The plant's interval file recorded 96 quarter-hours above 500 kW in a year, and what mattered was not how high they were. It was that they arrived in 41 separate events, and that one event runs for nine consecutive intervals. A battery that covers the tallest peak but not the longest event is billed for the event.
What the interval file says about this plant
Worked example. A two-shift manufacturing site with a recurring monthly peak of 620 kW, a demand target of 500 kW, and a demand charge of EUR 14.50 per kW per month.
- 96 intervals above the target in a year, which is 24 hours of over-limit time
- 41 separate events, so the average event is 2.3 intervals, about 35 minutes
- The longest event is 9 intervals, 2 hours and 15 minutes
- Energy above the line: 228 kWh in that longest event, 1,860 kWh across the year
The shape matters more than the height. A site whose over-limit time is concentrated into a few long events needs energy, and a site with many short spikes needs power and very little energy. This plant is the first kind, which is why the sizing was decided by the ninth interval of one event rather than by the 620 kW reading.
What the control has to decide
Peak shaving is delivered by a controller, not by a datasheet, and the controller has five decisions to make before any energy is stored.
The calculation, in the order the controller runs it
At the peak interval the load exceeds the target by 120 kW, which is inside the cabinet's 125 kW. The power requirement is therefore satisfied by one unit, and the rest of the calculation is about energy.
The longest event needs 228 kWh delivered above the line. The cabinet's usable window is 238.74 kWh at 90 percent depth of discharge, which leaves 10.74 kWh of margin, or 4.0 percent of the 265.26 kWh rated energy. A margin of four percent is thin enough that it belongs in the control settings rather than in a footnote.
Charging that event back costs 259 kWh at the meter, since the delivered energy has to be replaced through a round trip of 88 percent. That load has to be scheduled, because the recharge adds to the same demand the controller is trying to hold down.
Where control becomes a sizing decision
Ruibit quotes two figures for the 125 kW cabinet, 265.26 kWh rated and 238.74 kWh usable at 90 percent depth of discharge, and the reserve calculation has to come out of the difference between them rather than being added on top.
That is the point at which an operating choice turns into capital. A reserve of 10 percent of rated energy withholds 26.53 kWh, which leaves 212.21 kWh for shaving, and 212.21 kWh does not cover the 228 kWh the longest event needs. The event goes unshaved, and because that event is the one that sets the monthly peak in the month it occurs, the demand charge for that month is billed at the unshaved level regardless of how well the controller performed on the other 40 events.
The arithmetic sets the ceiling: 238.74 kWh minus 228 kWh leaves 10.74 kWh, which is 4.0 percent of rated energy. A site that needs a reserve larger than that has two honest options, and only two. It buys a second cabinet, or it accepts an annual bill that includes the one event that matters.
Why a schedule rule gives up part of the saving
Worked example. With a demand limit at 500 kW, the peak falls from 620 kW to the target, and the saving is 120 kW at EUR 14.50 per kW per month, which is EUR 1,740 a month, or EUR 20,880 a year.
Now suppose the controller discharges on a fixed schedule instead, from 10:00 to 14:00, which is the pattern most factory loads follow on paper. Six of the 41 events fall outside that window, and in two of the twelve months those events set the billed peak. Those two months are billed at the unshaved 620 kW, which removes EUR 3,480 from the annual saving, or 16.7 percent of it.
Two criteria separate the two approaches. If the over-limit events cluster inside a window narrower than four hours and stay there in every month of the year, a schedule rule is defensible. If the events that fall outside the window set the peak in more than one month, only demand-limit control avoids the charge, because the controller follows the load rather than the clock.
What to ask, and what the control cannot recover
Ask what the reserve is for, and whether the function it protects needs to be inside the same unit. An answer that the reserve is a safety margin, without a duty attached to it, is a reserve that is costing the site its longest event.
Ask what the controller does when the load exceeds the target by more than the battery can supply in power. An answer that the system shaves what it can is an answer about a peak that will still be billed.
Ask how the recharge is scheduled, and against which limit. An answer that the battery recharges after the event, without naming the window, usually means it recharges during the afternoon peak.
Three things are set once the controller is commissioned and cannot be recovered later. The reserve percentage is the sizing decision described above, and it is fixed by the unit that was ordered. The recharge window is fixed by the tariff structure the site is billed under. And the result is only verifiable after a full billing cycle, which is why the demand limit and the deadband belong in writing before the first discharge rather than in a settings file nobody reopens.
FAQs
1. What load data does factory peak shaving need?
Twelve months of 15-minute interval data. A monthly bill gives the maximum demand but not how long it lasted, how often it occurred, or how many separate events produced it. The event structure is what decides whether the site needs power or energy.
2. How is the discharge power sized?
From the largest gap between the load and the demand target, not from the peak itself. If the plant peaks at 620 kW against a 500 kW target, the requirement is 120 kW, and a 125 kW unit covers it with a small margin.
3. How is the battery energy sized?
From the longest event above the target, not from the annual energy. In the example one event runs for nine intervals and needs 228 kWh, which fits inside a 238.74 kWh usable window with 10.74 kWh left over.
4. Why is demand-limit control better than a time schedule?
Because factory peaks do not follow the clock. In the example, six of 41 events fall outside a fixed four-hour window, and in two months those events set the billed peak, which removes 16.7 percent of the annual saving.
5. How much reserve can a peak-shaving battery hold?
Only what is left after the longest event. A reserve comes out of the usable window rather than being added to it, so if the longest event needs 228 kWh of a 238.74 kWh window, the reserve is capped at 4.0 percent of rated energy.
