OZ-ENERGY-ANALYSIS.ORG   -   open science for the new millennium

THE STORIES   |   DATA   |   ANALYSIS   |   MODELS   |   LITERATURE   |   DISCUSSIONS

Status: archival - initial work circa May 2010

Gross Variability of SA Wind Power

ABSTRACT

Looking at the SA electricity demand and supposing 100% Wind Power supply. This supply is constructed by scaling up current wind resources so that, over an entire year (2009 in this case), the total power supply matches the total power consumption. All data is processed at the week-by-week level to examine the gross scale variability.

DATA and METHODS

I have taken the SA demand data for 2009, in weekly chunks (red curves), and divided by the mean so the data is around 1 (mean of this demand data is 1.5 GW).

Have also taken the Wind Farm output data , incorporating 5 SA wind farms and 2 others (1 Vic = YAMBUKWF, 1 Tas = WOOLNTH1), again in weekly chunks. The averege output of the individual WFs ranges from 8 to 55 MW; the 5 SA WFs under consideration add to ~100 MW average (~7% of demand). In the work that follows WF output is scaled up to meet the demand, either individually (the blue curves in the individual plots), or in the combinations used.

A mild bilateral filter has been applied to all curves.

ANALYSIS

We look in a simple way at what would happen if all SA electricity was from wind. To do this, assume that unlimited and perfect storage exists to buffer the variations in supply. How wind generation matches demand at the gross level of weekly chunks is shown in the following plots:

	Figure 1. SA wind example (2009). Scaled up, week level data for the Mt Millar WF. See here for individual plots of all the WFs (traces also shown in following figures).
	Figure 2. SA wind overall (2009). This is the week level data for the 5 SA WFs under consideration (black traces are the individual WFs). It is seen that at this gross level, the variability is mostly in step across the 5 SA WInd Farms. We see also that, at the week level, the more productive weeks involve more than three times the generation obtained in the less productive weeks.

To further extend the geographic diversity of the wind power, we now construct the wind resource 1/3 from the above combined SA WFs, 1/3 from the Victorian WF (YAMBUKWF), and 1/3 from the Tasmanian WF (WOOLNTH1), to obtain Figure 3.

Figure 3. Three states combined. Much as above, albeit a mild improvement overall. It is noteworthy that the same large scale weather patterns are, mostly, seen across all three states.

I'm assuing (without checking) that the two peaks in demand correspond to heat waves. The one at the begining of the year shows no particular relationship with the wind. The Oct / Nov demand spike appears to be associated with poor WF output; I speculate that the heat and the lack of wind are related to a large high pressure system, and note that such situations present a substantial wrinkle in high wind energy scenarios.

Another way to look at this is to look at the running total of supply minus demand (assuming that there is a perfect and as-large-as-we-require storage system). This is shown, lightly filtered, in Figure 4 below.

Figure 4; supply deficit from 100% Wind. The left panel (A.) shows the 5 SA wind farms (scaled to meet demand over the year) fall into deficit in the earlier part of the year (some 25 days short by early June), and make this up in the latter part of the year. The total storage that would be needed is over 30 days worth (bottom to top). The right panel (B. shows 3 states combined, using the overall SA curve from A, plus the two interstate WFs (Tas and Vic respectively); the overall storage requirement is slightly reduced.

The seasonal nature of the SA wind resource, at least as occurred in 2009, is clearly observed; more wind in June - October, less wind in November through to May. The same is observed for the Victorian WF (YAMBUKWF) in our dataset, while it is of note that the Tasmanian WF (WOOLNTH1) appears to be a substantially more even resource.

Examining just the major drop in Figure 4A (weeks 10-23), we see that over a period of ~90 days, the system becomes ~25 days of power in deficit. Thus, a 30% overcapacity in Wind Farm infrastructure would, in this highly simplified case, solve the variability problem for analysis at the week-by-week level. Of course, this does not solve the problem of what to do when the wind does not blow for three days and the ice-cream in your freezer melts.

Overall this is an exploratory analysis; more years of data would need be examined before making any specific claims. Of course there is no proposal to run SA on 100% wind; any high penetration of renewables would evolve over a number of years and would certainly incorporate other states, and probably involve other renewable resources (e.g. solar). It seems at this point that wind could play a substantial role in meeting electricity demand, and that it is the variability issues at lesser time scales (i.e. day-to-day, hour-to-hour) that now need examination.

9 OzEA_AVWW0009

G.R.L. Cowan Subject: Re: analysis_variability_wind_week Date: 2010-07-01 (at 14:05:22)

I took the data in
http://www.oz-energy-analysis.org/analysis/working_data/Dem_5WFtot_SA2009.txt.gz
into an Open Office spreadsheet, and then somewhat photoshopped
the graph produced there, to show what I have in mind,
attached.

[Show Full Lists]

 

  TOP   |   REFRESH PAGE   |   ANALYSIS

fc - April 2010