Talk:Simulation Environment

Length of a timeslot, length of simulation
There's been some debate about the length of a timeslot. Energy trading markets typically use timeslots, but the length varies between an hour and 5 minutes. The 15-minute number is not decided, and in fact the whitepaper says 1 hour. We must make a tradeoff between simulation realism and the ability to run a competition.

The basic plan is that the "game" will run as an alternating series of Contracting and Execution phases, and that each Execution phase will be at least one week long in order to have both weekdays and weekends. So an Execution phase could be as short as 168 timeslots, or as long as perhaps 1500 timeslots.

Grampajohn 01:10, October 19, 2010 (UTC)

The choice of 15 minute time slots came naturally to me for two reasons:

1) The Thesis I used as base for the implementation of the Household Models ("Smart Home Vision") had 15-minutes timeslots as its quantum of time for the house appliances functions as well as for the person models.

2) In an implementation of Smart Meters API in a thesis of an undegrad student, 15-minutes timeslots were used as quantum of time for changing energy cost prices, sent to the meters, by the Energy Supplier for energy incentive shifting of the appliances.

Of course this is not permanent, it can be always changed if it's decided. I can create aggregated time slots of 1 hour for example. It will take some time to fix it but it's not difficult.

Chrysopoulos18 11:10, October 27, 2010 (UTC)

Individual models and Population models
According to the TAC Energy Architecture page, the simulation must handle populations of a few hundred thousand households - a small to medium-sized metro area. It's hard to see how we could get away with running over 100 000 individual household models during a simulation run. Instead, we have postulated that the tariff customers will be represented as "population models" -- each representing a cluster in some multi-dimensional space of attributes and preferences that affect power consumption, production, preferences, and price response. So the question: is the some way we can take these individual models and aggregate them into population models?

Grampajohn 01:15, October 19, 2010 (UTC)

Now that I have implemented successfully (I think) the household models, I can create an aggregated model of a village (with 100 houses) for example or that of a city (1000 houses). This work can be made off-line without encumbering the server. Of course we will have to agree on the data that will be extrected from that models and the size of the villages, cities etc. I am waiting for your feedback on this matter

Chrysopoulos18 11:18, October 27, 2010 (UTC)