Controllable Capacity

classic Classic list List threaded Threaded
24 messages Options
12
Reply | Threaded
Open this post in threaded view
|

Controllable Capacity

grampajohn
Administrator
Dear colleagues -

I have been working on a writeup on the treatment and representation of controllable capacities for Power TAC. This is a critical feature for our simulation model, and it's time we nailed down the details so we can start implementation. It is also a prerequisite for the next phase of the DU balancing market, which is supposed to be able to exercise capacity controls in order to achieve balance.

I invite your input on this issue, as well as your thoughts on how to get it implemented.

Thanks much.

John
Reply | Threaded
Open this post in threaded view
|

Re: Controllable Capacity

Mathijs de Weerdt
Hi John,

Thanks for putting this down so clearly. I'll take it into account in my edit of the balancing section in the spec.

Cheers,
  Mathijs


On Jan 25, 2012, at 4:58 AM, grampajohn [via Power TAC Developers] wrote:

> Dear colleagues -
>
> I have been working on a writeup on the treatment and representation of controllable capacities for Power TAC. This is a critical feature for our simulation model, and it's time we nailed down the details so we can start implementation. It is also a prerequisite for the next phase of the DU balancing market, which is supposed to be able to exercise capacity controls in order to achieve balance.
>
> I invite your input on this issue, as well as your thoughts on how to get it implemented.
>
> Thanks much.
>
> John
>

Reply | Threaded
Open this post in threaded view
|

Re: Controllable Capacity

kvalogianni
In reply to this post by grampajohn
Hi John,

indeed the implementation described covers the major aspects related to controllable capacities. I think the separation into two "virtual" customers is very important as it will allow us to handle separately with the controllable part of the customer's capacity.

Regarding the PowerType, I agree with you that the first option makes more sense. Another option would be the third by I don't know what problems could arise from that..

Finally, with respect to the tariff structure, I think the maximum amount of interruption and the maximum duration are for sure needed. Another attribute would be to prioritize the appliances and in the tariff declare  till which appliance's priority the DU is authorized to interrupt.

Best,
Konstantina
Reply | Threaded
Open this post in threaded view
|

Re: Controllable Capacity

grampajohn
Administrator
Konstantina wrote
Finally, with respect to the tariff structure, I think the maximum amount of interruption and the maximum duration are for sure needed. Another attribute would be to prioritize the appliances and in the tariff declare  till which appliance's priority the DU is authorized to interrupt.
It's an interesting idea, but I don't know what you would do with it in a population model. But perhaps we need to free up our thinking on tariff terms a bit? Currently, what can a broker do to encourage a customer to accept external control? The only option is to offer a lower rate. In my personal case, we have a heat pump and water heater on interruptible service, and we have a tiered rate for that: the base rate is a 10% discount over the non-interruptible rate; above 400 kwh in a month, it drops to half the standard rate.

Perhaps a broker could offer a rebate for each interruption? Or a lower rate for some period of time, or some amount of energy, following an interruption? Are there real-world examples anyone knows of that don't require serious bookkeeping?

Just a thought...

John
Reply | Threaded
Open this post in threaded view
|

Re: Controllable Capacity

kvalogianni
True, in higher level of abstraction it's possible that it won't be that effective..

In case the broker offers lower rate for some time-period following the interruption, the DU has to take into account both the demand reduction in the timeslot of interruption and the possible demand increase in the lower-rate timeslot..
It could be interesting but as a more sophisticated tariff-offering I guess..
My opinion is to have as base the first suggestion which will provide us direct observation of the tariff results and then elaborate more the "controllable" tariff to create more advanced scenarios..

Any other ideas/suggestions?

Best,
Konstantina
Reply | Threaded
Open this post in threaded view
|

Re: Controllable Capacity

grampajohn
Administrator
I think we should be wary of adding complexity in this area. Let's see if we can get along with the existing tariff structure until someone finds a compelling case for enhancement.

Make sense?

John
Reply | Threaded
Open this post in threaded view
|

Re: Controllable Capacity

kvalogianni
Indeed..I think the existing tariff structure is good for the beginning.

--Konstantina
Reply | Threaded
Open this post in threaded view
|

Re: Controllable Capacity

grampajohn
Administrator
In reply to this post by grampajohn
I think it's time to start implementation of the controllable-capacity feature. I have updated the writeup over on the GitHub wiki - see the section at the bottom on Tariffs and Rates. Basically the proposal is to add a "shiftability" ratio to the PowerType representation, and to add a "curtailment" ratio to the Rate structure, which would apply only to tariffs for controllable PowerTypes. That would be in addition to the two new events and the updates to the customer models.

Please get your comments in over the next few days; I would like finalize this and get started by this weekend.

Cheers -

John
Reply | Threaded
Open this post in threaded view
|

Re: Controllable Capacity

kvalogianni
Hi John,

I think the overall  description for the implementation of Scenario II is complete.
Just a clarification: the EconomicControlEvent is basically exercised by the broker, so as to have a balanced portfolio, and the BalancingControlEvent is exercised by the DU for balancing, right?

As for the notice interval, if we want to be realistic I think we should add it.  On the other hand, as the curtailment is a predefined percentage on the given timeslot, we can assume that the contracted customer doesn't need any notice and this capacity is available (it's part of the tariff to have it available, so no notice needed). Do we have any idea how this is working in the reality?


Cheers,
Konstantina
Reply | Threaded
Open this post in threaded view
|

Re: Controllable Capacity

grampajohn
Administrator
Konstantina wrote
I think the overall  description for the implementation of Scenario II is complete.
Just a clarification: the EconomicControlEvent is basically exercised by the broker, so as to have a balanced portfolio, and the BalancingControlEvent is exercised by the DU for balancing, right?
Right, although it's equally likely that the broker would decide to curtail during periods when the wholesale cost is unusually high.
As for the notice interval, if we want to be realistic I think we should add it.  On the other hand, as the curtailment is a predefined percentage on the given timeslot, we can assume that the contracted customer doesn't need any notice and this capacity is available (it's part of the tariff to have it available, so no notice needed). Do we have any idea how this is working in the reality?
I don't know what the practice is everywhere, but we have two controllable devices in our home. The water heater is mostly shiftable, and the heat pump is mostly discretionary (we also have a gas furnace and a wood-burning stove). There is no warning given for curtailment - if you notice it's getting chilly, you have to go to a website to see that it's curtailed. The gas furnace does not kick in until the temperature has dropped by 1 C for over 45 minutes. As far as I can tell, it's only curtailed as a cost-control measure, and not more than once or twice/month.

John
Reply | Threaded
Open this post in threaded view
|

Re: Controllable Capacity

Prashant Reddy
In reply to this post by grampajohn
Hi John,

Regarding this part:
---
[The Distribution Utility] must exercise cost controls before the
customer models run for that timeslot, and balancing controls after
the customer models run. This means that the result of operating the
controls must be communicated not only to the brokers, but also back
to the customers, which in turn will need to issue additional tariff
transactions for each affected tariff subscription, to correct for
over or under-reporting in the original transactions.
---

I don't understand the rationale for the last part of the technical
design here.  Why doesn't the DU itself issue the correcting tariff
transactions and simply notify the affected customer models about
them?

Thanks,
Prashant


On Wed, Feb 15, 2012 at 1:25 PM, grampajohn [via Power TAC Developers]
<[hidden email]> wrote:
>
> Please get your comments in over the next few days; I would like finalize
> this and get started by this weekend.
>
Reply | Threaded
Open this post in threaded view
|

Re: Controllable Capacity

grampajohn
Administrator
Prashant Reddy wrote
...  the result of operating the
controls must be communicated not only to the brokers, but also back to the customers, which in turn will need to issue additional tariff transactions for each affected tariff subscription, to correct for over or under-reporting in the original transactions.
---
I don't understand the rationale for the last part of the technical design here.  Why doesn't the DU itself issue the correcting tariff transactions and simply notify the affected customer models about them?
That's a good question, Prashant. I see no reason why it would not work that way, and it might simplify the customer models. However, the customer models already have to deal with economic controls, and they already have to have the ability to create the market transactions (through the TariffSubscriptions), and your suggestion would require that the DU also have that ability. Does that create additional, unnecessary coupling? Is there an offsetting efficiency gain? I'll have to think about it. Maybe it's time to work up some interaction diagrams.

Thanks -

John
Reply | Threaded
Open this post in threaded view
|

Re: Controllable Capacity

Prashant Reddy
On Fri, Feb 17, 2012 at 3:54 PM, grampajohn [via Power TAC Developers]
<[hidden email]> wrote:

> Prashant Reddy wrote
> ...  the result of operating the
> controls must be communicated not only to the brokers, but also back to the
> customers, which in turn will need to issue additional tariff transactions
> for each affected tariff subscription, to correct for over or
> under-reporting in the original transactions.
> ---
> I don't understand the rationale for the last part of the technical design
> here.  Why doesn't the DU itself issue the correcting tariff transactions
> and simply notify the affected customer models about them?
>
> That's a good question, Prashant. I see no reason why it would not work that
> way, and it might simplify the customer models. However, the customer models
> already have to deal with economic controls, and they already have to have
> the ability to create the market transactions (through the
> TariffSubscriptions), and your suggestion would require that the DU also
> have that ability. Does that create additional, unnecessary coupling? Is
> there an offsetting efficiency gain? I'll have to think about it. Maybe it's
> time to work up some interaction diagrams.
>

Additional two cents:

The economic controls can be passed along to the customer model as a
parameter during its timeslot activation and they simply become a
factor in computing the capacity for the timeslot.  If the balancing
controls require customer action, under the current design it would
imply reactivation of the customer model for the same timeslot at a
later phase.  If on the other hand, if it was just a notification, it
could be differed until the next timeslot activation and passed along
as another parameter in that activation.  So, from a customer model's
perspective, I'd simply take economic controls for this timeslot and
balancing controls from the last timeslot and factor them into my
capacity calculation -- I like that symmetry.  Of course, this would
require the TimeslotPhaseProcessor callback signature to change to
allow additional parameters.

Another option is to send the tariff transactions from balancing
controls via a separate notification callback to the customer model,
and then the customer model may choose to simply ignore the
notification if the balancing control has no impact on its future
capacity (i.e., pure controllability, no shiftability).  If it was
required to issue the transactions itself, it couldn't achieve this
simplification.

On the other hand, you do raise a good point about possible extra
coupling in the DU, but really all a customer model does to issue a
tariff transaction is "subscription.usePower()" -- it wouldn't be much
code duplication I think?  Maybe a little bit of added complexity in
the DU would be worth simplification in the customer models since
there will be many customer models vs. one DU.

Anyway, that's the extent of my thinking on the topic.  As you say,
it's a tradeoff and interaction diagrams may indeed help.  Thanks for
considering it.

Cheers,
Prashant
Reply | Threaded
Open this post in threaded view
|

Re: Controllable Capacity

achryso
Hello there. I would like to put some issues on the table if you don't mind from my bottom-up approach of customers point of view.

Generally speaking, interruptable / curtailable capacities are going to increase complexity a lot. In my models, there are certain appliances that can shift their loads. Some of them are easier to handle than others. Also, the shiftable load in order to remain realistic have to stick in with the comfort parameter where it is plausible. This must be done in the beginning of the timeslot but is not as easy as moving a number in another timeslot, because there are actually appliances operating and consuming.

So I would like to ask:

1) In both descriptions of Economic and Balancing Control Events, the customer is obliged to move the load to another timeslot or move it to the current one. How is the comfort going to be part of the customers decision if he is obliged to do the shifting?

2) After talking with Konstantina over this matter, I have suggested the priority of the appliances. This way the customer can move over the loads of the appliances that he care less about, rather than arbritrarily moving appliances over.

3) If we wanted to be as realistic as possible, this would be a maximization problem. Each time the customer has to shift loads, he would have to find which shiftable loads he should utilize to make the saving in cost and minimizing the taint in customers' comfort.

Which of these should be taken into consideration in the implementation of the computations?

Also, I would like to say that I agree with Prashant over the matter of who will handle the TariffTransactions. It would not be easy for the customers to do that in any way that I can think of.
I should note also that I don't know if this can be done on a later phase of the same timeslot in my modelling pattern...

Let me know what you think.

Thank you in advance.
Reply | Threaded
Open this post in threaded view
|

Re: Controllable Capacity

grampajohn
Administrator
In reply to this post by grampajohn
I have made some updates to the proposal that I hope will simplify the customer model's responsibilities, and I have added sequence diagrams to illustrate how I think this could all work. In our group meeting yesterday, we came up with essentially the same solution Prashant suggests, which is to have the Customer models process EconomicControlEvents for the current timeslot and BalancingControlEvents from the previous timeslot more or less together, during normal activation. As Prashant noticed, it's not the customer that issues TariffTransactions anyway; it's the TariffSubscription, and the DU already needs to know about TariffSubscriptions.

In order to simplify representation and processing, I am suggesting that the Tariff simply specify a maximum curtailment percentage for a given timeslot, and so a Broker could issue a few tariffs for controllable capacity that trade off discount against curtailment percentage. Obviously, if a tariff for a completely shiftable load is curtailed by 50% in three successive timeslots, then the second and third (and fourth) would see the shifted load as well as the original load, and the load after curtailment in the second and third timeslots would represent 50% of the original load plus 50% of the shifted load.

Does this make sense?

John
Reply | Threaded
Open this post in threaded view
|

Re: Controllable Capacity

grampajohn
Administrator
In reply to this post by achryso
achryso wrote
Hello there. I would like to put some issues on the table if you don't mind from my bottom-up approach of customers point of view.
Thanks for your input. Your point of view is very important on this, of course.
Generally speaking, interruptable / curtailable capacities are going to increase complexity a lot. In my models, there are certain appliances that can shift their loads. Some of them are easier to handle than others. Also, the shiftable load in order to remain realistic have to stick in with the comfort parameter where it is plausible. This must be done in the beginning of the timeslot but is not as easy as moving a number in another timeslot, because there are actually appliances operating and consuming.

So I would like to ask:

1) In both descriptions of Economic and Balancing Control Events, the customer is obliged to move the load to another timeslot or move it to the current one. How is the comfort going to be part of the customers decision if he is obliged to do the shifting?
To keep things simple, I have suggested that the tariff specify a maximum curtailment/timeslot. It would be up to the customer to evaluate their utility for comfort vs. price discount, given the expected impact of curtailments (which are seldom likely to reach the maximum).
2) After talking with Konstantina over this matter, I have suggested the priority of the appliances. This way the customer can move over the loads of the appliances that he care less about, rather than arbritrarily moving appliances over.
It might also be helpful to consider the tradeoff between curtailing shiftable vs. discretionary loads.
3) If we wanted to be as realistic as possible, this would be a maximization problem. Each time the customer has to shift loads, he would have to find which shiftable loads he should utilize to make the saving in cost and minimizing the taint in customers' comfort.

Which of these should be taken into consideration in the implementation of the computations?
This would only matter if the customer had a choice to curtail shiftable vs. discretionary loads. Otherwise, at least in terms of the visible consumption/production behavior of the customer, it's hard to see how it would matter. The loss in convenience is paid for, presumably, by the reduction in rate, otherwise the customer would not have subscribed. Is this not correct?

Cheers -

John
Reply | Threaded
Open this post in threaded view
|

Re: Controllable Capacity

grampajohn
Administrator
In reply to this post by grampajohn
I have created Issue #493 to track the changes needed to implement controllable capacity. Please use this issue in commit messages if you are working on the controllable capacity problem.

Thanks much.

John
Reply | Threaded
Open this post in threaded view
|

Re: Controllable Capacity

grampajohn
Administrator
In reply to this post by achryso
Antonios has suggested some potential problems with implementation of controllable capacities:
achryso wrote
Generally speaking, interruptable / curtailable capacities are going to increase complexity a lot. In my models, there are certain appliances that can shift their loads. Some of them are easier to handle than others. Also, the shiftable load in order to remain realistic have to stick in with the comfort parameter where it is plausible. This must be done in the beginning of the timeslot but is not as easy as moving a number in another timeslot, because there are actually appliances operating and consuming.
Curtailment represents externally-controlled shutoff of power to certain loads. It is not the same as price-based shifting, which I believe you have already implemented. Most importantly, it only applies to loads that can be shut down without warning and with minimal impact on customer convenience. If you have a reasonably large electric water heater (200l or more), you would most likely not notice that it was shut off for a timeslot, but if it gets shut off, then its consumption will increase in the following timeslot. This is a "shiftable" load. The same is true of a heat pump or air conditioner, although you might notice a small change in temperature. However, at least in my part of the country, in order to have a heat pump set up for remote curtailment, you have to have a backup heat source, in my case a gas-fired furnace. So if the heat pump is curtailed for long enough to drop the temperature to an uncomfortable level, the furnace comes on, and the heat-pump load does not get shifted. This is a "discretionary" load.
1) In both descriptions of Economic and Balancing Control Events, the customer is obliged to move the load to another timeslot or move it to the current one. How is the comfort going to be part of the customers decision if he is obliged to do the shifting?
As I said, the key is to limit curtailment to appliances that can be interrupted with minimal impact on customer convenience, and to limit the proportion of the time during which they can be interrupted. The interpretation of that proportion is that for a single household, it is the maximum "off" duty cycle, or for a village it is the proportion of appliances that can be off at once. So if the village has some number of water heaters that can be off 50% of the time, then if they get a control message specifying full curtailment of 50%, half of the water heaters would be shut off for that timeslot. If the next timeslot is also curtailed by 50%, then the other half of the water heaters would be shut off, and the half that we off in the previous timeslot would come back on. Their consumption after being turned back on would presumably be (almost all of) the consumption that was curtailed in the previous timeslot plus the normal consumption for the current timeslot.
2) After talking with Konstantina over this matter, I have suggested the priority of the appliances. This way the customer can move over the loads of the appliances that he care less about, rather than arbritrarily moving appliances over.

3) If we wanted to be as realistic as possible, this would be a maximization problem. Each time the customer has to shift loads, he would have to find which shiftable loads he should utilize to make the saving in cost and minimizing the taint in customers' comfort.

Which of these should be taken into consideration in the implementation of the computations?
As I said, the set of controllable appliances is typically a small subset of the appliances that a customer might shift due to prices. They would be connected to a different meter, and potentially be subscribed to different tariffs.
Also, I would like to say that I agree with Prashant over the matter of who will handle the TariffTransactions. It would not be easy for the customers to do that in any way that I can think of.
I should note also that I don't know if this can be done on a later phase of the same timeslot in my modelling pattern...
Yes, that's what we decided also.

Thanks for your input. Please let me know if this is more clear to you now.

John
Reply | Threaded
Open this post in threaded view
|

Re: Controllable Capacity

achryso
I think that now I can see more clearly what is going to be. So I will have to add an variable in my appliance models that will say how much time they can shut down due to external control (for a example the refrigerator can be shifted without spilling the customers convenience for half hour). Then if the appliance is switched off externally for this duration (or less) I will move the load acoordingly to the next time slots.

The question is, if it is wanted to be shifted for more time than that, I will not accept the whole duration but only the one that is in my convenience criterias, or I will not accept the shifting at all?

Also, the tariff evaluation is going to be kind of tricky. because you really don't know how much you will save from shifting the appliances since you don't know when and which is going to be shifted during the day.

In order to continue with my implementation of the customers' side of the controllable capacity I will need to have the tariff implementation ready in order to test my code. I guess that someone is working on it right now...
Reply | Threaded
Open this post in threaded view
|

Re: Controllable Capacity

grampajohn
Administrator
achryso wrote
I think that now I can see more clearly what is going to be. So I will have to add an variable in my appliance models that will say how much time they can shut down due to external control (for a example the refrigerator can be shifted without spilling the customers convenience for half hour). Then if the appliance is switched off externally for this duration (or less) I will move the load acoordingly to the next time slots.

The question is, if it is wanted to be shifted for more time than that, I will not accept the whole duration but only the one that is in my convenience criterias, or I will not accept the shifting at all?
My thought is that each tariff for controllable capacity will simply specify a maximum fraction that can be curtailed. For a population model, or even a model with perhaps 10 or more individual households, you could think of a curtailment affecting the specified fraction of the customer base during the timeslot, so a 10% curtailment would curtail 1/10 of the households for the entire timeslot. If a 10% curtailment were in effect for the following timeslot, it would affect a different 10%. This probably makes sense for Prashant's statistical models.

Another way to think about curtailment might make more sense for your models, and that is to curtail the last x% of the timeslot. So a 10% curtailment would shut off power for the last 6 minutes of the timeslot for every customer subscribed to that tariff. Naturally, if it's a shiftable load, that would presumably increase consumption during the following timeslot.

Once you have subscribed to a tariff that allows for curtailment, you cannot refuse it; the controllable power runs through a separate meter that's controlled remotely. If your refrigerator is plugged into that circuit, it will be curtailed when a control event is processed. But in the cases I know about, household refrigerators are never curtailable.

With my co-op, only heat pumps, water heaters, and resistance-heating units are allowed to be connected to such a meter for a residential customer. Farm and commercial/industrial customers have more flexibility, but in general only loads that have significant heat storage or that have alternate energy sources are allowed. This means, for example, that the curtailment rules for heat pumps are different in the summer than the winter, because in the winter you can use another heat source. So during the heating season, heat pumps can be shut off for up to 6 hours, but in the summer it's something like 30 minutes maximum.
Also, the tariff evaluation is going to be kind of tricky. because you really don't know how much you will save from shifting the appliances since you don't know when and which is going to be shifted during the day.

In order to continue with my implementation of the customers' side of the controllable capacity I will need to have the tariff implementation ready in order to test my code. I guess that someone is working on it right now...
First of all, for a heat-storage device, you get at least most of the heat you need regardless of curtailment, as long as you don't agree to curtailment amounts so high that you cannot maintain close the the desired temperature. So if your water heater runs at most 50% of the time even when you are using hot water at a reasonable rate (taking a shower, washing laundry), then a 50% curtailment won't be very noticeable, ever.

So my expectation is that brokers will offer two or three different tariffs for controllable capacities, specifying different curtailment ratios and different discounts. It's then up to the customer to evaluate the tradeoffs, recognizing that brokers are unlikely to use the entire curtailment amount very often because that will limit their ability to use curtailment for balancing. In fact, a simple model of broker and DU behavior would suggest that if the broker is using curtailment to avoid balancing charges, they might on average use half the curtailment amount, and that only when wholesale prices are high enough to make the discounted price unprofitable.

I am hoping to have time to work through the rest of the representation issues this weekend. I have already written and tested an update to Rate, but I'm still looking at PowerType.

Does this help?

John
12