How much battery do you actually need for a hurricane outage?
Scenarios Updated 2026-08-07
Count it in days, and the arithmetic stops being reassuring. A refrigerator, a box fan, a router and modem and a CPAP machine cost 1,917.4 Wh for one day on our load records, so a three-day outage is 5,752.2 Wh — and the largest machine in this library that publishes enough to prove a runtime holds 1,634.2 Wh of usable energy, which is 20 h of that set and 28% of the three days.
So a recharge is not an accessory to the plan, it is the plan: solar, a car, or a full charge taken while the grid is still up and the storm is still offshore. The battery decides how comfortable the first day is; the recharge decides whether there is a third one.
The other half of the problem is that a hurricane brings heat, water and drainage with it, and those loads are decided by start-up current rather than by capacity. Of the 12 published US records here, 7 can prove how long they last, 1 starts a window air conditioner, a well pump and a sump pump, and the overlap between those two groups is 0.
What one storm day actually costs
A storm day, priced off the load records
Four loads, each priced from its own maker’s figures. Every row is our multiplication or division of them, because we have found no published daily total for any of them; each row states the arithmetic. A load supporting neither route is left out rather than estimated in.
The largest of the sensible loads, and the one that decides how much of a storm you can sit through. It cycles rather than running flat out, which is why this is well under its watts times a full day.
Frigidaire’s specification sheet for this model states 410 kWh a year; this row is that figure divided by 365 days. The annual number is the maker’s — it is the FTC EnergyGuide figure — and the division is ours.
See the record →The maker’s own low-speed figure over one night. On high it runs to the upper end of the record’s band — the difference between a fan you can run all week and one you cannot.
Lasko publishes 50 W on low for this model; the eight-hour night it is multiplied by is ours, not the maker’s, so this row is our arithmetic on one published figure. This load also has no record page of its own — we have found no published start-up current for any fan maker, and a load page that cannot state one would mislead, so the daily energy is quoted here and the start-up demand is left blank everywhere on this site.
Run it in the calculator →Running watts times the 24 h a day this record states it operates for. Both figures are on the load record; neither is assumed here.
See the record →The maker’s own measured draw with humidification, over one night. An electrical figure and nothing more — what a device needs is a question for whoever prescribed it.
ResMed’s power-consumption table gives 22 W for this device with humidification, and ResMed’s own battery guide sizes on an 8-hour night. This row is those two figures multiplied together. Both terms are the maker’s; the multiplication is ours, and we have found no published per-night total from ResMed.
See the record →The sum of the four rows in this card, and the figure every division on this page is made against.
Not in the total, because it is the fallback rather than the plan. When the local network goes down with the grid, the satellite terminal that replaces the router costs several times what the router did.
Starlink’s 40 W — the top of its published running band, and the one end its two current documents agree on — times twenty-four hours. Our arithmetic, not a Starlink figure: the company publishes a draw and no daily total. The bottom of that band is disputed between two Starlink documents, so the conservative end is the one quoted here.
See the record →Derived by WattPairverified 2026-08-01storm-day load records
What that buys on the machines that can prove an answer
Usable energy, divided into that day
7 records here publish a usable capacity, which is what it takes to turn a battery figure into hours at all. The rest publish a nominal capacity only, so no honest hour figure exists for them and none is printed.
Not the number on the box. The maker’s own depth-of-discharge and inverter-efficiency figures are already out of it, which is why it is the one worth dividing by.
See the record →This record’s published usable energy of 1,634.2 Wh divided by the storm-day total in the card above, times twenty-four. Only records that publish a usable capacity are turned into hours at all.
See the record →This record’s published usable energy of 1,586.3 Wh divided by the storm-day total in the card above, times twenty-four. Only records that publish a usable capacity are turned into hours at all.
See the record →This record’s published usable energy of 1,011.7 Wh divided by the storm-day total in the card above, times twenty-four. Only records that publish a usable capacity are turned into hours at all.
See the record →This record’s published usable energy of 856 Wh divided by the storm-day total in the card above, times twenty-four. Only records that publish a usable capacity are turned into hours at all.
See the record →This record’s published usable energy of 829.4 Wh divided by the storm-day total in the card above, times twenty-four. Only records that publish a usable capacity are turned into hours at all.
See the record →This record’s published usable energy of 246.2 Wh divided by the storm-day total in the card above, times twenty-four. Only records that publish a usable capacity are turned into hours at all.
See the record →This record’s published usable energy of 230.4 Wh divided by the storm-day total in the card above, times twenty-four. Only records that publish a usable capacity are turned into hours at all.
See the record →On top of every figure in this card, for as long as the machine is on. Over a full day it is a load in its own right, and the term most sizing advice forgets.
Official spec + Derived from official coefficients + Derived by WattPairverified 2026-08-08Jackery Explorer 2000 Plus Portable Power Station · US record
A three-day outage on that set is 5,752.2 Wh. Nothing in this library holds it, and the largest record that can prove its own runtime covers 28% of it. That is not a reason to buy a bigger battery — it is the reason the recharge, not the capacity, is what you are actually planning. A machine that goes back to full every afternoon on a panel lasts the whole storm; one that does not lasts until it does not.
The three loads a storm makes you want
Cooling, water and drainage, computed on this build
The percentage is the running load as a share of continuous output, and on most rows it sits comfortably inside the machine — which is exactly why reading it alone would mislead. What decides each row is the line underneath it.
Refused at the start, not at the run. The compressor’s locked-rotor figure is several times its running draw, and we have found no published surge ceiling for this record, so the engine uses its continuous output as a conservative floor rather than assuming a multiple.
Show the working →A published surge ceiling, and still a refusal — the start-up demand of a fixed-speed room unit is above it. Cooling a room is the hardest thing a household asks of a portable battery, and it is hard on the start rather than on the watts.
Show the working →The rural half of the same problem. A submersible pump’s start is in the same class as the air conditioner’s, so a house on a well loses its water supply for the same reason it loses its cooling.
Show the working →The one that clears, and it is the smaller machine. Its published surge ceiling is above the pump’s start where the larger record’s conservative floor is not — the clearest case on this site of a verdict being computed rather than looked up.
Show the working →The only published US record here that starts all three of these motors. We have found no published usable capacity for it, so this site will not tell you how long it runs any of them — the trade-off this whole page is about, in one row.
Show the working →Side by side, those two lists are the honest state of the market rather than an opinion about it. 7 of the 12 published US records here can be turned into an hour figure, because they publish a usable capacity. 1 starts a window air conditioner, a well pump and a sump pump. The overlap contains 0 machines. You can buy a proven number of hours or a proven motor start; on today’s published data you cannot yet buy both in one box, and any page saying otherwise has filled in a figure its manufacturer did not.
The swap that changes the arithmetic
The same machine that runs a window air conditioner for 2.1 h — supposing it started at all, which it does not — runs a box fan for 16 h. Same battery, same inverter, same night; the only thing that changed was the appliance. Moving air is what a battery is good at and removing heat from a room is what it is not, and no amount of capacity fixes a start-up refusal. A window unit run off the grid before landfall, a well-sealed room, and a fan afterwards is a plan that works on these numbers. Cooling a house on a battery is not.
What to do while the grid is still up
- Charge to full early, from the wall. A machine at eighty per cent when the lights go out has lost a fifth of the plan.
- Pre-chill hard. Turn the fridge and freezer down a day ahead and fill the empty space with water containers — cold mass is stored energy that costs nothing to keep, and it buys compressor hours you would otherwise spend from the battery.
- Work out your own storm day rather than borrowing this one. Four loads describes a household with no well and no basement; add either and the total moves enough to change the answer.
- Decide the recharge now, not during. A panel needs somewhere safe to sit that still sees sky after the wind stops, and a car needs fuel nobody will be selling for several days.
- Label the extension leads and tell the household what plugs into what. The commonest failure of an outage plan is not a battery running out; it is a fridge nobody moved onto the battery until morning.
Where the storm facts come from
One source for what the storm is doing. A forecast repeated second-hand carries an unknown timestamp.
The primary source for Atlantic and eastern Pacific advisories: track, intensity, watches and warnings, each issued as a numbered advisory with the time on it. Read the advisory number and its issue time rather than a headline about it — that timestamp is what tells you whether what you are reading is current.
Open the source →Restoration sequence and estimated times come from the utility that owns the line, and from nowhere else. Find yours and save it now — every outage map is slowest at the moment you need it.