Three terms, used loosely in conversation and precisely in tariffs. The imprecision costs money, because a control strategy aimed at the wrong one delivers a fraction of the projected saving and nobody can explain why.
The three definitions
Facility demand — also called maximum demand, billing demand in simple tariffs, or non-coincident demand. The highest interval average recorded at any time during the billing period. Every hour of every day is a candidate. Nothing is excluded.
On-peak demand. The highest interval average recorded inside a window defined by the tariff — for example weekdays between noon and six in the afternoon during the summer season. Intervals outside the window are ignored entirely, however large.
Billing demand. The number the demand rate is actually applied to, after the tariff has finished adjusting. It starts from a measured figure and then applies whatever clauses exist: a ratchet floor, a contract minimum, a power factor adjustment. It is a calculated result rather than a measurement, which is why it can exceed anything the meter recorded.
| What it measures | When it is measured | Typical rate | |
|---|---|---|---|
| Facility demand | Highest interval in the period | Any hour, any day | Lower $/kW |
| On-peak demand | Highest interval in the defined window | Window hours only | Higher $/kW |
| Billing demand | The determinant actually charged | After all tariff adjustments | The rate above applies to this |
The failure mode
A site sees a summer afternoon peak, installs a control strategy that moves the heavy load to early morning, and expects the demand charge to fall substantially. It falls by roughly half.
The reason is that the tariff bills both determinants. The on-peak figure fell, exactly as designed. The facility figure did not — the load did not disappear, it moved, and the facility measure has no window, so it simply found the new peak at six in the morning and billed that instead.
Why half the saving arrives
Load shifted out of the on-peak window, tariff bills both determinants.
- Facility demand before1,150 kW
- On-peak demand before1,150 kW
- Facility rate$8.20 / kW-month
- On-peak rate$13.90 / kW-month
- (Demand charges before: 1,150 × 8.20 + 1,150 × 13.90)$25,415
- Facility demand after the shift1,090 kW
- On-peak demand after the shift640 kW
- (Demand charges after: 1,090 × 8.20 + 640 × 13.90)$17,834
Saving achieved$7,581 / month
The on-peak component fell by 44% and the facility component by 5%, because the load moved rather than shrank. Rates and quantities illustrative.
Seven and a half thousand a month is a good outcome. It is not the outcome that was promised if the projection assumed the whole demand charge would fall proportionally, and the gap between promise and delivery is what damages the credibility of the next project.
Getting the projection right
The fix is procedural rather than technical: model both determinants separately against the post-change load shape before committing.
- Establish from the tariff which demand determinants are billed, and at what rates.
- For each, extract the exact definition: the interval length, the window, the season, and the days of the week included.
- Take your interval data and compute each determinant as the tariff defines it, for the last twelve months. Confirm your figures reproduce the billed amounts.
- Apply the proposed change to the interval series — do not estimate it, apply it interval by interval.
- Recompute every determinant on the modified series. The new facility peak may be somewhere you did not expect.
- Apply any ratchet clause to the modified series, since a floor set before the change persists after it. See ratchet clauses.
- Only then convert to dollars.
Step five is the one people skip and the one that catches the failure above. When load moves rather than disappears, the new facility peak has to be found, not assumed.
When the window works in your favor
The window cuts both ways, and on the right tariff it is a genuine opportunity.
If a substantial part of your load has genuine timing flexibility — batch processes, charging, non-critical pumping, thermal storage — then moving it outside the on-peak window avoids the expensive determinant entirely. A tariff with a high on-peak rate and a low facility rate rewards that kind of flexibility heavily, and a site with real flexibility should be looking specifically for such a schedule: how to choose a rate schedule and time-of-use rates for commercial accounts.
Conversely, a site with no timing flexibility gains nothing from a windowed tariff and should not pay a premium to be on one.
The detail that decides it
The window definition is more specific than most summaries suggest. Check all four attributes in the tariff:
- Hours — the start and end times, and whether they differ by season.
- Days — weekdays only is common, and public holidays are frequently excluded by name.
- Season — summer and winter boundaries are set by the tariff, not by the calendar, and a bill spanning the boundary is split.
- Interval — the same length applies as for facility demand, but confirm rather than assume. See the demand interval.
A control strategy set to the wrong hours by thirty minutes, or one that treats a holiday as a working day, will produce peaks in precisely the window it was meant to avoid. That is an avoidable and surprisingly common cause of a control project underperforming: demand limiting controls covers the rest of the ways they fail.
Once the determinants are pinned down, the value of a kilowatt against each of them can be priced properly, which is the input every business case on this site depends on: what a kilowatt of avoided peak is actually worth.
