Half-Hourly Kilowatt-hours on One Side, an SI Model on the Other
A smart meter export is a long, narrow file: a timestamp and a kilowatt-hour figure, repeated every thirty or every fifteen minutes for a year. It is an excellent record and an awkward input. Building simulation engines, thermal network models and most physics-derived code count in joules, so the first thing an analyst does with an interval series is restate it — and, in the same pass, work out what average power each interval implies, because that is the quantity a load profile is actually plotted in.
Three Things Worth Fixing Before the Series Is Modelled
Kilowatt-Hours per Interval Is Not Kilowatts
Simulation Engines Count in Joules
The Overnight Floor Is the Base Load
Converting a Channel of Interval Readings Before It Reaches the Model
Whether you are checking one suspicious interval or setting the constant for a whole pipeline, the sequence is the same.
Take one representative interval out of the series
Pick a peak slot, a mid-afternoon slot and one from the overnight shelf. Values from an export often arrive with a comma decimal or padding spaces; both are accepted here, so a figure can be pasted in exactly as the file wrote it.
Read the kilojoule value the model will consume
The second field carries the same energy in kJ. Multiply by a thousand if the model wants joules outright — and if you are working at whole-day or whole-site scale, the unit list reaches MJ and GJ so the exponent never has to be handled by hand.
Divide by the interval length to recover average power
1 512 kJ over 1 800 seconds is 0.84 kW, the same answer as doubling the half-hourly kilowatt-hours. Doing it both ways once is a cheap sanity check on whether the file's interval really is thirty minutes.
Swap when the model hands a joule figure back
Simulation output arrives in J or MJ, and the only fair comparison is against the metered series. Reversing the fields brings a modelled daily total back to the kilowatt-hours the meter recorded, which is the residual you actually calibrate on.
Interval and Daily Load Profiles in Kilojoules and Average Kilowatts
Each row is a real slot out of a domestic interval series: the energy as the meter reports it, the same energy in SI, and the average power it implies over the length of that slot.
| Load and interval | Metered energy | Same energy (kJ) | Average power |
|---|---|---|---|
| Standing load, one half hour | 0.06 kWh | 216 kJ | 120 W |
| Kettle boils, one half hour | 0.30 kWh | 1,080 kJ | 600 W |
| Washing machine, one half hour | 0.55 kWh | 1,980 kJ | 1.1 kW |
| Heat pump, cold morning half hour | 1.80 kWh | 6,480 kJ | 3.6 kW |
| Electric shower, 15-minute slot | 2.25 kWh | 8,100 kJ | 9 kW |
| Home charger, one half hour | 3.50 kWh | 12,600 kJ | 7 kW |
| Fridge-freezer, whole day | 1.10 kWh | 3,960 kJ | 46 W |
| Whole home, whole day | 9.50 kWh | 34,200 kJ | 396 W |
The last two rows are the interesting ones. A fridge averaging 46 W over a day never draws 46 W for a single second — it cycles a compressor of a hundred watts or more for part of each hour — and the whole-home average of 396 W hides both the overnight shelf and a 9 kW shower. Averaging is where interval data quietly destroys information, and knowing at which resolution that happened is half of reading a profile correctly.
What Speeds Up the Trip From Export File to Model Input
Step Through a Day of Half-Hours Without Clearing
Every keystroke recomputes, so a suspicious morning can be walked interval by interval — type the next reading over the last one and the kilojoule column keeps pace with you.
Numbers the Simulation Reads Without a Unit String
The copy control returns digits with no unit and no thousands spacing, which is what a CSV column, a JSON payload or a notebook variable expects. Ctrl+C from inside the field gives the same clean value.
Read a Modelled Joule Figure Back as Metered kWh
Calibration runs the other way. Reversing the direction turns a simulation's joule or megajoule output into the kilowatt-hours the meter actually recorded, so the residual is computed in the units the data came in.
Aggregate Channels That Arrive in Wh, MJ or toe
Sub-metering rarely uses one unit across a portfolio. Searching either dropdown reaches all 23 energy units, so a circuit logged in watt-hours and a gas channel logged in megajoules can be lined up on one scale.
Questions About Interval Data, Average Power and SI Units
A half-hourly channel says 0.42 kWh — what is that in kilojoules and in kilowatts?
1,512 kJ of energy, and 0.84 kW of average power. The two answers come from the same figure treated differently: multiplying by 3,600 restates the energy, while dividing by the length of the interval gives the rate. Because a half hour is 1,800 seconds, 1,512 kJ ÷ 1,800 s lands on 0.84 kW, which is simply the original 0.42 doubled. On a fifteen-minute feed the same energy would imply 1.68 kW. Nothing about the interval length changes the joules; it changes only what power they represent.
Why does a building simulation ask for joules when the meter reports kilowatt-hours?
Because everything else in the model is already SI. Thermal capacitance is in J/K, specific heat in J/(kg·K), conduction in W/(m·K), and a solver stepping in seconds accumulates watt-seconds — which are joules. Introducing kilowatt-hours means a factor of 3,600 riding along in every energy term, applied in some places and forgotten in others. Converting the measured series once, at import, keeps the whole model dimensionally consistent and leaves the conversion in exactly one place where it can be tested.
How do I read a base load out of a year of interval data?
Take the fifth percentile of all intervals rather than the outright minimum, which is usually an outage or a gap. In a domestic series that percentile typically lands somewhere near 0.05–0.10 kWh per half hour, meaning 180–360 kJ and a standing draw of 100–200 W. Compute it separately for summer and winter: anything that appears in the winter floor but not the summer one is heating-related rather than truly constant. The gap between that shelf and the daily mean is the part of the profile worth disaggregating.
What do I lose going from one-minute data to half-hourly?
Peaks and switching events, but not energy. Totals over a day agree exactly because averaging conserves the integral; what disappears is everything shorter than the window. A 9 kW shower running for eight minutes shows as 9 kW in minute data and as roughly 2.4 kW in the half hour that contains it. Compressor cycling, kettle spikes and inrush all vanish the same way. If your question is annual consumption, half-hourly is ample; if it is peak demand, diversity or appliance identification, the resolution has already thrown away the evidence.
Can I pull individual appliances out of a single whole-home series?
Partly, and the resolution decides how far you get. At half-hourly granularity the large, long and distinctive loads separate reasonably well — a 7 kW charger, an electric shower, a heat pump tracking outdoor temperature — because each adds a step far above the shelf. Small overlapping loads do not, since three appliances inside one thirty-minute slot arrive as a single number. Serious appliance disaggregation works at seconds and looks for edges in the power signal, so the honest use of an interval file is grouping loads into a handful of end uses rather than naming them individually.
No comments yet. Be the first to comment!