buoy data was compared through the Taylor diagram in Figure 1. The validation of the Iceland and Greenland buoys had a correlation of above 0.8 and root mean square error of less than 1 ◦ C and Sustainability 2022,14, 5630 5 of 15 2 ◦ C. The standard deviation of the model was very similar to the observation in Iceland, but in Greenland, ERA5 showed greater variability than that of the buoy measurements, almost doubling its standard deviation from 2 to 3.6 ◦ C. Although worse validation data were expected nearer the Arctic (i.e., Greenland) because of the systematic error found by Wang et al. [40] at temperatures below − 25 ◦ C and the lack of data in the Artic area due to the harsh environmental conditions (see Section 4), validation was very good in both cases in terms of correlation and relative errors. This was an expected result, given that ERA5 performs better in terms of accuracy than previous models and regarding reanalysis, such as ERA-Interim, even in the remote weather stations of the Antarctic, with fewer data assimilation sources than those in the northern hemisphere and the Arctic. ERA5 is highly accurate, and its higher spatial and temporal resolution significantly reduces cold coastal biases identified in ERA-Interim, increasing accuracy representing the wind [41]. Greenland Standard deviation Standard deviation 0 1 2 3 4 5 0 1 2 3 4 5 1 2 3 4 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.95 0.99 Correlation Observed Iceland (59.95 ºN, 39.57 ºW) Standard deviation 0 1 2 3 4 5 0 1 2 3 4 5 1 2 3 4 5 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.95 0.99 Correlation Observed Iceland Greenland Figure 1. Taylor diagrams of t2m between ERA5 and Iceland and Greenland buoys. 3.2. Annual FTF Mean Data and Decadal Trends Annual FTF data at each grid point in high latitudes (50 ◦ N, 75 ◦ N) of the northern hemisphere during 1950–2019 were quantified, and the mean of the results is shown in Figure 2. The figure shows that freezing is higher over land than that over the sea, and increases at higher latitudes. In the Arctic Ocean, high values of annual FTF were observed. At below about 60 ◦ N latitude, annual FTF was zero in offshore areas, especially in areas affected by the transportation of energy due to the Atlantic Ocean Gulf Stream. Maximal annual FTF was found in inland areas in Greenland, where all hours of the year are below 0 ◦C. In relation to decadal annual FTF changes, for a domain covering all high latitudes (50 ◦ N, 75 ◦ N), a − 72.5 h/decade slope is shown in Figure 3. This slope had a large 95% confidence interval [ − 84.33, − 50.21] h/decade. Because the study area is large, it is difficult to assign a general behaviour to it. However, lower and upper limits of the slope’s confidence interval were negative, and this confirmed an overall descent in the historical evolution of annual FTF. This descent was higher in the last three decades (1990–2019). In order to identify zones with different behaviours, the slope of annual FTF at each grid point was calculated to obtain more accurate information (Figure 4). Sustainability 2022,14, 5630 6 of 15 20 40 40 60 60 60 60 60 60 60 60 80 180°E 0°E 60°N Figure 2. Mean value of annual FTF at 100 m for ERA5 in 1950–2019. Results in percentage of hours per year. 0 2500 5000 7500 1950−1959 1960−1969 1970−1979 1980−1989 1990−1999 2000−2009 2010−2019 hours/year AFF Figure 3. Decadal box plot of annual FTF in 1950–2019 from ERA5 data. Red dashed line, Theil–Sen estimator of median of annual values. Sustainability 2022,14, 5630 7 of 15 180°E 0°E 60°N −600 −500 −400 −300 −200 −100 0 100 200 slope (h/decade) Figure 4. Map of decadal trends in annual FTF for 1950–2019. Slope obtained using Theil–Sein estimator of decadal ERA5 data. White values, nonsignificant. Most of the results were significant at the 95% confidence level, but 7% were not significant and are plotted with white. The highest reduction up to − 621 h/decade is located on the southeast coast of Greenland. Steep declines are also observed in Iceland, Norway, the Baltic Sea and west of Alaska. In offshore areas where the mean annual FTF values (Figure 2) are zero, the trend is around zero or slightly positive. However, in the offshore areas of higher latitudes, where there are high annual FTF mean values, the negative slope is very pronounced, being maximum over the sea. Finally, increasing the detail over the Scandinavian area, location of wind farms and annual FTF information are combined. Figure 5shows the location of wind farms registered up to 2016, with a total capacity of 515 MW. These data were extracted from the Wind Power Database [ 42 ]. The annual FTF slopes are also shown in Figure 5. In the area shown (4.5 ◦ E–40 ◦ E, 52.5 ◦ N–75 ◦ N) an average slope of − 119 h/decade is obtained. The maximum slope, − 310 h/decade is located in southwest coast of Norway at 5.75 ◦ E, 62 ◦ N point. The decrease in the annual FTF at this point is analysed by taking the values of the annual hours of both extremes 1950 (3623 h) and 2019 (1447 h). A reduction of 60% is obtained in a place where wind farms are already placed. Sustainability 2022,14, 5630 8 of 15 −50 −50 −150 −150 −150 −150 −150 −150 −150 −350 −300 −250 −200 −150 −100 −50 0 slope (h/decade) 10° 20° 30° 40° 60° 60° 70° 70° Figure 5. Map of the decadal trends in annual FTF for period 1950–2019 obtained using Theil-Sein estimator and ERA5 data in Scandinavian area. The green dots indicate the location of wind farms. 3.3. AEP Losses Approximation and Gain The temporal evolution of the IEA ice categorisation approximation, see Table 1, for the analysed regions and during the analysed decades is presented in Figure 6. Each IEA ice category is highlighted in the maps in Figure 6with a different color. The changes of the colors over the decades is indicative of the transformation of the freezing severity decade by decade in the region. 180°E 0°E 60°N 180°E 0°E 60°N 180°E 0°E 60°N 180°E 0°E 60°N 1 2 3 4 5 IEA Ice class 1950−1959 1960−1969 2000−2009 2010−2019 Figure 6. Decadal distribution of IEA ice class according to Table 1. Categorisation based on annual FTF approximation data at 100 m. Sustainability 2022,14, 5630 9 of 15 Analysis of the temporal evolution of the temperature over the years mainly showed that it led to mild favourable conditions, especially in the western coast of Canada, the Baltic Sea, the United Kingdom, and the Atlantic Ocean near Iceland and Norway. This reduction in turn caused a reduction in environmental harshness for the operation and maintenance of wind turbines in cold climate conditions, and improved their energy production. In the last decade (2010–2019), these conditions were weakened. In the Baltic Sea and Scotland, freezing severity decreased from Category 5 to Category 4. The Norwegian Sea is likewise one of the most affected places, together with western Canada (Figure 7). This weakening would be accompanied by a decrease in AEP losses from wind turbines, that is, an AEP gain. 180°E 0°E 60°N 1->2 2->1 2->3 3->2 4->2 4->3 4->3 4->5 5->2 5->3 5->4 IEA Ice class transition Figure 7. Transitions in IEA ice class according to Table 1from first (1950–1959) to last decade (2010–2019). Categorisation based on annual FTF approximation data at 100 m. The temporal evolution of ice categorisation thus shows a clear pattern of a reduction in freezing over the decades, especially in areas with a cold climate due to their prolonged exposure to icing conditions. On the basis of the long-term trend of the meteorological icing, AEP losses of a wind turbine located in a cold climate area are thus expected to decrease. Given the results in the previous maps and the well-known wind energy potential of the locations, the Gulf of Alaska and Northern Europe were selected for deeper study. 3.3.1. AEP Gain in Gulf of Alaska A milder climate causes fewer icing events, and its effect on wind turbine performance is lower. For instance, in the Gulf of Alaska, in domain 180 ◦ W, 120 ◦ W, 50 ◦ N, 65 ◦ N, a mean reduction in AEP losses of 5.19% was observed by comparing decades from 1950 to 2019 (Figure 8). This is a clear indicator (under currently expected climatological trends) of the favourable long-term prospective for the installation of wind turbines in cold climate areas.