• No results found

4.5 The impact of different weather years on grid carbon intensity

4.5.1 The impact of wind generation on grid carbon intensity

Modern-Era Retrospective Analysis for Research and Applications (MERRA) dataset (Drew et al., 2019a), (Drew et al., 2019b) provides GB-aggregated wind and solar capacity factors in hourly resolution from 1985 to 2015. The annual average figures for both are shown in figure 4.4. For the purpose of this study, three wind “years” have been selected to represent a low, average and high scenario of wind. In specific, 1986 was selected as the high wind, 2011 as the average wind and 2010 as the low wind scenario. These wind capacity factors datasets have been converted to wind generation datasets using the current wind capacities (see table 4.2) and runs of the 25 unit MILP have been carried out with 2017 National Grid transmission system demand data. It is noted that as the MERRA data is in hourly resolution, grid carbon intensity datasets are in hourly resolution too (as opposed to half-hourly in the remaining sections). Wind and solar have been considered separately in the calculations in order to assess how solely wind and then solar variability affects grid carbon intensity.

(a) Carbon intensity (g/kWh) - Low wind year

(b) Normalised wind generation - Low wind year

(c) Carbon intensity (g/kWh)- Average wind year

(d) Normalised wind generation- Average wind year

(e) Carbon intensity (g/kWh)- High wind year

(f) Normalised wind generation - High wind year

Figure 4.5: Grid carbon intensity and wind generation for different wind years.

Spearman correlation coefficient ρ has been used:

ρ = 1 − 6

P

d2

i

n(n2− 1)

where d is the pairwise distances of the ranks of the variables xi and yi and n is the number of samples. Spearman’s ρ is a rank-based version of Pearson’s correlation

coefficient, which can be used for variables that are not normal-distributed, more volatile and have a non-linear relationship. Moderate, negative correlation has been noticed in all wind years ranging from −53% to −47% which can also be observed in figure 4.6. This, as expected, means that when wind generation increases grid carbon intensity deceases.

Figure 4.6: Linear fit for wind generation against grid carbon intensity (three MERRA weather years).

Figure 4.5 depicts the normalised wind generation with the respective grid carbon intensity for the three wind years in coloured array plots while table 4.6 presents some of the statistical characteristics for the three time series. The average grid carbon intensity is shown to range only from 318 g/kWh for the high wind scenario to 331 g/kWh for the low wind scenario, while all time series are shown to be similarly “spread” (similar standard deviation 30-33). This difference of 13 g/kWh between the annual averages for the different wind years appears to be negligible and can be misleading about the real impact of wind on grid carbon intensity. The annual average as a metric to understand grid carbon intensity masks the patterns of behaviour that can be noticed only if the whole time series is assessed in higher resolution.

A significant feature can be noticed among the different wind years in figure 4.5; In the low and average wind years, it can be noticed that grid carbon intensity was higher (orange to red-coloured half-hours) for bigger parts of the total year. In specific, it was measured to be higher than 350 g/kWh 2494 times (28% of the year) for the low wind

scenario, 2006 times (23% of the year) for the average wind scenario and 1279 times (14% of the year) times for the high wind scenario (table 4.6).

Figures 4.7 and 4.8a present the average grid carbon intensity per hour and per month. Figure 4.7 indicates that different wind generation does not affect the general pattern of average hourly grid carbon intensity which is consistent with the demand pattern (figure 4.11a). Different wind generation is shown to cause an average fluctuation of 13 g/kWh to average hourly grid carbon intensity. On the contrary, average monthly grid carbon intensity does not follow a consistent pattern for all years and the effect of the various wind generation is more evident. Since all three grid carbon intensity datasets have been built using the same demand profile, it is safe to assume that all visible variability between the three profiles on figure 4.8a is caused by the different wind output. For instance, although 1986 (yellow line) was selected as the year with the highest average wind generation, September of the same year displays the highest average grid carbon intensity compared to the other two years. Looking now at figure 4.8b ,which presents the average monthly wind generation, September indeed had the lowest average wind generation across the three scenarios. It is also noticed that the average monthly grid carbon intensity follows the same pattern with the average wind generation in all three wind years. For example, in the case of January, it is seen that an increase of approximately 2000 MW in average wind generation (5000 MW for the high wind year versus 3000 MW for the low wind year) results in a decrease of approximately 30 g/kWh for the corresponding average monthly grid carbon intensity (360 g/kWh for the high wind year versus 330 g/kWh for the low wind year).

Mean CI Median CI Standard

deviation Min CI Max CI

% of the year CI ≥ 350

Low wind 331 330 33 206 430 28

Average wind 326 325 31 211 417 23

High wind 318 317 30 207 423 15

Figure 4.7: Average hourly grid carbon intensity (g/kWh) for MERRA weather years.

4.5.2

The impact of embedded solar generation on grid carbon