How should an electricity bill be designed? Retail, dynamic and large-load tariffs
Bill a renter, a solar and EV home and a factory under four tariffs, see who pays for whom, then set a large load’s minimum bill
- Analyst · L2
- 25 min guided · about 30 min practice
- After lesson 8
- Completed
Three Prairie Town customers, a month of days like 9 June 2025 (illustrative)
| Customer | Cost to serve | Flat | Time-of-use | Demand charge | Real-time |
|---|---|---|---|---|---|
| Renter | $71 | $92 (+$22) | $87 (+$16) | $75 (+$4) | $88 (+$18) |
| Solar + EV home | $262 | $66 (−$196) | $115 (−$147) | $224 (−$38) | $104 (−$158) |
| Factory | $11,984 | $11,984 ($0) | $11,984 ($0) | $11,984 ($0) | $11,984 ($0) |
Homes' tariffs:
- Flat: $10 a month + 14.7¢ per kWh
- Time-of-use: $10 a month + 8.2¢ per kWh, 24.5¢ from 16:00 to 21:00
- Demand charge: $10 a month + 5.2¢ per kWh + $21.00 per kW of the busiest hour
- Real-time: $10 a month + the town's price each hour + 9.5¢ per kWh for plants and wires
- At 14.7¢ per kWh the renter pays $92 for a cost of $71; the solar and EV home pays $66 for a cost of $262. Nine renters, about $22 each above their cost, cover that home's $196.
- Charging at 01:00 cuts the time-of-use bill by $68 a month, while the model's cost of serving the home falls $126. Under the flat price the bill would not move at all.
- At midnight the demand-charge bill falls $3,000 a month, but the model's cost of serving the factory falls only $720: the batch's energy now costs $480 more. Under real-time pricing the bill rises $480.
- Real-time: the renter's bill jumps from $88 to $132 this month. Flat: it stays $92 while serving the renter costs $114. Here we assume the utility pays the gap now and recovers it from all customers later.
| Minimum billing demand | Data center pays | Other customers pay | Billed above its highest demand |
|---|---|---|---|
| No minimum | $7.5M | $7.5M | 0 MW |
| Minimum 80% of the contract (the median of 55 real tariffs) | $12.0M | $3.0M | 30 MW |
With JavaScript on, you bill each customer yourself, move a car's charging and a factory's batch, add a scarcity week and set a large load's minimum bill.
Illustrative: Prairie Town is fictional; its customers, the utility's unit costs, the four tariffs and the large load are invented.
Real hourly patterns: each hour's price is the town's running cost of one more MWh on 9 June 2025, from its dispatch on real Southwest Power Pool shapes (EIA-930, accessed 2 November 2026), as in lesson 9.
A month is 30 days like this one. Cost of serving is the model's monthly allocation of energy, plants, wires and meters.
About 25 guided minutes. Practice, about 30 more: the practice set under Check.
What you just did, as diagrams
- Cost of service
- What serving a customer costs in this model: energy each hour, plants for the town's tightest hour (most gas needed), local wires for its own busiest hour, and a meter.
- A tariff
- The rule that turns those costs into a bill: a fixed charge, prices per kWh, and charges per kW.
- Revenue-neutral
- Each tariff is set so that a class of customers pays its total cost before anyone changes what they do.
- Cross-subsidy
- When one customer pays above its cost so another can pay below it, inside the same class.
- What a tariff rewards
- A move pays off on the bill only if the tariff charges for what the move changes.
- Large-load tariff
- A minimum bill, a long contract and an exit fee keep a large load's upgrades from falling on other customers.
Go deeper · AnalystSetting prices so that a class pays its cost
The homes’ class is nine renters and one solar and EV home. Serving them costs $897.58 a month, and they use 5,416.7 kWh net.
| Tariff | How its price is set | Result |
|---|---|---|
| Flat | ($897.58 − 10 × $10 fixed) ÷ 5,416.7 kWh | 14.7¢ per kWh |
| Time-of-use | the same $797.58 ÷ (3,233.6 off-peak kWh + 3 × 2,183.2 peak kWh) | 8.2¢ off-peak, 24.5¢ peak |
| Demand charge | plants and wires, $515.59, ÷ the sum of the busiest hours, 24.55 kW | $21.00 per kW, plus 5.2¢ per kWh for energy |
| Real-time | plants and wires, $515.59, ÷ 5,416.7 kWh | a plant-and-wire charge of 9.5¢ per kWh on top of each hour’s price |
- Revenue-neutral. Every row collects the same $897.58 at the start; they differ in who pays it.
- Real rate cases add more classes, many hours instead of one tightest hour, and a year of weather.
Go deeper · AnalystNet metering, net billing and what a kWh sent back is worth
- Net metering. A kWh sent back earns what a kWh bought costs: 14.7¢ under the flat price.
- What it is worth. At midday on 9 June the town is spilling sun, so one more kWh is worth $0 to it.
- Net billing. Credit each kWh sent back at the hour’s running cost, and re-set the flat price so homes still pay their cost: it falls to 13.5¢.
- The result. The renter’s bill falls from $92 to $86; the solar and EV home’s rises from $66 to $127, still below its $262 cost.
- Why still below. Its car charges in the tightest hour, and kWh prices do not see kW.
Go deeper · ModelerA bill as a function of 24 hourly loads
Write the customer’s net load in hour \(t\) as \(x_t\) (kW, held for the hour), and \(n\) = 30 days. Every tariff in this lesson is
\[\text{bill}(x) = F + n \sum_{t} p_t\, x_t + d \max_t x_t\]
- \(F\) is the fixed charge ($ a month), \(p_t\) the price of a kWh in hour \(t\) ($ per kWh) and \(d\) the demand charge ($ per kW).
- What it rewards. One kW less in hour \(t\) cuts the bill by \(n p_t\). The demand charge falls only when the highest remaining hourly demand falls: all tied peaks must come down (the factory’s 13:00 and 14:00 are tied).
- What it should reward. The cost of service has the same shape, with the town’s price in \(p_t\), plants on the tightest hour’s \(x_t\) and wires on \(\max_t x_t\).
- Cost-reflective. A tariff whose slopes match the cost’s slopes sends the right signals for the costs this model represents; this lesson’s four each match only part.
- Next. City models with screening fairness outputs come in M12; who pays and who benefits in A13; flexible demand in A7.
Check yourself
Myth or fact? “One flat price per kWh charges every customer for the costs it causes.”
- Myth
- Fact
Answer
Myth. In the toy the renter pays $22 above its cost and the solar and EV home $196 below: kWh are not what drives plants and wires.
Myth or fact? “A demand charge rewards exactly what lowers the grid's costs.”
- Myth
- Fact
Answer
Myth. Moving the factory's batch saves $3,000 on its bill, but the model's cost of serving it falls only $720: the charge sees the factory's own peak, not the town's tightest hour.
Myth or fact? “Real-time pricing puts the risk of price spikes on the customer.”
- Myth
- Fact
Answer
Fact. A scarcity week raises the renter's real-time bill from $88 to $132. Under the other tariffs the utility pays first; here we assume it recovers that from all customers later.
Myth or fact? “A minimum bill can protect other customers when a large load's demand stays below its contract.”
- Myth
- Fact
Answer
Fact. With its highest demand at half its contract, other customers pay $7.5M a year of the upgrades with no minimum, and $3.0M with a minimum of 80% (if they cover what it does not pay).
Under which tariff does moving the car's charging from 19:00 to 01:00 save its owner the most?
- Flat
- Time-of-use
- Demand charge
- Real-time
Answer
Time-of-use: $68 a month, against $30 under the demand charge, $13 under real-time pricing and nothing under the flat price. The model's cost of serving the home falls $126.
Practice · about 30 minutes on your own
Bill the renter under time-of-use by hand. It uses 560 kWh a month, 192 kWh of them from 16:00 to 21:00; prices are 8.2¢ and 24.5¢ per kWh, plus $10 a month.
Answer
$10 + 368 kWh × 8.2¢ + 192 kWh × 24.5¢ ≈ $87 (the page's $87 uses unrounded prices).
Who pays for the solar and EV home under the flat price? Its bill is $66 and serving it costs $262.
Answer
The other homes: nine renters contribute about $196 in total, about $22 each: the home's $196 gap. Prices were set so the class pays its cost.
A lower busiest hour for the factory. Its demand charge is $15 per kW, and moving the batch cuts its busiest hour by 200 kW. What does its bill save, and how far does the model's cost of serving it fall?
Answer
$3,000 on the bill (200 kW × $15), but the cost falls only $720: the wires' share falls $1,200 and the batch's energy costs $480 more.
A large load whose highest demand is 60% of its contract. Contract 100 MW, minimum billing demand 80%, upgrades $15M a year. Who pays what, if other customers cover the rest?
Answer
It is billed for 80 MW at $12.50 per kW a month: $12.0M a year; other customers pay $3.0M.
Move the time-of-use window. The town's dearest hours on 9 June 2025 are 20:00 to 23:00, when the peaker runs. Is 16:00 to 21:00 a good window?
Answer
Only partly: at 16:00 sun is still spare (the town's price is $0), and 21:00 to 23:00 is left out. A window set months ahead cannot follow every day.
Read a claim critically. A story says: “Real-time pricing cut one customer's bill by 40%.” Name three things you would ask.
Answer
Against which tariff? Did the customer change when it used power, or was it a month without scarcity? And who paid for the risk in other months?
Numbers, sources and what’s next
Numbers to remember
| Number | What it means | Source |
|---|---|---|
| 17.23¢ per kWh | Average U.S. residential electricity price in 2025 (commercial 13.37¢, industrial 8.51¢). | EIA, Electric Power Annual, Table 2.4 · accessed 3 November 2026 |
| 76.8% | Share of U.S. electricity meters that were advanced meters in December 2023 (128.4 of 167.2 million). Time-of-use needs use split by price period; real-time pricing needs hourly readings. | FERC staff, 2025 Assessment of Demand Response and Advanced Metering, p. 2 and Table 2-1, p. 4 (EIA data) (PDF) · December 2025; data for December 2023 |
| 80% · 12 years | Across 55 large-load tariffs reviewed, the median minimum billing demand was 80% of contract demand; the median minimum contract rose from 5 years (tariffs proposed before 2025) to 12 years (since 2025). | Berkeley Lab and The Brattle Group, Electricity Rate Designs for Large Loads, 2026 Update, pp. 2, 4, 5 (PDF) · August 2026; tariffs as of March 2026 |
Next
In the real world
Large-load tariffs decide who carries the cost if a large load uses less than it promised.
Ratepayer exposure depends strongly on large-load tariff design.
The full case: case file CF7 (what waiting for a grid connection costs, and who pays).
Real large-load tariffs: minimum bills, contract terms, collateral and exit fees, compared in Berkeley Lab's 2026 review (PDF).
Sources and further reading
- EIA. Electric Power Annual, Table 2.4: Average price of electricity to ultimate customers by end-use sector, 2015 through 2025 (accessed 3 November 2026).
- FERC staff (December 2025). 2025 Assessment of Demand Response and Advanced Metering (PDF), p. 2 and Table 2-1, p. 4.
- Mims Frick, N., Cappers, P. (Berkeley Lab), Kavlak, G., Lam, L., Hledik, R., Grove, J. and Rotella, S. (The Brattle Group) (August 2026). Electricity Rate Designs for Large Loads: Evolving Practices and Opportunities, 2026 Update (PDF), pp. 2–5.
- Li et al. (2026). Accepted paper on the cost of delayed grid energization for large loads, 2026 North American Power Symposium (NAPS), presented October 2026; DOI pending.
- The day's hourly shapes: EIA. Hourly Electric Grid Monitor (Form EIA-930), Southwest Power Pool, 2025 (accessed 2 November 2026).
Spotted an error, or stuck? Write to hlli@ksu.edu. Corrections are made in place and dated here.
Text licensed CC BY 4.0, code MIT: teachers and students may reuse and adapt it, with credit to Honglin Li.