Two demand charges can appear on the same bill, priced differently, measuring different things, and requiring completely different responses. Confusing them is expensive in both directions: you can spend money reducing a peak that was never being billed, or leave a genuine exposure untouched.
The distinction
Non-coincident peak (NCP) bills your facility's own highest interval during the billing period, whenever it occurred. It is a measurement of you, in isolation. Almost all distribution-level demand charges work this way.
Coincident peak (CP) bills your demand at the moment the wider system peaked — the utility's system peak, a transmission zone peak, or the regional market's peak. Your own maximum is irrelevant; what matters is what you happened to be drawing during somebody else's maximum.
Why both exist
The cost causation differs by level of the network.
The distribution transformer, service conductors and local feeder serving your site have to be sized for your maximum, regardless of when it occurs. Nobody else's load shares that equipment at that moment. A non-coincident measure matches that cost.
The transmission network and the generation fleet, by contrast, are sized for the system maximum. Your contribution to that cost is what you were drawing when the system was at its peak. If your plant is idle at that instant, you imposed no burden on the system peak, however large your own peak was at three in the morning.
Hence the split. Distribution costs are recovered non-coincidentally; transmission and capacity costs increasingly are not.
The management problem is completely different
This is the practical consequence, and it is the reason the distinction is worth the trouble.
Managing NCP is a measurement problem. Your peak is in your own data. You can see it, attribute it to specific equipment, and control it with sequencing, scheduling or storage. The feedback loop is immediate: you change something, the next bill tells you whether it worked.
Managing CP is a forecasting problem. The interval that matters is defined by the system, not by you, and you do not know which interval it will be until afterward. You have to predict it, act on the prediction, and accept that some of your curtailments will have been unnecessary and some of the real peaks will have been missed.
That difference cascades:
| Non-coincident | Coincident | |
|---|---|---|
| The interval that matters | Your own maximum | The system's maximum |
| Known in advance? | You choose it | Must be forecast |
| Feedback | Next bill | Next annual reset, usually |
| Typical measure | Sequencing, controls, storage | Forecast-triggered curtailment |
| Cost of a wrong call | You paid for a peak | You curtailed for nothing, or missed it |
| Events per year | Every month | A handful |
Note the last row. A coincident regime often turns on a very small number of intervals across an entire year — four in the ERCOT four coincident peak method, a small set of summer hours in PJM's peak load contribution calculation. Getting those few right is worth a great deal; getting them wrong is worth nothing at all, in either direction.
The two you are most likely to meet
ERCOT four coincident peak. Transmission cost is allocated on your average demand across the single highest fifteen-minute system interval in each of June, July, August and September. Four intervals set a determinant that applies for the following year. See four coincident peak.
PJM peak load contribution. A capacity obligation is derived from your demand during the highest system peak hours of the preceding summer, and it drives your capacity charge for the following delivery year. See PJM capacity tags.
Other regions allocate differently again. The pattern to look for is the same: a determinant defined by somebody else's peak, set once, applied for a long period afterward.
What to do about each
If your charge is non-coincident, work from your own interval data. Find the peaks, attribute them, and treat it as an internal control problem: how to reduce peak demand charges covers the ladder of measures.
If your charge is coincident, the work is different in kind:
- Establish which intervals count and over what period — this is stated in the tariff or in the ISO's allocation methodology.
- Get access to a forecast. ISOs publish load forecasts, and commercial services exist that issue coincident peak alerts.
- Decide in advance what you will curtail and who has authority to call it. An event lasting an hour is not a decision to make by committee.
- Accept a false-positive rate. Curtailing on ten predicted peaks to catch four real ones is normal and is usually still economic, because the value of the four is large and the cost of the six is small.
- Log every event, whether it turned out to be real or not, so the strategy can be evaluated afterward.
If you have both, price them separately before choosing a measure. A battery sized for the coincident events may be entirely wrong for the monthly non-coincident peak, and vice versa: sizing a battery for peak shaving treats those as separate sizing cases.
Reading it off the bill
The give-away wording is usually explicit: demand at time of system peak, coincident demand, transmission demand, or a reference to a capacity or transmission obligation with a determinant that does not change month to month.
A determinant that stays constant across several bills while your own load varies is almost always a coincident one, set annually. If you see that, the intervals that set it happened months ago, and the way to change it is to be ready next season — which is a planning exercise, not a control one. The full vocabulary is in billing determinants, and the underlying rationale in why utilities bill for demand at all.
