4.2 Inherent variability

Power outputs from both solar and wind generating sources are known to vary considerably with varying irradiance and wind speeds respectively. Figure 40 shows the variation of wind power output from a single turbine over a period of 20 seconds with sub-second sampling time [25]. By looking at the turbine output between time = 8seconds and time = 9 seconds, a drop of power output from 1.5MW to 0.9MW, corresponding to a drop in power of 40%, can be seen to occur in just 1 second. It is clear that large rapid drops in power output can be experienced by wind generating systems, similar to that seen in solar PV systems, at the 1-second timescale.

Figure 40 Wind power output for one wind turbine (doubly fed induction generator) [25]

The variability of solar power over timescales of 1-minute and 10-minute for plants located in Las Vegas can be seen in Figure 41 [12]. Corresponding variability of a wind power generation system in California (CAISO) at 1-minute and 5-minute intervals is shown in Figure 42 [10]. Comparing the 1-minute deltas from both graphs, solar shows a maximum percentage change of 50% for individual plants and around 20% for all six plants in aggregate. Wind, on the other hand, (note Figure 42 would be considering the wind generation in aggregate) shows a maximum change (negative) of 50 MW or 0.4%. The installed capacity is 12500 MW. The maximum delta for a 5-minute interval is 200 MW (negative) or 1.6%. Although the variation of wind power in this comparison appears insignificant compared with that of solar power (which can change up to 20% in aggregate over a 1-minute interval), it should be noted that the level of variability is highly dependent upon the number of sites and spatial diversity of the solar and wind sites. In this case, 6 solar plants were considered and the number of wind plants is unknown.

285

Source: Yih-huei Wan, NREL. Adapted from presentation at the PV Variability Workshop.

Figure 41 Cumulative distributions (95th and 100th percentiles) of six individual PV plants within a ~200 square kilometre area in Las Vegas [12]

430

Figure 42 Wind variability (deltas) for CAISO - 12500MW capacity [10]

When variation is considered and analysed at the hourly level, greater variability is observed in wind systems. Figure 43 shows the duration curves for load, wind and solar for CAISO for a simulated scenario with 12500 MW of wind and 2600 MW of solar, for a total of 33% renewables penetration. These graphs are created by sampling the wind and solar power output every hour, then sorting the samples in descending order (highest output reading at hour 1, lowest at hour 8760). Comparing the duration curves for wind and solar shows that wind has a steeper gradient; indicating greater variation hour-to-hour. If a non-variable resource (coal fired power station for example) is plotted on the same graph, an almost horizontal line would be observed due to its minimal hourly variation in output.

Figure 43 Load, wind and solar duration curves for CAISO (2010X scenario) [10]

Analysing all the graphs illustrated in this section would indicate that in the second-to-second timeframe the output of individual wind turbines is similarly variable in nature to individual solar arrays. When wind and solar output are taken in aggregate and analysed at 1-minute to 10-minute intervals, wind seems to benefit more from the smoothing effect associated with aggregation, showing less variability than solar. Finally, when behaviour is observed at 1-hour intervals, again in aggregate, wind is shown to be more variable.

The difference in correlation versus distance for both wind and solar systems was studied in [26] for sites in Sweden spread over approximately 700 square kilometres. The correlation between wind sites is seen to decrease with distance at a greater rate than for solar, as shown in Figure 44. This suggests the smoothing effect due to aggregation will be greater for wind than solar. The sample rate for this correlation analysis is given as one hour for both wind and solar. It is not specified whether the irradiance is averaged over the hour or an instantaneous measure.

Figure 44 Correlation vs. distance for wind and solar [26]

The smoothing effect due to aggregation of output from the Swedish wind and solar sites is shown using duration curves in Figure 45. The decrease in gradient of the aggregated curve is an indication of reduced variability. Again, it can be seen that the smoothing effect due to aggregation is greater for wind than for solar. This hourly load duration curve shows a different picture of the relative variability of wind and solar when compared with the hourly load duration curve shown in Figure 43 for the CAISO system. A steeper gradient for wind is observed for the CAISO system, implying greater variability, while a steeper gradient is seen for solar power output in the Swedish study, implying solar is more variable. From the results of these studies, one could deduce that variability of both solar and wind resources depends on factors such as geographical location, spatial diversity and size of the renewable generation system.

Figure 45 Duration curves wind and solar [26]