Peak Demand vs. Peak Usage: Why They're Billed Differently

Energy and power are different quantities

A kilowatt-hour is a quantity of energy. A kilowatt is a rate. Your bill's usage line counts the first: every kWh that passed through the meter in the billing period. Your bill's demand line records the second: the highest rate at which energy passed through the meter, averaged over one short interval. Santee Cooper's plain definition is that demand "is how much electricity you are using at one time. It is measured in kilowatts (kW)" [santee-cooper-demand].

An analogy that holds up: usage is the miles you drove this month, demand is the fastest you drove. Two fleets can cover the same miles; the one that hits 90 on the highway needs a different vehicle, different tires, and different insurance. Utilities price the same way. Generators, transformers, and conductors are sized for the fastest rate, so the customer who creates the rate pays for it, and NREL's survey of more than 10,000 tariffs found that demand can be 30 to 70 percent of a commercial bill [nrel-demand-charge-survey].

The demand charges page walks through the charge itself. This page is about the two things underneath it: how the meter turns a continuous draw into a billed number, and why there is more than one kind of peak.

How interval metering turns a draw into a billed number

The meter does not record your instantaneous peak. It integrates energy over fixed windows, usually 15 minutes and sometimes 30, and reports the average power in each window. Your billed demand for the month is the highest window. Santee Cooper's example is exact on this: a customer whose instantaneous draw reached 160 kW was billed 130 kW, because 130 kW was the average over the highest 30-minute interval [santee-cooper-demand].

That mechanic is the whole basis of peak management. A 350 kW lighting step that lasts a full 15-minute interval registers as 350 kW. The same step, if half the fixtures switch at minute 0 and half at minute 15, registers as 175 kW in each of two intervals plus whatever else is running. The energy is identical. The billed demand is not.

Three consequences for a grow:

• The interval length matters. A utility that bills on 30-minute intervals is more forgiving of a short spike than one that uses 15.
• Ramping works. LED fixtures that dim up over 15 to 30 minutes spread a step across intervals.
• Averaging cuts both ways. A short shutdown inside an interval barely helps, and a long plateau is billed in full.

Interval meters also produce the data that shows you which interval set the peak. If yours is an AMI meter, the utility's portal or a data request gets you 15-minute history. The interval data page covers how to get it and what to do with it.

Non-coincident peak: your own worst interval

The utility's distribution demand charge is set by your non-coincident peak, your highest interval regardless of what the rest of the grid was doing. It does not matter whether it happened at 3 a.m. in February or 4 p.m. in July; the wires serving your building had to carry it.

For a cultivation facility the non-coincident peak is almost always the lights-on transition of the largest photoperiod group, stacked with the HVAC and dehumidification that respond to it. Lighting was 66 percent of electricity use in the Northwest Power and Conservation Council's survey, cooling 15 percent, and dehumidification most of the rest [nwpcc-cannabis], and all three step together at lights-on. Because photoperiods are scheduled, the peak is the same interval every day, which is convenient: you know exactly when it is and exactly what causes it.

Many tariffs add a ratchet, so that billed demand is the greater of this month's peak or a percentage of the highest peak in the past 11 months; Santee Cooper uses 30 percent [santee-cooper-demand]. A ratchet turns one bad interval into a year of charges.

Coincident peak: your draw when the grid peaks

The second kind of peak is not about your worst moment. It is about your load at the grid's worst moment, and it drives two different charges.

Capacity in PJM (5CP). Every load-serving entity in PJM must buy capacity for its customers, and each customer's share is set by their peak load contribution. ComEd's PJM tariff attachment defines the PJM Five Peaks as "the five hours occurring on different calendar days in a summer during which the electric system served by PJM experiences its five highest daily summer demands," with summer defined as June 1 through September 30 [pjm-oatt-m2-comed]. A customer's average load in those five hours, adjusted for losses, becomes their capacity peak load contribution for the period from June of the following year through May of the year after [pjm-oatt-m2-comed]. The same document sets a separate network service peak load contribution, used for transmission, from the customer's load at the ComEd zone's single highest hour of the summer [pjm-oatt-m2-comed].

Transmission in ERCOT (4CP). Texas allocates transmission cost by the four coincident peaks, which ERCOT defines as "the highest-Load 15-minute settlement intervals in each of the four summer months (June, July, August, September)" [ercot-dr-overview-2023]. ERCOT estimated that 1 MW of load reduction during those four intervals was worth about 38,000 dollars a year for a transmission-connected customer on Oncor's system, and counted 41 near-CP event days in 2022, which is how many afternoons a customer had to watch to be sure of catching four [ercot-dr-overview-2023].

The coincident peaks share three properties. They occur on hot summer weekday afternoons, typically between 2 p.m. and 6 p.m. They are set by hours you cannot know for certain until the season is over, though forecasts narrow it to a handful of days. And they are billed with a lag: what you do this summer shows up next year. That lag is why a facility that changed its schedule in August still sees the old capacity charge through the following May [pjm-oatt-m2-comed].

For a grow the coincident-peak exposure depends on one question: what is lit at 4 p.m. on a July weekday? A facility whose flower rooms go dark at 3 p.m. has a low coincident peak and a low capacity charge even if its non-coincident peak at 3 a.m. lights-on is enormous. A facility that lights up at noon has the opposite profile. The demand response page covers the programs that pay for reductions in exactly these hours.

Load factor: the number that summarizes all of this

EIA defines load factor as "the ratio of the average load to peak load during a specified time interval" [eia-glossary-l]. Compute it from any bill:

Load factor = kWh ÷ (peak kW × hours in the period)

A load factor of 100 percent means you drew exactly the same power every hour. A load factor of 25 percent means your average draw was a quarter of your peak, which is what a building with a big afternoon spike and quiet nights looks like. Since delivery demand, capacity, and transmission are all priced on some kind of peak, a higher load factor means those per-kW costs are spread over more kWh, and your blended cost per kWh falls.

Grows have naturally decent load factors because the lights run 12 hours a day and HVAC runs continuously. Staggering photoperiods pushes load factor higher still, because the lit canopy stays roughly constant around the clock.

A worked example

Every number below is an input we chose. Substitute your own.

Assume a facility with 10,000 square feet of flowering canopy in two rooms, lighting at 35 W per square foot, HVAC and dehumidification at 50 percent of lighting kW when lights are on and 30 percent when dark, 25 kW of other load, and a 30-day month. Assume a distribution demand charge of 12 dollars per kW and a capacity price of 8 dollars per kW-month on the PJM peak load contribution.

Scenario A: one photoperiod, lights on noon to midnight.

Quantity	Calculation	Result
Lighting kW	10,000 × 35 W	350 kW
Lit-hour load	350 + 175 HVAC + 25 other	550 kW
Dark-hour load	105 HVAC + 25 other	130 kW
Monthly kWh	(550 × 12 + 130 × 12) × 30	244,800 kWh
Non-coincident peak	Highest interval	550 kW
Load factor	244,800 ÷ (550 × 720)	62 percent
Load at 4 p.m. on a July weekday	Lights on	550 kW
Distribution demand charge	550 × 12	6,600 dollars
Capacity charge next year	550 × 8	4,400 dollars per month

Scenario B: two photoperiods, room 1 lit noon to midnight, room 2 lit midnight to noon.

Quantity	Calculation	Result
Lighting kW at any time	5,000 × 35 W	175 kW
Load at any hour	175 + 88 HVAC (lit room) + 53 HVAC (dark room) + 25 other	about 340 kW
Monthly kWh	340 × 24 × 30	244,800 kWh
Non-coincident peak	Highest interval	340 kW
Load factor	244,800 ÷ (340 × 720)	100 percent
Load at 4 p.m. on a July weekday	Room 1 lit	340 kW
Distribution demand charge	340 × 12	4,080 dollars
Capacity charge next year	340 × 8	2,720 dollars per month

Scenario C: same as B, but room 1 lit 8 p.m. to 8 a.m. and room 2 lit 8 a.m. to 8 p.m.

The non-coincident peak and load factor are the same as Scenario B. But at 4 p.m. on a July weekday, room 2 is lit, so the coincident load is still 340 kW. Now flip it: if instead room 2 ran 4 p.m. to 4 a.m. and room 1 ran 4 a.m. to 4 p.m., the 4 p.m. hour sits at a transition, and a facility that adds a 30-minute gap where both rooms are dark, say 3:45 to 4:15 p.m., could see its coincident load fall to roughly 130 kW in the hours most likely to be PJM peaks, while its non-coincident peak stays 340 kW. The capacity charge would then be 130 × 8 = 1,040 dollars a month instead of 2,720.

Whether a 30-minute dark gap is agronomically acceptable is a question for your cultivation lead, and the answer may be no. The point of the example is that the usage line is identical in all three scenarios, 244,800 kWh, while the demand and capacity lines differ by thousands of dollars a month, and the two kinds of peak respond to different decisions. Scheduling rooms against each other lowers the non-coincident peak. Scheduling rooms against the grid lowers the coincident peak. You can try your own inputs in the demand charge estimator.

What to do with this

1. Compute load factor from each of the last twelve bills, then get interval data and find the exact interval that sets each month's peak.
2. Note what is lit between 2 p.m. and 6 p.m. on summer weekdays. That is your coincident-peak exposure.
3. Decide which peak is worth more: with a high delivery demand charge, chase the non-coincident peak first; in a PJM zone with high capacity prices or in ERCOT, the coincident peak may be the bigger number. Then confirm your rate class matches the peak you actually have.