DATA VIEWER

5 OzEA_DVSA90005

Barry Brook Subject: WF and demand correlation Date: 2010-05-20 (at 12:17:19)

Okay, to try and get an initial handle on this, what sort of correlation are we talking about here? Could you try this - take a 3 month (say) segment of the data and randomly subsample 1000 WF/Demand pairs (to avoid autocorrelation), calculate rho, and then rinse and repeat 10,000 times and plot up the resulting histogram?

First step, I want to get a handle on the strength of the relationship. Next step might be to repeat the above but block the sampling by times of the day (e.g. draw all subsamples from a 1 or 2 hour fixed window in each day).

7 OzEA_DVSA90007

sam Subject: Wind farm output and demand Date: 2010-05-20 (at 12:55:48)

Can we do some simple time-series analysis to check for cross-correlation between the WF output and demand? It definitely looks like the wind farms go off during times of peak demand, which is odd no matter what the role of the wind farms are.

I'm afraid I don't know enough about this to provide a very informed hypothesis-- others on here can no doubt provide a better guess as to what's happening. However my first thought was that wind farms are probably more limited by the amount of wind rather than demand if they are providing base load to the grid. Obviously if they are installed for peak load then this would not be the case and we'd expect to see peaks in wind output where we get peaks in demand.
What are the current roles of wind farms in SA? Are they peak providers or do they contribute to regular base load?

My other guess is that if wind farms are providing baseload power to the grid and their output is unreliable, then other power sources need to increase their output. That's going to take some time-- is it possible that something in the grid can be made to wait for power while peak loading ramps up? In a nutshell I'm wondering if demand can actually lag WF output due to the unreliable nature of wind. A quick google search on load management looks promising, but again, this is something for someone more expert than myself.

fc - wind farms make electricity when the wind blows. I'm doing some characterisation work on the data before any 'analysis'; but I welcome the analysis of others.

9 OzEA_DVSA90009

douglas wise Subject: electricity price, demand and windfarm output Date: 2010-05-22 (at 21:26:06)

In the preamble, it is stated that the default viewing data are for one week but I only see 6 days with Sunday missing. I would have supposed that weekend demand would have been be less than that during the week. However, judging by the Saturday figure that I can see on my screen, my supposition was wrong. What do you guys get up to on a Saturday that demands more energy than when you\'re at work? Is it extra air conditioning?

Just to check that I am interpreting the data correctly, could someone confirm that the following conclusions that I have drawn are correct? On 6th Feb 2009, when wind peaked, WF is shown as 21.76. At that point, it was providing at 83.7% capacity (21.76 times 100 divided by 26). At the same time, actual demand was 63.9% of mean demand (31.96 times 100 divided by 50). Finally, wind at 100% would have the capability of providing 52% of mean demand (26 times 100 divided by 50).

fc: [31st May 2010] - these numbers were read correctly by Douglas Wise, but were wrong - the viewer has since been corrected

11 OzEA_DVSA90011

douglas wise Subject: data viewer Date: 2010-05-23 (at 15:22:44)

Francis, thanks for your reply. Since yesterday I can no longer view the data on my screen and get a message informing me that some script is making it run so slowly that it will shut down before download is complete.

That aside, I had a few thoughts on why the grid seems to take more wind when demand is low (presumably the easiest way to establish whether the wind farms have any large (significant) off-grid contracts is to ask them.) Does this grid that you are using as an exemplar have any hydro with pumped storage? When demand is higher one might expect that the grid manager wouldn't want to risk pushing water uphill if, at any moment, he might be met by the need to send a load down (I'm assuming there aren't separate up and down pipes for the whole length of the run).

Would it not be helpful to have data from some representative anemometers? If wind speed doesn't correlate well with wind power uptake by the grid, it would suggest that the grid manager daren't use wind at times of high (anticipated) demand. Should such be the case, it would suggest that wind's typically measured capacity factor is greater than its useful (or practicably useable)percentage.

Is there any reason why you can't get your hands on other grid data? If you knew, for example, how much electricity each type of energy generator (coal, gas and hydro plus that imported from another grid) was contributing to the grid at any time, you would have a much fuller grasp of what actually happens. If on top, you could get a handle on what is being paid for each type of power at any point in time, you would nearly have cracked the problem. It should be possible then to model a situation in which wind provided from, say, 0-50% of total demand and calculate effects on electricity cost and CO2 emissions.

Should these data not be available, perhaps you should create a virtual but realistic grid as a model and invite a range of grid managers to operate it (until you became sufficiently skilled yourselves). As a matter of fact, it could then be made into an entertaining and educational game show on the telly and your research would become self financing! Franchise it round the world.

The huge spikes in price at certain periods are presumably associated with gas or hydro being switched in. (Alternatively, are certain large industrial users being compensated for not using power at these times?)

One might need long term data to enable one to work out what the mean price of grid used power is for each type of generator and what are their used capacity factors.

16 OzEA_DVSA90016

Francis Subject: Re: data quality on WF output Date: 2010-05-26 (at 14:06:13)

Yes, there is a problem. Have now convinced myself of that. Now I need to work out what has happened, and then fix it.

18 OzEA_DVSA90018

Neil Howes Subject: demand and wind output Date: 2010-05-28 (at 11:54:37)

Francis,
The data from the small number of wind farms(311MW)now looks very similar to the entire aemo data set for the 18 wind farms(1608MW).
Looking at 2009 I count 30 days when wind output peaks above WF=20( 70% capacity) compared to 20 peaks for the entire wind farms.
Of those 30 peaks SA demand is >35(1070MW) in all but 5 occasions(when output peaks at night).

If it is reasonable to expect load shedding above 70% capacity for 6-12h, 5 days/year, that would be a loss of <0.1% of production.
It would allow 100/70x (1070MW+ 520MW export) =2270MW wind capacity, providing all other generation could be turned off during high wind output periods. This is 794MW average( at 35% capacity factor) or a little over 52% of average demand that could be accommodated without any additional interstate grid upgrades. Thats similar to what I calculated(59%) using the aemo data of 18 wind farms and total demand.

Thus 50% energy from wind power in SA(2200MW capacity) may be a good starting point. That is less than the present planned 3200MW capacity.

20 OzEA_DVSA90020

Neil Howes Subject: low wind output Date: 2010-05-31 (at 11:43:54)

Francis,
The low wind output periods appear to be due to high pressure systems moving over the SE of Australia. It is interesting to look at individual outputs around these events, for example 12March,2009, Woolnorth accounts for 80% of the 154MW( of the 9 farms on the NEM grid), 13-14 April, 2009 Woolnorth accounts for 50-65% of output, and May2-4 again Woolnorth accounts for 60-70% of output.
I interpret this to indicate that because Woolnorth is the most extreme SE location(in 2009) it is on the edge of the low wind system.
More sites, covering a larger geographical area are going to reduce the duration and increase the low power output but probably will need to have >90% back-up capacity for those extremely large high pressure systems( unless wind sites extend to WA and to QLD).

I dont see a problem in having >90% OCGT back-up( or a mixture of OCGT and pumped hydro) if it is only required to operate at 10-15% capacity. Its more of a problem if the OCGT is required to operate at 50% capacity as in the Hazelwood replacement scenario, ie not enough wind capacity is available and it is too concentrated in one small region. With a higher wind capacity the problem becomes one of how much peak output has to be load shed or used for pumped hydro storage. Wind peaks tend to be much shorter duration than low wind periods because high wind fronts cover a smaller geographical area and move rapidly, so storage(MWh) requirements are less, but may need higher pumping power capacity(MW).