How these numbers are produced

Why ranges instead of a number

A refrigerator does not have "a wattage". It has a compressor that cycles on and off, and how much energy it uses over a day depends on the unit's age and efficiency, the temperature of the room, and how often the door opens. Across common household units that is a spread of several times over.

A calculator that reports a single runtime figure has resolved that uncertainty by picking one value and not telling you. We show the whole range, and we size on the higher number.

The sizing rules

• Running watts are summed across every load you select, treating them as concurrent.
• Starting watts add only the single largest startup surge on top of the running total. Modelling every motor starting simultaneously would be unrealistically pessimistic; ignoring startup entirely is how undersized systems get recommended.
• Energy is running watts × hours × duty cycle, computed at both ends of the band.
• Usable battery energy is 85% of the nameplate rating where the manufacturer does not publish a usable figure, because inverter losses, cell cutoffs and cold weather all take a share.
• Runtime is usable energy divided by your concurrent draw. Higher draw gives the shorter runtime, so the short end of the runtime band pairs with the high end of the load band.

How our buying guides and comparisons work

Each guide is one plan (the appliances, the hours, indoors or out) run against every unit we track. The picks and the verdict come first on the page. This is the method behind them, kept here so each guide can stay short.

The three checks

• Energy: does the usable battery cover the plan’s energy, at the easy end and at the hard end?
• Running load: does the continuous output carry everything that runs at once?
• Starting surge: does the peak output clear the biggest motor’s start? A compressor pulls several times its running watts for a split second, and that spike is what shuts an undersized unit down.

“Clears the numbers” means a unit passes at the hard end. “Clears the best case only” means it passes at the easy end and not the hard one. Neither is a promise about your own night.

Hours on one charge

Hours = usable battery divided by the appliance’s average draw (running watts times the share of the time it runs). The one figure on a model page, and its chart, uses the middle of our range for that appliance; the tables show both ends. Nothing else is on the battery, and there is no solar.

Dollars per usable kWh

The Amazon price of the bare unit divided by its usable energy (85% of the maker’s battery size unless the maker publishes a usable figure). For an expandable unit we do the same sum for the maker’s own extra battery, using the pack size the maker states. We show a price only on the day we checked it, with that date, and keep no price history.

How we check a listing

A search for any model returns bundles, sibling models and, often, the same unit under two listings at different prices. We link only a listing whose own title states the bare unit and its exact battery and output. Every listing we turned down is on the model page with the reason. Check the seller is the maker or Amazon itself, and that the model number on the label matches the one on the page. Photos come from the listing we matched.

Hooking it up

• Plug-in: the station sits indoors and the appliances plug into its outlets. Nothing touches the house wiring.
• Pass-through: with the station between the wall and the appliance, the appliance keeps running when the grid drops. The switchover time, where the maker publishes it, is on the model page. A fridge tolerates a brief gap; a desktop computer may not.
• Wired loads (a well pump, a furnace, a whole circuit) need a transfer switch or power inlet installed by a licensed electrician. Never feed a station or generator into a wall outlet.
• Output quality: a compressor wants a pure sine wave inverter; avoid modified sine.
• Medical equipment: before you rely on a unit at night, check its manual for how fast it switches over when wall power fails and whether it turns its outlets off at a very low draw. Your equipment supplier is the person to ask.

What we will never rank

A product is shown for research but never ranked, never recommended and never given a buy link when:

• its recall or model identity is unresolved
• a required specification is missing, so it cannot be sized honestly
• it would need building wiring that we cannot verify for your situation

The planner enforces this before it does any sizing, so a suppressed product cannot reach a result page by being a good fit on paper.

How we decide a fact is solid

• A specification is recorded only when it appears on a manufacturer page or manual that we fetched, with the URL and capture date stored alongside it.
• Figures seen only in a search result summary, a retailer listing or a review are not recorded.
• We do not borrow a figure from a similar model in the same product family. A missing field stays missing.
• Regional applicability matters: a European manual is not a source for a US-market unit, because voltage and outlet configuration differ.
• Accessory compatibility records the manufacturer's own wording, because "sold on the same page" is not the same claim as "compatible with this model".

Current coverage across the 12 anchor stations: 128 of 144 required fields captured, 12 needing confirmation, 4 not listed by the maker.

What this site does not do

• Not an electrical design, and not a code or permit determination.
• Not a life-safety or medical guarantee.
• Not a promise that a system will keep your home safe.
• Not a substitute for reading the nameplate on your own appliances.

Sources

• Product specifications — manufacturer product pages and user manuals, captured with URL and date.
• Recall data — the US Consumer Product Safety Commission recall database.
• Appliance load ranges — category ranges across common household units, given as bands rather than point estimates, with the confirmation step you should take stated on each load page.

Corpus built 2026-08-13. The planner in your browser runs the same logic as our reference implementation, held together by generated parity tests, so the published calculator and the tested calculator cannot drift apart.