Do Brokers need to see total net load/timeslot?

classic Classic list List threaded Threaded
6 messages Options
Reply | Threaded
Open this post in threaded view
|

Do Brokers need to see total net load/timeslot?

grampajohn
Administrator
Issue #431 says brokers are supposed to be informed of total net load/timeslot, and that it's called out in Section 3.2 of the spec. But it's not in Section 3.2. I don't know what I was smoking back in November when I entered this issue, but unless someone makes a convincing argument why brokers really need to know this, I'm going to close it.

I invite your feedback. Thanks.

John
M
Reply | Threaded
Open this post in threaded view
|

Re: Do Brokers need to see total net load/timeslot?

M
It is customary that following information is publicly available:
a) total net load/timeslot (transmitted after the timeslot closes or with a one timeslot delay).
b) DUs forecast of the total net load for all of the timeslots for the day-ahead (e.g. 24 timeslots ahead or 36 timeslots ahead refreshed at every noon)

a) because brokers use the ratio of their load/total load to determine their change in market share (in reality all the tariffs are not publicly available, e.g. all negotiated contracts with large customers)
a.2) brokers can use the total load to forecast their future consumption
a.3) brokers can forecast future total load based on the present one & weather forecast and it can give insights on the directions of both wholesale and balancing markets on which brokers can profit.

b) DUs forecast can be more reliable than the one brokers make because DU has more information available. It is public due to public awareness of the consumption in the area of the DU, not favoring the market participants that are already present in the market, check-ups for various cost calculations and other.
Reply | Threaded
Open this post in threaded view
|

Re: Do Brokers need to see total net load/timeslot?

grampajohn
Administrator
M wrote
It is customary that following information is publicly available:
a) total net load/timeslot (transmitted after the timeslot closes or with a one timeslot delay).
b) DUs forecast of the total net load for all of the timeslots for the day-ahead (e.g. 24 timeslots ahead or 36 timeslots ahead refreshed at every noon)
I can see how to do (a) very easily. The question is how sophisticated should the predictions be? It's easy to do some sort of time-series prediction, for example using some variant of exponential smoothing. Another alternative would be to construct a 1x168 array to represent the hours for a full week, and accumulate mean consumption values to use as predictions. We could then keep a current deviation from the long-term mean using something like exponential smoothing. Is there a preferred method for making this prediction?

Thanks.

John
Reply | Threaded
Open this post in threaded view
|

Re: Do Brokers need to see total net load/timeslot?

grampajohn
Administrator
In reply to this post by M
M wrote
b) DUs forecast of the total net load for all of the timeslots for the day-ahead (e.g. 24 timeslots ahead or 36 timeslots ahead refreshed at every noon)
I wonder if we really need this in the simulation environment. Due to the relatively brief period represented by a simulation, all brokers are present from the beginning. At that point they have the actual usage data for the previous two weeks. If we give them the actual net usage in every timeslot, they would have the freedom to choose their own prediction models. I assume brokers in the real world would also do this once they have accumulated enough data.

For now, I think I will implement the total-load notification, and wait for further discussion on the need for predictions.

John
Reply | Threaded
Open this post in threaded view
|

Re: Do Brokers need to see total net load/timeslot?

grampajohn
Administrator
In reply to this post by grampajohn
The server now broadcasts a DistributionReport in each timeslot, after all the customer models have run, giving the total production and total consumption reported by the customer models in that timeslot.
M
Reply | Threaded
Open this post in threaded view
|

Re: Do Brokers need to see total net load/timeslot?

M
There is no preferred method for making that prediction.
Brokers in real-world should be able to create their own forecasts for their own purposes. Some brokers will not see the need to use total load forecast at all. Some will develop methods to do it with higher accuracy than DU does it.
Total load prediction ahead depends in which direction you want to emphasize agent development. If it is done very good by DU, or if it is available, competitors will more likely concentrate on other things (whatever improves their chances of winning).