Difference between revisions of "Talk:Game Design"

From powerwiki
Jump to: navigation, search
(Prices for external balancing)
(Prices for external balancing)
Line 46: Line 46:
  
 
We propose to model the possibility of a DU to complement (or reduce) production (in order top prevent blackouts) by a marginal cost function such as in the example. In principle the DU plans to operate at the most efficient production level, but in case of imbalances the DU needs to adjust its production (see figure).
 
We propose to model the possibility of a DU to complement (or reduce) production (in order top prevent blackouts) by a marginal cost function such as in the example. In principle the DU plans to operate at the most efficient production level, but in case of imbalances the DU needs to adjust its production (see figure).
 +
 +
[JEC:] Some quick thoughts:
 +
* Regarding #2 - the actual imbalance is not the broker's private information, it's computed by the Distribution Utility. The cost curve is the important bit of private information, I believe.
 +
* Regarding #3 - The other "ancillary service" available is to reduce or increase imports through the wholesale market. The cost curves for that should be available in the records. Derivation of those cost curves is covered pretty well in Shahidehpour Ch. 9.
 +
* Regarding #5 - Social welfare is more important than budget balance, because the Distribution Utility is presumably the one running the market, and it is charging the brokers for services anyway. Any imbalance can be transferred to that account, I would think.

Revision as of 15:59, 21 December 2010

Sequence of days for timeline

I suggest that we think about the pattern of weekdays/weekends when deciding on which days to simulate. If we skip by a multiple of 7, then every day would be the same day of week; if we skip by 8, we would have a cycle of 5 weekdays and 2 weekend days. If we skip by 12, we would have a cycle of M-Sat-Th-Tu-Sun-Fri-W so weeks would alternate between 4 days and 3 days. Would it be better to just pre-select the days to get something consistent like 4-day weeks with one weekend day? Grampajohn 18:09, November 3, 2010 (UTC)

I do like the idea of having a consistent number of week days and weekends periodically. Skip by 8 seems like a good way to do this. The only drawback, compared to skipping by 12 days, is that we will only be able to simulate about 1.3 years instead of 2 years, so we may not capture seasonal dynamics as well. We could also skip by 15 and get the same 5+2 pattern and simulate a longer time overall. But, would only two days per month be too sparse a sample? PReddy 18:19, November 3, 2010 (UTC)
A cycle of 12,12,12,13,12,12,12,13,... gives you a 49-day cycle that could run FWMSFWMS...
Grampajohn 21:04, November 3, 2010 (UTC)

Blueprints needed for each of the "Major Scenarios"

Once we understand how the game progresses in enough detail to understand how the game works, we need to take each of the scenarios and detail them in enough technical detail to begin writing an implementation. As these are written, the scenarios need to link to the blueprints so we know they are done. Grampajohn 14:20, November 4, 2010 (UTC)

What is the initial state of the market?

I have heard two proposals for the "initial state" of the market:

  1. All customers are subscribed to "generic" fixed-price tariffs, which are evenly distributed among the competing Brokers. Each Broker must immediately begin serving its portfolio, while it develops and offers new tariffs, and negotiates contracts with larger customers.
  2. Brokers begin with no tariffs, but instead the Distribution Utility holds fixed-price tariffs with all customers. This is a typical state for a new retail market - see Joskow's "Lessons Learned" paper for examples. The key is that the "default" tariffs are fairly unattractive.

In the first case, the Broker may need some leadtime, perhaps 24 hours, to retrieve forecasts and build up a market position in the wholesale market before it must serve its load base.

Grampajohn 23:53, November 7, 2010 (UTC)

After discussion among Wolf, Carsten, Prashant, John, and the Minnesota group, we have decided that for the first year's competition we will take the second approach - brokers start with empty portfolios, and must attract customers away from the "incumbent."

--Grampajohn 05:26, 22 November 2010 (CET)

RTD and RTC modeling

(JDcosta asks: Is the Execution (shorter time interval) and the Contracting (relatively longer look-ahead time horizon) identifiable with the Real Time Dispatch (RTD) over 5-10 minute interval) and Real Time Commitment (RTC over a 1-3 hr ) in a Real Time (as well as a Day -Ahead scheduling optimized over hourly intervals over a 24 hr period ) Market operation of a market design referenced on (for example) page 1964 of the reading material assigned for today's meetup (Chow et al)? (except perhaps for convolution with a time contraction/dialation operator to adjust for the relative differences in the magnitudes of the time periods involved in the proposed Game Design and the Market Design presented in the paper) [1]

No, we are abstracting away this distinction for the simulation. Because we are concentrating on the distribution side of the market, the only effect of RTD is (potentially) some short-term price variations. However, the simulation is running a one-hour timeslot in 5 seconds, so a resolution of 10 minutes is unrealistic. We treat the one-hour timeslots as discrete quanta, and only balance overall supply and demand over the full timeslot.

Grampajohn 04:04, November 8, 2010 (UTC)

Prices for external balancing

[mdeweerdt:] For a good "mechanism design" we first need to come up with some properties that we would like to the mechanism and the imbalance penalty/fee to have. Do you agree with the following?

  1. Truth telling should be a dominant strategy in the real-time balancing game, because this will reduce complexity for the brokers: they cannot do better than submit their private information truthfully. We want to avoid having a complete secondary market.
  2. Private information in this balancing game consists of: the actual (im)balance and a (marginal) cost curve for diverting from this planned consumption/production.
  3. For repairing the imbalance we need to make sure that the aggregate consumption/production is taken into account. This aggregate needs to add up to zero, possibly using some of the spinning reserve of the ancillary service (in our case of the distribution utility (DU)).
  4. This balancing game is repeated every time slot (e.g. 15, 30 or 60 minutes). In principle these games can be dependent because of contracts that allow for shifting load/production to another time slot, or the use of batteries. However, we feel that it may be worthwhile to start under the assumption that this does not happen (that often). Possibly the repeated (sometimes also called dynamic) setting could be scheduled for the second version of PowerTAC. Reference: Dynamic-VCG by Ruggiero Cavallo (2006-2009).
  5. Other nice properties would be cost-efficiency of the system as a whole (i.e. social welfare) and budget balancedness, but these cannot all be attained together with truthfulness.

We propose to model the possibility of a DU to complement (or reduce) production (in order top prevent blackouts) by a marginal cost function such as in the example. In principle the DU plans to operate at the most efficient production level, but in case of imbalances the DU needs to adjust its production (see figure).

[JEC:] Some quick thoughts:

  • Regarding #2 - the actual imbalance is not the broker's private information, it's computed by the Distribution Utility. The cost curve is the important bit of private information, I believe.
  • Regarding #3 - The other "ancillary service" available is to reduce or increase imports through the wholesale market. The cost curves for that should be available in the records. Derivation of those cost curves is covered pretty well in Shahidehpour Ch. 9.
  • Regarding #5 - Social welfare is more important than budget balance, because the Distribution Utility is presumably the one running the market, and it is charging the brokers for services anyway. Any imbalance can be transferred to that account, I would think.