Smart Meter Interval Data: What the Hourly Record Shows
Why this matters
A monthly bill gives you one number for thirty days. A customer calling about that number wants to know which machine did it, and a monthly total cannot say. Most smart meters record use every hour or every fifteen minutes, and that record usually sits in the utility's online usage portal unread. Fifteen minutes with a day of it can rule out half the theories a tech would otherwise spend a site visit chasing, and it often points at the right one before anybody drives out.
What the meter actually records
A smart meter, also called an AMI meter (advanced metering infrastructure), totals the energy that passed through it in each interval and stores one figure per interval. The common intervals are 15 minutes and 60 minutes, and the utility decides which. Each value is energy, in kWh, not power.
The conversion that makes the record readable:
- An hourly kWh value is the average kW across that hour. 2.4 kWh in one hour means the house averaged 2.4 kW for the hour. It does not mean anything ran at 2.4 kW; a 4.8 kW load running half the hour gives the same figure.
- A 15-minute value times 4 is the average kW for that quarter hour. 0.9 kWh in 15 minutes is 3.6 kW average.
- The lowest values in the day are the floor. The house never drops below its continuous loads, so the quietest overnight hours are an upper bound on everything that runs all the time.
The record is whole-house. It cannot tell you which appliance drew the energy, only when and how much. Everything you learn from it comes from shape: when the energy rises, how high, how long, and whether it repeats.
Getting the data
The customer retrieves it. Most utilities with smart meters let the account holder see interval data in their online portal and download it, often as a spreadsheet file or in a standardized format many utilities label Green Button (an industry download standard, not a product). Some utilities offer only daily totals online and the hourly record on request.
Do not ask for the customer's password and do not log in on their behalf. Ask them to download a date range and email it, or to share their screen while they open it. See related: How to Get a Customer's Interval Usage Data Without Their Password, which owns the steps.
Ask for at least two weeks that include the complaint period, plus a similar span from a month the customer considers normal. One day proves nothing by itself; a pattern across days does.
A day from the record
Here is one cold January day from an all-electric house with a heat pump and electric backup strips. Hours are hour-beginning (the row labeled 05 covers 5:00 to 6:00), which you must confirm in the utility's file header, because some utilities label the hour by when it ends.
HOUR KWH HOUR KWH HOUR KWH
00 0.6 08 2.4 16 1.3
01 0.6 09 1.6 17 2.2
02 0.7 10 1.2 18 2.6
03 0.6 11 1.1 19 2.3
04 0.7 12 1.0 20 1.9
05 6.8 13 0.9 21 1.5
06 7.4 14 0.9 22 1.0
07 5.2 15 1.0 23 0.8
DAY 46.3 KWH
FLOOR 0.6 KWH/H = 0.6 KW CONTINUOUS
05-07 19.4 KWH = 42 PCT OF THE DAY
Read it in three passes.
The floor. Midnight to 5:00 sits at 0.6 to 0.7 kWh per hour. All five of those hours read 0.6 or 0.7 with nothing climbing, so roughly 0.6 kW runs all the time. Across a day that is 0.6 x 24 = 14.4 kWh, which is 14.4 / 46.3 = 31 percent of this day's total. Across a 30-day bill it is 0.6 x 24 x 30 = 432 kWh before anybody turns anything on. Whether 0.6 kW is normal depends on the house; a home with a well pump, a second refrigerator and a dehumidifier can sit there honestly. See related: Standby and Phantom Loads: What Is Worth Chasing.
The morning spike. At 5:00 the house jumps from 0.7 to 6.8 kWh and holds near 7 for two hours. Three hours, 6.8 + 7.4 + 5.2 = 19.4 kWh, are 19.4 / 46.3 = 42 percent of the whole day. An average of 7.4 kW for a full hour is more than a heat pump compressor draws on its own in a typical house. Something large ran most of that hour.
The evening hump. 17:00 to 20:00 runs 1.9 to 2.6 kWh per hour: cooking, laundry, lights, heat. It has the shape of people being home, and it is the part of the day a customer usually blames.
Reading the morning spike
A step to 7 kW for two or three hours on a cold morning is the classic signature of electric backup heat during recovery from a thermostat setback. The customer sets the house to 62 overnight and 68 at 5:00. The thermostat sees a 6 degree gap and calls for auxiliary heat to close it fast, and the strips, often 5 to 15 kW depending on the air handler, run alongside the compressor until the gap closes. The spike starts at the setback end time to the hour, which is the tell.
It is not the only thing that draws 7 kW for a few hours. A Level 2 EV charger on a 40 A circuit charges at up to 32 A continuous, which at 240 V is 7.68 kW. That is the same height. The two separate on shape and on what moves them:
- EV charging runs flat at nearly the same kWh every hour, then stops abruptly when the battery fills. It often starts at a time-of-use off-peak hour and does not care about the weather.
- Backup heat tapers as the house closes on setpoint (7.4, then 5.2, then 2.4 here), and it scales with outdoor temperature. Compare a mild morning: if the same house on a 45 F day shows 5:00 to 7:00 at about 2 kWh per hour with no step, the load is weather-driven.
Ask the customer one question before you reach any conclusion: what is the thermostat schedule, and does anyone in the house charge a car? Then confirm the runtime directly. See related: How to Confirm Backup Heat Runtime From the Thermostat and a Clamp.
From one day to the fortnight
Put the same 5:00 to 8:00 window for fourteen days side by side with each day's overnight low from any public weather history. In this house, say, six of the fourteen mornings show the step above 6 kWh, and all six had lows below about 25 F; the other eight, with lows in the 30s and 40s, show 5:00 to 8:00 at 1.5 to 2.5 kWh per hour. A load that appears on exactly the cold mornings and on none of the mild ones is answering to the weather. If two of the six spikes had fallen on mild mornings instead, the weather story breaks and something else is switching on.
What the record reveals that a total hides
- A constant baseload. The floor across a week of nights. A floor that jumped from 0.3 to 0.8 kW on a specific date is a new continuous load, and the date is your best interview question: what came into the house that week?
- A nightly cycling load. A block that appears every night at the same hour and lasts the same length is a timer: a pool pump, a block heater, an irrigation pump. A load that comes and goes in short pulses all night, never long, at a steady height, points at something cycling on a pressure or temperature switch, such as a well pump or a sump pump in a wet season.
- Cold-morning backup heat. Spikes that appear only on mornings below a certain outdoor temperature and grow as it falls.
- A load that never shuts off. A day with no floor dip at all, where midnight reads like noon, is a heating element or a motor that is running when it should be satisfied.
- When the change happened. A monthly bill tells you that use rose. Daily totals tell you the day it rose, which separates a new load from a weather change across the whole period.
Traps in the file
- Hour-ending versus hour-beginning. If the file labels by hour-ending and you read it as beginning, the spike moves an hour and no longer matches the thermostat schedule. Read the header or the utility's description.
- Daylight saving days. The spring day has 23 hours and the fall day has 25. A missing or doubled row on those two dates is the calendar, not the meter.
- Estimated intervals. When the meter fails to report, the utility fills the gap with an estimate, usually flagged in a quality column. A flat, perfectly smooth stretch of identical values is often an estimate, not a load.
- Solar houses. A net meter records two channels, energy delivered from the grid and energy received back from the house. A file that shows only one, or nets them together, will make a sunny afternoon look like near-zero use. See related: Net Metering + Solar Billing Reference.
- kWh read as kW. A daily total of 46.3 kWh is not a 46 kW house. Divide by the hours before you talk about power.
- One day treated as the month. The day above is a cold day. A month of milder days might average 28 kWh. Multiply the worst day by 30 and you have invented a bill.
Taking it from the record to the house
The worked day gives you a prediction you can check: the strips run at setback recovery on cold mornings. What would confirm it is the thermostat's auxiliary-heat runtime for that date matching the spike, or a clamp on the air handler supply reading the strip current during a recovery. What would kill it is the spike appearing on a mild morning too, or the thermostat showing no aux runtime on the day of the spike, at which point the next suspect is a load the customer has not named. See related: The Loads a Customer Forgets They Own.
What changes the reading entirely is the fuel. In a house with a gas furnace, a 7 kWh hour at 5:00 is not heat at all, because the furnace burns gas and its blower draws well under 1 kW. In that house the same spike sends you to the EV, a well pump, an electric water heater recovering from morning showers, or a shop on the same meter.
Never quote a cost off the record as a promise. If the 5:00 to 8:00 window used about 17.6 kWh above the floor on the worst day (19.4 kWh less 3 hours x 0.6 kW), the cost of that morning's extra is 17.6 x the rate per kWh printed on the bill, and the month depends on how many mornings were that cold. Write it that way. See related: Reading Energy Data Without Misleading the Customer.
References
- See related: Reading Customer's Electric Bill Reference
- See related: Utility Bill Spike - Source Isolation Decision Tree
- See related: Time-of-Use Rates and Equipment Settings
- See related: How to Run a Whole-House Baseline Test at the Electric Meter
- Green Button data standard documentation, as offered by the customer's own utility