Skip to content
Tech SolarEnergy

How we size solar systems

This page explains where our numbers come from, how we turn your answers into a system, how we choose and check the parts, and where our method has limits. We publish it so you can check our reasoning. Terms are explained in our glossary.

Data sources

Our source of solar radiation and temperature data is PVGIS (Photovoltaic Geographical Information System), developed and maintained by the European Commission Joint Research Centre (JRC). We use hourly data for every available year (2005 to 2023, or 2005 to 2015 for NSRDB), already calculated for the angle and direction of your panels and including shading by hills and mountains on the horizon:

We store the data we fetch so that the same location is not requested twice. We look up data for a grid square of about 5 km that contains the point you give us, never for an exact address. Points outside the United States and the United Kingdom, or over the sea, are not supported.

Credits

PVGIS and the European Commission do not endorse this site or its recommendations.

From your answers to a system

The quiz asks at most four questions you must answer: what the system is for, where (a ZIP code or postcode, or for vehicles the regions and months of your trips), when you need power (or how long outages last), and what you want to power. Everything else has a default that the result lists under "What we assumed", and you can change any answer with "Refine my answers". The defaults:

You can describe your energy use as a list of appliances (power, hours per day and how much of it happens in daylight) or as kWh from your electricity bills. When you use bills, we turn monthly kWh into a daily figure with the real number of days in each month and size for the heaviest month in your season (a winter bill is best). Unless you tell us otherwise, we assume most of it is used in the evening and at night, because night-time energy has to come from the battery. Bills often include heating, air conditioning or charging an electric car, which off-grid systems rarely cover, so you can choose to cover all of it, about half or only the essentials. A bill tells us how much energy you use, not how much power you need at once, so we assume a peak of three times your average power and a starting surge of twice that.

Simulation

We do not size systems with a rule-of-thumb formula. For each candidate system (a combination of panel power and battery capacity) we run a day-by-day energy balance over every year of weather in the record for your grid square and season, and count the days on which the system would run short. On each day the panels first serve what you use in daylight; any surplus charges the battery; the battery covers the rest of the day and the night. Days outside your season have no load, and the panels keep charging the battery.

When the battery stays outside or in an unheated space and a self-heating battery exists in your country's catalog, we also size with that battery, and use the result when it needs fewer panels or reaches more targets without leaving a part open.

The simulation accounts for:

Three options by reliability

For cabins, sheds, homes, vehicles, boats and portable systems we show three options:

For each option we look for the smallest panel power that reaches the target for a range of battery sizes (half a day to seven days of your use), and choose the combination that balances the two. The result also says how many days a year you would typically run short, and how many in the worst year of the record. These percentages describe historical weather; a future season can be worse than any in the record.

For vehicles, boats and portable systems you pick up to three regions and the months you spend in each. We size for the weakest of your trips, so the system also works in the darkest place and month you named.

Where winters are dark, such as in much of the UK, adding ever more panels is often poor value. In those cases we suggest a hybrid setup with a generator, mains hook-up or grid charging instead. A target counts as out of reach when more panels stop helping: a quarter more panels adds less than one point of the days and the target stays far away, or a load of up to 3 kWh a day would need more than about 5 kW of panels. Minimum then shows the best system found. Medium shows the best system found when that gives at least one point more of the days than Minimum, otherwise the Minimum system. With margin shows the system of the option below when its target is out of reach, when it needs more than twice the panels of Medium, or when a change of panel type would gain less than one point for more than one and a half times the panels. The result says so each time. Each option has at least the panels and storage of the one before it and is at least as reliable.

Plug-in (balcony) solar

Plug-in kits feed a socket in your home, so the question is not "will I run short?" but "how much will I use myself and how long does the kit take to pay for itself?". We simulate every hour of the weather record: what the panels give, capped at the legal limit for your country or state after the microinverter (the small box that turns the panels' DC into household AC), against your base load (the power your home always uses). By default we assume most of that use is in the evening, which is the cautious choice for savings. Energy you do not use goes to the grid; we assume it is not paid for unless you say it is.

With a battery, surplus energy charges the battery and the battery serves your home when the sun does not; the kit's output still never exceeds the legal limit and the battery is never charged from the grid. Where batteries may be combined with plug-in kits, we show the kit with and without a battery side by side; in the UK the current rules do not allow it, so UK results show the kit without a battery. We show up to three kit sizes. The main one sends the largest share of its output into your home; on equal shares, the one that saves more. A larger kit can still save more money a year. Savings use the default electricity price (sources under Credits) unless you give your own. Payback is shown only when the kit has an official price in our catalog.

United Kingdom: only kits on the Energy Networks Association plug-in register, up to 800 W, one per household; the rules cover Great Britain, so the quiz does not offer plug-in kits in Northern Ireland yet. United States: plug-in solar is allowed only in some states. We keep a table of every state's rule, each with its official source (legislature, governor or regulator), and re-check it at least every 90 days. A US visitor sees plug-in kits only when the state's plug-in law is in force and the kit is certified to UL 3700 (for the kit as sold, or for its microinverter with panels certified to UL 61730 or UL 1703). Where a state allows more power only with an electrician, we use the lower limit and show the higher one separately. Otherwise the quiz says why it shows no kit (no law, bill pending, law not yet in force, not confirmed). This is not legal advice: utilities, landlords and homeowners' associations can add their own rules.

Backup power for outages

For backup power the question is how many outages of the length you choose the system rides out. We start an outage on every day of the weather record at midnight, 6 am, noon and 6 pm, with the battery full at the start, and count how many the system survives. The three options cover about 90 %, 97 % and 99.5 % of those outages. Without panels, the battery is sized for the worst outage on the coldest day at the end of the battery's life. With panels, each option takes the smallest battery that reaches its target with no more panel power than an average day of the darkest month needs to cover one day of use, and never more than 6,000 W.

We also show a second option without panels: a power station charged from a socket before the outage. It cannot recharge during a long outage, and we say so. Connecting a backup system to your home wiring (a transfer switch or interlock) needs a qualified electrician. If you list a medical device, we size for it like any other load and say plainly that a single battery system is not a safe sole supply for life-supporting equipment.

Snow

For every 5 km grid square in the US and the UK we know, from the ERA5-Land reanalysis for 2005-2024, how many days per month the ground usually has at least 2.5 cm (1 inch) of snow, how much snow falls and how cold and humid the month is. A published model (Townsend and Powers, 2011) turns this into the share of a month's output lost to snow on the panels. Steep panels shed snow faster than flat ones; vertical panels lose little. We put that loss on the coldest days of each month, because snow and cold come together, which makes the battery size cautious.

This is an approximation. Snow on a dark, tilted panel melts and slides off sooner than snow on the ground, and how much sticks depends on wind, the frame and how often you clear it. ERA5-Land is a model of about 9 km and can miss snow in hills or along coasts; we checked it against weather stations and describe the result in our documentation. When we have no snow data for a place we use your answer to the snow question instead.

Products and compatibility

Specifications come from manufacturer documents and certification registers; where we could not confirm a safety limit, the product is not recommended. Every value is stored with its source and the date we read it, and the values must pass physics and plausibility checks. A safety value needs one manufacturer document or register entry for the exact model, or two agreeing independent sources. Where sources disagree on a safety limit, we hold the product back until the question is settled. We have not tested these products ourselves; we rely on published specifications, standards and registers.

How we confirm a safety value depends on the kind of part. For electronics (panels, batteries, power stations, charge controllers, DC-DC chargers and inverters) it counts as checked only from the manufacturer's own document for the exact model (a datasheet or a manual), a certification register, or two agreeing independent sources. For fuses, breakers, cables, connectors and other passive protective parts, the rating table on the manufacturer's own product page is also accepted. A few electronics checked before 23 October 2026 from the manufacturer's product page are being checked again against the manufacturer's documents.

A manufacturer's document counts the same wherever we find it: on the maker's site, or as the original file published by a distributor, a shop, a manual archive or the Internet Archive, as long as it is recognisably the maker's document for the exact model. When no manufacturer document for a model exists, we accept the specification table on a distributor's page that names the exact model, and say so next to the part: "Specifications from the distributor (no maker document found). Check them before you buy." A manufacturer document always takes precedence over distributor data. Many products are built for the brand by another factory; that factory's own datasheet counts as a manufacturer document when a certificate or register names both models, when the brand prints the factory model, or when at least three figures match (safety settings of the battery protection only with one of the first two), and we label it next to the part. Marketplace listings, reviews, forums and videos are never used for figures.

Checked and provisional. A product is checked when every safety value is confirmed as above. A panel, power station, charge controller or DC-DC charger whose safety values are all complete, without conflicting figures and taken from the manufacturer's own website, but not yet confirmed by a document or a register, is provisional. Checked products always come first and provisional ones after them, and we say so next to the part and on the whole option: its specifications come from the manufacturer's website (or from a manufacturer manual that does not yet settle every value), so check them before you buy. Many makers of portable panels do not publish the temperature coefficients. For such a panel we use the least favourable coefficient among at least five checked panels of the same type (rigid, flexible or folding, and the cell type where it is known), mark it as a typical value next to the figure, and keep the panel provisional; it is never shown as checked, and the maker's own figure replaces it as soon as we have it. When a maker does not publish a panel's maximum series fuse rating, we connect at most two strings of that panel in parallel, because three or more parallel strings need a fuse sized to that rating. A battery whose maker states only "Li-ion" or "semi-solid state" is never treated as lithium iron phosphate, and we accept its cold charging only from the charge temperature in the manufacturer's document. Sodium-ion cells are treated the same way: only their own printed charge temperature decides. When an inverter maker prints the surge only as a multiple of the rated power (for example "1.5 times of rated power, 5 s"), we store that multiple of the rated power with its duration and use it only for the start-up peak of a load, never as continuous power.

A few values have a cautious rule when the maker prints nothing, always labelled next to the part and never giving a product the checked status: the rated charge current as the limit of a controller's solar input current; 0 °C as a power station's lowest charging temperature; 5 V above the battery voltage as the start voltage of a controller, or the margin its own printed operating range gives; pure sine for a hybrid inverter whose maker prints only the distortion figure. When two documents of the same maker disagree on a limit we use the stricter value. A printed value always replaces such a rule. Batteries, fuses, breakers, cables and plug-in kits are never provisional, and products the manufacturer has stopped selling are not offered.

A set the station's own maker sells as one product can be shown as provisional when only its solar current is above the printed input limit; the note says so. For ready kits, the cold checks run with the temperature of your place, so a kit can appear in one place and not in another.

Every system we show passes these checks:

Some notes appear on the result without stopping a system. When a power station's solar input uses a different plug from the panels (for example XT60 against MC4), we say an adapter or cable may be needed. When a self-heating battery warms itself only above a charge current its manufacturer names, and the system cannot deliver that current, we say it may not charge on cold days until it warms up. When the panels' voltage on a hot day may be too low for the controller to start, or a panel is not the usual voltage class for a PWM controller, the system stays with a warning and comes after the systems without one. A mobile system with a lot of panels on a roof gets a note that so much solar rarely fits.

How we put a system together

The numbers on each option describe the system shown under it: its panels and batteries added up. What the simulation asked for is listed under "Why this system"; parts come in fixed sizes, so the system can be larger. The lines about days, outages and hours describe it too: we run the simulation again for that system, so they can be better than the target of the option.

We build each option only from products that passed all checks, for your country's catalog. A system is never smaller than the simulation needs. We avoid large oversizing (at most about 25 % more panel power and 35 % more battery than needed; if nothing fits within that, we take the smallest that meets the need and say so). Among systems that fit equally well we prefer fewer batteries (one or two, then three or four, then five to eight), then, for a bank above 6 kWh, a higher system voltage (48 V, then 24 V, then 12 V: lower currents and thinner cables; a small load stays on 24 V when a 48 V inverter would be more than twice its size), then the one closest to the need (compared in 10 % steps), then the one with fewer parts, then an MPPT charge controller over a PWM one, then the lower official price. The simulation assumes an MPPT controller; a PWM controller uses only about three quarters of the panel power, so a system with one gets about 29 % more panel power. For the inverter, the charge controller and the DC-DC charger we look only at sizes up to one and a half times the smallest one that is big enough and prefer the brand of the battery. Among panels of the same type that give the same layout and almost the same power, we also prefer the brand of the battery or power station. Provisional products follow the checked ones (see above) and are marked, with the whole option. The panel type of your mounting answer comes first; another type appears only when no complete system of your type exists, with a note. A system without an open combiner box comes before one with it. A system has at most 30 panels and one charge controller. An inverter with a built-in solar charge controller needs no separate controller; we use one only when no complete system can be built otherwise. Its solar inputs get the same checks as a charge controller.

When a maker does not state how many batteries can be connected in parallel, we count no more than four. A limit printed by the maker takes priority.

A battery with no disconnect of its own always comes with a separate battery management system chosen for the voltage and current of the system. When a battery with its own battery management system also fits, it comes first, because the system needs one part fewer.

Where a ready kit or one of our reference systems fits, you can choose it instead of separate parts. Reference systems are labelled "Assembled from parts sold separately": each part is bought from its own shop. Under "More complete systems" we show other panel and battery combinations of about the same size, where possible from another brand, and a ready kit when one fits. Where no product we have checked fills a part, we say so and list what to look for in plain words.

The price class (for example "up to $500") comes only from manufacturer or official-store prices with the date we read them, never from shop listings; when some parts have no official price, we say "at least" or show no class.

With "Choose another" you can pick a different part. We run the same checks on your pick with the rest of the system and tell you what passed, what failed and what could not be checked. Products from shops that we have not checked ourselves are marked "We have not checked compatibility of this item"; provisional products are listed after the checked ones with the same note as in the result; missing data means a check could not run, and such a product is never part of our own suggestions. Your result is calculated again each time its link is opened, so a shared link can show newer products than when it was made.

Catalog pages

Our pages for part types, uses, brands and products are generated from the catalog numbers. Rows are ordered by checked status, then by fit for the use, then by the lower official price. Labels such as "Best fit for this use" go to checked products only. "Works with" and "Does not work with" come from the same compatibility checks at fixed conditions (-10 °C coldest, panels at 60 °C, up to three panels in series). Pairs we could not check are not listed and are never shown as compatible.

Buying

Our recommendations are based on the products' technical specifications. Prices change, and the price at the shop is the one that counts. Some links on this site are affiliate links. If you buy through one, we may earn a commission at no extra cost to you. See our affiliate disclosure.

How we check the engine

Limitations

Safety and the role of an electrician

Our results are estimates for planning. They do not replace advice from a qualified electrician. Electrical rules differ between countries, for example the National Electrical Code (NEC) in the US and BS 7671 in the UK. Have any fixed installation, and any connection to a building or vehicle electrical system, designed or checked by a qualified person.