SizingKitIEC 61215 STC · PVWatts loss set
Solar Panel Output Calculator
This estimates what an array actually delivers by walking the whole distance between its nameplate and the meter, one factor at a time: ten loss categories, the cell temperature the modules run at, the inverter, and how old the array is. It returns kilowatt-hours per day, per year, and the specific yield in kWh per kW of DC per year — the figure the industry compares systems on — with every derate left visible instead of collapsed into a single number.
- 100% free
- No signup
- 10 loss factors, each editable
- Specific yield
- STC and NOCT stated
The array and the sun on it
Module wattage is the number on the back of the panel. Peak sun hours is the daily average energy arriving on the plane the panels sit in, in kilowatt-hours per square meter — which is numerically the hours of full 1000 W/m² sun the day is worth.
The STC power stamped on the label, not the module’s area.
8 kW DC — the figure a system is sold and permitted under.
Default 4.5 — A placeholder in the middle of the contiguous US range, which runs roughly 3.5 kWh/m²/day in the Pacific Northwest and upper Midwest to about 6.5 in the desert Southwest. This is the single largest input on the page and a default for it is close to useless. Take the annual plane-of-array figure for your own site, tilt and azimuth from NREL's PVWatts or the National Solar Radiation Database — it is free, it is hourly, and it already contains the orientation. A global horizontal figure is a different and smaller number; entering one here is the most common way this calculation comes out wrong.
How hot the cells get, and what that costs
This is kept out of the loss stack below because it is the only term that changes sign. A cell at 25 °C makes its plate rating; every degree above that costs, and every degree below it pays.
25 °C · 77 °F
Default 45 °C. Nominal operating cell temperature, the module's own cell temperature at 800 W/m², 20 °C ambient, 1 m/s wind and an open back — the IEC 61215 NOCT condition. It is printed on the module's data sheet.
Default -0.35. Runs about −0.29 to −0.45 %/°C across crystalline products, with n-type and heterojunction cells at the flat end and older p-type multicrystalline at the steep end. Thin film is different again — CdTe sits near −0.25 %/°C. This is the one loss on the page that reverses: below 25 °C the coefficient works in your favor, which is why a cold clear morning can push an array above its plate rating.
At 25 °C ambient and 1000 W/m², the NOCT model puts the cell 31.3 °C (56.3 °F) above the air around it, at 56.3 °C — a 10.9% loss against the plate. The rise is proportional to irradiance, so it is largest at exactly the moment the array is producing most.
The derate stack, one line each
Every figure here is editable and every one is a default rather than a measurement. Two of them — shading and soiling — are worth more attention than the other eight put together.
| Loss | % | What it is |
|---|---|---|
| Soiling | Dust, pollen, bird droppings and agricultural or traffic film. The one loss on this list that is a hose and a ladder. It is seasonal and site-specific — a long dry season with no rain to rinse the glass can cost several times this, and a wet climate less than half of it. | |
| Shading | The default assumes a nearly unshaded array. It is worth almost nothing on a real site: a plumbing vent that clips two cells at 9 a.m. can pull a whole string down, which is what module-level power electronics exist to prevent. Measure this with a shade analysis, do not take 3%. | |
| Snow | Zero by default because most of the US does not have it. Where it matters it is not a percentage of the year's output evenly spread — it is a handful of days at zero, in the season when the array was making least anyway. | |
| Module mismatch | No two modules in a string have exactly the same current-voltage curve, and a series string is held to its weakest member. Binning at the factory narrows this; aging widens it again. | |
| DC and AC wiring | Resistive loss in the string conductors, the home run and the AC run. This is the one term you can design away by going up a conductor size, and it is why a long roof-to-panel run is worth calculating rather than assuming. | |
| Connections | Contact resistance at every MC4 pair, terminal and lug. A connector crimped with the wrong die is not this number — it is an arc and a fire, and it is the single most common cause of a rooftop DC fault. | |
| Light-induced degradation | The step a crystalline module takes in its first hours of sun and then stops taking. It is already served by the time anyone reads a production report, which is why it belongs here rather than in the annual degradation rate. | |
| Nameplate rating | The allowance for a module not quite making its stamped power. Modern flash-test tolerances are one-sided — 0 to +5 W — which is why this default is small and would have been larger twenty years ago. | |
| Age | Zero for a first-year estimate. Set it to the degradation rate times the age in years to model an array already in service; the field below does the same arithmetic for a chosen year. | |
| Availability | The fraction of the year the system is down — inverter faults, grid outages, maintenance, a tripped breaker nobody noticed. Three percent of a year is eleven days, which sounds generous until you count how long a failed inverter waits for a replacement. | |
| Combined | 14.1 | The ten add to 15% and multiply to 14.1%, because each loss acts on what the one before it left rather than on the nameplate. |
Default 96%, weighted not peak.
6.67 kW AC of inverter behind this array.
1 is a new array. 25 is the end of a production warranty.
Default 0.5%/year, compounded.
What the array delivers
Per day
26.4 kWh
95.2 MJ
Per year
9,653 kWh
9.65 MWh
Specific yield
1,207
kWh per kW DC per year
| Array at standard test conditions20 × 400 W at 1000 W/m² and a 25 °C cell | ×1.0000 |
|---|---|
| System losses, combined10 categories summing to 15% but multiplying to 14.1% | ×0.8592 |
| Cell temperaturecell at 56.3 °C (133 °F), 31.3 °C off the test condition, at -0.35 %/°C | ×0.8906 |
| Inverter96% weighted efficiency, DC to AC | ×0.9600 |
| Degradation to year 1first year — nothing lost to age yet | ×1.0000 |
| Nameplate to delivered | ×0.7347 |
8 kW of module on the roof behaves like 5.88 kW of module in the meter, and 26.5% of the plate never becomes a kilowatt-hour. The single largest term is the loss stack, which is worth knowing before anyone proposes better modules as the fix.
Tilt and orientation, checked at noon
The peak sun hours above already contain the orientation, so this panel is a check rather than a multiplier: it reports how squarely the sun strikes the modules at solar noon on the three days that bound the year.
Negative south of the equator.
A 4:12 roof is 18.4°, 6:12 is 26.6°, 12:12 is 45°. Flat is 0.
0 is due south, −90 due east, +90 due west.
Winter solstice
38.4°
cos θ = 0.783 — the beam arrives at 78.3% of what a normal surface would collect
Equinox
15.4°
cos θ = 0.964 — the beam arrives at 96.4% of what a normal surface would collect
Summer solstice
8.45°
cos θ = 0.989 — the beam arrives at 98.9% of what a normal surface would collect
Beam only, at one instant. Between a fifth and half of what a module collects on a given day is diffuse light off the whole sky dome, and diffuse does not obey a cosine — which is why an array facing 45° off south loses far less than the cosine of 45° suggests, and why this page will not turn a horizontal irradiation figure into a plane-of-array one. That conversion is a transposition model over a year of hourly data, and NREL publishes it for free.
The default system-loss categories and values documented for NREL's PVWatts, which are the figures the tool applies when a user does not change them.
Defaults, not measurements of your array. Two of them — shading and soiling — vary by more than an order of magnitude between sites and are the two worth replacing with a real figure. The stack is applied multiplicatively, so replacing a value here moves the answer by slightly less than the change itself.
IEC 61215 / IEC 60904-3 standard test conditions: 1000 W/m² irradiance, 25 °C cell temperature, AM1.5 spectrum. Every watt printed on a module's plate was measured here.
An estimate, not an interconnection application. What a utility will let you build, what a net-metering or export tariff pays for a kilowatt-hour, whether the roof takes the load and how the array is bonded and rapid-shutdown compliant are separate decisions, and the drawing is stamped by somebody licensed to stamp it.
How to estimate what an array will produce
One input dominates this calculation and it is not the module. Get the irradiance right and the rest is bookkeeping; get it wrong and no amount of care with the derates will rescue the answer.
Get plane-of-array peak sun hours for the actual site
Not a national average and not a horizontal figure. NREL's PVWatts and the National Solar Radiation Database will give you the annual average irradiation in the plane of your own tilt and azimuth, free, from a full year of hourly satellite data. A global horizontal figure is a different and smaller number for any tilted array, and entering one here is the single most common way this estimate comes out low.
Replace the two derates that are really yours
Shading and soiling are defaults standing in for measurements. A shade study with a horizon scan replaces the first; how long your dry season runs and whether anything rinses the glass replaces the second. The other eight defaults are reasonable for almost any installation and are worth leaving alone until something specific says otherwise.
Read the multiplier column, not just the total
The chain prints each factor separately for a reason. If cell temperature is costing more than the entire loss stack, the fix is airflow behind the modules rather than better modules. If availability is the largest single line, the fix is monitoring that tells somebody the inverter has been off for three weeks. A single blended derate figure hides both of those and sells a panel upgrade instead.
Technical specifications
| The chain it computes | Nameplate DC × plane-of-array peak sun hours × ten system losses × cell temperature × inverter efficiency × compounded degradation. Every factor is printed as its own multiplier to four decimals, and the product is printed beside them so you can see which one did the damage. |
|---|---|
| Loss categories | Soiling, shading, snow, module mismatch, wiring, connections, light-induced degradation, nameplate tolerance, age and availability — the PVWatts default set, prefilled at 2, 3, 0, 2, 2, 0.5, 1.5, 1, 0 and 3 percent, all editable. |
| How the losses combine | As a product of (1 − loss), never as a sum. The ten defaults add to 15 percent and multiply to a 14.08 percent derate, because each one acts on what the one before it left. That gap is under a percentage point on a clean site and several points on a shaded, dirty one. |
| Cell temperature | Held outside the loss stack, because it is the only term that reverses sign. Cell temperature comes from the NOCT model — ambient plus (NOCT − 20 °C) scaled by irradiance — and is applied at the module's own Pmax coefficient, prefilled at −0.35 %/°C and editable across the −0.29 to −0.45 range crystalline products span. |
| Reference conditions | 1000 W/m², a 25 °C cell and the AM1.5 spectrum — the IEC 61215 and IEC 60904-3 standard test conditions the number on the module's label was measured at. NOCT is the separate 800 W/m², 20 °C, 1 m/s condition, which IEC 61215:2016 replaced with NMOT. |
| The conversion it refuses | Converting a horizontal irradiation figure into a plane-of-array one. That is a transposition model over a year of hourly data, and a fifth to a half of what a tilted module collects is diffuse light that obeys no cosine. The tilt and azimuth inputs report solar-noon incidence on the solstices and the equinox instead — exact geometry, honest about covering one instant. |
| Degradation | 0.5 percent a year compounded from year two, the median measured rate for fielded crystalline modules in NREL's degradation review. First-year light-induced degradation is separate and already sits in the loss stack, so the two are not double-counted. |
| Where the array data stays | In the browser. Nothing about the site, the array or the bill it is meant to offset is uploaded or retained, which matters when the numbers are a customer's rather than your own. |
Frequently asked questions
Why does my array never reach its rated wattage even at noon in June?
Mostly because the cells are hot, and the rating was measured at 25 °C. Standard test conditions hold the cell at 25 °C under full 1000 W/m² sun, a combination that essentially never occurs outdoors — full sun heats the module, and the NOCT model puts a typical cell about 31 °C above the surrounding air at that irradiance. On a 30 °C day the cells are near 61 °C, and at −0.35 %/°C that is roughly a 12 percent loss before wiring, inverter or dirt has taken anything. The corollary surprises people: a cold clear February morning with snow reflecting can briefly push a string above its plate rating, which is why string voltage is calculated at the record low temperature rather than the average one.
Do the loss percentages add up or multiply?
Multiply, as (1 − L₁)(1 − L₂)…, and this page does it that way. The ten defaults add to 15 percent but the product of the ten factors is a 14.08 percent derate, because soiling takes 2 percent of what arrives and mismatch then takes 2 percent of what soiling left rather than 2 percent of the original. On a well-behaved array those two figures are less than a percentage point apart and nobody would notice. On an array with 25 percent shading and 8 percent soiling, adding instead of multiplying is out by several percent of annual production, and it is out in the optimistic direction.
Should I use the panel's rated watts or its watts at NOCT?
The rated watts, which is the STC figure — this page applies the temperature correction itself, so entering the NOCT power would apply it twice. Module data sheets print both because they answer different questions: STC power is what the module is sold and permitted as, and the NOCT power is a courtesy figure showing roughly what it does at a realistic operating temperature. Use the STC number here, use the STC number on the permit, and treat the NOCT number as a sanity check on the temperature step this page performs.
How much does facing southwest instead of due south really cost?
Much less than the cosine of the angle suggests, which is why this page will not multiply by one. Between a fifth and half of the energy a module collects on any given day arrives as diffuse light from the whole sky dome, and diffuse does not care which way the module faces. A 45° azimuth error on a moderately tilted roof typically costs a few percent of annual production rather than the 29 percent a cosine implies — and a west-facing array can be worth more than a south-facing one where the tariff pays more for afternoon energy. Get the plane-of-array irradiance for the orientation you actually have, and the question answers itself with real numbers.
What is a DC-to-AC ratio above 1.2 doing to my production?
Clipping the tops off the clearest days and buying morning and evening energy in exchange. An oversized array relative to its inverter spends most of the year below the inverter's limit, where the extra modules produce freely, and a few dozen hours a year above it, where the inverter simply holds its ceiling and the surplus is discarded. The trade is usually favorable up to about 1.25 to 1.35 because inverter efficiency is poor at low load and good near full load, so a fatter array keeps the inverter in its efficient band for more hours. Modeling the clipping itself needs an hourly time series rather than a daily average, so this page reports the ratio and the risk rather than pretending to a number it cannot compute.
Is specific yield a better comparison than total kWh?
Yes, whenever two systems are different sizes, and it is the number to ask an installer for. Specific yield is annual kilowatt-hours divided by kilowatts of DC nameplate, so it strips the array size out and leaves the site, the orientation, the equipment and the workmanship. It lands roughly between 1,000 and 1,600 kWh per kW per year across the contiguous US, and the spread within one city is mostly shading and orientation. A proposal quoting only total production hides whether the number came from a good design or simply from selling more panels.
Why is availability set at 3 percent when my inverter has never failed?
Because 3 percent of a year is eleven days, and the question is not whether the inverter fails but how long it stays failed. A string inverter that trips on a grid disturbance and does not restart, a breaker somebody switched off during other work, a communications gateway that stopped reporting six months ago — none of those are failures anybody notices quickly on a system without monitoring. The default is an allowance for downtime of every kind including planned maintenance and utility outages. If you have monitoring with alerts and a maintenance contract, lowering it to 1 percent is defensible; if nobody looks at the system, 3 percent is generous.
Nameplate, plane of array, and the ten places the watts go
Every module carries a number measured once under conditions that do not occur: 1000 W/m² of light with the cell held at 25 °C and a defined AM1.5 spectrum, which is the IEC 61215 standard test condition and the whole basis on which panels are sold, permitted and compared. Outdoors, one of those two conditions can be met at a time. Full sun heats a module tens of degrees above the air around it, so a cell reaching 1000 W/m² is never at 25 °C, and a cell at 25 °C is not seeing 1000 W/m². The gap between the label and the meter is not a defect and it is not a manufacturer's exaggeration — it is the distance between a laboratory condition and a roof, and the only honest way to estimate production is to walk it factor by factor rather than to multiply by a single number somebody heard was about right.
The ten loss categories on this page are the set NREL documents for PVWatts, and keeping them apart is the point of the exercise. Two of them are yours to change and eight are not: soiling depends on your dust and your rain, shading depends on your chimney and your neighbor’s oak, and mismatch, wiring, connections, light-induced degradation and nameplate tolerance are properties of hardware that vary little between competent installations. Availability is the sleeper — the default of 3 percent is eleven days a year of the system being off for any reason, and on an unmonitored array a single unnoticed inverter fault dwarfs every other term on the list. Cell temperature is deliberately kept out of that stack, because it is the one factor that can improve the answer: below 25 °C the temperature coefficient works in your favor, which is why cold climates with clear winters do better per kilowatt than their latitude suggests.
What this page will not do is guess at irradiance. A design temperature is looked up per station rather than shipped in a table, and solar resource is the same kind of number — it is local, it is measured, and NREL gives it away for free in the plane of your own array. What comes out the far end is an estimate of energy, which is not yet an estimate of money: what a kilowatt-hour is worth depends on a tariff, an export rate and a time of day, and the electricity cost calculator takes it from there. Energy that arrives at noon and is wanted at eight in the evening is a battery capacity question, and a household that wants to ride through an outage rather than shave a bill usually ends up comparing this against a generating set, which answers a different question at a different price. None of this is an interconnection application or a structural review, and the drawing is stamped by somebody licensed to stamp it.
Where the site data stays
Every number on this page is worked out by JavaScript running in the tab you are reading it in. Nothing you type — loads, lengths, nameplate ratings, the rates your utility charges you — is uploaded, logged or kept, which is also why the calculators carry on working in a mechanical room with no signal.
An address, a roof orientation and a consumption figure together describe somebody’s household, and none of it leaves this tab — useful when the array being modeled belongs to a customer rather than to you. The DC run from the array to the inverter is a conductor sizing problem of its own, and a house that is chasing a smaller array rather than a bigger one usually finds more in the assembly R-value of its walls, or in whatever the propane usage calculator says the heating is burning.