Retired Firefighter    ■    Power Dispatcher    ■    Husband    ■    Daddy    ■    Grandpa    ■    Crazy Man

Friday, January 13, 2012

The Smart Grid, Part 1 - First Tell Me About LMP

I already know this won't be brief. No matter how much I say I'll try to keep it short, I can't. I talk too much and over-explain everything. Having accepted that, I sleep OK at night at peace with who I am.

With that out of the way....

Why do we need a "smart grid"?

If you've been following me for a while and have read the tutorials, you know the basic truth that AC power is generated exactly as it is consumed, with no practical storage applications. As demand goes up, frequency drops, and power plants increase output to match demand, and vice-versa. It is fluctuating every second. Thankfully, the smoothed out impacts of millions and millions of people turning lights and toasters and Blu Ray players on and off don't amount to much instantaneous fluctuation relative to the size of the interconnected grid. If you were trying to run a power grid of only five houses with a plant just barely big enough to do it, and they all happened to turn on a bunch of stuff at once, everyone would see a big dip and maybe low voltage damage - assuming it didn't trip the generator outright.

Following demand is not an efficient game. During high demand periods, power plants with expensive fuel sources have to run to keep supply up, raising the aggregate cost of power. During low periods, only the cheapest units run, lowering the aggregate cost, but when coal plants run below efficiency (usually under 80% or so), they create more pollution relative to power produced. Running thermal plants up and down every day for years wears them out, as the boilers and systems fatigue and break down from being heated and cooled over and over and over again. Overall energy costs go up not just from fuel costs, but repair and maintenance costs.

Despite real-time hourly changes in energy prices based on what power plants are running at that time of day, almost everyone at the consumer level pays a flat per-kilowatt hour charge for electricity. This is tricky for power companies, as they need to make their rate case with magic math that figures out how much revenue will be coming in based on which hours their systems are used and what energy costs actually are, with hopes that they get enough money to run the thing, but not so much that the consumers scream foul. It is frankly very easy to screw this up, and a couple of disastrous power market days with real-time energy prices increased by 1000% or more can take years for a company to recover from with their rate case locked-in usage fees.

The first step towards efficiency is to get all power companies in a large area to cooperatively control their power plants instead of playing the hourly market game. The Midwest Independent System Operator (MISO), a quasi-governmental reliability agency with authority over a huge footprint of power companies in the midwest, was the first to do this on a really large scale. Their coordinated control concept was based on a model called Locational Marginal Pricing (LMP).

Don't let these acronyms scare you, I'll try to keep it simple.

Let's say you have a self-contained power system with four power plants and no ties to the outside world. Your transmission lines are overbuilt and there are no restrictions on power flows anywhere in your system. Your customers pay for their electricity based on hourly usage instead of a flat round-the-clock average cost. Your power plants use different fuels and therefore have different costs. You run the cheapest unit until it reaches capacity, and then start the next cheapest unit. When #2 peaks, you start the third unit, etc. Figuring out your hourly energy costs are easy, it's basic math based on how much was produced from each plant for any particular hour, and everyone in your system is charged the same thing for any particular hour.

In this model, location means nothing. Pricing is flat across all areas by the hour. The wrinkle that screws everything up is constraints on the electric transmission system.

Back to the same model. Self-contained system with four plants, one in each quadrant. But this time, the transmission lines between the quadrants do not have unlimited capacity. As long as no lines are at their limits, the flat hourly pricing works fine. Let's say the cheapest unit is in quadrant 1. Now, as long as demand is low, that cheap unit can supply everyone fine, but as the day picks up, demand in quadrant 2 rises to the point that if the cheap unit in quadrant 1 keeps going up to cover, the power line between them will overload. Since the cheap plant cannot supply the demand in quadrant 2, the next-cheapest plant somewhere else will have to come online to make up the shortage.

At this point, you have locational marginal pricing, based on transmission (transportation) constraints. Back in quadrant 1, energy stays cheap, and only the people in quadrant 2 who did not have the foresight to increase their transmission line's capacity have to pay the averaged costs of what they could get from the cheap plant and the remainder they had to get from somewhere else. It's no different than living next to a bakery and getting cheap donuts, but over the hills where they have a small bakery that can't keep up and they can't ship enough donuts in from elsewhere on the only dirt road into town, the value of a donut logically goes up.

Now, the MISO did exactly this, back in about 2003 or so, but on a GRAND scale. Covering all or parts of 11 U.S. States and the Canadian Province of Manitoba, with 35 power companies owning transmission assets and another 98 companies owning generation and/or serving load, they coordinate over 130,000 megawatts of energy generation from hundreds of power plants. In real time, pulsing power plants to move up and down to live within the cheapest-power-possible-to-where-you-are LMP model every few seconds.

How do they set prices? The generation owners bid the units into the MISO system, with costs based on output. They even have allowances for efficiency bandwidths, so you can say your power plant costs 'x' from 200-400MW, 'y' from 401-475Mw, 'z' 476-510MW, etc. The MISO super AGC computer sees real-time demand across their entire footprint, determines when generation needs to increase, and finds the next cheapest unit anywhere in their entire area to go up. Except that the MISO super AGC computer also knows what the real-time flows on all of the transmission lines are, and knows how moving any generator up or down will impact every transmission line. So it doesn't just grab the next cheap unit in the stack, it grabs the next cheap unit that can increase without overloading any power lines anywhere else.

This creates hundreds of pricing bubbles in the MISO footprint based on LMP restrictions. The MISO super AGC computer measures consumption at thousands of points on the system and knows which power company is the consumer at every given point. Knowing how much energy you used at any given consumption point every hour, and knowing the energy cost in that bubble for that hour, it is no great leap to figure out what to charge you for the energy based on your usage at all of your consumption points. So, in effect, the MISO "buys" the energy using cheapest-possible source from the bidding process and influenced by the LMP model, and then sells the energy back to the members with pricing based on their usage and location, keeping a cut to operate their massive bureaucracy.

This effectively killed the hourly power market in MISO, as it completely ended the hourly guessing game. Your units ran if you priced them low enough to get picked up and had transmission availability out of your system, and you know you always got the cheapest energy possible that was offered. If your costs seem too high, build more power lines into your system or help your neighbors build lines that will help you get more cheap stuff. The day of the energy trader making hourly deals like working Wall Street ended in MISO.

It was gigantic gain in efficiency, as you no longer run your own units up and down to meet only your local demand. Cheap units run pretty much full power all day, and the load increases and decreases at scales large enough that it is unlikely that the stack will land on any one unit and run it up and down for hours. Increased demand flies through the stack quickly enough that your unit running at minimum waiting get picked up gets the signal to go to full and then stays there for hours.

MISO is not the only one playing the LMP game in the U.S., but they are the largest and did it best first (though not without some noteworthy hiccups worth another post another day). And while efficiency was vastly improved, it did not address any of the problems of meeting demand when things really go bad on the system, or how to manage the somewhat uncontrollable "green power" resources such as wind and solar. They still flex the system to work around those, and stress the system when a big contingency hits somewhere. MISO's LMP manages generation pretty well, but what it doesn't really do at all is manage the other side.... it doesn't manage load. Imagine how much nicer it would be if you could prevent donut riots on the other side of the hills by magically influencing how badly those people wanted donuts on any particular day? That is the foot in the door for the Smart Grid.

Another post, another day.

So..... did any of this make sense?


No comments:

Post a Comment