Offshore wind power around the Iberian Peninsula: variability, complementarity and added value for the power system

The advances in floating offshore wind energy are opening deep sea areas, like the coastal waters of Iberian Peninsula (IP), for the installation of wind farms. The integration of this new energy source in a semi-closed power system with an already high share of variable renewable energies would be facilitated if the potential contribution of offshore wind energy shows reduced variability and limited seasonal variations, as the power demand in IP shows two maxima in winter and summer. The aims of this study are the analysis of temporal variability and spatial complementarity of the potential installation sites, and the identification of an optimal combination of installation areas that minimizes the temporal variability of the aggregated offshore contribution. In order to better capture the marked mesoscale features of winds around the IP, wind data from a very high resolution reanalysis (COSMO-REA6) are used. The analysis considers allowed areas for installation, delimited by the maritime spatial planning. Northern coast areas are characterized by high capacity factors (CFs) and high seasonality, while the lower CFs at the western and southern coasts are compensated by a limited seasonality. Pairwise correlation between the potential areas shows outstanding results, with several negative correlation values within a synoptic scale region, in contrast to other mid-latitude regions like the North Sea or the Eastern USA coast. An optimal aggregation of areas includes at least one area at each of the four main Iberian coasts. A strong reduction of hourly variability is obtained through the resulting combinations, and the seasonality of the aggregated CF is clearly below the values for other offshore areas. Therefore, offshore wind energy can indeed offer an added value for the Iberian power system beyond the high resource amount, reducing the need for storage or backup plants.


Introduction
Renewable energy sources are crucial for the decarbonization process in next decades.In this challenge, renewable technologies such as solar photovoltaic, onshore and offshore wind can be the cornerstone, but high installation rates are needed [1] for addressing ambitious targets that could mitigate climate change impacts.The European Union plans to increase investment in renewable energy sources at both continental and national levels to become the first climate-neutral continent by 2050 (Renewable Energy Directive 2018/2001).In the Iberian Peninsula (IP), variable renewable energy technologies like onshore wind and solar photovoltaic are currently by far the leading technologies in the way towards a decarbonized power system [2,3].The Iberian power system is semi-closed, with rather limited interconnections to France and Morocco, and the share of renewables is already high, approaching 50% in Spain and 60% in Portugal [4].
In addition to onshore wind and photovoltaic technologies, currently with a very high rate of power installation, the IP has a large potential for the development of offshore technologies, due to the great length of its coasts [5][6][7].Offshore wind energy has certain advantages over traditional onshore wind energy, such as an increased capacity factor (CF) or the higher availability of new areas for the development of future wind farms [8][9][10].The IP presents several areas with significant offshore wind resources, most notably at the northwestern coast, near the Gulf of Lion or around Gibraltar Strait [9,[11][12][13][14][15][16].However, there are spatial planning limitations for the deployment of floating offshore wind towers, due to other uses of the marine areas or the existence of protected areas [17].In this context, the Spanish and Portuguese governments have established a legal framework (Royal Decree 363/2017 and Decree-Law No. 38/2015, respectively) to facilitate the installation of offshore wind farms through maritime spatial planning (MSP).In the Spanish MSP, areas of high potential are delimited for the development of offshore wind energy, while the Portuguese MSP identifies existing and potential areas for offshore renewable energy development.Information from both MSPs is publicly available and updated in the INFORMAR [18] and PSOEM [19] geoportals, respectively.
The integration of offshore wind energy in the semi-closed Iberian power system, that has already a high share of variable renewable energy sources, would be facilitated if the contribution of the potential installation sites shows reduced temporal variability.Among other aspects, seasonal variations are particularly important in this case, as the seasonal demand shows two maxima in winter and summer, and summer demand is expected to increase in the future due to the growing frequency of heat waves [20].Previous studies have analysed temporal variability of the wind resource at different time scales [15,[21][22][23].Spatial complementarity among the different installation sites can be used to smooth out the temporal variability of the combined power supply [24][25][26][27][28][29][30][31][32].
Offshore observations are very limited and not sufficient to analyze the wind power production variability and its regional balancing effects for electricity systems.Atmospheric reanalyzes are a common tool which can provide gridded wind speed (WS) at hub-height [33].Among them, the high-resolution reanalysis product COSMO-REA6 (with hourly frequency and a horizontal resolution of approximately 6 × 6 km) [34] has previously been analyzed as a suitable database with good performance compared to both wind mast observations [35] and the widely used ERA5 and MERRA reanalyzes [33,[36][37][38].The much higher spatial resolution of COSMO-REA6 compared to ERA5 and MERRA may be particularly adequate for analyzing offshore wind resources at Iberian coasts, as winds around IP show marked mesoscale features, like the wind channeling around Gibraltar Strait [39] or the strong horizontal gradients of WS associated to the Tramontane wind at northeastern IP [40].
Due to the bathymetric characteristics of the Iberian coast, floating offshore turbines are needed as they can operate in deep waters [41,42].Currently, these turbines have hub heights of between 70 m and 100 m and wind power outputs of between 3 MW and 8 MW.However, rapid advances in offshore wind technology are expected to allow the future installation of turbines with hub heights of around 150 m [43,44] and power outputs of 15 MW (e.g.Vestas236-15.0MW or Siemens-Gamesa14-222 DD).
The main objective of this paper is to analyze the temporal variability and the spatial complementarity of the wind energy potential in allowed areas for the installation of floating offshore wind farms around the IP.An optimal aggregation of potential sites that minimizes the overall temporal variability is obtained.These analyses are performed at 105 and 150 m hub height on different time scales using the high-resolution reanalysis COSMO-REA6.Publicly available information from the Spanish and Portuguese MSPs is used to locate the offshore wind potential areas (OWPAs) included in the analysis.This work presents a first view of the added value of the aggregation of offshore wind energy for the Iberian power system, taking into account the current potential sites delimited in the legislation of the Iberian countries.

Data source
The high-resolution reanalysis COSMO-REA6 [34] used in this work, is based on the COSMO numerical weather prediction model.It has a spatial resolution of 0.055 • and covers the CORDEX EUR-11 area for 24 full years, from 1995 to 2018.This time period limitation is due to the boundary conditions provided by ERA-Interim reanalysis, which was discontinued since August 2019.Wind components in hourly, daily and monthly resolution are provided for the six lowest levels of COSMO model corresponding to heights of 10 m, 35 m, 69 m, 116 m, 178 m and 258 m.
Hourly WS was employed to compute the wind power output at 105 (WS 105 ) and 150 m (WS 150 ).Zonal (u) and meridional (v) wind components from model levels between 35 m and 178 m were used for obtaining WS 105 , through a vertical interpolation between these levels by a cubic polynomial function using a traditional least squares fit [36,45].WS 150 was obtained directly from the COSMO database.The spatial domain corresponds to ocean points around the IP.The time period covers the available full years, i.e. 24 years from 1995 to 2018.

Estimation of offshore wind energy capacity factor
Wind energy CF was obtained from WS at 105 and 150 m hub height following the piece-wise definition of the power curves proposed by [46].The authors provide the parameters for different wind turbines in their appendix C, in particular for Vestas V164-10.0 MW (figure n) and Haliade-X 13 MW (figure a), which were used as representative power curves for 105 and 150 hub height, respectively.105 m was chosen because it is a hub height that is representative for present offshore wind turbines, while 150 m was chosen as a likely future hub height on the basis of different offshore wind farm projects submitted to Spanish Ministry for the Ecological Transition and the Demographic Challenge [47].

Offshore wind potential areas
A total of 15 OWPAs were chosen based on the zoning of the High Potential Areas for the Development of Offshore Wind Energy included in the Spanish MSP [18]: 8 areas in the North Atlantic (NOR) MSP zone, three areas in the Levantine-Balearic (LEBA) MSP zone and 2 areas in the Estrecho and Alborán (ESAL, near Gibraltar Strait) MSP zone.Moreover, two Portuguese areas (POR) were included (near to Viana do Castelo (POR2) and Sines (POR1) cities) based on the publicly available information from the Direção-Geral de Recursos Naturais, Segurança e Serviços Marítimos of the Portuguese government [48] (figure 1).For a balanced representation of all areas, a maximum of five grid points were chosen per area for the calculations.The CF for each of the 15 OWPAs was then obtained as the average of the CF of the corresponding grid points.

Statistical characteristics
A U-shaped Beta distribution is used for representing CF values.Previous works have shown that this distribution is suitable for studying aggregate energy production [49][50][51].The suitability of the parameters of this distribution has been assessed by a Monte-Carlo method for each OWPA, with positive results both for the mean and for the standard deviation in all cases (not shown).This distribution allows inferring the mean and standard deviation of temporal series from its shape parameters ( [52], see section 4.4.4).Using these quantities, three main statistics are computed at annual and seasonal timescales for the period 1995-2018: hourly mean, hourly coefficient of variation (CV = standard deviation/mean) and interannual CV.Seasons are considered as follows: winter, December to February; spring, March to May; summer, June to August; and autumn, September to November.These measures were computed at each OWPA.Moreover, to investigate the pairwise spatial complementarity between OWPAs, a Kendall's τ correlation matrix was obtained through the R package corrplot [53].This correlation coefficient has previously been used in several complementarity assessments of wind resources [54,55] because it is a robust and resistant alternative to the conventional Pearson correlation, as it allows the assessment of the monotonic relationship between two variables when they are not normally distributed.

Optimal aggregation of potential installation areas
An optimal aggregation of Iberian OWPAs that minimizes the overall temporal variability was obtained following [26].This method provides consecutively the combination of 2, 3, 4 and more separated locations with the lowest hourly coefficient of variation (CV) of the aggregated wind power output, in order to dampen variations in wind power generation.In our case, we use 24 years time series of hourly wind CF for the 15 OWPAs, and the method is applied to find the combination of at most seven OWPAs with the minimum hourly variability of the aggregated CF.Both the OWPAs order in the combination and their geographical location are then taken into account in the following analysis.

Main offshore wind power features of the individual areas
Figure 2 represents the main individual characteristics of the 15 Iberian OWPAs at annual and seasonal scales for both hub heights (105 and 150 m).There are large differences between the different OWPAs.As shown in figures 2(a) and (b), most of the areas at the northern coast (NOR2-NOR7) show high annual CF values, in agreement with previous studies [13][14][15].The seasonal CF variations in that areas are large, with a marked winter maximum and with winter-/summer CF ratios between 1.5 and 2 for a 105 m hub height, similar to the winter/summer ratios observed at individual sites in the North Sea for a 100 m hub height [56,57].In contrast, CFs at the western and southern IP coasts (NOR1, POR and ESAL areas) are clearly lower, but with a much smaller seasonality.Remarkably, at POR1 (south of Lisbon) there is even a summer maximum of CF for a 150 m hub height.This is consistent with the results of [58] for southwestern Portugal.LEBA1 area (northeastern IP corner) stands out with a combination of high annual CF and low seasonality.CF values for 150 m hub height are qualitatively similar, but consistently higher than for 105 m.
The hourly scale variability analysis (figures 2(c) and (d)) shows that winter is the most stable season (CV ranging from 0.5 to 0.8) at most OWPAs, with two exceptions: POR1 and ESAL1 where the CV reaches its lowest value respectively in summer and spring.The highest hourly variability is found usually in summer, except for western coast areas (NOR1, POR1 and POR2) where autumn shows the highest hourly CV and the differences between seasons are small.Hourly variability decreases consistently when the hub height rises to 150 m hub height.
The interannual CV (figures 2(e) and (f)) of annual CF shows values of about 0.06 at most OWPAs, with maximum values at the Portuguese coast and a minimum value at LEBA1.These values are rather low, if we compare them with the global interannual variability values shown in [21], which is a sign of the good quality of the offshore wind resource around IP. Winter and summer interannual variability is rather similar at many areas, with the exception of western and northwestern OWPAs.POR1 stands out with much higher interannual CV in winter than in summer.

Pairwise complementarity of the different potential areas
Hereafter, an overview of the pairwise correlation structure of CFs for all OWPAs is shown in figure 3.In general, a similar behavior is observed at 105 (lower triangle) and 150 m (upper triangle of the correlation matrix).
A highly remarkable result is the appearance of several negative correlation values at annual scale (figure 3(a)).In other mid-latitude regions, pairwise correlation values are positive in almost all cases.In the study of [59] about onshore wind power in the United Kingdom, only one slightly negative correlation at 900 km distance is reported among all correlations of 66 sites.A study about offshore wind energy [24] at the eastern USA coast, considering sites as far as 2500 km apart, reports a minimum correlation of zero.[31] who analyze the characteristics of several interconnected production sites in the North Sea over a maximum distance of 1100 km, show that the correlation coefficient never drops to zero or below zero.Even for onshore winds over IP, correlations between 20 wind regions only show one slightly negative correlation value of −0.02 ( [60]).Therefore, clearly negative annual correlation values are very uncommon for a synoptic-scale size region (the distance between opposite Iberian coasts is about 1000 km).
Another noteworthy aspect is the occurrence of negative or zero annual correlation values at short distances: NOR1 and NOR8 show a correlation of −0.1 (105 m) or −0.02 (150 m hub height) at about 400 km distance, while LEBA1 and LEBA3 show a correlation of −0.07 (105 m) and −0.03 (150 m hub height) at less than 300 km distance.For comparison, previous studies [24,31,59] indicate that pairwise correlation is clearly positive at distances of several hundreds of km, and diminishes with distance approaching a positive value of 0.1 only at a large distance (1000 km).
The highest complementarity at annual scale (considering both hub heights) is found between POR1 (western coast) and several areas at the northern coast.The Balearic Islands areas (LEBA2 and LEBA3) show also negative or near zero correlation values with northern and southern coast areas, particularly for 105 m hub height.Therefore, OWPAs with relatively low mean CF (POR1, LEBA2 and LEBA3) show a clear complementarity to OWPAs with high mean CF (northern coast).This is relevant for obtaining a combination of areas that smoothes out temporal variations of wind energy output, as will be analyzed later.At seasonal scale, complementarity in summer is much higher than in winter.Western coast areas are negatively correlated in summer to many northern and southern coast areas.POR1 stands out, with several correlation values between −0.3 and −0.4.NOR1, also at the western coast, shows a high complementarity to northern coast areas, which is remarkable due to the short distance among them.This points to coastal orientation as a relevant feature for complementarity, apart from distance.
What are the reasons for the high spatial variability at short distances that is behind these exceptional complementarity characteristics?The IP is characterized by several high altitude mountain ranges and deep valleys, that cause marked mesoscale modifications of synoptic winds that are linked to fixed orographic features.This results in strong horizontal gradients of WS, as can be seen e.g. for the Tramontane wind at northeastern IP [40] or in the wind channeling effect around Gibraltar Strait [39].Another reason are certain summer flows like the western coast low-level jet, which is related to the Iberian thermal low [61] and to the oceanic upwelling linked to the Iberian-Canary current [62].All these flows are restricted to certain areas around IP, and react differently to the synoptic-scale meteorological situation, which can explain the uncommon correlation values.

Optimal aggregation of potential offshore areas
The spatial complementarity characteristics described in the last section are fundamental for an optimum combination of offshore sites that reduces aggregated wind power output variability, which is obtained following the method described in section 2.5.The characteristics of the combination of up to seven interconnected OWPAs with the minimum CV among all possible configurations of different sites are shown in figure 4 both for 105 and 150 m hub height.Hourly variability decreases first rapidly and then more slowly, but consistently, when the output of more areas is aggregated.This result is in agreement with the findings of [31] for interconnected sites in the North Sea.At the same time, the aggregated CF tends to decrease, with some jumps, with an increasing number of combined areas.For a combination of seven areas, the hourly CV decreases (≈ 40%) much more than the CF mean (≈ 15%) with respect to the first selected site.This behavior varies seasonally, with the hourly CV decreasing the most in winter and spring (≈ 45%) and the CF mean decreasing the most in summer (≈ 24%).Note also that at higher altitude, the resource is larger and more stable.Finally, a combination of seven areas yields winter/ summer CF ratios of 1.35 (105 m) and 1.25 (150 m hub height), which is a very favorable result when compared to the larger seasonality found for aggregated offshore output at other regions: 1.47 for UK [29], above 1.6 for eastern USA [24] and about 1.7 for North Sea [31].The limited seasonality of aggregated offshore IP output represents also a substantial improvement with respect to aggregated onshore wind energy output in Spain, which shows a 1.6 winter/summer ratio (data from the Spanish transmission system operator for the period 2016-2022, www.ree.es/en/datos/todate).
Looking at the combination of smaller number of areas, the aggregated CF reaches its maximum when only LEBA1 and NOR2 are considered, at the same time that the hourly variability diminishes strongly.The optimum combination of four areas includes one for every coastal orientation, highlighting the important role of this geographical aspect for complementarity.
On the other hand, the interannual CV of the annual CF decreases less than the hourly CV with an increasing number of combined areas.The interannual CV of the seasonal CF shows a highly variable behavior as sites are aggregated.Summer output shows the smallest interannual CV values if four or more areas are combined.Summer power production values that are rather stable from year to year are very relevant due to the increasing importance of summer demand in IP.

Conclusions
This paper performs an analysis of the offshore wind resource around the IP with a realistic spatial distribution, focused on the possible installation areas delimited by the Spanish and Portuguese governments.The Iberian power system is semi-closed (with limited interconnections to France and Morocco), with an already high variable renewable energy share and two demand maxima in winter and summer.Offshore wind energy resources are known to be large, but the needed floating wind turbines are more costly than onshore wind and solar technology or fixed offshore turbines [9].If offshore wind energy can offer an added value for the Iberian power system in terms of a stable contribution, with low temporal variability and reduced seasonality, the case for its development would be strengthened.Therefore, the analysis focuses on temporal variability and spatial complementarity characteristics of the different potential sites, and an optimum aggregation of sites that minimizes the temporal variability of the combined production is suggested.The study is performed using COSMO-REA6 high-resolution reanalysis (0.055 • ) at hourly frequency in the available period (1995-2018) at 105 and 150 hub heights (representative of present and expected future offshore wind turbines).
Two major distinct zones are identified in the IP regarding offshore wind resource characteristics.
Northern coast areas show high CF values but large differences between the winter maximum and the summer minimum, while western and southern coast areas are characterized by lower CF values but also by a limited seasonality that matches better the seasonal variations of power demand.The contrast between northern and western coasts shows similarities to the results obtained by [56] in their study comparing different offshore sites over the North Sea.They found that the sites which have the lowest mean power production showed similar hourly variability across all seasons, while the sites with higher mean output showed significantly low variability during the winter.They justify this by the fact that at higher WS sites during the winter, the turbine will frequently operate above rated power, thus reducing variability compared to lower WS sites, where a much larger fraction of the time elapses between the cut-in speed and the rated WS.Negative correlation values between certain pairs of areas, that are particularly marked in summer, reveal a high spatial complementarity of the Iberian offshore wind resource, even at relatively short distances, that stands out in comparison to other midlatitude regions like the North Sea or the eastern USA coast.The effects of a marked orography and summer thermal flows can explain the large mesoscale deviations from the synoptic-scale winds that lead to such correlation values.
The outstanding spatial complementarity characteristics are the basis for plausible combinations of the potential installation areas in a way that smoothes out the temporal variability.A key aspect of such optimum aggregations is the need to include at least one area at each of the four main Iberian coasts with different orientations (eastern, northern, western and southern coasts).In contrast to the requirement of interconnecting wind farms at distances of about 1000 or more km to dampen aggregate variability in other regions [24], distance between sites is here less important.
For the optimal aggregation of seven areas, a strong reduction in hourly variability is obtained, much larger (around 40%) than the reduction in the CF (around 15%) linked to the inclusion of lower CF areas.The reduced CF seasonality for the analyzed combinations is clearly below the values found for the offshore resource in areas like the North Sea or the eastern USA coast and for onshore wind energy in Spain.This implies a better adaptation of the Iberian offshore resource to the seasonal demand profile, that shows two maxima in winter and summer.The optimal aggregation also reduces the interannual variability, approaching values typical for Iberian PV plants, which are characterized by a stable output from year to year [63].The results show that expected advances in offshore floating turbine technology leading to increased hub heights of 150 meters will improve the offshore wind resource both in terms of output amount and of temporal variability.This work underscores the importance for stakeholders to not only consider the amount of offshore wind resource when siting offshore wind farms, but also to incentivize the installation of complementary sites that, in combination, provide a stable contribution to the power system, reducing the need for storage or backup plants.The added value for the power system of including areas with lower CF but favorable variability and complementarity characteristics demands an adequate design of the expected auctions, incorporating non-price criteria as it is beginning to occur in certain European offshore auctions [64].

Figure 1 .
Figure 1.Location of the Iberian offshore wind potential areas (OWPAs) based on the public information available in the Spanish and Portuguese governments' Maritime Spatial Plans.

Figure 2 .
Figure 2. Statistical characteristics of offshore wind potential areas for the period 1995-2018 represented by a radarcharts composition at annual (grey shading) and seasonal (colored lines: winter (DJF, blue), spring (MAM, green), summer (JJA, red) and autumn (SON, orange)) scales at 105 m (left column) and 150 m (right column).The mean is represented in the first row (a)-(b), the hourly coefficient of variation in the second row (c)-(d) and the interannual coefficient of variation in the third row (e)-(f).

Figure 3 .
Figure 3. Kendall's τ correlation matrix of the hourly capacity factor across the offshore wind potential areas considered at 105 m (lower triangle) and 150 m (upper triangle) at annual (a) and seasonal (b)-(e) scales for 1995-2018.For a clearer interpretation, the elements of the matrix showing a different sign between heights are dotted.

Figure 4 .
Figure 4. Statistics of capacity factor (CF) combinations of the seven offshore wind potential areas (OWPAs) at 105 m (left column) and 150 m (right column) hub heights: hourly mean (first row), hourly CV (second row) and interannual CV (third row) at annual (black line) and seasonal scales (DJF; blue; MAM, green; JJA, red and SON, orange) for the period 1995-2018.