[For those looking for "How to Peak Around a Corner," give me a couple more days. It's a bit harder than I first thought.]
I literally just got out of a seminar given by Marija Ilic, of Carnegie Mellon, on her work and prognostications for making the delivery of electric power more efficient. I was expecting the typical drab droning that usually confronts me from invited speakers, but I was able to pull several salient and cogent points out of her talk. This is a fairly extemporaneous post, so please forgive my stream-of-consciousness style.
Utility companies, in general, try to provide uninterrupted service at minimal cost to their subscribers. We could go on and on about the scruples of utility companies and their guaranteed monopolies, but that is not really very productive at this point, so let us take their cost-minimization desires as a given. While "minimal cost to subscribers" is a very soft term, I never realized how hard the uninterrupted service constraint really was. Imagine this scenario: Town X has a daily power consumption of no more than 90 MW for 350 days a year. Given this information, constructing a 100 MW power plant to supply Town X seems like a good idea. However, for 15 days per year, Town X demands 120 MW (due to heat waves, cold snaps, etc.). If the utility company really had only constructed a 100 MW power plant, all of the directors would be fired after the first year because they couldn't supply the town sufficiently for 4% of the year. So instead, the utility has to lay out the cash to build more power generation capacity that just sits around for up to 350 days a year.
As Prof. Ilic stated, "cost management is all about the spikes." She was referring to the spikes---hourly, daily, and yearly---in demand that utility companies see. If it weren't for the spikes in demand, the total demand would have far less variance, and less superfluous generation capacity would be necessary.
What can we do about these spikes? This is where the system intelligence comes into play. Prof. Ilic claims that currently, most utility companies have extensive records to estimate their long-term average demand, but they use only something like the past 15 minutes of data to predict upcoming surges. I would really like to see a "look-ahead" filter (or model-predictive controller, as Prof. Ilic calls it) on the power companies' demand estimates. This information should then obviously be distributed out to the consumers, who then are incentivised to decrease demand when faced with surging costs.
Obviously this incentive scheme can only be implemented with some form of "smart meter" at the consumer side. It really baffles me why these haven't gone past pilot programs across America. The basic functionality of a smart meter is that it records the power drawn by a consumer, and logs this information over time. More dynamic implementations could go so far as to incorporate an automatic controller. This would be a household appliance (there's an app for that!) that lets the user list and prioritize power-hungry processes as well as input a desired maximum cost of electricity each day. The meter/controller would then use the predicted power costs to schedule the user's processes to run as time---and money---allow.
The funny thing is: This "scheduling" of "processes" already happens in every office and nearly every home in the country. Computers and servers operate on the very principle that some processes have higher priorities than others, and that they will all (hopefully) get done in due time. Except that computers decide between processes thousands of times every second. Should it really be that hard to decide to run your laundry dryer at 1:00 AM and your dishwasher at 3:00 AM?
This pseudo-rant so far has only touched on the "one-way" current model of the power grid. The sexiest topic in smart grid research is how to (best) incorporate distributed power generators. On the one hand, the concept is quite simple: Whereas before a utility user could be viewed solely as a power sink, now there is the possibility of seeing a negative sink (commonly referred to as a source). However, as Prof. Ilic is quick to explain, things are not quite so simple.
All electric power these days travels into our homes on endless sine waves of current. In the US, if you were to stand at one point on the transmission wire, the peaks of these waves would pass you 60 times every second. This standard has allowed for the universal (well...almost) wall plug that all your household devices use. Correspondingly, the power companies transmit and transform the power so that this is exactly what you get at the "tap." When people start generating their own power, there is currently (haha...get it?) no way of knowing that the phase of the user-generated signal will match the phase of the utility-generated one. This phase-matching problem is fairly serious and expensive to control.
The alternative, it has been postulated, is to discourage end users to feed back into the grid. Rather, users should endeavor to store as much of their locally generated power as possible. Now people are thinking inter-disciplinary. Now if you have a plug-in hybrid vehicle, you instantly have a massive battery in which to store your excess power. It's like a two-for-one! The problem is that without dynamic smart meters, everyone comes home from work and at 6:00 PM, plugs in their hybrid and create one of those damned spikes. Now you see how all of these problems are interconnected.
I'll wrap things up with a little utopian vision. My dream house will have several acres of woodland (and be under an hour on public transportation to a major airport) on which I can erect a wind turbine or two. These will provide power to my house as allocated by my self-designed power controller. On windy days, I'll have so much electricity that I'll be able to pump water out of a little stream into a tank uphill from my house. Then, when I need hot water for my bath, I'll suck it out of this tank, through a parabolic reflector and then through a generator, into my tub and use the electricity to run my reading light. Too easy.
Stay smart,
Clay
