Neighborhood Power Simulations

I'm taking a class on smart grids this semester and a good chunk of the classwork is actually working on a large project. I'm working with a good friend of mine who specializes in networking and we are trying to study the effect of residential power customers (normal people like you and me) and how we all might be more involved in helping the power grid run more efficiently and economically. One of the ways this can be done is to limit the amount of energy being consumed during the periods of peak demand, specifically those hot summer days when everybody's air conditioner is running like mad.

To study this more in detail we are using a relatively new simulator called Gridlab-D.  This simulator is, in a word, quite amazing.  It was written by Pacific Northwest National Lab  under a grand from the DOE specifically to address the problems with simulating in a smart grid environment.  Smart grids are tricky because they break a lot of the traditional thinking about how the grid works.  With the old power system, the power flows from the generators to the customers and whenever the customer wants more power, it is available on demand.  The new smart grid paradigm allows for customers to become something closer to peers in the relationship: solar panels on the roof or electric vehicles in the garage putting energy back onto the grid, customers changing their consumption to save money as the prices or energy changes throughout the day, customers voluntarily limiting their AC use during peak periods for a price break the rest of the year, charging of electric vehicles being co-ordinated among owners (by computers) so that large surges in power don't occur at in-opportune times, all of these can allow the grid to operate more efficiently but require much planning and development if they are to become a reality.

And that's where the academic nerds like me come in.  With Gridlab-D, we can simulate these kinds of situations and see what happens on the power grid.  Specifically, if somebody has a promising idea about one particular aspect of the smart grid concept, it is likely that it can be programmed into Gridlab-D and we can run a simulation to evaluate the idea.

To do this, Gridlab-D has been provided with incredibly detailed models of neighborhood power grids. These models include things like:

• Number of houses in the neighborhood
• Size of houses
• Insulation level of the houses
• Thermostat set-points for the houses
• Efficiency of the air-conditioners
• Number and size of windows in each house
• ...

Almost any type of house you could imagine can be set-up in this simulator.  The houses that come "pre-built" have all of these variations built in so that they mimic the actual variations of houses in a neighborhood.

And the weather.  Gridlab-D allows the user to define which set of recorded weather data to be used so that the neighborhood can be simulated as if it is in Wichita or Chicago or Miami.  You can run the simulation for a day or for a year. How much sunlight hits the houses, how the wind blows, all of it is there.

I'm just getting my feet wet with this simulator and managed to get a simple simulation up and running today.  I took one of the built-in neighborhoods and subjected it to the summer weather here in Wichita, running the simulation four times, each with a different thermostat set-point applied to all the houses in the neighborhood.  To see what happened, I looked at the amount of power the neighborhood consumed throughout the day and I looked at the number of air-conditioners running at any given time.  Here are the pretty pictures:

As you can see from the graph above, the power consumption in the neighborhood goes up as the heat of the day sets in and then backs off as the evening cools down.  Also note that more power is consumed if the entire neighborhood set their thermostats to 73 'F (blue line) vs (81' F).  Going further, we can see that the biggest difference in energy consumption between these two cases is from 8am to noon.  That's kind of interesting.

Looking at the number of running AC units we can see a similar phenomenon.  As the day heats up, more AC units are running and that lowering the thermostat causes more units to run, particularly after 8am. 

Just to make the nerd case clear, this is all being run in simulation.  The simulator is taking the weather data and the house data, calculating how fast the house is heating up, calculating when the air-conditioning turns on, calculates how much energy it is going to consume and does this all for all the houses for the entire day.  That I, with my limited programming ability, could get this all up and running in a week or so is a great indication of just how powerful this simulator is.

Now that I've validated that I've got the simulator running, its time for the real work to begin on our project.  I'm looking forward to this.