How to Calculate Your Home's Energy Load for an Off-Grid System
Updated 8 October 20267 min readNext-Gen & Off-Grid
Sizing an off-grid system starts with a question that sounds simple and is really three wearing one coat. How much energy does the house use in a day? How much power does it draw at once? And what is the largest instantaneous demand anything in it can make? Those size different parts of the system, and satisfying one can fail badly on another.
Key takeaways
- Energy per day sizes the array and the battery; peak power sizes the inverter; surge sizes the inverter's overload capability.
- Nameplate ratings are worst-case numbers, and most appliances spend most of their time well below them — so measurement beats arithmetic.
- Anything with a thermostat or a motor is really a duty cycle, and its average draw is what matters, not its running draw.
- Usable capacity is always smaller than nominal: depth of discharge, round-trip efficiency and temperature each take a share first.
- An off-grid array is sized for the worst month, not the average one — averages describe a year nobody actually lives through.
On this page
Three numbers, three different jobs
Before any arithmetic, separate the quantities. They are measured differently, they behave differently, and each one sizes a different part of the system.
Energy is power multiplied by time, in watt-hours (Wh) or kilowatt-hours (kWh). It is the area under the day's curve, and it sizes the battery and the array. Continuous power is the highest level the house sustains for minutes at a time, in watts (W), and it sizes the inverter. Surge is the momentary demand when a motor starts — potentially several times its running power for a fraction of a second — and it is measured against the inverter's short-term overload rating rather than its continuous one.
A house with modest daily energy use can still have a punishing surge, and one with large daily use can have a gentle one. Neither number predicts the other.
Building the load table
The load table is the whole exercise. Every device, what it draws, and how long it actually runs.
Anything with a thermostat or a motor spends much of its life switched off, so the duty cycle is what matters: a compressor running a third of the time contributes roughly a third of its running power on average. Wrong in the optimistic direction, the battery empties overnight; wrong the other way, you have sized for a house nobody lives in.
| Device | Running power | Hours or duty | Energy per day | Notes |
|---|---|---|---|---|
| Lighting, whole house | 40 W | 5 h | 200 Wh | Constant while on |
| Refrigerator | 90 W | 33% duty, 24 h | 710 Wh | Duty cycle, not run time |
| Laptop and router | 45 W | 10 h | 450 Wh | Router continues overnight |
| Water pump | 500 W | 0.5 h | 250 Wh | Surge at every start |
| Washing machine | 400 W avg | 1 h, 3 days a week | 170 Wh | Averaged across the week |
| Standby loads, everything | 15 W | 24 h | 360 Wh | The total nobody expects |
| Daily energy | ≈ 2140 Wh | Sizes battery and array | ||
| Peak continuous | ≈ 700 W | Sizes the inverter | ||
| Largest surge | pump start | Sizes overload capability |
Illustrative arithmetic, not typical values — your own devices and hours are what belong in these columns. The point is the shape of the calculation and the three totals it produces.
Two rows deserve comment. Standby is the line that surprises people: a handful of watts drawn continuously by devices that appear to be off becomes a meaningful share of a small system's daily total, because it runs every hour of every day. And peak continuous is not the sum of everything in the table — it is the largest combination that realistically runs at the same moment, which takes judgement rather than addition.
Measure rather than declare
The single most valuable step in this whole process costs nothing but patience: measure the devices instead of reading their labels.
A plug-in energy meter left in place for several days gives an energy total that already includes duty cycling, standby and the way the household actually behaves — three things arithmetic tends to get wrong in the same direction. Where metering is impractical, err towards the nameplate for resistive loads and towards measurement-based estimates for everything with a motor or a thermostat.
From daily energy to a battery
Here is where most sizing goes astray: the battery's nominal capacity is not the energy available to the load. Several things take a share first.
Start with daily energy and multiply by days of autonomy — how long the pack must carry the house with no meaningful generation. Then divide by the usable depth of discharge, because taking a pack to empty is how it is destroyed rather than how it is used; the trade between depth and cycle life is the subject of depth of discharge and it is a design decision, not a constant. Divide again by round-trip efficiency, since more energy goes in than comes back out. Finally apply a temperature derating if the pack lives somewhere cold, because capacity falls when it is cold — the effect described in how temperature affects battery performance.
The pack also needs to satisfy the *power* question, not just the energy one: it must be able to deliver the peak continuous draw within the limits its battery management system enforces. A pack big enough in kilowatt-hours can still be current-limited at exactly the wrong moment.
From daily energy to an array
The array calculation has the same shape and one crucial difference: it is sized for the worst month rather than the average one.
Take the daily energy, divide by the peak sun hours available at the site in the worst month, and then divide again by the cumulative system losses between the array and the load — soiling and temperature at the modules, wiring, the charge controller, the battery's round trip, and the inverter's conversion. Each stage takes a percentage, and they multiply rather than add.
Peak sun hours vary enormously with latitude, climate and shading; they are site data rather than something to assume, and irradiance data sets published by bodies such as NREL exist for exactly this. What is reliable is the shape: winter days are shorter, the sun is lower and cloud more persistent, so the same array yields far less — for the reasons set out in how panels behave under cloud. Heating loads peak in those same months, worst of all an air-source heat pump in deep cold; micro-hydro, where a stream allows, keeps running through the winter.
Where the estimate goes wrong
Four failure modes account for most disappointed off-grid systems, and all of them are errors of method rather than arithmetic.
Growth. The table describes the house on the day it was written. Households acquire devices, and a system sized exactly to today's table has no room for next year's freezer or heat pump.
Simultaneity. Adding every device's running power gives a peak that will never occur; assuming nothing coincides gives one that certainly will be exceeded. Judgement, not addition.
Forgotten continuous loads. Standby, a router, a controller, a fan — small, invisible, running every hour of the year.
Margin stacked on margin. Rounding up at each stage compounds into a system that spends its life lightly loaded, where its own standby consumption becomes a large share of what it actually delivers.
The discipline that avoids all four is the same: write the table honestly, measure what you can, size each stage against the number that governs it, and put the margin where you chose to put it.
Frequently asked questions
Should I use nameplate ratings or measured values?
Measured, wherever you can. A nameplate is a worst-case declaration, and most appliances draw well below it. The exceptions are things that genuinely run at their rating — resistive heaters, kettles, hobs. For everything else a plug-in energy meter left in place for a few days answers it properly, because it catches duty cycles and standby too.
What are days of autonomy?
How many days the battery alone can carry the load with no meaningful generation — the cloudy-week question. More autonomy means a larger pack that spends its life partly full, which is gentler on the cells but a great deal more storage to own. It is a judgement about how a bad week should feel, not something the load table can give you.
Why does surge matter if it only lasts a moment?
Because a motor starting can demand several times its running power for a fraction of a second, and an inverter that cannot supply it shuts down instead — the house goes dark because a pump started, even though the average load was comfortable. Inverters publish a continuous rating and a short-term overload rating; surge is measured against the second.
Can I just add a safety margin to everything?
Margin in the right place is prudent; margin everywhere compounds. Oversizing every stage produces a system that spends its life lightly loaded, where standby consumption becomes a large fraction of what is actually delivered. It is better to be honest about the load and deliberate about where the margin goes.
How do I account for seasonal differences?
Build the load table twice, for a representative winter day and a representative summer day — both the load and the available generation move, usually in opposite directions. Size the array against whichever month has the worst ratio of demand to sunlight, which for most sites is a winter one.
Sources
Named organisations whose published material underpins this article. Where no link is given, the source is named rather than linked.
- National Renewable Energy Laboratory (NREL)Solar resource data and photovoltaic system performance modelling.
- U.S. Department of EnergyBackground on appliance energy use and standalone power systems.
- Equipment data sheets — inverters, charge controllers, batteriesContinuous and surge ratings, efficiency curves and temperature derating are published per product and govern any sizing calculation.
Editorial Team
Research, drafting and review
Articles are drafted from primary engineering and physics references with AI-assisted tools, then reviewed and fact-checked line by line by a human editor before publication. We publish explanations, not recommendations: no products, no pricing, no country-specific rules, and no invented author personas.
Last reviewed 8 October 2026. How we research and review