Estimated supercharger costs and margins. Calculations and analysis. Group effort

[insaneoctane]

This is a spin-off from a question I asked in the investor thread HERE.
I'll kick off the first question for folks who have first hand knowledge of how much demand fees might be (specify what State your knowledge is from) for:

10 SC stalls @250kW = 2.5MW theoretical peak load * 80% = 2.0 MW actual peak load
20 SC stalls @250kW = 5.0MW theoretical peak load * 80% = 4.0 MW actual peak load

I've reduce the theoretical values since the likelihood that all stalls are really hitting the relatively short charging "sweet spot" of 250kW is very low. Even if we assume the SC has a line and all stalls are busy, they'd not all be at 250kW at once. I'll use 80% de-rating for this phenomena, but I'm open to change that.

 

[MC3OZ]

how much demand fees might be

We also need to know the tariff for a bulk plan with demand charges...
For example the tariff might drop from $0.25 per KWh to $0.08 per KWh the demand charge may be daily or monthly...

We also don't know if Tesla throttles Supercharging to stay below a target peak demand.

Other variables are if the site has solar and batteries, the SOC and the charging curve on each car...
It is hard to know how long the average customer hangs around after the charging rate drops below the peak..

I'll repeat my post form the main thread below:-

IMO for paid Supercharging we need to make some guesses and assumptions about fixed costs and margins,..

Here are my guesses, which may be the right order of magnitude.

Supercharger V2 - fixed costs: (per day per stall) Equipment $3 Site $5 margin $0.10 kWh - kWh to breakeven. = 80
Supercharger V3 - fixed costs: (per day per stall) Equipment $7 Site $5 margin $0.15 kWh - kWh to breakeven. = 80

Both coming out at 80 kwh per day sold to break-even is a coincidence, we do need to multiply that by the number of stalls to get the site figure...
Underutilization may be a bigger problem than overutilization in some sites.

What evidence to I have for a margin?
.These are rough Australian figures Residential $0.25 kWh, Supercharging ?? ($0.40-$0.50 kWh) - I paid $0.42 last time).
Residential demand charges (proposed by utility) $0.08 kWh + peak monthly usage charge..

I can't remember all the details, but a bulk deal with peak charges normally has a signicantly lower per KWh cost than a standard residential tariff.

Tesla also knows the utilization at any time and can throttle charging to stay within peak usage limits.

From a customer point of view low utilization is best, from a ROI point of view high utilization is best. The optimal outcome is a happy compromise where charging is at least break-even in the quieter periods, but can handle the busy holiday loads...

The Tesla fleet is growing rapidly other cars using chargers are likely to be a rounding error...

But there is little doubt bigger Supercharger V3 sites with some solar and lots of stalls, give the best combination of margin, utilization and customer service.

This is an interesting topic and a separate thread is a good idea...

EDIT: If the site has solar and batteries, they may use them to stay below a target peak load, say the target was 1 GW and solar and batteries could contribute 100 MW the effective cap is 1.1 GW maybe all stalls are throttled to (1.1GW/no of active stalls). There is an obvious trade off in terms of time and money. And a trade off between solar+batteries and demand charges.

 

[MP3Mike]

10 SC stalls @250kW = 2.5GW theoretical peak load * 80% = 2.0 GW actual peak load
20 SC stalls @250kW = 5.0GW theoretical peak load * 80% = 4.0 GW actual peak load

First, you are off by a factor of 100. (That would only be MegaWatts, not GigaWatts.) Second, stalls do not have a dedicated 250kW supply. 4 stalls share a cabinet that only supplies 350kW of power. (Though they can borrow power from other cabinets or battery storage.)

So if all stalls are in use with empty Teslas, and there is no battery storage, each stall is limited to ~90kW. (Unless the number of stalls is not a multiple of 4 in which case the limit would be slightly higher.)

• 10 V3 SC stalls = 3 cabinets = 1.05MW theoretical peak load
• 20 V3 SC stalls = 5 cabinets = 1.75MW theoretical peak load

 

[RandyS]

A recent 16 stall V3 installation has a 1.5 MVA transformer installed, so there are definitely some load diversity calculations in play.

 

[NickInLab]

In your attempt to be overly dramatic you have designed an example that could never occur. There is NO WAY a 50 stall, 250 kW Supercharging location could have a peak draw of 12,500 kW. First of all, [...] I think at PEAK draw, a 50 stall, 250 kW station could, at most, draw up to around 30% of your estimate of 12,500 kW demand charge. This would be when there was a line of cars waiting to plug in.

Building an overly simplified model in R (since I happened to have the REPL open... and I'm a dumb physicist who prefers not to do math), I get the following time-averaged charge power per car (all starting at 5% SoC):

Final SoC	Charge Time (Hrs)	Avg Power (kW)	Percent of Peak
0.5	0.15	214	85.6%
0.6	0.21	195	78.0%
0.7	0.28	174	69.6%
0.8	0.37	149	59.6%
0.9	0.54	117	46.8%

If we make a spherical black cow assumption: cars arrive at a uniform rate to a very large number of chargers, all cars charge to the same SoC, then depart with zero downtime before the next car starts charging; then the station-average should equal the time-average of the cars (in other words, we're assuming ergodicity).

So under this model it seems like the demand charge would be between approximately 47% and 86% of the peak charging speed, depending on the assumed ending SoC.

It would be interesting to see how this changes for more realistic scenarios... I might try doing that next.

charge_power <- function(soc){
    if (soc < 0.2 ) 250 else (1.0 - soc) * 250/0.8
}

charge_time <- function(start_soc, end_soc, kwh){
    time_sec <- 0.0
    soc <- start_soc
    kjoules <- khw * 3600
    while (soc < end_soc){
        time_sec <- time_sec + 1
        soc <- soc + charge_power(soc)/(kjoules)
    }
    time_hr <- time_sec / 3600
    avg_pwr <- kwh*(end_soc-start_soc)/time_hr
    c(time_hr, avg_pwr)
}

print("Final SoC,  Charge Time,   Avg Power")
for (i in 5:9){
  print(c(i/10, charge_time(0.05,i/10, 75)))
}

 

[SOULPEDL]

I thought we were talking about using batteries for peak demand at superchargers, then I brought up my powerwall, but obviously it was an analogy as MegaPacks would be needed and would provide some payback I think. But more importantly, control over the power, costs, and reliability vs the utilities. That has high value I think.

 

[StealthP3D]

Building an overly simplified model in R (since I happened to have the REPL open... and I'm a dumb physicist who prefers not to do math), I get the following time-averaged charge power per car (all starting at 5% SoC):

Final SoC	Charge Time (Hrs)	Avg Power (kW)	Percent of Peak
0.5	0.15	214	85.6%
0.6	0.21	195	78.0%
0.7	0.28	174	69.6%
0.8	0.37	149	59.6%
0.9	0.54	117	46.8%

If we make a spherical black cow assumption: cars arrive at a uniform rate to a very large number of chargers, all cars charge to the same SoC, then depart with zero downtime before the next car starts charging; then the station-average should equal the time-average of the cars (in other words, we're assuming ergodicity).

So under this model it seems like the demand charge would be between approximately 47% and 86% of the peak charging speed, depending on the assumed ending SoC.

It's not disclosed where your Average Power data for different SoC's came from but it's not even close to being correct for the Model 3 and Model Y which should comprise the bulk of the fleet. I think assuming the average Supercharger user starts at 5% SoC is the real error here. Almost no one starts charging at 5% SoC and assuming they do will greatly skew the Average kW numbers to levels unrealistically high. Also, in many situations the power level ramps down much more quickly than the numbers indicate. It also doesn't include turn around time when no charging is happening. And some models can only charge at a maximum of 150 Kw and ramp down from there.

 

[insaneoctane]

Sorry for the initial GW mistake, yes it should be:
10 SC stalls @250kW = 2.5MW theoretical peak load * 80% = 2.0 MW actual peak load
20 SC stalls @250kW = 5.0MW theoretical peak load * 80% = 4.0 MW actual peak load

Maybe I watched ”Back to the Future” too recently?

ok, I think we start by estimating the current SC economics first, so ignore the battery and solar initially. We can then justify the ROI of solar and batteries after we see what Tesla is likely paying today on the grid.

I am fairly confident that v3 SC DOES have dedicated 250kW feed. I’ll dig around to confirm that. Assuming that’s true, what do you guys think the simultaneous de-rating factor for the calculation ought to be? I’ve suggested 80%, but overlaying 10 M3 charging curves at “random” offsets might suggest a different number…

 

[insaneoctane]

So, while several early photos of cabinets seemed to have conflicting data about whether v3 has dedicated 250kW per stall, it seems like we should just assume that based off this:

About Supercharging V3:

• New architecture of chargers
• 1 MW power cabinet, which can support up to 250 kW per car (four stalls per power cabinet)
• no power sharing between the Superchargers (previously two stalls shared 120 kW of power)
• cuts charging time by up to 50% (when battery is warm up)
• expected typical charging time to drop to around 15 minutes
• expected to serve more than twice the number of customers per hour (than V2)

source

and this:

Faster Charging, No More Power Sharing
V3 is a completely new architecture for Supercharging. A new 1MW power cabinet with a similar design to our utility-scale products supports peak rates of up to 250kW per car.

Source

So for calculations, let’s assume 1 MW cabinet per 4 stalls?

 

[MP3Mike]

I am fairly confident that v3 SC DOES have dedicated 250kW feed. I’ll dig around to confirm that.

I provided a picture of the plate from a V3 cabinet that clearly shows the 350kVA rating and outputting to 4 stalls. It is from a couple years ago but I don't think Tesla has changed them since then.

and this:
Source

So for calculations, let’s assume 1 MW cabinet per 4 stalls?

The "1 MW power cabinet" includes the 575kW DC bus. 350kW AC input + 575kW DC bus = 925kW. (So almost 1MW.)

 

[insaneoctane]

I provided a picture of the plate from a V3 cabinet that clearly shows the 350kVA rating and outputting to 4 stalls. It is from a couple years ago but I don't think Tesla has changed them since then.

The "1 MW power cabinet" includes the 575kW DC bus. 350kW AC input + 575kW DC bus = 925kW. (So almost 1MW.)

I’m starting to come around ….thanks for explaining the “almost 1MW” and why the cabinet rating doesn’t mean all AC input….
So, I think we stick to your earlier suggestion:

• 10 V3 SC stalls = 3 cabinets = 1.05MW theoretical peak load
• 20 V3 SC stalls = 5 cabinets = 1.75MW theoretical peak load

I don’t think we need any 80% de-rating factor since it’s totally possible, IMO, to have those cabinets fully loaded when there is a line….

From a per stall calculation standpoint, using 100 kW/ stall average probably works fine for most common size installations….

Stalls	Cabinets reqd	AC load (kW)	Avg/stall (kW)
1	1	350	350.0
2	1	350	175.0
3	1	350	116.7
4	1	350	87.5
5	2	700	140.0
6	2	700	116.7
7	2	700	100.0
8	2	700	87.5
9	3	1050	116.7
10	3	1050	105.0
11	3	1050	95.5
12	3	1050	87.5
13	4	1400	107.7
14	4	1400	100.0
15	4	1400	93.3
16	4	1400	87.5
17	5	1750	102.9
18	5	1750	97.2
19	5	1750	92.1
20	5	1750	87.5
21	6	2100	100.0
22	6	2100	95.5
23	6	2100	91.3
24	6	2100	87.5
25	7	2450	98.0
26	7	2450	94.2
27	7	2450	90.7
28	7	2450	87.5
29	8	2800	96.6
30	8	2800	93.3
31	8	2800	90.3
32	8	2800	87.5
33	9	3150	95.5
34	9	3150	92.6
35	9	3150	90.0
36	9	3150	87.5
37	10	3500	94.6
38	10	3500	92.1
39	10	3500	89.7
40	10	3500	87.5
41	11	3850	93.9
42	11	3850	91.7
43	11	3850	89.5
44	11	3850	87.5
45	12	4200	93.3
46	12	4200	91.3
47	12	4200	89.4
48	12	4200	87.5
49	13	4550	92.9
50	13	4550	91.0

 

[insaneoctane]

So almost starting over…my original question is now:

I'll kick off the first question for folks who have first hand knowledge of how much demand fees might be (specify what State your knowledge is from) for:

10 stall with peak demand of 1.05MW
20 stall with peak demand of 1.75MW

 

[MP3Mike]

So almost starting over…my original question is now:

I'll kick off the first question for folks who have first hand knowledge of how much demand fees might be (specify what State your knowledge is from) for:

10 stall with peak demand of 1.05MW
20 stall with peak demand of 1.75MW

Here is an example of a fee schedule for PGE in Oregon: https://assets.ctfassets.net/416ywc...705987e64860ddd58d10a8f20b16cf0/Sched_085.pdf

For 1.05MW:

Fee	Amount	kW:	1005
Base	$ 470.00
Transmission	$ 763.80
Distribution (First 200kW of Capacity)	$ 620.00
Distribution (Remaining Capacity)	$ 1,529.50
Distribution (Demand)	$ 2,542.65
	$ 5,925.95

For 1.75MW:

Fee	Amount	kW:	1750
Base	$ 470.00
Transmission	$ 1,330.00
Distribution (First 200kW of Capacity)	$ 620.00
Distribution (Remaining Capacity)	$ 2,945.00
Distribution (Demand)	$ 4,427.50
	$ 9,792.50

Then there is the charge per kWh based on TOU.

But it sounds like they could qualify for the EV TOU option which doesn't have demand charges, just a charge per kWh 11.77 - 13.277 cents/kWh: https://assets.ctfassets.net/416ywc...86028119708cc516188e872f8ab1388/Sched_038.pdf

 

[insaneoctane]

I think SCE in Southern California has your demand charges beat. Someone needs to check my interpretation of the Schedule TOU-8 tariff

Demand Charge - $/kW of Billing Demand/Meter/Month = $14.98

So, the demand charge (in addition to the other fees) for the:
10 stall = 1,050 kW * $14.98 = $15,729
20 stall = 1,750 kW * $14.98 = $26,215

plus $571.50/meter fee
plus ~$0.12/kW gen/dist fee

Am I reading this right???

 

[cpa]

I'll let you smart people carry on with the discussion of electricity charges.

I'm staying by the river with the luggage to add the following (pursuant to the title that wants costs and margins):

In addition to the direct cost of electricity there are a number of additional direct and indirect costs with Supercharging

--Depreciation of the construction and acquisition costs
--Possible rent expense to property owners
--Real property taxes at locations owned or leased if taxes are part of the lease payment or possessory interest taxes at locations owned by the government
--Unsecured property taxes on the cabinets, stalls, and hardware
--Repairs and maintenance
--Business licenses required by local jurisdictions
--Corporate overhead allocated to Supercharging

No doubt the electricity charges are the lion's share of Tesla's costs. But those items listed above will add to the overall cost to Tesla to determine if Tesla is losing money, breaking about even, or making money.

Then no complete analysis would ignore the income tax benefits of constructing and installing Superchargers. Companies that install electric vehicle charging stations receive can receive up to a $30,000 federal income tax credit per location. Whether Tesla currently pays any federal income taxes is subject to debate, but these tax credits can be carried forward to reduce taxes for twenty years.

 

[MP3Mike]

EVgo has released details on the cost of electricty at one of their sites in NY: DCFC Cost Components: Much More than Electricity!

So for that 800kWh it cost them $2.90/kWh. (And that is with the peak demand being only 102.5kW.)

I think this is why Tesla has said opening their networking and getting more usage will actually lower their costs. For example at this site if they doubled the usage, but kept the peak demand the same, the cost would have dropped to ~$1.80/kWh. (I just doubled the total non-demand portion, which probably causes more taxes to be in that price than it really would be.)