The Demand Interval — Why Fifteen Minutes Decides Your Bill

The demand charge is described as pricing your peak. It does not, quite. It prices your peak average over a defined window, and the definition of that window does more work than most people expect.

Averaging is the whole mechanism

Suppose a 300 kW load runs for the first three minutes of an interval and nothing else is on. The average over fifteen minutes is 300 × (3 ÷ 15) = 60 kW. The meter records 60, not 300.

Now suppose the same 300 kW load runs for the full fifteen minutes. The average is 300 kW, and 300 is what gets recorded.

Identical equipment, five times the demand determinant, entirely because of duration. This single arithmetic fact explains most of what is counterintuitive about demand charges:

• Motor starting inrush rarely matters. It lasts seconds. Fifteen minutes of averaging reduces it to near-invisibility.
• Simultaneity matters enormously. Three 200 kW loads that run in sequence never produce more than a 200 kW average. The same three running together produce 600.
• Short interruptions do not help much. Cutting a load for two minutes inside a fifteen-minute interval removes about 13% of its contribution to that interval.

Fixed block against rolling

The second definition to check is how the window is positioned.

A fixed-block interval runs between set clock times: 10:00 to 10:15, 10:15 to 10:30, and so on. A rolling or sliding interval is recalculated continuously, taking the average of the most recent fifteen minutes of data at every step.

The difference shows up with a spike that straddles a boundary. Under fixed blocks, a heavy load running from 10:12 to 10:27 is split across two blocks and diluted in both. Under a rolling interval, the fifteen minutes from 10:12 to 10:27 are evaluated as a window in their own right and the full average is captured.

Rolling is stricter. It is also harder to manage against, because there is no boundary to hide behind and a demand limiting controller has to predict a continuously moving average rather than a block that resets on a clock. See demand limiting controls.

The practical difference shows up in how a control strategy has to be written. Against fixed blocks, a controller knows exactly how much time remains in the current window and how much headroom is left in the accumulating average, so the decision to shed is arithmetic: at minute eleven of fifteen, with the average tracking above target, the required reduction over the remaining four minutes is calculable. Against a rolling window there is no end of window to work toward. Every new minute both adds a value and drops one, and the controller has to hold the average down continuously rather than rescue it before a deadline.

Sites that install a demand controller configured for the wrong convention typically find it performs well in testing and then permits peaks in service. It is a one-line check in the tariff and it is worth doing before the controller is specified rather than after.

Older thermal demand meters

Some services are still billed on thermal demand meters, which do not average over a discrete window at all. They respond to heating in a sensing element with an exponential lag characterized by a time constant. The result approximates a heavily damped rolling average: brief loads register weakly, sustained ones fully.

If your tariff references a thermal demand meter or a response time constant rather than an interval length, this is what it means, and analysis based on fifteen-minute blocks will not reproduce the billed figure exactly.

Why this changes your analysis

The practical consequence is about data, and it catches people out routinely.

If your tariff bills on fifteen-minute intervals and you analyze hourly data, every hourly figure is the average of four fifteen-minute figures. A sharp quarter-hour spike is diluted to a quarter of its size. You will conclude your load is flatter than it is, that your peak is lower than it is, and that a smaller intervention is needed than actually is.

Ask the utility for data at the billing interval. If only hourly is available, treat every conclusion as a lower bound. How to get your interval data covers making that request specifically.

What to confirm in your tariff

The interval definition sits in the demand section of the rate schedule, usually under a heading like determination of demand. Three things to extract:

1. Length — fifteen minutes, thirty, or a thermal time constant.
2. Positioning — fixed block or rolling.
3. Scope — every interval in the month, or only those inside an on-peak window, or both as separate determinants. See facility, on-peak and billing demand.

With those three, your own interval data reproduces the billed demand figure. If it does not, the discrepancy is a ratchet, a power factor adjustment or a contract minimum — see billing determinants — and knowing which is the first step in demand charges explained becoming actionable rather than theoretical.