A methodology for evaluating the spatial variability of wind energy ...

Renewable Energy 69 (2014) 147e156

Contents lists available at ScienceDirect

Renewable Energy journal homepage: www.elsevier.com/locate/renene

A methodology for evaluating the spatial variability of wind energy resources: Application to assess the potential contribution of wind energy to baseload power F.J. Santos-Alamillos a, b, D. Pozo-Vázquez a, *, J.A. Ruiz-Arias a, c, V. Lara-Fanego a, J. Tovar-Pescador a a b c

Physics Department, University of Jaén, Spain Center for Wind Energy Research (Forwind), University of Oldenburg, Germany National Center for Atmospheric Research (NCAR), Boulder, CO, United States

a r t i c l e i n f o

a b s t r a c t

Article history: Received 19 July 2013 Accepted 3 March 2014 Available online

We propose a method for analyzing the potential contribution of wind energy resources to stable (baseload) power within a region. The method uses principal component analysis (PCA) to analyze spatiotemporal balancing of wind energy resources and then assesses the optimal wind farm location to reduce wind power fluctuations. The ability of different reference wind turbines, alone or interconnected, to provide stable power is ultimately evaluated at selected locations. The method was tested in the southern Iberian Peninsula, including offshore areas. We used hourly wind energy estimates from the WRF mesoscale model at 3-km spatial resolution for the period 2008e2010. First, results show a valuable spatial balancing pattern between the wind energy resources in the northeast study region and Strait of Gibraltar area. The pattern was found to result from the interaction of mesoscale zonal flow with the complex topography of the region. Second, the results indicate that by taking advantage of the spatial balancing pattern, the optimal allocation and interconnection of wind farms across the region, can substantially reduce wind power fluctuations. This optimal allocation can in some cases generate stable power, thereby contributing to baseload power. Ó 2014 Elsevier Ltd. All rights reserved.

Keywords: Wind energy variability PCA Andalusia Firm capacity Balancing

1. Introduction In the next decades, wind energy is expected to play a major role in the replacement of fossil fuels by renewable energy sources [1,2]. Given the fluctuating nature of wind resources and their sensitivity to weather patterns, the integration of major wind yields into the existing energy supply infrastructure will be a challenge [3,4]. One way to reduce this power fluctuation of wind energy is to take advantage of the strong spatial variability of this resource. The idea is that since the spatial correlation of wind speed diminishes with distance, the combined output of numerous widely-spaced wind farms should be smoother than the simple output of an individual wind farm. There are numerous studies in the literature that analyze the spatiotemporal variability of the wind energy resource and its consequence for wind power integration [5e9]. In addition,

* Corresponding author. Physics Department, University of Jaén, Campus las Lagunillas, E23071 Jaén, Andalusia, Spain. E-mail address: [email protected] (D. Pozo-Vázquez). http://dx.doi.org/10.1016/j.renene.2014.03.006 0960-1481/Ó 2014 Elsevier Ltd. All rights reserved.

numerous studies have analyzed the benefits of combining wind and hydropower to provide stable power [10e12], but only a few have addressed if, by interconnecting advantageously-distributed wind farms, it may be possible to guarantee a certain amount of wind power output all the time, eventually transforming wind energy into a reliable supply. There is no clear consensus in the literature for evaluating the contribution of wind toward reducing the conventional capacity of power systems. In general, two parameters for describing this contribution have been used: capacity credit and firm capacity. Capacity credit is a measure of the ability of a wind farm or a solar plant to contribute to the peak demands of a power system, maintaining the same security of supply as conventional power plants [13]. This parameter is usually expressed in %. Obtaining the capacity credit is no easy task, since full modeling of the power system is necessary. Several studies have analyzed the wind energy capacity credit, in individual countries and for all Europe [13e15]. The main conclusion was that wind power has capacity credit but is highly dependent on the supporting electric system, wind load factor and penetration level. On average, for a penetration around 20%, the wind energy capacity credit could

148

F.J. Santos-Alamillos et al. / Renewable Energy 69 (2014) 147e156

approach 20% of installed power [13]. The firm capacity is defined as the fraction of installed wind capacity that is available with the same probability as that of a thermal plant [14]. With this definition, the firm capacity of wind energy can be used to analyze the potential contribution of wind to provide baseload power, that is defined as the minimum amount of power that must be available to the final customers in all moment. Unlike capacity credit, firm capacity can be assessed relatively easily from pure wind data [6]. In contrast to the capacity credit, the ability of wind energy to provide firm capacity in central and northern Europe has been shown to be limited [16e18]. Underpinning these results are the relatively homogeneous weather conditions and limited topographic features across this area. Studies in other regions of Europe and the world are more encouraging. For instance, Archer and Jacobson [19] showed the existence of wind power firm capacity in the Midwestern United States, Cassola et al. [20] in the isle of Corsica (France) and Kempton et al. [21] in offshore areas along the east coast of the United States. In this study, a method for analyzing the potential contribution of wind energy to stable power in the Andalusian region of southern Spain is proposed and evaluated. The method is based on Principal Component Analysis (PCA) to obtain spatiotemporal variability patterns of wind energy in the region. Based on analysis of these patterns, the best wind farm locations to reduce wind power fluctuations through their interconnection are then assessed, to optimally reduce the wind power fluctuations. Data for the analysis consist of 3-km spatial resolution gridded wind energy time series across the Andalusian region, including offshore areas, derived from mesoscale Numerical Weather Prediction (NWP) model simulations. Finally, the firm capacity provided by different combinations of offshore and onshore wind farms is estimated. Although relatively small, the study region is characterized by varied topographic, geographic and weather conditions. This fact anticipates the existence of considerable wind speed spatial variability and, ultimately, spatial balancing of wind energy resources. This paper is organized as follows: Section 2 deals with the study area, data and method. Section 3 describes the results. Finally, in Section 4, a summary and conclusions are presented. 2. Methods and study area 2.1. Study area

topographic point of view, the region may be split into two different parts. The western part is a nearly homogeneous flat area, open to the Atlantic Ocean. The eastern part has very complex topography and is isolated from the Atlantic influence by the Sierra Nevada and Cazorla mountain ranges (Fig. 1). Atmospheric circulation across the study region is dominated by a semi-permanent subtropical high-pressure centre over the Azores islands. The position and intensity of this centre changes through the year [22]. The study region is bounded on the south by the Atlantic (western part) and Mediterranean (eastern part), with about 900 km of coastline. One of the most important features regarding wind energy is the existence of strong, semi-permanent winds near the coast of the Strait of Gibraltar. Low-speed winds are channeled and accelerated through this 11 km-wide topographic feature. Despite the relatively small size of the region, there is a remarkable surface wind speed spatial variability across it, caused by interaction of mesoscale circulations with the topographic features [23]. The geographic, topographic and climatological characteristics of the study are fully discussed in Refs. [24,25]. The study region has areas of great significance for offshore wind energy, especially near the Strait of Gibraltar [26,27]. Although development of offshore wind farms in these areas is in its early stages, the offshore wind energy resource is addressed herein. 2.2. Data Region-wide wind energy resources at 80 m above ground level (m.a.g.l.) were obtained from integrations with the WRF (Weather Research and Forecasting) NWP (Numerical Weather Prediction) model [28]. Simulations were conducted for three years (2008e 2010). The model configuration included three nested domains, with 27, 9 and 3 km spatial resolutions (Fig. 2). We used 36 vertical levels, 8 levels within the lowest 1000 m.a.g.l. The five lowest vertical levels, which are important for wind energy applications, were approximately located at 4, 18, 43, 64 and 80 m.a.g.l. Hourly data from this last level in the third nested domain, with 3-km spatial resolution and centered on the study region, were used in the study. The analysis included offshore areas as much as 20 km from the coast, a distance considered feasible for offshore wind energy projects [29]. The physical configuration of the model was selected based on results of an evaluation study of WRF model performance in the study region [25], and we used initial and

The study region (Fig. 1) of Andalusia is in the southern Iberian Peninsula (IP), and covers an area of 87.000 km2. From a

Fig. 1. Location (inset, bottom right) and principal topographic and geographic characteristics of study region. Gray scale colors indicate elevation based on the 3-km spatial resolution Digital Elevation Map used in the model simulation to derive study data. Scale at right indicates elevation in meters above sea level.

Fig. 2. Spatial configuration of domains for numerical simulation: three nested domains with 27, 9 and 3 km horizontal resolutions. Data from the inner domain (3 km) were analyzed.


boundary conditions from the Climate Forecast System Reanalysis [30] with spatial resolution 0.5 and updates every 6 h. Following Santos-Alamillos et al. [25], analysis nudging toward the reanalysis was used for the first domain and for the last 15 vertical levels. Additionally, Sea Surface Temperature (SST) analyses were interpolated to ocean grid points in each model domain, using the NCEP (National Center for Environmental Prediction) version 2.0 global SST dataset [31]. This dataset was updated daily, based on a 0.05 0.05 grid. Integrations were carried out for eight-day periods over the three years, taking the first of these eightday periods as spin-up [32]. With this setup, onshore wind speed standard deviation error at 40 m.a.g.l. in the study region was about 2e 2.8 ms1, and the bias was from 0.4 to 0.3 ms1 [25]. 2.3. Wind energy spatial variability analysis PCA [33,34] is a widely-used method to analyze spatiotemporal variability of an environmental variable. The technique allows reduction of initial dataset size to a few representative variables, called principal components (PCs) and often referred to as modes of variability, which are obtained as linear combinations of the initial variables. These combinations are obtained such that the new variables account for the maximum fraction of variance contained in the original dataset. Coefficients of the linear transformation, known as loading factors, range from 1 to 1 and represent correlations between the original field variables and corresponding PC at each point. A map of the loading factors of a particular PC can be used to analyze spatial balancing of the variable studied. Areas where loading factors have opposite sign (dipolar behavior) are indicative of balancing, and it is expected that wind energy lacking positive/negative loads can be compensated by an excess of negative/positive loads. Obviously, the relevance of this balancing is related to the magnitude of the loads and the compensating effect is more notable with stronger dipolarity. PCA has been used for spatial decomposition of wind power generation in Germany [35]. In this work, PCA was used to analyze spatial variability of wind energy resources. Given that the focus of the present study is wind energy, PCA was conducted solely for areas with significant wind energy resources, rather than for the entire study region. To this end, hourly capacity factor time series for reference wind turbines were estimated at each grid cell of the third domain in Fig. 2. The WRF provides horizontal wind components at different pressure levels. Following Santos-Alamillos et al. [25], horizontal wind components corresponding to the first eight pressure levels (all below 1000 m above ground level) were interpolated to 80 m using cubic splines. For onshore locations, the selected wind turbine was the VESTAS V90-2.0, with rated nominal power 2 MW and hub height 80 m.a.g.l. For offshore locations, the selected turbine was the VESTAS V90-3.0, with nominal power 3 MW and the same hub

149

height. Fig. 3a shows power curves for both turbines. To obtain the hourly capacity factors, the wind power time series were normalized to the respective nominal power. Finally, only cells with yearly mean capacity factor greater than 0.25, represented in Fig. 3b (masked cells), were considered in the PCA. Next, PCA was conducted for daily integrated wind energy resources computed at the masked cells:

Edaily wind ¼

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi3 24 X 1 r ðui Þ2 þ ðvi Þ2 1h 2 i¼1

(1)

where r is air density and ui and vi are hourly zonal and meridional wind components respectively at 80 m.a.g.l. Daily, rather than hourly, time resolution was chosen to focus on mesoscale circulation balancing features. Wind energy rather than wind power was used for the PCA to keep results more general and independent of specific wind turbine power curves. Only the power fluctuation analysis was done with hourly capacity factors. The PCA was applied independently for each season and the annual period, and the number of significant modes was selected based on the rule of North et al. [36]. Not all the statistically significant modes were further analyzed, but only those relevant to the wind energy balancing analysis. Finally, to understand the nature of the modes resulting from the PCA, a brief description of associated mesoscale atmospheric circulations is given. To this end, we produced composite maps of daily mean sea level pressure (SLP) and horizontal surface wind speed (10 m.a.g.l.) anomalies associated with tails (values above percentile 90% and below percentile 10%) of PC series distributions. The composites were derived from the ERA-Interim reanalysis [37] dataset, covering a 4000 4000 km region centered on the IP. For a more detailed view, composite maps of daily mean horizontal wind component anomalies at 80 m.a.g.l., derived from WRF 3-km integrations, were also analyzed. 2.4. Wind power balancing analysis The relevant modes of variability derived from the PCA were further examined to assess potential reduction of the wind power fluctuations. To this end, four locations (grid points) were selected for each mode of variability: two onshore (those with the highest positive and negative loading factors resulting from the PCA, called Onþ and On hereafter), and two offshore (also with the highest positive and negative loading factors, called Offþ and Off hereafter). At these four locations, hourly capacity factors were computed for corresponding reference wind turbines (Fig. 3b), based on the 80 m.a.g.l. WRF 3-km wind speed estimates. These four capacity-factor time series summarize the spatial variability of wind energy resources accounted for by the corresponding mode of variability. In a further step, we analyzed the gain of wind power

Fig. 3. a) Power curves of Vestas-V90 2-MW (dashed line) and Vestas-V90 3-MW wind turbines. The first turbine was used to assess onshore locations and the second for offshore locations. b) Cells (gray shade) corresponding to the third simulation domain with yearly mean capacity factor greater than 0.25. Only these cells were considered in the PCA.

150


reliability attained by different combinations (connections) between the selected onshore and offshore sites. Three types of connections were evaluated: Onshore (denoted as On, from the connection of Onþ and On locations); Offshore (denoted as Off, from the connection of Offþ and Off); and connection of four sites (denoted as All). To characterize power output fluctuations for both standalone and connected wind turbines, we used the mean and standard deviation of the hourly capacity factor throughout the corresponding study period. The firm capacity, that is, the fraction of installed wind capacity that is online with the same probability as a conventional thermal power plant, was also evaluated. This probability can be highly variable, depending on the technology (nuclear, coal, gas). We used the reference value 87.5% in Archer and Jacobson [19], which is the average availability factor reported for thermal power plants in Spain [38,39]. Again, this firm capacity was computed for both individual and connected wind turbines, and expressed as a percentage of installed nominal power. It has been observed that simple spatial aggregation of wind farms may substantially reduce total power fluctuations [19,21]. To evaluate the performance of our proposed methodology, we conducted 100 random experiments for each study period. In each experiment, four locations (2 onshore and 2 offshore) were randomly selected in the masked area of the study region (Fig. 3b). Then, the mean capacity factor, standard deviation and firm capacity for interconnection of the four locations in these experiments were computed. These mean values, called hereafter Ref, can be used to test the significance of (All) wind farm connection corresponding results. Note that use of four connections for testing the performance of the proposed method is considerably more stringent than the use of only two (as used here in the Off and On connection cases). This is mainly because as noted in Archer and Jacobson [19], the power fluctuations decrease considerably just by increasing the number of aggregated wind farms.

3. Results Table 1 summarizes the PCA results for each study period, namely, the four seasons and annual period. In general, the first mode accounted for a substantial part of the total variance, from 35% in summer to 61% in autumn. This first mode accounts for a spatially coherent variability across the study region, that is, the loading factors have the same sign across the region (positive or negative). This mode can be considered as the background coherent temporal mode of variability in the entire region, to which the temporal variability represented by the other modes should be added. In contrast, the second and subsequent modes generally show areas of positive and negative loadings (dipolar pattern). Consequently, these modes may be relevant for our balancing study. In the following sections, we have only analyzed the second mode for all study periods except autumn, for which both the

Table 1 Summary of principal component analysis (PCA) results for wind energy resources in study region. The table shows, for seasonal and annual analysis, the explained variance associated with the five leading modes. Last column shows variance accounted for by these five modes together. Only modes highlighted in bold were considered in wind energy spatial balancing analysis. Study period

Winter Spring Summer Autumn Annual

Mode 1

Mode 2

Mode 3

Mode 4

Mode 5

Total

(%)

(%)

(%)

(%)

(%)

(%)

58 59 35 61 59

15 16 21 11 12

8 8 12 7 9

3 4 6 5 4

3 4 4 3 3

87 91 78 87 87

second and third modes were analyzed. This selection was based on the power fluctuation reduction obtained. Only the aforementioned modes showed a power fluctuation reduction considerably greater for the (All) connection than for the (Ref) connection. 3.1. Winter The second mode of the winter PCA accounts for 15% of total variance (Table 1), and corresponding loading factors and PC are displayed in Fig. 4a and f, respectively. A dipolar mode is clearly evident. Positive loads (0.3e0.8) are observed in the east, including an offshore area. Negative loads (0.3 to 0.8) are found in the southwest, namely the Strait of Gibraltar area (including adjacent offshore areas) and Grazalema mountains. This dipolar pattern indicates the existence of spatial balancing between the areas of high positive and high negative loads, which merits further analysis. Fig. 4bee indicate that this balancing is associated with a meridional gradient of SLP anomalies. In particular, high positive anomalies in the PC series (Fig. 4f) are associated with positive SLP anomalies in the southwestern IP and negative ones in the northeastern IP (Fig. 4d). This distribution of SLP anomalies produces notable westerly wind anomalies across the entire region, which are especially intense in its eastern part (Fig. 4b). This enhancement of wind speed anomalies appears associated with the unique topographic features of this area (Fig. 1). First, westerly winds are channeled through the Guadalquivir River basin, which is open to the Atlantic Ocean, creating strong surface wind speeds in the upper valley. Second, these winds are further enhanced at the Cazorla mountain range, which has steep slopes and a perpendicular aspect to these westerly winds. Valleys in this range, channel and accelerate these winds. All these effects explain the positive loading factors seen in the eastern study region in Fig. 4a. Although wind speeds under the mesoscale atmospheric circulation conditions represented in Fig. 4d are also relatively strong in the Strait of Gibraltar area (Fig. 4b), anomalies are negative (negative loading factors in Fig. 4b). Negative anomalies in the PC series (Fig. 4f) are associated with counterpart SLP anomalies (Fig. 4e). These SLP anomalies generate region-wide easterly winds, which are significantly intensified by the channeling effect of the Strait of Gibraltar (Fig. 4c) and by hill/channeling effects of nearby mountain ranges (Grazalema). At the same time, relatively weak easterly winds are observed in the eastern part of the region. In summary, the spatial balancing between wind energy resources in the eastern study region and Strait of Gibraltar area, highlighted by this PCA mode, appears to result from the interaction of mesoscale zonal flow and topography in the study region. On one hand, westerly winds are channeled by the Guadalquivir River basin, giving rise to strong anomalies in the upper valley. On the other hand, easterly winds are greatly intensified in the Strait of Gibraltar area, the only open area this flow encounters along its trajectory. The next step is to evaluate the potential impact of this spatial balancing on the stability of wind power production in the region. Fig. 4g presents the mean and standard deviation of hourly capacity factor along with the firm capacity, for wind turbines sited at each of the four selected locations in Fig. 4a. Values corresponding to various combinations (interconnections) of these turbines and to the reference case (described in Section 2.4 are also represented. The mean and the standard deviation of the capacity factor reach values greater than 0.4 for both onshore locations (Onþ and On). In contrast, values are substantially smaller at the Offþ location and substantially higher at the Off location. This location (Gibraltar strait offshore area) shows a capacity factor approaching 0.55. A clear reduction of standard deviation is obtained for the set of connections between the various locations. For instance, for the connection of onshore sites (On), the standard deviation declines


151

Fig. 4. Second mode of winter principal component analysis (PCA). a) Loading factors map. Only statistically significant values (above 0.3 or below 0.3) are displayed. b) Composite of anomalies of daily-mean horizontal wind components at 80 m.a.g.l. from the 3-km spatial resolution simulation, corresponding to values above 90% percentile of PC series distribution. Data are taken from inner domain of the simulation (Fig. 2). c) As in b), but for values below the 10% percentile of PC series distribution. d) Composite of anomalies of daily mean sea level pressure (in HPa) and horizontal wind components at 10 m.a.g.l. from the ERA-Interim reanalysis, corresponding to values above 90% percentile of PC series distribution. In all figures, wind speeds are scaled according to the reference at top or bottom, which indicates maximum wind speed anomaly in ms1. e) as in d), but for values below the 10% percentile of PC series distribution. f) Principal component time series. Values are standardized anomalies. g) Mean (first column) and standard deviation (second column) of hourly capacity factor and firm capacity (third column) for: four selected sites in Fig. 4a (Onþ, On, Offþ and Off); various connections between them (Onshore connection (On), Offshore connection (Off)) and connection between the four sites (All)); and the connection of four randomly-located wind farms across the study region (Ref).

from more than 0.4 to about 0.25. Similar reduction is attained by connecting the offshore sites (Off). Additionally, when the four locations are connected (All), the standard deviation is reduced to less than 0.25. Moreover, connections give rise to firm capacity: relatively modest for connection of the offshore (Off) locations (0.05, 5%), noticeable for the onshore (On) connection (12%), and remarkable when all four locations (All) are connected (15%). Finally, the (All) connection provides substantially better results in terms of power fluctuation reduction than the (Ref) one. Notably, a lower standard deviation of the hourly capacity factor value (0.25 versus 0.31) and substantially higher firm capacity (15% against 6%) are obtained. 3.2. Spring The second mode of the spring PCA accounted for 16% of the variance (Table 1). Corresponding loading factors and PC series are displayed in Fig. 5a and f, respectively. The loading factor map and the balancing pattern resemble that obtained for the second winter mode. Notably, Fig. 5a shows a dipolar pattern, with positive loads (0.3e0.6) in the northeast, west and small areas in the region center. There are negative loads (0.3 to 1) in the Strait of Gibraltar, Grazalema mountains and south Mediterranean coast. Atmospheric circulation conditions associated with this balancing mode (Fig. 5bee) are qualitatively similar to those associated with the second winter mode (Fig. 4bee), but the anomaly patterns are

displaced southward. In summary, the spatial balancing pattern observed in the second spring PCA mode is similar to that associated with second winter PCA mode one, since they have similar associated mesoscale atmospheric circulation patterns. Similar to Fig. 4g, Fig. 5g shows results of the power fluctuations analysis. As in the winter case, the wind farm connection reduces the standard deviation and provides firm capacity. As observed in Fig. 5g, the connection of offshore wind farms (Off) results in considerable reduction of the standard deviation and a notable firm capacity (8%). Results of connecting all the wind farms highlighted in Fig. 5a (All) are very similar to those obtained for just the offshore connection (Off). Finally, the (All) connection gives substantially better power fluctuation reduction than the (Ref) one. Notably, a lower standard deviation of hourly capacity factor value (0.21 against 0.26) and a higher firm capacity (8% against 5%) are achieved. 3.3. Summer The second mode of the summer PCA accounted for 21% of the variance (Table 1). Corresponding loading factors and PC series are portrayed in Fig. 6a and f, respectively. The loading factor map and balancing pattern share some properties with those of the second winter and spring modes, but loading factors values are higher. Positive loads (0.6e0.8) are observed in the northeast, west, certain small areas in the region center and Mediterranean coast close to

152


Fig. 5. As in Fig. 4, but for spring analysis.

Fig. 6. As in Fig. 4, but for summer analysis.


the Gibraltar strait. Strong negative loads (0.6 to 0.8) are revealed in the Strait of Gibraltar area and Grazalema mountains. Atmospheric circulation conditions associated with this balancing mode (Fig. 6bee) are qualitatively similar to those associated with the second spring mode (Fig. 5bee), but the anomaly patterns are displaced northward. This displacement is associated with seasonal migration of weather patterns in this area [22]. Strong positive anomalies in the PC series (Fig. 6f) are associated with a centre of low pressures over the British Islands, which generates northwesterly wind anomalies across the region (Fig. 6b and d). These winds are enhanced by topographic features of the Cazorla Mountains in the east, the western study region, and Mediterranean area east of Strait of Gibraltar. This leads to the positive loadings in Fig. 6a. Under such a mesoscale atmospheric circulation, wind speed is also relatively strong in the strait area but anomalies are negative, causing the negative loads shown by Fig. 6a in this area. Negative anomalies in the PC series (Fig. 6f) are associated with counterpart SLP anomalies (Fig. 6e). A positive anomaly centre is observed south of the British Isles, which generates mesoscale easterly winds in the study region. These winds are significantly intensified by the channeling effect of the Strait of Gibraltar and by hill/channeling effects of nearby mountain ranges (Grazalema) (Fig. 6c). At the same time, relatively weak easterly winds are found in the northeastern and western parts of the region. Fig. 6g shows results of the power fluctuation analysis. As in the previous cases, wind farm connection reduces the standard deviation and provides firm capacity. For instance, the onshore (On) and offshore (Off) connections reduce the standard deviation to around 0.22 and produces firm capacity (6% and 5%, respectively). The (All) connection shows slightly better results for reduction of the power fluctuations than the (Ref) one. In fact, a lower standard deviation

153

of the hourly capacity factor value (0.22 against 0.24) and a higher firm capacity (6% against 4%) are obtained. 3.4. Autumn Unlike the previous cases, the autumn PCA revealed the existence of two modes relevant to the balancing study. The second mode of the Autumn PCA accounted for 11% of the total variance (Table 1). The corresponding loading factor map and PC series are represented in Fig. 7a and f, respectively. The loading factor pattern (Fig. 7a) resembles those obtained in the prior three analyses. In particular, the map shows balancing between wind energy resources in the northeastern study area (positive loads from 0.3 to 0.8) and Strait of Gibraltar area (negative loads from 0.3 to 0.8). Nevertheless, the loading factor map depicts unique features, such as an area with positive loads on the Mediterranean coast and the fact that the negative loads extend through the Gulf of Cádiz (unlike the prior cases). The associated atmospheric circulation patterns are significantly different to those of the previous analyses. Notably, high positive anomalies in the PC series (Fig. 7f) are associated with a center of positive SLP anomalies west of the IP in the central Atlantic and a center of negative SLP anomalies east of the IP, south of France (Fig. 7d). This distribution of SLP anomalies causes strong northwesterly winds across the study region, which are significantly modified by the topographic features (Fig. 7b). The northwesterly anomalies are strongly enhanced in the Cazorla mountain area northeast of the study region, leading to the positive loads in this area shown by Fig. 7a. Negative anomalies in the PC series (Fig. 7f) are associated with counterpart SLP anomalies (Fig. 7e). The center of negative SLP anomalies in the central Atlantic generates southwesterly winds across the study region. These winds are

Fig. 7. As in Fig. 4, but for autumn analysis.

154


particularly strong in the Gulf of Cádiz area (Fig. 7c). At the same time, relatively weak easterly winds are observed in the northeastern part of the region under these atmospheric conditions. Fig. 7g shows results of the power fluctuation analysis. Both the onshore (On) and offshore (Off) connections considerably reduce the standard deviation, from about 0.4 to 0.3, and provide firm capacity (around 3%). The (All) connection gives better results than the (Ref) one, i.e., a lower standard deviation of hourly capacity factor (0.27 against 0.3) and greater firm capacity (5% against 3%). The third mode of the Autumn PCA accounted for 7% of total variance. The corresponding loading factor map and PC series are represented in Fig. 8a and f, respectively. The loading factor pattern shows balancing between the (mainly) offshore areas south of the region and some areas in its interior. Negative loads (0.3 to 0.6) are found west of the Strait of Gibraltar area, Grazalema mountains and the Mediterranean coastal area south of the study region. On the other hand, positive loads (0.3e0.6) are at the far west of the study region, in interior areas. Atmospheric circulation conditions associated with this balancing mode (Fig. 8bee) closely resemble those associated with the second spring mode (Fig. 5bee). Therefore, this balancing pattern can be described from a meteorological standpoint in a similar way as the pattern in Fig. 5a. Fig. 8g depicts results of the power fluctuation analysis. A marked power fluctuation reduction is obtained for the onshore (On) connection; standard deviation declines from about 0.42 to 0.28, and firm capacity reaches 9%. The offshore (Off) connection shows a similar reduction of standard deviation and a 6% firm capacity value. The (All) connection delivers substantially better results than the (Ref) one, i.e., a lower standard deviation of hourly capacity factor value (0.25 against 0.3) and a higher firm capacity (11% against 4%).

3.5. Annual Only one mode of the PCA for the annual data (the second) was relevant for the balancing study. This mode accounted for 12% of total variance (Table 1). The loading factor map (Fig. 9a) clearly resembles the dipolar pattern obtained in the spring, summer and, especially, winter analysis. Similarly, the associated atmospheric pattern (Fig. 9bee) closely resembles those associated with the winter mode (Fig. 4bee). Therefore, this balancing pattern can be described from a meteorological standpoint in a similar way as the pattern in Fig. 4a. Fig. 9g shows results of the power fluctuation annual analysis. A marked power fluctuation reduction is attained for both the onshore (On) and offshore (Off) connection. Remarkably, the (On) connection reduces the standard deviation from 0.4 to 0.27, and provides a 5% firm capacity. The (Off) connection shows a similar reduction of standard deviation, with firm capacity of 3%. The (All) connection furnishes substantially better results than the (Ref) one. A lower standard deviation of hourly capacity factor value (0.24 against 0.28) and a higher firm capacity (7% against 4%) are obtained. 4. Summary and conclusions In this work, we proposed and evaluated a method for analyzing the potential contribution of wind energy resources to stable (baseload) power in a region. The method consists of two steps. In the first, spatial variability of the wind energy resource is analyzed based on PCA. This method facilitates recognition of spatial balancing of this resource. In the second step, PCA results are used to assess optimal locations (onshore and offshore) of wind farms in the region for provision of more stable wind power.

Fig. 8. As in Fig. 4, but for third mode of autumn analysis.


155

Fig. 9. As in Fig. 4, but for annual analysis.

The method was tested in the Andalusian region of Spain, in the southern Iberian Peninsula. First, wind energy resources in the study region were estimated based on the WRF mesoscale model. In particular, wind speed at 80 m above ground level, with 3-km spatial resolution and 1-h temporal resolution, were estimated throughout 2008e2010 across the entire region. This region included offshore areas as far as 20 km from the coast. Second, based on these wind estimates, PCA was conducted and various relevant balancing patterns were identified and described in terms of their associated mesoscale atmospheric circulations. The final step was to evaluate potential usefulness of the wind energy spatial balancing analysis results for the stability of wind power production in the region. To this end, as PCA results, four optimal locations (two onshore and two offshore) were selected for each balancing pattern, and power production of reference (onshore and offshore) wind turbines at these locations was computed. The ability of these wind turbines to provide stable power was then assessed. To this end, mean and standard deviations of hourly capacity factor and the firm capacity were computed for both standalone and connected wind turbines. First, the results showed the existence of a valuable spatial balancing pattern between the wind energy resources in the northeastern study region and Strait of Gibraltar area. With some differences, this pattern was found in the annual and all seasonal analyses, being stronger in winter. Analysis of atmospheric circulation conditions showed this balancing to result from the interaction of mesoscale zonal flow and the topography of the study region. On one hand, westerly winds are channeled by the Guadalquivir River basin, giving rise to strong anomalies in the upper valley (northeast of the study region). On the other hand, easterly winds are greatly intensified in the Strait of Gibraltar area, the only open area this flow encounters along its trajectory. In a recent work, we evaluated balancing

between the solar and wind energy resources in southern Spain [24]. We obtained valuable balancing patterns, resembling in many cases those presented here. Interestingly, the wind-solar balancing patterns in that work appear to be dominated by the wind energy spatial balancing presented herein. Second, our results showed that wind power fluctuations in the study region can be substantially reduced by siting and interconnecting wind farms that takes advantage of the balancing pattern derived from the PCA analysis. In particular, the standard deviation of the capacity factor decreased substantially upon connecting the wind farms. The reduction was dependent on season and the sites being interconnected, but an overall decrease of about one third resulted with interconnection of both offshore and onshore sites. This reduction has important consequences for wind energy grid integration. First, other generators within the power system can take advantage of this smoother fluctuation to compensate deviations, allowing time to ramp up or down. Second, since standard deviation is a measure of the reserves necessary for wind energy grid integration, interconnection appears to significantly diminish these necessities. In addition, it was found in all analyzed periods that the interconnection of wind farms creates firm capacity. Values were highly dependent on the season and interconnected sites, but ranged from 15% during winter to 6% during summer. The annual value was 7%. Finally, in the annual and all seasonal analyses, the power fluctuation reduction attained following our proposed methodology was substantially greater than those obtained by interconnecting randomly-located wind farms across the study region. Our proposed procedure can be easily applied to areas of arbitrary size. The method facilitates planning of the locations of future wind farms within a region, so that total electricity production will fluctuate less. This may result in greater use of wind energy as reliable electric power.

156


Acknowledgments The Consejería Economía, Innovación, Ciencia y Empleo, Junta de Andalucia (Spain) and the Ministerio de Economía y Competitividad (MINECO, Spain) financed this research through a postdoctoral scholarship associated with the Project P07-RNM02872. This work has been also supported by the Solar Radiation and Atmosphere Modeling Group (TEP-220 Junta de Andalucía) at the University of Jaén (Spain), and by the energy and meteorology group at the Center for Wind Energy Research (Forwind), University of Oldenburg, Germany. References [1] Mathiesen BV, Lund H, Karlsson K. 100% Renewable energy systems, climate mitigation and economic growth. Appl Energy 2011;88(2):488e501. [2] Jacobson MZ, Delucchi MA. Providing all global energy with wind, water, and solar power, Part I: technologies, energy resources, quantities and areas of infrastructure, and materials. Energy Policy 2011;39(3):1154e69. [3] Lund H. Large-scale integration of wind power into different energy systems. Energy 2005;30(13):2402e12. [4] Lund H. Large-scale integration of optimal combinations of PV, wind and wave power into the electricity supply. Renew Energy 2006;31(4):503e15. [5] Kahn E. The reliability of distributed wind generators. Electr Power Syst Res 1979;2(1):1e14. [6] Landberg L. The availability and variability of the European wind resource. Int J Sol Energy 1997;18(4):313e20. [7] Holttinen H. Hourly wind power variations in the Nordic countries. Wind Energy 2005;8(2):173e95. [8] Heide D, Von Bremen L, Greiner M, Hoffmann C, Speckmann M, Bofinger S. Seasonal optimal mix of wind and solar power in a future, highly renewable Europe. Renew Energy 2010;35(11):2483e9. [9] Roques F, Hiroux C, Saguan M. Optimal wind power deployment in Europeda portfolio approach. Energy Policy 2010;38(7):3245e56. [10] Jaramillo O, Borja M, Huacuz J. Using hydropower to complement wind energy: a hybrid system to provide firm power. Renew Energy 2004;29(11): 1887e909. [11] Denault M, Dupuis D, Couture-Cardinal S. Complementarity of hydro and wind power: improving the risk profile of energy inflows. Energy Policy 2009;37(12):5376e84. [12] Kapsali M, Kaldellis J. Combining hydro and variable wind power generation by means of pumped-storage under economically viable terms. Appl Energy 2010;87(11):3475e85. [13] Milborrow D. Wind power on the grid. In: Renewable electricity and the grid: the challenge of variability; 2007. pp. 31e45. [14] Giebel G. Wind power has a capacity credit a catalogue of 50þ supporting studies; 2005. [15] Giebel G. A variance analysis of the capacity displaced by wind energy in Europe. Wind Energy 2007;10(1):69e79. [16] Oswald J, Raine M, Ashraf-Ball H. Will British weather provide reliable electricity? Energy Policy 2008;36(8):3212e25. [17] Østergaard PA. Geographic aggregation and wind power output variance in Denmark. Energy 2008;33(9):1453e60.

[18] Weigt H. Germany’s wind energy: the potential for fossil capacity replacement and cost saving. Appl Energy 2009;86(10):1857e63. [19] Archer CL, Jacobson MZ. Supplying baseload power and reducing transmission requirements by interconnecting wind farms. J Appl Meteorol Climatol 2007;46(11):1701e17. [20] Cassola F, Burlando M, Antonelli M, Ratto CF. Optimization of the regional spatial distribution of wind power plants to minimize the variability of wind energy input into power supply systems. J Appl Meteorol Climatol 2008;47(12):3099e116. [21] Kempton W, Pimenta FM, Veron DE, Colle BA. Electric power from offshore wind via synoptic-scale interconnection. Proc Natl Acad Sci 2010;107(16): 7240e5. [22] Trigo RM, Osborn TJ, Corte-Real JM. The North Atlantic oscillation influence on Europe: climate impacts and associated physical mechanisms. Clim Res 2002;20(1):9e17. [23] Andalusian Energy Agency (AAE). Wind Energy resources in Andalusia (Spain). Tech. Rep.; 2009 [24] Santos-Alamillos F, Pozo-Vázquez D, Ruiz-Arias J, Lara-Fanego V, TovarPescador J. Analysis of the spatio-temporal balancing between wind and solar energy resources in the southern Iberian Peninsula. J Appl Meteorol Climatol 2012;51:2005e21. [25] Santos-Alamillos F, Pozo-Vázquez D, Ruiz-Arias J, Lara-Fanego V, TovarPescador J. Analysis of WRF model wind estimate sensitivity to physics parameterization choice and terrain representation in Andalusia (Southern Spain). J Appl Meteorol Climatol 2013;52:1592e609. [26] Sempreviva AM, Barthelmie RJ, Pryor S. Review of methodologies for offshore wind resource assessment in European seas. Surv Geophys 2008;29(6):471e97. [27] Todt O, González M, Estévez B. Conflict in the Sea of Trafalgar: offshore wind energy and its context. Wind Energy 2011;14(5):699e706. [28] Skamarock WC, et al. A description of the Advanced Research WRF version 3. NCAR Tech Note NCAR/TN-47511STR; 2008. [29] Fichaux N, Wilkes J. Oceans of opportunity: Harnessing Europe’s largest domestic energy resource. EWEA; 2009. [30] Saha S, Moorthi S, Pan HL, Wu X, Wang J, Nadiga S, et al. The NCEP climate forecast system reanalysis. Bull Am Meteorol Soc 2010;91(8):1015e57. [31] Reynolds RW, Rayner NA, Smith TM, Stokes DC, Wang W. An improved in situ and satellite SST analysis for climate. J Clim 2002;15(13):1609e25. [32] Hahmann AN, Rostkier-Edelstein D, Warner TT, Vandenberghe F, Liu Y, Babarsky R, et al. A reanalysis system for the generation of mesoscale climatographies. J Appl Meteorol Climatol 2010;49(5):954e72. [33] Preisendorfer RW, Mobley CD. Principal component analysis in meteorology and oceanography, vol. 17. Amsterdam: Elsevier; 1988. [34] Wilks D. Statistical methods in the atmospheric sciences, vol. 100. Academic Press; 2011. [35] Von Bremen L, Saleck N, Heinemann D. Enhanced regional forecasting considering single wind farm distribution for upscaling. In: Journal Phys Conference Series, vol. 75. IOP Publishing; 2007. p. 012040. [36] North GR, Bell TL, Cahalan RF, Moeng FJ. Sampling errors in the estimation of empirical orthogonal functions. Mon Weather Rev 1982;110(7):699e706. [37] Uppala S, Dee D, Kobayashi S, Berrisford P, Simmons A. Towards a climate data assimilation system: status update of ERA-Interim. ECMWF Newsletter 2008;115(115):12e8. [38] Gómez A, Zubizarreta J, Dopazo C, Fueyo N. Spanish energy roadmap to 2020: socioeconomic implications of renewable targets. Energy 2011;36(4): 1973e85. [39] Red Elétrica Española (REE). Annual Report 2010. Tech. Rep.; 2010