Data Collection

Table 2 provides a description of sampling sites and data collection methods used in this study. Weather monitoring stations that are located in continental and coastal environments belong to the Estación de Monitoreo Ambiental Costero (EMAC) network, which was implemented by researchers from IADO-CONICET. The offshore station belongs to the Bahía Blanca Port Consortium and is installed on a tower structure.

Data on electricity demand load for the three cities of interest (Figure 1) were collected from reports issued by the Ministry of Energy and Mining (Argentina). Unfortunately, these data are available at monthly intervals and refer to electricity consumption, but do not report peak consumption values; power consumption data by sectors are available only for annual periods.

Wind

Table 2 provides a description of the wind speed measured. Data were averaged over 10 min for the statistical analysis. Wind speed (V) increases with height; wind speed at a given height was calculated applying the power law:

where α is the power law exponent, V_h is measured wind speed (m s^-1), and V_ref is estimated wind speed (m s^-1) at height h (m). The exponent was adjusted for each windward face (every 45°) in each site according to terrain characteristics.

The distribution of wind speed is a key factor in wind source assessment for a given site. Different theoretical models, such as Weibull, Rayleigh, Log-normal, and Gamma, are applied to fit the distribution of wind speed data. Several studies (e.g., ^{Rehman et al., 2012}; ^{Ramos and Iglesias, 2014}) showed that the Weibull distribution was the best model due to its great flexibility and simplicity. The general form of the Weibull distribution function for wind speed data is:

where f(V) is the probability density function of observed wind speed data, and k (dimensionless) and c (m s^-1) are shape and scale parameters, respectively. The maximum likelihood method was selected to calculate k and c, and is expressed by the followings equations (^{Stevens and Smulders, 1979}):

where V_i is wind speed in time point i, and n is the number of data points.

Wind power density based on the Weibull probability density function can be expressed as follows (^{Ohunakin,2011}):

where p_v is wind power density (W m^-2) and Γ is the Gamma function. In order to characterize the wind power potential in the study area, wind power density was graded according to ^{Elliott et al. (1987)}, ranging from poor to very outstanding.

Solar PV

Incident solar radiation was measured using pyranometers (Apogee, model SP-110; spectral range: 300-1100 µm). In this study, the periodicity of radiation data collection was variable (Table 2). Cloudiness is the most important factor affecting the solar radiation on the earth’s surface. Hence, in order to characterize solar power potential in the study area, cloud cover was considered. Cloud cover (Cc, dimensionless) data are usually unavailable, but it can be determined from daytime using the following equation (^{Crawford and Duchon, 1999}):

where K↓ is incident solar radiation and K↓_t is the theoretical clear-sky incoming solar radiation. Cc values close to 0 imply a clear sky, whereas values close to 1 indicate a fully cloudy sky.

Wave

Table 2 provides the description of wave monitoring. The equipment used included a pressure sensor that measured water level at 10 Hz for 30 minutes.

Under a random sea, wave power per unit crest (P, kw m^-1) is given by:

where ρ is sea water density (ρ = 1030 kg m^-3), g is acceleration due to gravity (g = 9.81 m s^-2), H_s is significant wave height (m), and T is wave period (s). Frequency analysis and descriptive statistics for wave data were conducted.

Data handling and statistical analyses were carried out in Microsoft Excel and R. Finally, the relationships between electricity load demand and a given renewable source (wind, solar, and wave), as well as between sources, were explored using the Pearson correlation coefficient.

Wind Potential

Figure 2 shows hourly and monthly mean wind speed data for all study sites at 10 m height. On an hourly time scale, sites GC, MH, and LS exhibited a similar temporal pattern, with wind speed increasing from the early morning to the late afternoon hours (Figure 2a); OT showed a pattern opposite to that of continental/onshore sites, especially from 8:00 am to 12:00 am. On a monthly time scale, for all sites, wind speed reached peak levels from October to February, while the lowest values occurred during winter (Figure 2b). These results indicate that, among the continental/onshore sites, LS and MH attained the maximum and minimum values, respectively. On the other hand, compared to continental/onshore sites, the offshore site showed higher wind speed values in all months. Monthly mean wind speed at 10 m height varied between 4 and 7.7 m s^-1 and between 6.1 and 8.7 m s^-1 in continental/onshore and offshore sites, respectively.

Figure 2 Hourly (a) and monthly (b) mean wind speed (m s-1) for study sites at 10 m above ground level. 

Table 3 shows monthly shape and scale parameters and wind power density for each site at 30, 60, and 90 m above ground level. Weibull distribution parameters usually vary from site to site, as in this case. These results indicate that the mean shape parameter for the region lies within the range of 1.57 (GC) and 2.49 (OT). The k parameter exhibited a slight inter-monthly variability, although discontinuous peaks occurred. Monthly k values ranged from 1.30 to 1.87 for GC, from 1.57 to 1.78 for MH, and from 1.60 to 1.95 for LS. For the offshore site, monthly k values were always above 2, ranging from 2.16 to 2.81.

The scale parameter showed a clear temporal pattern characterized by lower values in winter and higher in summer (Table 3). At an intermediate height (60 m), monthly c values (m s^-1) ranged from 5.8 to 8.5 for MH, from 6.4 to 10.2 for LS, from 6.7 to 9.5 for GC, and from 8.5 to 12 for OT.

According to ^{Elliott et al. (1987)}, mean wind power density (calculated by averaging monthly values) was graded as fair/moderate for MH (244 W m^-2 at 30 m, 393 W m^-2 at 60 m), good for sites GC (404 W m^-2 at 30 m, 552 W m^-2 at 60 m) and LS (451 W m^-2 at 30 m, 556 W m^-2 at 60 m), and excellent for the offshore site at both height levels (636 W m^-2 at 30 m, 783 W m^-2 at 60 m), (Table 3). Wind power density for sites GC, LS and OT was graded as good and excellent during winter and summer, respectively, at both 30 and 60 m; for the remaining site (MH), wind power density was fair in winter and good in summer. At 90 m, mean wind power density reached 508 W m^-2 for MH, 620 W m^-2 for LS, 652 W m^-2 for GC, and 876 W m^-2 for OT.

Solar PV Potential

Figure 3 shows monthly mean solar radiation and mean cloud cover for the three sites monitored. Solar radiation ranged from 163 W m^-2 (June) to 556 W m^-2 (December) for GC, from 195 W m^-2 (June) to 467 W m^-2 (January) for MH, and from 212 W m^-2 (June) to 530 W m^-2 (January) for LS. In this study area, with a latitude ranging from 38°38 to 39°30S, mean sunshine duration varied from 8.28 h in June to 13.85 h in December. Peak values occurred at about 13:00 local time.

Figure 3 Monthly mean solar radiation (watt hours per square) and monthly mean cloud cover for sites GC (a), MH (b), and LS (c). 

Cloud cover varied between sites. Mean cloud cover was 0.218, 0.225, and 0.253 for LS, GC and MH, respectively. The intra-annual behavior of cloud cover for all sites followed a pattern opposite to that of solar radiation intensity (Figure 3). Thus, in winter, and to a lesser extent in intermediate seasons, the monthly mean cloud cover was close to 0.30, whereas in summer it did not exceed 0.20.

Figure 4 shows the evolution of cloud cover and ONI (Oceanic Niño Index) for MH between 2010 and 2016. MH was selected for this analysis because it had the longest solar radiation recording period. During that period, two opposite events (El Niño and La Niña) varying in intensity level were identified. The most significant episodes were one moderate La Niña (from September 2010 to January 2011) and one strong El Niño (from November 2014 to February 2015). Extreme cloud cover and ONI values were seemingly out of phase with each other. As regards the moderate La Niña episode, clear-sky conditions were observed with cloud cover values ranging from 0.12 to 0.18 only; after three months of the strong El Niño episode, a peak of 0.45 was recorded.

Figure 4 Monthly Cc and time series corresponding to SST anomalies in the Niño-3.4 region for MH during 2010-2016. 

Wave Potential

The study area is exposed to waves generated by local winds (Figure 2), mostly in spring. In winter, the longest periods were mainly associated with waves from the Atlantic Ocean. Mean H_s for the study period was 0.78 m, with a standard deviation of 0.14 m (monthly values). Figure 5 shows monthly averages of H_s, T and wave power for the study site. Peak monthly H_s values were recorded in spring-summer; in May, H_s was also high. In contrast, winter recorded the lowest H_s values, with more than 50 % of waves smaller than 0.5 m in height, and with long periods of more than 8 s.

Figure 5 Monthly averages of Hs, T and wave power for OT. 

Mean T for the whole period was 8 s. In general terms, T seemingly followed a trend opposite to that of H_s (Figure 5). This is an important observation, since the resulting wave power is a function of both H_s and T (Eq. 7). Thus, the monthly mean wave power peaked in September and November (4.6 kW m^-1), followed by May (3.8 kW m^-1) and January (3.3 kW m^-1) (Figure 5). Mean wave power was 2.8 kW m^-1. Figure 6 shows the percentage of wave power at various intervals of H_s (0.5 m) and T (2 s). The maximum wave power was recorded at 1-1.5 m (25.8 %) and 4-6 s (23.3 %).

Figure 6 Percentage of wave power generated at various intervals of H_s (a) and T (b). 

Power Load Demand versus Renewable Sources

Table 4 shows the average monthly consumption for the three cities of interest (Bahía Blanca, Monte Hermoso and Pedro Luro). These cities differ in many ways, including main economic activity, population size, etc. The Study area section and Table 1 provide details on electricity consumption (e.g., annual consumption level and number of users for each main economic sector) for the three cities. In all cases, the highest consumption occurs during summer, mainly in January, with increases of 15, 55 and 17 % over the minimum values for Bahía Blanca, Monte Hermoso and Pedro Luro, respectively (Table 4). This behavior is most pronounced in the coastal city of Monte Hermoso, resulting from the seasonality of tourism (sun and beach product).

Figure 7 shows monthly wind, solar, and wave power for all sites already considered, to summarize power availability. There is an overall pattern of increasing power levels during summer, particularly when considering wind and solar sources, reaching mean values up to 1.06 kW m^-2 (wind at 30 m height, OT) and 0.56 kW m^-2 (solar, GC). On average, wind and solar power showed variations of 65 % and 63 % between minimum and peak annual values, respectively. On the other hand, wave power showed a higher (intra-annual) variability.

Figure 7 Monthly wind (30 m height), solar, and wave power for all sites. 

The Pearson correlation coefficient, which quantifies the similarity between two data sets, was applied to measure the association between electricity load demand and a given power source (wind, solar, and wave) on a monthly basis. The key point is that when considering associations between load demand and power available from renewable sources, the identification of positive correlations is essential.

Figure 8 shows seasonal and annual correlation coefficients between electricity load demand and each source for the three cities. Regarding the wind source, the annual correlation coefficient ranged from 0.05 (Pedro Luro) to 0.52 (Monte Hermoso). However, as expected, some seasonal values yielded higher correlations. For instance, in summer, when consumption rises considerably, correlations were high (and positive) for Bahía Blanca (r = 0.83) and Pedro Luro (r = 0.95) (Figure 8a,c). In contrast, in the case of Monte Hermoso, the correlation was low (Figure 8b) due to the fact that the disproportionate consumption, which is highly concentrated in January (Table 4), differs from the trend of wind power (Figure 7). As regards the solar source, the annual correlation between load demand and solar power ranged from -0.14 (Pedro Luro) to 0.63 (Monte Hermoso). In summer, correlations were high (and positive) for Bahía Blanca (r = 0.88) and Pedro Luro (r = 0.99) (Figure 8a,c), and low for Monte Hermoso (r = 0.15) (Figure 8b), similar to those observed for wind. Finally, wave power was poorly correlated (r < ±0.2) with load demand in annual terms, for the three cities. Seasonal correlations were high, with some exceptions.

Figure 8 Seasonal and annual values of the correlation coefficient between electricity load demand and wind, solar, and wave sources for Bahía Blanca (a), Monte Hermoso (b), and Pedro Luro (c). 

Integration of Renewable Sources

Table 5 shows the Pearson correlation coefficients between the different sources at multiple time scales (hourly, daily, and monthly) for all monitoring sites. This coefficient was determined based on mean values. When considering associations between the different sources in search of complementarity options, the identification of uncorrelated or negatively correlated results is essential.

In general, a clear pattern of associations is observed. The correlation between wind (continental/onshore) and solar sources was high and positive for all monitoring sites on hourly and monthly time scales, but weak on a daily time scale. In this study area, both wind and solar power followed a very similar temporal trend at hourly and monthly time scales, meaning these cannot complement each other for combined exploitation. Most of the remaining power source pairs, i.e., wind (continental/onshore) versus wave, solar versus wave, wind (offshore) versus wind (continental/onshore) and wind (offshore) versus solar, showed a high negative correlation at the hourly time scale and a weak correlation (approximately from 0.3 to 0.7) at a monthly time scale, indicating that these may be complementary. In the particular case of OT, correlations were high (and positive) and weak at hourly and monthly time scales, respectively.