Off-Grid Solar Sizing Calculator: Array, Battery and Inverter by State

Updated September 3, 2026. Quick answer: an off-grid system is sized by the worst month you intend to use it, and for a latitude-tilted array that month is December in all fifty states and the District of Columbia. A cabin in Vermont gets 1.81 peak sun hours a day in December against 3.79 as an annual average, so a system sized on the annual figure is roughly half the system it needed to be. Panels flat on an RV roof are worse again: 1.08 in the same state and month. Everything below runs from published climatology and published equipment datasheets, both named.

If you are pricing the kit route rather than designing it, this site’s solar kit matcher sorts by project type, utility situation, load range and storage and returns a starting category. Sizing still has to come first: a matcher cannot tell you what you need, only where to look once you know.

Why December, not the annual average, decides the size

A grid-tied system can be sized on an annual average, because the grid quietly covers every shortfall and banks every surplus. An off-grid system has no such counterparty. On the worst day of the worst month, the array either replaces what the loads took out of the battery or it does not, and if it does not for long enough, the lights go out.

The size of that gap is not small. At each state’s 2020 Census center of population, using the NASA POWER climatology for 2001 to 2020, a fixed array tilted to the local latitude sees these December figures against its own annual average:

• Washington: 1.80 peak sun hours in December, 4.23 as an annual average. The December array is 2.4 times the average-sized one.
• Vermont: 1.81 against 3.79.
• Michigan: 1.94 against 4.08.
• Arizona: 4.90 against 6.52, a ratio of only 1.3. Sunny places are forgiving of this mistake; northern ones are not.
• Alaska: 0.52 against 2.92. At half a peak sun hour a day, the arithmetic still returns an array size, and the answer is that solar alone is not the design.

December is the worst month at latitude tilt everywhere in the United States. It is not always the worst month at other tilts: at a steep winter tilt of latitude plus fifteen degrees, the worst month becomes January in California and June in Florida, because a steep panel gives up so much summer sun that a cloudy Florida June beats its own December. Florida’s June and July are in fact tied at 4.49, and the tool names June. If you use the winter-tilt option, read the month the tool names rather than assuming December.

Where the efficiency numbers come from, and the one we will not invent

Between the sunlight and the load there are four separate losses. Three of them are published specifications and are used here as such. The fourth is not, and this page declines to pretend otherwise.

Charge controller. The SmartSolar MPPT datasheet states plainly: “Maximum efficiency exceeds 98 %.” The tool uses 98 percent.

Battery round trip. The same manufacturer’s lithium datasheet gives “Round trip efficiency 92 %” for LiFePO4 and states that “the round-trip energy efficiency (discharge from 100 % to 0 % and back to 100 % charged) of the average lead-acid battery is 80 %”. Those two figures are the reason a lead-acid bank needs a visibly larger array for the same load, before any question of usable capacity arises.

Inverter. The Phoenix 1200 VA to 5000 VA datasheet gives maximum efficiency by system voltage: 93 percent at 12 V, 94 percent at 24 V and 95 percent at 48 V on the 3000 VA model. The tool uses those three. Note the word maximum: it is the efficiency at a favourable load, not an all-day average.

Soiling, shading, wiring and hot panels. This is the one we do not have. No source read for this page fixes a loss figure for an arbitrary site, and the obvious source for it, NREL’s PVWatts and its published default losses, was unreachable for the whole time this page was built: the nrel.gov domain did not resolve from any of three independent name services, while other federal domains answered normally on the same queries. Rather than borrow a number from memory and present it as sourced, the tool exposes it as a design margin you set, defaulted to 25 percent and labelled as a margin. If you know your site is clean, unshaded and cool, take it down. If a spruce shades the array at ten in the morning in December, no margin fixes that and you should move the array.

What the inverter costs you before it powers anything

Every sizing exercise counts loads. Very few count the inverter itself. From the same datasheet, zero-load power, meaning what the inverter draws with nothing plugged into it, runs from 8 W to 35 W across the range: 8 W at 12 V on the 1200 VA model, 10 W at 24 V, and 25 W at 48 V on the 3000 VA. Leave a 10 W inverter powered up for sixteen hours a day and it has taken 160 Wh out of the battery before a single appliance runs. On a small cabin that is a larger line item than the lighting.

That is why the tool asks how many hours the inverter is on, not how many hours it is used. The same datasheet lists a search mode drawing 2 W to 15 W, which is the design answer to this problem where the loads tolerate it.

The second inverter figure worth reading is the temperature one. The 1200 VA model is rated 1000 W continuous at 25 C, 900 W at 40 C and 600 W at 65 C. The 3000 VA is 2400 W, 2200 W and 1700 W on the same three rows. An inverter chosen to exactly match a peak load on a bench will not hold that load in an uninsulated summer cabin, so the tool divides your peak by the 40 C ratio rather than the 25 C one.

Usable capacity is not rated capacity

A 100 Ah battery does not give you 100 Ah, and the two chemistries fail to give it to you for two entirely different reasons.

Depth of discharge. The lithium datasheet publishes cycle life against depth of discharge: 2500 cycles at 80 percent, 3000 at 70 percent, 5000 at 50 percent, all measured to the point where capacity has fallen to 80 percent of nominal. The AGM datasheet publishes 400 cycles at 80 percent discharge, 600 at 50 percent and 1500 at 30 percent. Those two ladders are why this tool plans 80 percent of a lithium bank as usable and 50 percent of a lead-acid one: both are the manufacturer’s own roughly-2500-and-600-cycle points, not a rule of thumb. If you want a decade out of a lead-acid bank rather than a few winters, plan 30 percent and size accordingly.

The discharge rate, which only lead acid punishes. The AGM datasheet is explicit that its rated capacity is a 20-hour figure, and publishes what you get at other rates:

An off-grid evening load empties a bank in something closer to five hours than twenty, so a 100 Ah AGM asked to run an evening delivers about 85 Ah, and the datasheet adds that the reduction “will be even faster in case of a constant power load, such as an inverter”. This is on top of the depth-of-discharge limit, not instead of it, and it is the single most common reason a lead-acid bank sized on paper disappoints in the field. The tool does not apply this second reduction automatically, because it depends on your load shape rather than on your state, but if your bank empties over an evening rather than over a day, add roughly 15 percent to the lead-acid answer it gives you.

Peak sun hours by state, worst month and annual average

These are the numbers the calculator reads. Each row is one point, the state’s 2020 Census center of population, which is where its people actually are rather than its geometric middle. A cabin two hundred miles north of that point in a cloudier valley will do worse, and one on a south-facing ridge will do better.

Where an RV or a van differs from a cabin

Only two things change, and neither is the arithmetic.

The first is the roof. Panels bonded flat to an RV or van roof cannot be tilted, and the flat-roof column above is what that costs: 1.08 peak sun hours in Vermont in December against 1.81 at latitude tilt, or 3.10 in Arizona against 4.90. Across the 51 jurisdictions the median December loss from lying flat is 39 percent, with the middle half of states between 36 and 44 percent; Hawaii loses the least at 19 percent and Alaska the most at 73 percent. In a stationary installation, tilting the same panels buys more than adding panels does, which is why the tool asks about the mount before it asks about anything else.

The second is that a vehicle moves, and the state you are parked in is the state that matters. Someone wintering a van in Arizona is sizing an Arizona system regardless of where it is registered. Size for where the vehicle will actually be in its worst month, and if that is genuinely unknown, size for the worst place you expect to sit still in.

What does not change is the load side. An inverter idling in a van costs the same 8 W to 25 W it costs in a cabin, and a lead-acid bank in a van is punished by the discharge rate exactly as it is in a cabin, usually more so because the loads are concentrated into an evening.

What this tool does not do

It does not price anything. It returns an array size in watts, a bank size in watt hours and amp hours, a controller current and an inverter rating; what those cost depends on a market this page does not track.

It does not design a system. Wire gauge, fusing, disconnects, grounding, series and parallel string layout, controller voltage ceilings and the electrical code that applies to a dwelling are all outside it, and all of them can make a correctly sized system unsafe.

It does not model your site. It samples one coordinate per state and a twenty-year climatology. Local cloud, elevation, snow cover on the panels and shading are all real and none of them are in the number.

It does not know your loads. Every figure it returns is downstream of two numbers you typed. If you are guessing at the watt hours, measure them for a week with a meter before buying anything, because that guess dominates every output here.

If what you are actually comparing is a financed grid-tied system, two other pages matter more than this one, because the 30 percent federal credit ended on 31 December 2025 and the dealer fee inside a solar loan is usually larger than any sizing decision you will make.

Scope, method and sources

Sun hours come from the NASA POWER Climatology API, parameters SI_TILTED_AVG_LATITUDE, SI_TILTED_AVG_LAT_PLUS15 and SI_TILTED_AVG_HORIZONTAL, community RE, queried on September 3, 2026 once per jurisdiction. The API returns its own provenance with every response: a twenty-year monthly and annual climatology covering January 2001 to December 2020, source SYN1DEG, in kWh per square metre per day, which is the same quantity as a peak sun hour. Coordinates are the U.S. Census Bureau’s 2020 Census centers of population by state. Puerto Rico is in that file and is excluded here; fifty states and the District of Columbia remain.

Equipment figures are read from published manufacturer datasheets: Victron Energy’s Lithium Battery Smart sheet for round-trip efficiency and lithium cycle life, its GEL and AGM Batteries sheet for lead-acid cycle life and the effective-capacity table, its Phoenix Inverter 1200 VA to 5000 VA sheet for inverter efficiency, zero-load power and the temperature derating, and its SmartSolar MPPT sheet for charge-controller efficiency. One manufacturer’s sheets were used throughout so the numbers are internally consistent and every one of them is checkable against a single named document. They are that manufacturer’s published figures for its own products, not universal constants, and naming them is not a recommendation to buy them; substitute your own equipment’s datasheet where you have it.

The arithmetic, in full: energy the battery must deliver each day equals the DC load, plus the AC load divided by inverter efficiency, plus zero-load power multiplied by the hours the inverter is on. Bank size equals that figure multiplied by days of autonomy and divided by the planned depth of discharge. Array size equals the daily battery energy divided by round-trip efficiency, divided by charge-controller efficiency, divided by the sizing month’s peak sun hours, multiplied by one plus your design margin. It assumes every watt hour passes through the battery, which is true of the worst month and conservative in the best. Controller current is array watts divided by nominal system voltage, which is deliberately the conservative form.

General information about sizing, not an electrical design, not an installation instruction, and not advice about the code that applies where you live.