Measuring Existing Load Before You Sell a Service Upgrade

Why this matters

A service upgrade is often the largest and slowest line on an electrification job: utility scheduling, a permit, sometimes a new meter base and trenching. Quote one the house did not need and the customer either walks or pays for capacity they will never use. Skip one the house did need and the main trips on the first cold night the heat pump, the backup strips and the car all run together. The way to know which house you are standing in is to measure what it actually draws, and the NEC has a method for exactly that.

This card owns the measured-load method. Panel upgrade mechanics belong to the Electrical Panel Upgrades for Electrification Reference; what to do when the numbers come up short belongs to Alternatives to a Service Upgrade.

The method in one paragraph

NEC 220.87 (numbered that way in the 2020 and 2023 editions, as adopted by your jurisdiction; confirm the number if your jurisdiction has moved to the 2026 edition) lets you size the existing part of a service or feeder load from actual maximum demand instead of from a calculation. Take the highest demand from a year of data, multiply it by 125 percent, add the new loads at their calculated values, and the total must not exceed the ampacity of the service. Where a year of data is not available, a recorded study of at least 30 days can stand in, under conditions the section sets. The service also has to have overload protection in accordance with 230.90. This is a code method, not a rule of thumb, so the inspector can accept it, but only if you can show the data.

Steps

1. Get the utility's interval data, with the customer's authorization. Ask for 12 months of interval demand, at 15-minute resolution if the utility has it. Most utilities release it only to the account holder or with their signed consent, often through the customer's online account. Reason: the year captures both seasonal peaks without you having to guess at either. What breaks if you skip it: you fall back to a 30-day study, which costs a site visit, a month of waiting and an open panel.

2. Check the resolution before you trust the peak. Fifteen-minute data gives you the maximum demand the code is talking about. Hourly data gives you the average over an hour, and the short burst inside that hour, the range and the dryer and the AC starting together, is averaged away. An hourly peak reads lower than the 15-minute peak inside it. Reason: the error runs in the direction that makes the service look bigger than it is. What breaks if you skip it: a margin that exists only on paper. If the utility offers only hourly data, ask the authority having jurisdiction whether they accept it before you build a quote on it.

3. Check that the year is the house you are wiring. Look for months of vacancy, a space heater or a hot tub that has since gone, an occupant change, or solar. A meter on a house with rooftop solar records net load, so the peak it shows is not what the house draws. Recent editions bar the 30-day study where the service has solar, wind or peak shaving behind it; read the section in your adopted edition before you use the data on such a house. Reason: 220.87 assumes the data describes the load. What breaks if you skip it: you size a service off a year the household no longer resembles.

4. Find the single highest interval and convert it to amps. On a 120/240-volt single-phase service, kilowatts times 1,000, divided by 240, gives amps. The utility figure is the whole service, so this is an average of the two legs; a house with most of its 120-volt load on one leg has a worse leg than the average shows. Reason: the service is rated in amps and the comparison has to be in the same unit. What breaks if you skip it: comparing kilowatts to an amp rating, or missing an unbalanced leg the main will feel first.

5. Multiply by 125 percent. This is the code's margin on measured load, and it is not optional. Reason: one year's peak is a sample, not a ceiling. What breaks if you skip it: the calculation fails inspection, and it deserves to.

6. Add each new load at its calculated value under Article 220, not at what you think it will draw. Heat pumps and air handlers at nameplate. Resistance heat at its full rating. A household range per 220.55. An EV charger per 220.57, at the larger of 7,200 watts or the charger's nameplate, unless an energy management system governs it. Use the actual equipment's nameplate, not a typical value. Reason: the measured method relaxes how you count the existing load, not the new one. What breaks if you skip it: a passing total built on a guess the inspector will replace with the nameplate.

7. Do not credit a removed load unless your authority having jurisdiction agrees. The data contains the old AC and the old furnace blower. The heat pump replacing the AC will never run alongside it, but 220.87 says nothing about subtracting, and many inspectors will not allow it. Ask; do not assume. Reason: it is the most common place a borderline calculation gets argued. What breaks if you skip it: a design that passes only with a credit the inspector refuses.

8. Compare against the service rating and the weakest piece of it. The service conductors, the main breaker and the panel bus each have a rating; the smallest governs. Reason: a 100 amp main on a panel bus rated lower, or on conductors that were never upgraded, is not a 100 amp service. What breaks if you skip it: a total that passes against the breaker and fails against the bus.

9. Write down what you used. Data source, resolution, date range, the timestamp and value of the peak, the conversion, the 125 percent line, each new load with its nameplate, and the total. That page is what the inspector reviews and what the next tech reads.

When there is no year of data: the 30-day study

The exception path in 220.87 allows the maximum demand, the highest average kilowatts reached and maintained for a 15-minute interval, recorded continuously over at least 30 days on the highest-loaded phase of the service. The recording has to be taken while the building is occupied and has to include, by measurement or by calculation, the larger of the heating or cooling load and any other seasonal load. In practice that means a study in April records neither season, and you add the larger of the two by calculation.

The recorder goes on energized service conductors, and that is electrical work. Installing or removing current clamps on service entrance conductors means taking the dead front off a panel whose line side cannot be de-energized from inside the house. That is work for a qualified person as OSHA defines it at 29 CFR 1910.399 (general industry) and 29 CFR 1926.449 (construction, which a service upgrade usually is), with the work practices of 29 CFR 1910.333 or 1926.416 and arc-flash PPE chosen under NFPA 70E Article 130, in the current edition (2024 at this writing). Removal at day 30 is the same exposure as installation, and it is the half that gets rushed. If nobody on your crew is a qualified electrician, the study goes to one.

Worked example: the arithmetic, laid out

A 1,700 square foot house, 100 amp service at 120/240 volts, bus and conductors confirmed at 100 amps, no solar. Gas heat and water heating today, central AC. The owner wants a heat pump with 5 kilowatts of resistance backup, a 240-volt heat pump water heater, and a Level 2 charger. Nameplates below are illustrative; use the ones on the actual equipment. The utility supplies 12 months of 15-minute data, and the peak is a July afternoon interval at 7.2 kilowatts.

MEASURED
  peak 15-min demand         7.2 kW   (Jul, 4:45 pm)
  7,200 W / 240 V          = 30.0 A
  x 1.25                   = 37.5 A
NEW LOADS (Article 220, nameplate)
  heat pump                  20.0 A
  5 kW backup, 5,000/240     20.8 A
  HPWH element, 4,500/240    18.8 A
  subtotal                   59.6 A
TOTAL  37.5 + 59.6         = 97.1 A   vs 100 A   PASS
  margin                      2.9 A
ADD EV, 32 A charger
  7,680 W > 7,200 W, so     32.0 A
TOTAL  97.1 + 32.0         = 129.1 A  vs 100 A   FAIL

What it says. The heat pump, its backup and the water heater fit, with 2.9 amps to spare. The charger does not fit at its full calculated value. That is a load-management or a lower-setting question, or an upgrade, and the choice is the alternatives card's subject.

The conservatism you can see. The 7.2 kilowatt peak is a summer interval with the old AC running. The backup strips will only ever run in winter, when that AC load is gone. The method still adds them to the summer peak. That is deliberate margin in the code's favor, and it is why measured results that pass usually pass for real.

What the hourly data would have said. Suppose the utility had offered hourly data only, with a peak hour of 6.1 kilowatt-hours. That is an average of 6.1 kilowatts: 6,100 divided by 240 is 25.4 amps, times 1.25 is about 31.8 amps, 5.7 amps lower than the 15-minute figure. Every downstream total would have looked 5.7 amps better than the house really is, which is more than the 2.9 amp margin above. This is the reason step 2 exists.

What the removed-load credit would change. Say the AHJ accepts crediting the old AC at its illustrative 15 amp nameplate. The four-load total drops from 97.1 to 82.1 amps, and with the charger it is 114.1 amps, still over 100. The credit improves the margin and does not rescue the charger. Build the quote on the uncredited number and treat any credit as an improvement the inspector may or may not grant.

Why the standard dwelling calculation often says otherwise

Run the same house through the standard or optional calculation and the existing load usually comes out higher than 37.5 amps. Those calculations count general lighting by floor area, count every small-appliance and laundry circuit, and apply demand factors written for an assumed household, not this one. A gas-heated house whose family cooks little and dries on a line can measure well under what the calculation assigns it.

That gap is the reason this method sells fewer upgrades, and also the reason to respect its limits. The measurement describes one year of one household. It can understate a house where the next owner runs a hot tub, or where the data year had a mild winter and a resistance space heater sat unused. A calculation is a ceiling for an imagined family; a measurement is a sample of a real one. When the two disagree by a wide margin, ask the customer what changed, or what is about to.

How to verify you got it right

Re-read the data page against the arithmetic. The peak value and its timestamp appear on the utility file. The resolution is stated. Kilowatts were converted at 240 volts and multiplied by 1.25. Every new load matches a nameplate or a stated Article 220 rule, and none was entered at a guessed running current. The total was compared against the smallest of bus, main and conductors. And no removed load was subtracted without the inspector's agreement in writing.

References

• NFPA 70, National Electrical Code, 220.87, 220.55, 220.57 and 230.90 (2023 edition numbering), as adopted by your jurisdiction
• 29 CFR 1910.332, 1910.333 and 1910.399 (general industry) and 29 CFR 1926.416 and 1926.449 (construction), qualified persons and electrical work practices; NFPA 70E, Article 130, current edition
• See related: Electrical Panel Upgrades for Electrification Reference
• See related: Alternatives to a Service Upgrade
• See related: The Whole-House Electrification Sequence