Residential Demand Charges: What Solar Cannot Reduce
Georgia Power's residential demand tariff defines the number it bills you for in one sentence: "Maximum kW shall be the highest 60-minute kW measurement during the current month." Read it again for what it leaves out. It does not say on-peak. It does not say summer. It does not say weekday. On Schedule TOU-RD-12, effective with bills rendered for the billing month of June 2026 and read on 26 September 2026, that single unrestricted hour costs $12.44 per kilowatt.
Your array will cut the kilowatt-hours on your bill. Whether it cuts the kilowatts is a separate question with a separate answer, and on a tariff like that one the second number can be more than half of what you pay.
Two limits belong on the table before any of the arithmetic below. The first is scope: residential demand charges are still the exception rather than the rule, and the three schedules used here were picked because their clauses are unusually explicit, not because they describe a national practice — the opening task in the checklist at the end is finding out whether any of this reaches your bill at all. The second is an assumption. None of the three sheets says anywhere that billing demand is computed net of what your panels produce, and every calculation below assumes it is. That is the one thing worth making your own utility confirm in writing, and the section on array output returns to it.
The sentence that defines the number
Every rate schedule with a demand charge contains a clause telling you exactly how the kilowatt figure is produced. It is usually headed Determination of Demand or Determination of Billing Demand, it is usually one or two sentences, and it is usually nowhere near the price table. Find it before you look at the price.
Three specimens, all read on 26 September 2026:
| Tariff | Determination clause | Interval | Restricted to on-peak? |
|---|---|---|---|
| Georgia Power TOU-RD-12 | "the highest 60-minute kW measurement during the current month" | 60 min | No |
| APS R-3 | "highest amount of demand (kW) averaged in a one-hour On-Peak period for the billing cycle" | 60 min | Yes, 4–7 p.m. weekdays |
| APS R-2 | same wording, "for the billing month" | 60 min | Yes, 3–8 p.m. weekdays |
| SRP E-27 | "the maximum thirty-minute integrated kW demand occurring during the on-peak periods of the billing cycle" | 30 min | Yes, 2–8 p.m. summer; 5–9 a.m. and 5–9 p.m. winter |
Interval length is not a technicality. Integrated and averaged mean the meter spreads whatever you drew across the whole window. A 4.8 kW electric dryer running for fifteen minutes registers as 2.4 kW on a thirty-minute meter and 1.2 kW on a sixty-minute meter — same appliance, same fifteen minutes, twice the billed demand on one of them.
Simultaneity is the other half. Two 3 kW loads that overlap for an hour produce a 6 kW reading. Run them in separate hours and the reading is 3 kW, with the identical number of kilowatt-hours consumed. At Georgia Power's $12.44 that separation is worth $37.32 in the month, and a $20 appliance timer does it. Nothing on the roof competes with that on a cost-per-kilowatt-avoided basis.
One kilowatt, priced in kilowatt-hours
The reason demand charges break payback estimates is that they are priced on a scale nothing else on the bill uses. Converting them into energy-equivalent terms makes the mismatch legible.
Georgia Power's TOU-RD-12 off-peak energy is 1.5569¢ per kWh, plus 3.7441¢ from Schedule TOU-FCR-7 for a secondary-distribution customer, giving 5.3010¢ all in before environmental, demand-side-management, and franchise adders. Divide the demand price by it:
$12.44 per kW / $0.053010 per kWh = 235 kWh
One extra kilowatt in one hour of one month prices the same as 235 off-peak kilowatt-hours. Against the base energy rate alone, before fuel recovery, it prices the same as 799.
The same arithmetic on the other two:
- APS R-3, summer. Demand $19.585 per kW, on-peak energy $0.14227 per kWh. One kilowatt equals 138 on-peak kWh.
- SRP E-27, July and August. Demand is tiered — $11.90 for the first 3 kW, $19.97 for the next 7, $36.05 for everything above 10 kW. On-peak energy is $0.0823. That top marginal kilowatt equals 438 on-peak kWh.
The SRP tiers are worth sitting with. A household whose worst thirty-minute on-peak interval in August is 9.4 kW pays 3 × $11.90 + 6.4 × $19.97 = $163.51 in demand charges that month. Push that interval to 12 kW and it becomes $247.59. The extra 2.6 kW costs $84.08 and moves no more energy than a well-timed evening would have.
Then look at what SRP charges for the energy itself: 8.23¢ on-peak and 6.13¢ off-peak in the summer peak season. Cheap energy and expensive kilowatts is the whole design of the plan. A proposal that values your solar production at a blended average of that bill is valuing it against a number the array cannot earn.
Why a south-facing array barely touches it
The array's output at the specific interval your tariff measures is the only production figure that matters for this charge. Annual kWh is irrelevant to it. So is offset percentage.
On SRP's E-27 the summer on-peak window runs 2 p.m. to 8 p.m. A Phoenix roof is producing well at 2 p.m. and producing very little at 7:30 p.m., and the binding thirty-minute interval in an August billing cycle is far more likely to be the late one — air conditioning at its hottest indoor setpoint, oven on, someone home. The array shaves the early part of the window and leaves the part that sets the bill.
The winter case is starker. From November through April, E-27's on-peak hours include 5 a.m. to 9 a.m. A resistance-backup heat pump coming on at 6:30 a.m. in December sets a demand peak in total darkness. No array orientation changes that, and no amount of capacity does either.
Georgia Power's version removes the window entirely. With billing demand taken from any 60-minute interval in the month, the peak can land on a Sunday night in February. October through May is entirely off-peak on that tariff, which means that for eight months of the year the energy your panels displace is worth 5.3010¢ per kWh while the demand charge sits there untouched.
What you need from your production model is therefore hourly, not annual. PVWatts will give you an hourly series for your own site and orientation; the method is in checking a quote against PVWatts by hand. Line that series up against the hours your determination clause names and read off the kW your array actually contributes there. Usually it is smaller than people expect, and in a winter morning window it is zero.
Here is the caveat flagged at the top, in the place where it bites. Only one of the three sheets ties the demand figure to the meter in so many words: E-27 calls the billing demand the maximum thirty-minute integrated kW "as measured by the meter." Georgia Power says "measurement" and leaves the instrument implied. Neither APS schedule mentions a meter in its demand clause at all.
What points toward netting sits in a different document. Item 9 of APS's Rate Rider EPR-2 requires a partial-requirements customer to have "an AMI meter that will register and accumulate the net electrical requirements of the customer," and item 1 of the same rider describes generation "configured so that the energy generated first supplies its own electric requirements." What points the other way is E-27's Condition F, which spells netting out in careful detail and then applies it only to kilowatt-hours — energy delivered to SRP is subtracted from energy delivered from SRP, and the demand clause three pages later is left out of that sentence entirely. A schedule that defines netting for one unit and not the other has probably not left the second one out by accident.
So on the single point the arithmetic below rests on, the sheets do not speak. Put the question to your utility in writing: is billing demand computed net of on-site generation? The answer decides whether an array can move this charge at all, not merely how much.
The March bill, done twice
Here is the arithmetic a proposal usually skips. Prices are from TOU-RD-12 and TOU-FCR-7 as read on 26 September 2026. The load figures are assumptions standing in for your own interval data, and you should replace every one of them.
Inputs
Billing days 31 (March)
Purchased kWh, no solar 1,050 (all off-peak: March is off-peak)
Highest 60-minute kW in the month 7.2 (oven + dryer + water heater, 6:30 p.m. Sunday)
Load factor 1,050 / (7.2 x 31 x 24) 19.6%
Basic service charge $0.4603 x 31 $14.27
Off-peak energy, all in $0.053010 per kWh
Demand $12.44 per kW
Before solar
Energy 1,050 x 0.053010 $55.66
Demand 7.2 x 12.44 $89.57
Basic $14.27
Subtotal $159.50
After an 8 kW-dc array: 880 kWh produced, 600 self-consumed
Energy 450 x 0.053010 $23.85
Demand 7.2 x 12.44 $89.57 <- 6:30 p.m. in March is after sunset
Basic $14.27
Subtotal $127.69
Actual saving $31.81
Now the number a spreadsheet produces instead. Blended rate is the whole bill divided by the whole consumption, $159.50 / 1,050 = $0.15190 per kWh. Multiply by the 600 kWh the array displaced and you get $91.14, which is 2.87 times the $31.81 the tariff actually gives back. The gap is not an error in the multiplication. It is the demand charge, sitting inside the blended average as though solar removed it.
The 280 kWh exported in that month settles under a separate schedule at whatever rate that schedule sets, which is a different document and a different calculation — the three structures it could be are laid out in net metering versus net billing versus buy-all/sell-all.
Do not annualise this one month. SRP's winter demand tiers are $4.93 / $7.02 / $11.00 against summer peak's $11.90 / $19.97 / $36.05, so the distortion is heavily seasonal and can invert. Twelve months of your own interval data priced against your own sheet is the only version of this that means anything.
The load-factor cap solar customers do not get
Both APS residential demand schedules contain a safety valve, and both take it away from anyone with panels.
Both sheets open the clause with the same words, and the first four are already doing the work: "For full requirements Customers, billing demands are limited to a kW no higher than that which would result in a 15% load factor, based on the Customer's kWh usage during the month." R-3 keeps going from there, as below; R-2 stops at that full stop. Both spell out the formula in Service Details item 6:
Monthly Load Factor = Billed kWh / (Billed kW x Billing Days x 24 hours)
Rearranged, the cap on billed demand is billed kWh divided by 3.6 times the billing days. A 30-day month at 600 purchased kWh caps billing demand at 5.56 kW. At R-3's summer price of $19.585 per kW, being billed 5.56 kW instead of a metered 9.0 kW is worth $67.46 in that one month.
Then comes the sentence directly after it, identical in both schedules: "This limitation is not available to Partial Requirements Customers." And Service Details item 1 of both schedules says that customers who self-provide some of their requirements from on-site generation are billed under a Partial Requirements rider. Installing solar is what makes you one. EPR-2 confirms there is no relief in the rider either — "All terms and charges in the Customer's rate schedule continue to apply."
R-3 narrows the valve further even for those who qualify: three instances per billing year, with only one of them allowed in the summer season. R-2 attaches no such counter, which is the single respect in which it is the better of the two sheets.
It is also the one you cannot have. R-2 was frozen as of 1 December 2021, and its availability section is explicit that no customer connecting service after that date is eligible to choose it. It appears throughout this article because its wording is the plainest of the three and its demand price is flat across both seasons — not because it is a plan you can shop for.
The mechanism behind the exclusion is worth understanding, because it is the same effect that produced the March example above. Solar lowers your purchased kilowatt-hours without lowering your peak kilowatts, which lowers your load factor. Low load factor is precisely what the cap was written to protect. Solar customers are both the group most likely to need it and the group explicitly denied it.
That is two sentences in a five-page PDF, and neither of them is in the charge table. The exclusion sits in running text on page 3, below the prices; the Service Details item that turns a solar customer into a Partial Requirements Customer is on page 4. A rate schedule prints what it charges in a grid and puts what it withholds in prose, and the grid is the part that gets quoted.
Five clauses sitting next to the price
Once you have the determination clause and the price, keep reading the same document. The terms around them change the arithmetic, and each of these is one line.
A lockout on leaving. SRP's E-27 Condition C: cancel and elect another plan, and "the customer may not subsequently elect service under this price plan for at least one year after the effective date of cancellation." Georgia Power's TOU-RD-12 runs the other direction with a one-year minimum term that auto-renews annually unless terminated with 30 days' notice before the anniversary. Either way, the plan you pick when the system energises is the plan you are modelling for at least a year.
The season turns on your meter read date, not on the calendar. SRP's customer page for the plan states that at a seasonal transition all usage and monthly demand charges on the bill are calculated at the rate in effect on the meter read date, so a bill read on 14 May prices April's demand peak at summer rates.
A minimum bill. E-27's minimum is the Monthly Service Charge. Georgia Power's is the basic service charge plus environmental, demand-side-management, and franchise components. No quantity of generation goes below it.
One line on this bill grows when the array does. APS R-3 adds a Grid Access Charge of $0.215 per kW-dc of nameplate generation for distributed-generation customers; R-2's version is $0.250. On a 7.6 kW-dc system that is $1.63 or $1.90 a month. Small, but it is the one line on the bill that grows when you make the array bigger.
The tariff expires before a twenty-year model does. E-27's availability section says the plan "will be eliminated as of the November 2029 billing cycle," and that anyone still on it then moves to E-16. SRP's page attributes the retirement to its publicly elected Board on 27 February 2025. A 20-year model built on E-27 prices is assuming something the sheet contradicts on its first page.
One more, from the service-charge structure rather than the demand clause: E-27 sets its Monthly Service Charge tier by dwelling type and service amperage together, and Tier 3 at $40 applies to "any residence with service of more than 225 amps" against $30 for Tier 2. A panel upgrade to support a heat pump or a second EV charger can therefore move you a tier and add $120 a year, entirely separate from the demand charge. The therms-to-kilowatt-hours conversion that makes electrification look good is done on energy prices, and this is not an energy price.
Who can add one, and where that record lives
Whether a demand charge might arrive on your bill in year six of a twenty-year model depends on who sets your rates, and the answer changes where you have to look.
For a commission-regulated utility the lineage is printed on the sheet. Each APS page carries an A.C.C. number, the number it cancels, a revision number, and the decision that made it effective: R-3 is A.C.C. No. 6150, Revision 3, cancelling 6121, effective 8 March 2024 in Decision No. 79293. Those four numbers are the handle for a docket search at the Arizona Corporation Commission's Utilities Division, and a proposed change appears there, with a comment period, before it reaches a bill. Georgia Power's tariffs carry the same kind of note in different words: the bill is "subject to change in such an amount as may be approved and/or amended by the Georgia Public Service Commission."
SRP's sheet has no commission number anywhere on it. It says instead that service is governed by "SRP's Rules and Regulations, as they may be amended or revised by SRP from time to time," and the plan retirement traces to a board vote rather than a regulatory decision. Different forum, different record, different notice — and no commission docket to search. Find out which of the two your utility is before you go looking, and if it is the second kind, the thing to monitor is the board meeting agenda.
There is also a lagging-document problem to watch for. Georgia Power's consumer page for this plan footnotes its figures as "based on TOU-RD-11" while the tariff library is serving TOU-RD-12. The prices happen to match at $12.44. That will not always be true, and the tariff is the document that bills you. The same discipline applies to everything in your spreadsheet: schedule name, revision, and the date you read it, in the cell beside the number. Reading the rule that governs your connection has the same requirement, which is why the interconnection tariff post belabours sheet headers.
Pull your worst hour out of twelve months
A battery is the only hardware that acts on this charge, and it is sized in kilowatts at one interval rather than kilowatt-hours across a day. Holding a 7.2 kW peak down to 4 kW through a sixty-minute measurement window means roughly 3.2 kW of discharge for that hour, every month, including the month nobody was home to notice. That is a different sizing question from the usual one, and the interval data it needs is the same data described in sizing a battery from your own load profile.
So download the interval data first. Most utilities publish fifteen-minute or hourly history through the account portal, and twelve months of it is what this requires.
Then do four things with it.
- Find the determination clause in your own rate schedule and write down the interval length and whether it is restricted to on-peak hours. If there is no such clause, you are not on a demand tariff and none of this applies to you yet.
- For each of the twelve months, pull the single worst interval that your clause would count. Twelve numbers, not an average.
- Price those twelve against the demand table, seasonal tiers included, and add it up. That annual figure is the part of your bill that solar addresses only in the intervals where the sun is up and the meter is looking.
- Ask whoever wrote your proposal what rate schedule and what demand figure their savings line assumed, and compare it with your twelve.
If the answer is a blended cents-per-kilowatt-hour rate, you now know what it contains. Your rate code is on the bill, in the place described in the four numbers on your electric bill, and the schedule it names is a free PDF with the determination clause in it.
Frequently asked questions
Does solar reduce a demand charge?
Only in the metering intervals where the array is actually producing, and only if the tariff measures demand in those intervals. Georgia Power's TOU-RD-12 takes the highest 60-minute kW in the whole billing month with no on-peak restriction, so the binding interval can land at 6 a.m. in January. Salt River Project's E-27 restricts it to on-peak hours, but in the November-to-April season those hours include 5 a.m. to 9 a.m., when a south-facing roof produces nothing. Where the measured interval is dark, the charge is untouched no matter how large the system is.
How is billing demand actually measured on a residential tariff?
Look for a clause headed Determination of Demand or Determination of Billing Demand. It states an averaging interval and sometimes a restriction. SRP's E-27 sheet says the maximum thirty-minute integrated kW demand occurring during the on-peak periods. Both APS R-2 and R-3 say the highest demand averaged in a one-hour on-peak period. Georgia Power's TOU-RD-12 says the highest 60-minute kW measurement during the current month. Interval length alone changes the billed figure: a 4.8 kW dryer run for fifteen minutes registers 2.4 kW on a thirty-minute meter and 1.2 kW on a sixty-minute one.
Is there a cap on how high my billed demand can go?
Some tariffs have one and it is often unavailable to solar customers. APS schedules R-2 and R-3 limit billing demand, for full requirements customers, to a kW no higher than one producing a 15 percent load factor, using the formula billed kWh divided by billed kW times billing days times 24. Both sheets then add that this limitation is not available to Partial Requirements Customers, and Service Details item 1 of each puts anyone with on-site generation onto a Partial Requirements rider. Read your own schedule, because the exclusion is a single sentence and it is not in the charge table.
Can a battery remove the demand charge?
It can cut it, and it is the only piece of hardware that acts on it directly, but the sizing question is kilowatts at one interval rather than kilowatt-hours over a day. To hold a 7 kW household peak down to 4 kW across a one-hour measurement window you need roughly 3 kW of discharge for that hour, and the battery has to do it every month including the month you were away. Size it against your own interval data and against the measurement clause, not against a brochure.