How many batteries to go off-grid? Sizing beyond backup

Backup sizing and off-grid sizing look like the same problem and are not. Backup asks how you survive one bounded event. Off-grid asks how you survive every day of the year, including the worst week your climate produces, with nothing behind you if the answer is wrong.

Our backup sizing guide works the first question: essential loads, a coverage window in hours, and a margin. This page works the second and does not repeat that arithmetic. Off-grid replaces an essentials subset with full household consumption, replaces hours with days, and adds a generation problem the backup case does not have. Those three substitutions are why the same house lands on one battery for backup and six or seven to leave the grid.

Every figure below traces to a published source. No dollar amounts appear anywhere on this page, because we have no manufacturer-published prices for these configurations and inventing them would be worse than leaving them out.

Start from consumption, and pick the right consumption number

The US Energy Information Administration reports that the average annual amount of electricity sold to a US residential electric-utility customer in 2022 was 10,791 kWh, an average of about 899 kWh per month, with Louisiana highest at 14,774 kWh and Hawaii lowest at 6,178 kWh. That page was last updated January 8, 2024. Source: EIA, How much electricity does an American home use?, read August 6, 2026.

Divide 10,791 by 365 and you get about 29.6 kWh a day. That is the anchor for everything below.

EIA attaches a caveat to its own figure that happens to matter more for off-grid sizing than for any other use of it. The 10,791 kWh number measures electricity purchased, not electricity consumed, and EIA notes that homes with net-metered solar consume more than they purchase. Its Residential Energy Consumption Survey measures household-level consumption instead, and for 2020 it estimated 7,976 kWh in Hawaii against 6,446 kWh of purchases, and 14,779 kWh in Louisiana against 14,407 kWh of purchases.

For an off-grid design, consumption is the correct basis and purchases are the wrong one, because there is no meter and no grid to buy from. If you already have solar, your bill understates what your house actually uses. Pull your own consumption data rather than your bill total, and pull a summer month and a winter month, because that spread is the thing off-grid sizing has to survive.

The three multipliers

Days of autonomy is the big one. This is the number of consecutive days the bank must carry the house with no meaningful generation, and no manufacturer publishes it because it is a fact about your weather, not about their product. It is also strictly linear: every additional day adds a full day of consumption to the bank.

At 29.6 kWh a day:

• One day of autonomy: about 29.6 kWh of usable capacity.
• Two days: about 59.2 kWh.
• Three days: about 88.8 kWh.

Compare that to the worked example in our backup guide, where essential loads averaging 0.5 kW for 20 hours land near 12 to 13 kWh usable with margin, which is one typical unit. Three days of off-grid autonomy at the national average consumption figure is roughly seven times that answer. That multiple is the honest headline of this page.

Round-trip efficiency sizes the generation, not the bank. Usable capacity is generally the energy you can draw out, so a day of autonomy needs a day of consumption in usable kWh. What the round-trip figure tells you is how much generation has to go in to refill it. At a 90 percent AC round trip, replacing 29.6 kWh of consumption requires about 32.9 kWh reaching the battery, every day, in December as well as June.

Use the published figure for the model you are actually considering, and read the basis rather than the number. Among the models we track, FranklinWH publishes 90 percent grid to battery to load, Enphase publishes 90 percent AC round trip with a separate 96 percent DC figure, Anker publishes up to 90.2 percent measured AC to battery to AC at 77 degrees F, Panasonic publishes 89 percent AC coupled and 94 percent DC coupled, and Generac publishes 88 percent system round trip. Three figures in our set are not comparable to those and should not be dropped into this arithmetic: Tesla's 89 percent is a solar round trip, meaning solar to battery to home, and is not the AC to AC number the others quote; Fortress Power's 98 percent is a cell-level figure measured at 0.5C rather than a certified AC round trip; and BYD publishes greater than or equal to 96 percent with no test conditions stated. The concept is unpacked in our round-trip efficiency explainer.

Depth of discharge decides whether the nameplate is the number. Size with usable kWh, never rated kWh. Among the models we track, only eight publish a depth-of-discharge figure at all: five publish 100 percent (SolarEdge, sonnen, EG4, Fortress Power, BYD) and three publish 90 percent (Panasonic, Qcells, Savant). The other nine publish nothing, which is the more common answer and the reason to ask. See depth of discharge.

Which tracked products even scale that far

Off-grid sizing runs into a limit that backup sizing never meets: the manufacturer's own published maximum system size. Here is what each maker publishes, set against the 88.8 kWh three-day target above.

• FranklinWH aPower 2: up to 15 units, listed at 225 kWh, per aGate. At 15 kWh usable each, six units cover 90 kWh. Comfortably clears it.
• Fortress Power eVault Max 18.5: up to 370 kWh, 20 modules in parallel. Clears it, though this is a DC module whose delivered power depends entirely on the paired inverter.
• Savant Power Storage 20: stacks up to 10 units for up to 125 kW and 200 kWh, supporting services up to 800 A. Clears it.
• Panasonic EverVolt 2.0: 9 to 18 kWh per cabinet, up to 72 kWh across four units. Covers two days and falls short of three.
• LG Energy Solution Home 8: up to 57.6 kWh across four units on one SE Box. Covers not quite two days.
• sonnen sonnenCore+: 10 or 20 kWh per unit, up to 3 systems, listed at 60 kWh per home. Covers two days.
• Anker SOLIX X1: up to 40 kWh per power module, up to 6 systems in parallel. Depends heavily on configuration, and note that Anker publishes nominal module capacity rather than a usable figure, so a usable-kWh target cannot be checked against it directly.
• Tesla Powerwall 3: power scalability up to 4 Powerwall 3 units, energy scalability up to 3 Expansion units, for a published maximum total of 7 units. At 13.5 kWh each that is 94.5 kWh, which is our arithmetic from Tesla's published figures rather than a number Tesla prints. Three days of autonomy at the national average consumption figure sits at roughly seven units, meaning at the published ceiling of one product line.
• BYD Battery-Box Premium HVS: up to 38.4 kWh by paralleling three identical HVS stacks. Roughly one day.
• Generac PWRcell: two cabinets reaching up to 36 kWh. Roughly one day.
• SolarEdge Home Battery 400V: up to 3 batteries per inverter, about 29 kWh usable. Roughly one day.
• Qcells Q.HOME CORE: up to 20 kWh per system.

None of that ranks the products. A 20 kWh ceiling is irrelevant to a buyer sizing backup and decisive to a buyer sizing off-grid, which is the entire point of separating the two questions. Figures are as recorded in our model data from manufacturer datasheets; the full set is on our comparison pages.

Off-grid operation is a published capability, and some makers do not publish it

Most home batteries are grid-tied products with a backup mode, which is not the same thing as an off-grid product. Here is what each datasheet actually says.

States or rates off-grid operation:

• Tesla Powerwall 3 publishes a configurable maximum continuous discharge power off-grid of 15.4 kW, qualified as PV only and rated from -20 to 25 degrees C, with a footnote stating it is available only if the on-grid rating is 11.5 kW and that the unit must be installed with an 80 A breaker and appropriately sized conductors.
• FranklinWH aPower 2 states the system is off-grid ready and designed to operate independently from the main power grid.
• Enphase IQ Battery 5P and 10C carry a footnote stating the rating is supported in both grid-tied and backup or off-grid operations, with the 5P adding an explicit exception: backup and off-grid are not supported for 208 V single-phase operation.
• Anker SOLIX X1 publishes a separate AC Output (Off-Grid) table with its own rated and peak power figures, distinct from the on-grid table.
• Savant Power Storage 20 publishes an off-grid power rating of 52 A and separate off-grid surge figures of 24 kW for 400 ms, 19.1 kW for 1 second, and 16 kW for 10 seconds, all lower than the on-grid surge figures on the same sheet.
• Panasonic EverVolt 2.0 publishes two off-grid continuous backup power rows, one with solar and one without, and they differ: on the 10 kWh system, 5.5 kW without solar and 7.6 kW with it.
• BYD Battery-Box Premium HVS lists applications as on grid, on grid plus backup, and off grid.
• Fortress Power eVault Max 18.5 names off-grid among the applications it is built for.

Does not state off-grid on the datasheet we read:

• SolarEdge Home Battery 400V describes on-grid and backup power applications and does not use the word off-grid anywhere on the sheet.
• Generac PWRcell 2 describes whole and partial home backup and pairing with a Generac standby generator, and does not claim off-grid operation.
• Qcells Q.HOME CORE states microgrid supported on the Q.VOLT inverter, which is adjacent to but not the same claim as off-grid.
• EG4 PowerPro WallMount is a battery module rather than a system, so off-grid capability is a property of the paired inverter and not of this datasheet.

Two of those published details are worth more than the rest. Savant's lower off-grid surge ratings and Panasonic's split between off-grid with and without solar both say the same useful thing: a system's off-grid behavior is not simply its on-grid behavior with the utility removed. If off-grid is your plan, the on-grid power figure on the front page of a brochure is not your number.

The generation half of the problem

Storage without generation is a countdown timer. An off-grid design has to refill the bank every day, and the day it must survive is not the average day.

The tool for this is NREL's PVWatts Calculator, which is free, published by a national laboratory, and gives month-by-month estimated production for a specific location, system size, tilt, and azimuth. Run it for your address and your intended array, then read the December output against the June output. That ratio is your seasonal deficit, and it decides how much array you need, because an array sized to the annual average will not carry the house in the month it matters. We are not going to print a national figure for that ratio: it varies too much between locations for a single number to be useful, and a system sized to a national average is a system sized to fail somewhere.

The consequence, and the reason most people who look at this seriously stay grid-connected: sizing the array to the worst month means overbuilding it for the other eleven, and sizing the bank to three days of autonomy means most of that capacity sits unused most of the year. Both are correct engineering for a house with no fallback, and both are why off-grid multiplies the hardware rather than adding to it. Grid-tied storage gets to be sized to the useful case because the grid absorbs the tail. Off-grid storage has to be sized to the tail.

If what you want is resilience rather than independence, the grid-tied answer is usually the better buy, and our calculator will run that version with your real usage and rate. If you want a battery without solar and without leaving the grid, that is a different question again, covered in home battery without solar.

Everything here is arithmetic from published figures, with sources and read dates named. We have not tested any of these products, and no configuration described here is a design. An off-grid system is engineered by a qualified professional against your own consumption data and your own site, not from a national average.