Michael Giberson
A reader points out a news release by Baryonyx Corp. announcing its success in a recent Texas lease offer for two offshore wind concessions. Baryonyx intends not only to build the offshore wind power plants but also to co-locate a high-reliability data center with the wind farms.
File the idea as “just so crazy it might work.” At first wind farms and data centers seem like strange bedfellows. Wind farms offer variable power and data centers demand reliable power. But at least part of the idea is that by co-locating a large load with wind power, it can reduce the need for significant transmission upgrades to accompany the wind farm.
I suppose the power system issue is whether the net load at the point that the project connects to the grid is more or less variable with the data center added to the wind farm than without it. If the data center power usage is positively correlated with wind farm output, then the answer is yes. My first guess would be a negative correlation, but I don’t know much about data center load profiles and haven’t looked at any data.
(The press release makes an odd claim about providing “indigenous low-carbon energy to offset imported energy.” Wind power in Texas tends to offset natural gas and coal consumption, and while a small amount of natural gas is imported, most of it comes from Canada. Well, press releases are not known for the rigor of their claims.)
You can use selective data center operation to create artificial negative correlation between the price of selling your wind power and data center power load. Did you see that Google was planning to build a data center that turned off (data load sent elsewhere) when it got too hot to cool naturally? That seems to be a strategy that would work here.
When there is no (or not enough) wind at the data center and power is cheaply available use the power lines to import power
When there is too much wind power to export on the existing power lines (or exporting it is less valuable than running the center) then run the center.
When the price of power is more valuable to send away then to use for the data center then shut down.
This could minimize the negative power price situation we’ve seen in Texas too.
Ah ha, the ‘world is flat’ in this case! Selective data center operation!
If power expense is large and variation in power cost high from hour to hour, it may make sense to maintain multiple data centers and switch the load around as you indicate.
Wonder if it is easy to find places with negatively correlated power prices, so that when the price is high at one location it is likely low at the alternative.
Fine if data currency and availability are of no importance. However, if that is the case, why bother collecting and storing the data in the first place?