What a Load Calculation Is Actually Asserting

Why this matters

Two documents can sit on the same job and both be labeled "the load," and they are not making the same statement. One is a calculation that models what a building of this description could coincidentally demand. The other is a recording of what this particular building actually demanded over a window of time. A tech who treats them as interchangeable will either tell a customer their service is fine when the permit says otherwise, or condemn a service that has never come close to its limit. Knowing which assertion you are holding is the difference between an argument with an inspector and a conversation with one.

A sibling reference in this library teaches how to run the calculation itself, method by method. This card is about what the result claims, and what it refuses to claim.

The calculation asserts a modeled coincidence

A load calculation takes ratings, not measurements. It starts from nameplate and code-fixed values, applies demand factors, and produces a number in amperes that the service must be able to serve.

The demand factors are the interesting part, because they are the whole model. When a method lets you count a portion of a category rather than all of it, the code is asserting something about coincidence: that in a building of this type, these loads are not all at maximum at the same instant. That assertion was made once, generically, by people who never saw your building. It is a design rule that has held up across a very large number of buildings, which is a different kind of true from "it holds in this house."

So the calculation asserts: under the code's model of how loads coincide, a building described this way requires at least this much service. It is a floor for the equipment, expressed as a demand.

What it deliberately does not assert

• It does not predict your building's actual peak. It is not trying to.
• It does not say the building will never exceed the number. A model of coincidence is not a promise about behavior.
• It says nothing about the instant. Starting current, inrush and the sag that comes with them live outside the calculation entirely, which is why a service that passes the calculation can still dim the lights every time a compressor starts.
• It says nothing about power quality, imbalance between legs, or how the load is distributed across the panel.

That last set is where most nuisance complaints actually live, and none of them are answered by re-running the arithmetic.

The recording asserts an observed maximum, averaged

A demand recording asserts something narrower and more concrete: over this window, on this instrument, at this averaging interval, the maximum was this.

Every word in that sentence is load bearing, and the averaging interval is the one that catches people. Demand instruments integrate over a block of time, commonly 15 minutes, and report the highest block. A motor start that lasts a couple of seconds and pulls several times its running current disappears almost completely into a 15-minute average. That is not a defect in the instrument; it is what demand means. If your question is "will the main hold," the recording answers it. If your question is "why do the lights dip," the recording cannot see the event you are asking about, and a sibling card on what an averaging meter hides covers the instrument side of that problem.

Windows matter as much as intervals. A recording taken in a shoulder season, with neither heating nor cooling working hard, is an honest measurement of a period that does not represent the year.

Where the code lets the two meet

The code does allow a measured history to substitute for a calculated existing load, and the conditions it attaches are the best short summary of everything above. Under NEC 220.87, in the edition your authority having jurisdiction has adopted, an existing load may be determined from maximum demand data over a one-year period, or from a continuous recording of at least 30 days where a year of data is not available. Two conditions ride along:

• The recorded value is multiplied, not used raw. Recent editions apply 125 percent, which is the code buying back some of the headroom that averaging and a short window gave away.
• The result must account for the greatest demand of the heating or air-conditioning equipment, by measurement or by calculation. In other words, if your window missed the season, you do not get to pretend the season does not exist.

Confirm both against the adopted edition before you rely on them, and confirm with the authority having jurisdiction that the method is accepted at all. Some accept it routinely for an existing service; some want the full calculation regardless.

Worked example: one building, two documents, one added load

An existing dwelling with a 200 A service, 240 V. The customer wants a 48 A continuous charging load added. Continuous loads are counted at 125 percent, so that addition is 60 A on either route.

Route one, the calculation. Run by the optional method for an existing dwelling, the calculated load comes to 148 A. It already includes the heating equipment, because the method requires the larger of heating or cooling to be counted whether or not it happened to run this month. Add the new load: 148 plus 60 is 208 A against a 200 A service. On the calculation, the job fails.

Route two, the recording, taken naively. A 30-day recording in October at 15-minute intervals shows a peak demand of 76 A. Apply the code's 125 percent multiplier and you get 95 A. Add 60 and you land at 155 A, comfortably inside 200. On this reading the job passes with 45 A to spare, which is 22 percent of the service.

The two documents now disagree by more than 50 A, and the temptation is to pick the one the customer will like.

Route two, corrected. October in this climate meant the heat pump and its supplemental heat never ran hard, so the recording is missing exactly what the code says it must account for. Take the heating demand by calculation - say it comes to 34 A - and add it to the recorded peak, giving 110 A as the existing maximum demand on the same basis the calculation uses. Then apply the multiplier: 110 times 1.25 is 137.5 A. Add the new 60 A and the total is 197.5 A against a 200 A service, leaving about 2.5 A of headroom, or a little over one percent.

Three things fall out of that, and they are the reason to do the correction rather than take the friendlier number.

The disagreement was never between the methods. It was between a corrected figure and an uncorrected one. The calculation's 148 A already carried the heating equipment; the raw 95 A did not. Comparing them was comparing two different bases, and the corrected recording lands within about 10 A of the calculation rather than 50 A away from it. Any time you correct one side of a comparison, correct the other side or say plainly that it is uncorrected.

Both answers are now honest and they still point different directions. 197.5 A is under 200 and 208 A is over it. That is not a contradiction to resolve with arithmetic; it is a policy question about which assertion the jurisdiction accepts, and it is answered by the authority having jurisdiction before you order anything.

One percent of headroom is a finding regardless of which route wins. It means the next added load, whatever it is, requires a service change. Tell the customer that now, while it is a planning conversation, rather than in a year when it is a surprise.

What changes the answer

If the occupancy is not a dwelling, the optional methods that carry the friendliest demand factors are not available to you, and the standard calculation's assumptions about coincidence are different. Do not carry a residential intuition into a commercial building.

If the added load is not continuous, the 125 percent continuous multiplier does not apply to it, and a load that runs at full output for less than three hours at a time is counted at its rating instead. That single distinction moves numbers more than most arguments about method do.

If the service is fed by, or shared with, another source or another building, the calculation you need is not the one described here and the answer belongs with a design professional rather than a service tech.

Checking your own reading before you hand it over

1. Name the assertion out loud on the ticket. "Calculated load, optional method, 148 A" and "recorded 30-day maximum demand, 15-minute interval, corrected for heating, 137.5 A" are two different sentences and neither of them is "the load is X."
2. Check the averaging interval on the instrument, not on the report header. A logger set to a one-minute interval and a utility demand record at 15 minutes will disagree, and the disagreement is not an error.
3. Check the window against the calendar. Write the dates on the record. A reader six months from now cannot tell a January recording from a May one unless you did.
4. Re-run the fixed portion of the calculation by hand. The general lighting and small-appliance portion is arithmetic that does not depend on judgment, so if two people disagree about a total, that portion is where you confirm you are both starting from the same building.

If you cannot state, in one sentence, what your number is asserting and under what conditions it was produced, you do not yet have a result you can defend to an inspector or to a customer.

References

• NFPA 70 (National Electrical Code), Article 220 including 220.87 for determining existing loads, in the edition adopted by your authority having jurisdiction
• Utility service rules and metering practice for demand averaging intervals, which vary by utility and are stated in that utility's own tariff
• See related: Electrical Load Calculation Reference; Averaging and Spikes: What Your Meter Hides; How to Check Whether a Panel Can Take What You Are Adding