Planner: Incorporation of feedin/charge tradeoff

[kalaws]

If I have scheduled planner to complete charging at X o’clock, say Monday 6 AM and time is now Sunday 10 AM, sometimes it is better (also from the point of view of co2) to export pv surplus rather than charge. For example, feedin price may presently be 0,15, and around midnight grid cost may be 0,05. Hence it is better to postpone charging to midnight hours, and export the existing surplus now. As far as I can see, planner does not currently consider this, but only considers differences in grid cost.

I don’t know whether it would actually be possible to implement this behavior into planner, but perhaps there are others who would like to speculate on how it might be done?

One potential obstacle I can think of is that pv production forecasting might have to be included for this to make sense. Or am I wrong here?

[andig]

I don’t know whether it would actually be possible to implement this behavior

What exactly would that be? A cost function that is a combination (sum or difference) of tariffs?

/cc @premultiply

[kalaws]

I’m not sure how it is properly expressed. I suppose if surplus>charge minimum, then calculate difference between hr 1 feedin price and hr 2 grid cost (hr1price-hr2cost). If difference is positive (above some defined threshold, to not get silly) then schedule charging for hr 2, not hr 1. Something along those lines?

[andig]

I don't understand :/

[kalaws]

Ok, so I'll try to illustrate my thinking.

Let's say 5 hrs of charging is needed to reach a specified SOC in 24 hrs.
Two tariff lists, Grid and Feedin, exist. Whether there is PV Surplus or not determines which one reflects the True cost of energy. Out comes a single tariff list, True cost. This allows us to rank the hours by price. The existing planner logic applies from there, and decides whether to Charge. The difference from the current functioning of the planner is that some charging will be done when there is surplus, some charging will be scheduled for when there is not surplus.

	Grid	Feedin	Surplus	True	Charge
Hr	tariff	tariff	Y/N	cost	Y/N
1	1	0,8	N	1	N
2	1	0,8	N	1	N
3	1	0,8	N	1	N
4	1	0,8	N	1	N
5	1	0,8	N	1	N
6	0,8	0,64	N	0,8	Y
7	0,8	0,64	N	0,8	Y
8	1	0,8	N	1	N
9	1,3	1,04	Y	1,04	N
10	1,2	0,96	Y	0,96	Y
11	1,5	1,2	Y	1,2	N
12	1,5	1,2	Y	1,2	N
13	1,6	1,28	Y	1,28	N
14	1,6	1,28	Y	1,28	N
15	1,5	1,2	Y	1,2	N
16	1,3	1,04	Y	1,04	N
17	1,4	1,12	Y	1,12	N
18	1,2	0,96	Y	0,96	Y
19	0,9	0,72	Y	0,72	Y
20	1,1	0,88	N	1,1	N
21	1,3	1,04	N	1,3	N
22	1,4	1,12	N	1,4	N
23	1,3	1,04	N	1,3	N
24	1,1	0,88	N	1,1	N

In reality, this may be complicated by the fact that it is not necessarily easy to know the non-present state of Surplus, without forecasting. But perhaps this logic can be used on the present hour, when surplus is known?

[kalaws]

Below table illustrates the same solution but without assuming that we can forecast surplus. In this case, 4 hrs are needed to reach the wanted SOC. (Of course, hrs 1-12 before NOW are not really relevant at all.)

	Grid	Feedin	Surplus	True	Charge
Hr	tariff	tariff	Y/N	cost	Y/N
1	1	0,8	N	1	N
2	1	0,8	N	1	N
3	1	0,8	N	1	N
4	1	0,8	N	1	N
5	1	0,8	N	1	N
6	0,8	0,64	N	0,8	N
7	0,8	0,64	N	0,8	N
8	1	0,8	N	1	N
9	1,3	1,04	N	1,3	N
10	1,2	0,96	N	1,2	N
11	1,5	1,2	N	1,5	N
12	1,5	1,2	N	1,5	N
----------------------------------------------
13 NOW	1,6	1,28	Y	1,28	N
----------------------------------------------
14	1,6	1,28	N	1,6	N
15	1,5	1,2	N	1,5	N
16	1,3	1,04	N	1,3	N
17	1,4	1,12	N	1,4	N
18	1,2	0,96	N	1,2	Y
19	0,9	0,72	N	0,9	Y
20	1,1	0,88	N	1,1	Y
21	1,3	1,04	N	1,3	N
22	1,4	1,12	N	1,4	N
23	1,3	1,04	N	1,3	N
24	1,1	0,88	N	1,1	Y

[kalaws]

Perhaps, with the second suggestion, we have stepped outside of what is the point of the planner? I.e., the planner is not concerned with continually evaluating the present hour, but only ranks one time (when a departure hour is set) a list of hours by grid tariff, which is the only tariff that can safely be assumed as long as we cannot forecast pv surplus?

[andig]

Afaikt the planner would have made the exact same plan with grid information only. Still not clear what you‘re asking. You also must not forget that the planner will- since cheap- always charge at max speed. All other use cases are PV mode.

[kalaws]

Ok, so I've done a small Excel experiment on scenarios that can actually happen. It compares the result of factoring in feedin/charge tradeoff vs basing planner on grid tariff as is currently the case.
In one scenario, there is a relatively small tariff difference during the hours. In the other, there is a relatively greater difference. In the first case, the 'new' planner returns a better result, even though the charge cost is actually higher (feedin profits reduces the overall sum); in the second case, results are in favor of the 'old' planner. However, the sum of the differences in these two scenarios is a small profit (6,61-2,54=4,07, roughly 5%) owing to the effects of factoring in the feedin/charge tradeoff.

I still don't know how to do this in practice, but the experiment shows that there is in principle a difference from doing so. I realize that in reality, it may not be possible to achieve.

Edit: Updated tariffs to include fixed charges (less linear behavior)

Scenario 1: PLAN TO CHARGE 5 HRS, SMALL TARIFF VARIATION 
PROFIT NEW vs OLD PLANNER: 6,61			Fixed charges, Grid: 0,945  Feedin: 0,6344  

- New Feedin/charge tradeoff planner
		Grid	Feedin	Surplus	True	Charge	Charged	Feedin	Charge	Feedin	Hourly
Hr	Spot	tariff	tariff	kWh	cost	1/0	kWh	kWh	cost	profit	total
----------------------------------------------------------------------------------------------
1	0,76	1,90	1,39	4,60	1,39	NO	0,00	4,60	0,00	6,41	6,41
2	0,76	1,90	1,39	4,60	1,39	NO	0,00	4,60	0,00	6,41	6,41
3	0,8	1,95	1,43	0,00	1,95	YES	4,60	0,00	-8,95	0,00	-8,95
4	0,8	1,95	1,43	0,00	1,95	YES	4,60	0,00	-8,95	0,00	-8,95
5	0,8	1,95	1,43	0,00	1,95	YES	4,60	0,00	-8,95	0,00	-8,95
6	0,9	2,07	1,53	0,00	2,07	YES	4,60	0,00	-9,52	0,00	-9,52
7	0,9	2,07	1,53	0,00	2,07	YES	4,60	0,00	-9,52	0,00	-9,52
----------------------------------------------------------------------------------------------
						SUM:	23,00	9,20	-45,89	12,83	-33,06


- Old Grid tariff planner
		Grid	Feedin	Surplus	True	Charge	Charged	Feedin	Charge	Feedin	Hourly
Hr	Spot	tariff	tariff	kWh	cost	1/0	kWh	kWh	cost	profit	total
----------------------------------------------------------------------------------------------
1	0,76	1,90	1,39	4,60	1,39	YES	4,60	0,00	-6,41	0,00	-6,41
2	0,76	1,90	1,39	4,60	1,39	YES	4,60	0,00	-6,41	0,00	-6,41
3	0,8	1,95	1,43	0,00	1,95	YES	4,60	0,00	-8,95	0,00	-8,95
4	0,8	1,95	1,43	0,00	1,95	YES	4,60	0,00	-8,95	0,00	-8,95
5	0,8	1,95	1,43	0,00	1,95	YES	4,60	0,00	-8,95	0,00	-8,95
6	0,9	2,07	1,53	0,00	2,07	NO	0,00	0,00	0,00	0,00	0,00
7	0,9	2,07	1,53	0,00	2,07	NO	0,00	0,00	0,00	0,00	0,00
----------------------------------------------------------------------------------------------
						SUM:	23,00	0,00	-39,67	0,00	-39,67

Scenario 2: PLAN TO CHARGE 5 HRS, LARGER TARIFF VARIATION
LOSS NEW vs OLD PLANNER: -2,54			Fixed charges, Grid: 0,945  Feedin: 0,6344  

- New Feedin/charge tradeoff planner
		Grid	Feedin	Surplus	True	Charge	Charged	Feedin	Charge	Feedin	Hourly
Hr	Spot	tariff	tariff	kWh	cost	1/0	kWh	kWh	cost	profit	total
----------------------------------------------------------------------------------------------
1	0,2	1,20	0,83	4,60	0,83	NO	0,00	4,60	0,00	3,84	3,84
2	0,2	1,20	0,83	4,60	0,83	NO	0,00	4,60	0,00	3,84	3,84
3	0,7	1,82	1,33	0,00	1,82	YES	4,60	0,00	-8,37	0,00	-8,37
4	0,7	1,82	1,33	0,00	1,82	YES	4,60	0,00	-8,37	0,00	-8,37
5	0,7	1,82	1,33	0,00	1,82	YES	4,60	0,00	-8,37	0,00	-8,37
6	0,8	1,95	1,43	0,00	1,95	YES	4,60	0,00	-8,95	0,00	-8,95
7	0,8	1,95	1,43	0,00	1,95	YES	4,60	0,00	-8,95	0,00	-8,95
----------------------------------------------------------------------------------------------
						SUM:	23,00	9,20	-43,01	7,68	-35,33


- Old Grid tariff planner
		Grid	Feedin	Surplus	True	Charge	Charged	Feedin	Charge	Feedin	Hourly
Hr	Spot	tariff	tariff	kWh	cost	1/0	kWh	kWh	cost	profit	total
----------------------------------------------------------------------------------------------
1	0,2	1,20	0,83	4,60	0,83	YES	4,60	0,00	-3,84	0,00	-3,84
2	0,2	1,20	0,83	4,60	0,83	YES	4,60	0,00	-3,84	0,00	-3,84
3	0,7	1,82	1,33	0,00	1,82	YES	4,60	0,00	-8,37	0,00	-8,37
4	0,7	1,82	1,33	0,00	1,82	YES	4,60	0,00	-8,37	0,00	-8,37
5	0,7	1,82	1,33	0,00	1,82	YES	4,60	0,00	-8,37	0,00	-8,37
6	0,8	1,95	1,43	0,00	1,95	NO	0,00	0,00	0,00	0,00	0,00
7	0,8	1,95	1,43	0,00	1,95	NO	0,00	0,00	0,00	0,00	0,00
----------------------------------------------------------------------------------------------
						SUM:	23,00	0,00	-32,79	0,00	-32,79

[kalaws]

@andig @premultiply would you agree that there are cost savings to be found here? I also did the same comparison using the 7 actual consecutive spot rates between 1pm and 7pm today, meaning my assumed pv surplus figures could apply. Resulting difference for that period was around 20 percent.

[andig]

As said before: the current planner charges at maximum rate. A solar forecast might make a difference but is not available today. Overall I‘m still unclear, what exactly the proposed algorithm (or cost function) is.

[kalaws]

Yes, I will have to agree that so, too, am I. I have no idea how one might get a final schedule out of it all, nor do I know how a yes/no to charging during the present hour could be produced. But hopefully the experimental tables are at least useful for thinking around the problem.

[kalaws]

Here's the table for spot rates in Sweden, area SE3, 1–7PM 21/7 2023. Savings 17 %.

PLAN TO CHARGE 5 HRS IN THE NEXT 7 HRS, SMALL TARIFF VARIATION 
PROFIT NEW vs OLD PLANNER: 4,59			Fixed charges, Grid: 0,945  Feedin: 0,6344  

- New Feedin/charge tradeoff planner
		Grid	Feedin	Surplus	True	Charge	Charged	Feedin	Charge	Feedin	Hourly
Hr	Spot	tariff	tariff	kWh	cost	1/0	kWh	kWh	cost	profit	total
----------------------------------------------------------------------------------------------
1PM	0,3151	1,34	0,95	4,60	0,95	NO	0,00	4,60	0,00	4,37	4,37
2PM	0,311	1,33	0,95	4,60	0,95	NO	0,00	4,60	0,00	4,35	4,35
3PM	0,3098	1,33	0,94	0,00	1,33	YES	4,60	0,00	-6,13	0,00	-6,13
4PM	0,3232	1,35	0,96	0,00	1,35	YES	4,60	0,00	-6,21	0,00	-6,21
5PM	0,353	1,39	0,99	0,00	1,39	YES	4,60	0,00	-6,38	0,00	-6,38
6PM	0,3553	1,39	0,99	0,00	1,39	YES	4,60	0,00	-6,39	0,00	-6,39
7PM	0,3664	1,40	1,00	0,00	1,40	YES	4,60	0,00	-6,45	0,00	-6,45
----------------------------------------------------------------------------------------------
						SUM:	23,00	9,20	-31,55	8,72	-22,84


- Old Grid tariff planner
		Grid	Feedin	Surplus	True	Charge	Charged	Feedin	Charge	Feedin	Hourly
Hr	Spot	tariff	tariff	kWh	cost	1/0	kWh	kWh	cost	profit	total
----------------------------------------------------------------------------------------------
1PM	0,3151	1,34	0,95	4,60	0,95	YES	4,60	0,00	-4,37	0,00	-4,37
2PM	0,311	1,33	0,95	4,60	0,95	YES	4,60	0,00	-4,35	0,00	-4,35
3PM	0,3098	1,33	0,94	0,00	1,33	YES	4,60	0,00	-6,13	0,00	-6,13
4PM	0,3232	1,35	0,96	0,00	1,35	YES	4,60	0,00	-6,21	0,00	-6,21
5PM	0,353	1,39	0,99	0,00	1,39	YES	4,60	0,00	-6,38	0,00	-6,38
6PM	0,3553	1,39	0,99	0,00	1,39	NO	0,00	0,00	0,00	0,00	0,00
7PM	0,3664	1,40	1,00	0,00	1,40	NO	0,00	0,00	0,00	0,00	0,00
----------------------------------------------------------------------------------------------
						SUM:	23,00	0,00	-27,43	0,00	-27,43

[kalaws]

Just read that Tibber is offering this precise algorithm. I don’t have tibber as a supplier myself (just using a friends token to get prices).

https://support.tibber.com/sv/articles/6363886-faq-smartladdning

Google translate:

Within smart charging, you now have a function for solar charging, how does it work?

With solar charging activated, in addition to what was already mentioned, we will also check whether, according to the forecast, you will have any overproduction in the coming hours. Then our algorithm will calculate what you lose or gain by consuming the electricity yourself compared to if you were to sell and export the electricity. Your cost or profit will then be compared against if the charging session had taken place later in the day during times when the electricity price is low = Most price efficient wins.

For example. you need to charge 10kWh and usually charge at 10kWh/h:
The electricity price this day is 40 öre and you are currently producing 10kW/h.
If you sell that electricity now, you earn 40 öre + 60 öre in tax reduction/kWh but get 0kWh charged in the car. You therefore earn (40öre + 60öre) x 10 = SEK 10 for the 10kWh but have to charge the car and pay for that electricity. If you do it now, you get electricity for the car for SEK 0, but you lose income of SEK 10.
If the price at night, which is usually lower, is 20 öre/kWh, we could charge the car then and calculate it from the profit of your sold production. This means that you still earn SEK 8 for your sold electricity, even though you fully charged the car and paid for it.
The alternative would have been to earn nothing because you chose to charge the car with your produced electricity.
This choice makes our solar charging for you. And it's always what you save/earn the most on that wins.

[kalaws]

Further info on the algorithm:

https://tibber.com/se/magazine/power-hacks/smartladdning-med-sol

[kalaws]

@andig @premultiply
I've suggested a logic below for future reference, when forecasting is in place. However, it seems to me that a basic implementation of 'forecasting' is to leave an option Prioritize hours between sunset and sunrise for charging. Implementation might multiply price of sun hours, i.e., daytime hours (surely time of sunset and sunrise at user location can be easily pulled from some source?), with a factor of X when scheduling (see tables below for reference). A sort of 'daytime tax' is added to the planner functioning. If user has chosen 07.00 AM as time of departure, this will tilt the schedule toward night hours. If at 08.00 AM user sets departure time to the following afternoon, 16.00 PM, this 'tax' does not in practice affect the schedule output. The alternative way to add this 'tax' could be point b) further below. Whether such a tax is to be added (useless on rainy autumn days) could simply be marked by a tick box or similar option: Maximize savings from export of PV surplus during daytime (beta).

Anyhow, for future reference, when forecasting is in place, I think perhaps the logic is the following, assuming we can compare two sets of hours, one marked PvHour, one marked nonPvHour.

When two hours are compared, for PvHour Set charge to 1 if 2*feedinPrice of PvHour is less than gridPrice of noPvHour, else sell (do not charge)
The following tables point in this direction. The strongest argument for a NEW PLANNER is that it should pick up on the first kind of case which can leave a major difference in net cost, because OLD PLANNER does not shift charging to a nonPvHour that has a marginally higher price than a PvHour.

NEW PLANNER
		grid/	Charge	Sell	NET
Hr #	PV hour	PVPrice	cost	profit	COST
1	Yes	200	0	200	
2	No	201	-201	0	
--------------------------------------------
			-201	200	-1
					
OLD PLANNER
		grid/	Charge	Sell	NET
Hr #	PV hour	PVPrice	cost	profit	COST
1	Yes	200	-200	0	
2	No	201	0	0	
--------------------------------------------
			-200	0	-200

NEW PLANNER
		grid/	Charge	Sell	NET
Hr #	PV hour	PVPrice	cost	profit	COST
1	Yes	100	0	100	
2	No	201	-201	0	
--------------------------------------------
			-201	100	-101
					
OLD PLANNER
		grid/	Charge	Sell	NET
Hr #	PV hour	PVPrice	cost	profit	COST
1	Yes	100	-100	0	
2	No	201	0	0	
--------------------------------------------
			-100	0	-100

b) These lists I guess are composed by setting Price of PvHour to feedinPrice and Price of nonPvHourto gridPrice.
A rudimentary implementation of PV forecasting might leave just True/False for an entire hour, so that even if forecast suggests production will be 3kWh for a 10kWh system, such an item could still be marked as a nonPvHour given that max charge power for 1 hour would exceed 3kW. A more granular planner might schedule in intervals of 20 minutes, for example, to solve that problem?

The greater effect I think will be that a lot of charging will be shifted to night hours, which seems very unproblematic in cases where one sets the time of departure to 07.00 AM, for example.

[kalaws]

If basic forecasting is implemented through a simple 'daytime tax,' could planner keep a shadow schedule based on the old planner logic, literally for a rainy day? Let's say that old planner schedule would have suggested charging between 13–14 PM, but new planner with a 'daytime tax' added to daylight hours did not include 13–14 PM. It's raining and because of dark clouds there is no surplus. The charging function could use the inverse logic of pv mode, and proceed until there is a surplus. Should the sun start shining, the user wants to export the surplus per the new planner, and therefore charging halts. A benefit would have been added given that during this particular hour the grid price was actually lower than during the hours scheduled by new planner.

I.e., at the time of planning, two schedules are produced. If maximize export is ticked, new planner schedule applies, and if during daytime surplus becomes 0 then old planner applies until non-0?

[kalaws]

I think the below is a workable suggestion for how a netPlanner, as it can be called, might function:

1. Allow netPlanner to be activated through tick-box: E.g., Consider export savings during daytime (beta)

2. Identify exportHours, meaning a consecutive series of daylight hours during which export savings might possibly accrue. For example, produce a list of daylight hours at user location. However, it may be important to subtract hours just after sunrise and before sunset which in principle count as daylight hours but are invariably insufficient for pv inverter to initiate; these hours are often relatively cheap, and we don't want to have them 'taxed' when sending them through planner. Or, use a source which gives general insolation given user's tilt/azimuth of PV installation. Or, simply let user select a set of exportHours (E.g., 07–19) as response to the query "At what time of day do you expect your PV production to initiate/terminate (above/below household baseload)?". This set of hours could also be set by mqtt if a user has another relevant source of data.

3. Produce full list of netPlanner hours and prices to feed into planner for the purpose of producing a charging schedule. For exportHours use feedinPrice*2 as price. For all others, use gridPrice as price. The result is a list where daytime hours are 'taxed,' likely to be excluded from scheduled charging.

4. When box for netPlanner is ticked and user sets a select departure time, produce two schedules: one using old planner list of grid prices, one according to netPlanner merged list of prices.

5. Until the selected departure time has arrived or planning is canceled, keep the netPlanner schedule active. If pv surplus falls to 0 during exportHours, revert back to old planner schedule and then back again if surplus goes above 0. If pv surplus falls to 0 outside exportHours, stay with netPlanner (otherwise no charging hours may be left).

Edit: Adding again tables to support the above logic. If surplus falls to 0 during exportHours, and if this lapse occurs during an hour that was listed for charging by the old planner, it is likely beneficial to revert momentarily to old planner schedule and then back to netPlanner schedule when surplus goes above 0. During exportHours, PV or feedinPrice is set to feedinPrice*2 (realPrice/calcPrice, e.g., 200/400, in below table), to block charging at these hours.

NEW PLANNER					|  OLD PLANNER					
		grid/	Charge	Sell	NET	|  		grid/	Charge	Sell	NET
Hr #	PV hour	PVPrice	cost	profit	COST	|  Hr #	PV hour	PVPrice	cost	profit	COST
1	Yes	200/400	0	200		|  1	Yes	200	-200	0
2	No	201	-201	0		|  2	No	201	0	0
--------------------------------------------	|  ------------------------------------------
			-201	200	-1	|  			-200	0	-200
- With new planner, surplus is sold and charging during the exportHour is blocked compared to old planner.

Below indicates the logic of setting price of exportHours to feedinPrice*2. If netPlanner simply prioritizes export at all times, it loses to old planner on net cost whenever, instead of charging, it exports at a feedinPrice less than half of the gridPrice it schedules for charging.

NEW PLANNER					|  OLD PLANNER					
		grid/	Charge	Sell	NET	|  		grid/	Charge	Sell	NET
Hr #	PV hour	PVPrice	cost	profit	COST	|  Hr #	PV hour	PVPrice	cost	profit	COST
1	Yes	100/200	-100	0		|  1	Yes	100	-100	0
2	No	201	0	0		|  2	No	201	0	0
--------------------------------------------	|  ------------------------------------------
			-100	0	-100	|  			-100	0	-100
- feedinPrice*2 is less than gridPrice. Surplus is therefore charged according 
to the netPlanner logic, or there would be a loss (-101 vs -100) compared to 
old planner, as shown by the net cost difference. As surplus is now charged 
and not sold, net cost becomes -100, as would have been the cost with old planner.

NEW PLANNER					|  OLD PLANNER					
		grid/	Charge	Sell	NET	|  		grid/	Charge	Sell	NET
Hr #	PV hour	PVPrice	cost	profit	COST	|  Hr #	PV hour	PVPrice	cost	profit	COST
1	Yes	102/204	0	102		|  1	Yes	102	-102	0
2	No	201	-201	0		|  2	No	201	0	0
--------------------------------------------	|  ------------------------------------------
			-201	102	-99	|  			-102	0	-102
- Surplus is sold, charging during the exportHour is blocked compared to old planner.

[kalaws]

@andig I think above comment states the case quite clearly as well as what the feature suggested is. Please consider re-opening the issue/feature case.