Guide

How much energy will my solar panels actually make?

A walk through solar performance modeling for homeowners.

By SolarTI83September 4, 202618 min read

Friends often ask me to review their solar quotes. Every one of them comes with a number on it like “11,400 kWh per year.” Where does that number come from? Is it real? Can I check it?

Getting to that specific number is challenging, but there are basically four things that drive it, and all four are simple to understand.

1. Where the sun is. Known exactly, for any moment, past or future.
2. What the atmosphere and weather do to the sunlight before it reaches your roof.
3. Your panels: which way they face, how many there are, how dirty they get, what is shading them, and how well they turn light into DC power.
4. Your electrical system: the inverter that turns DC into the AC your house uses, and a battery if you have one.

Every professional solar model in the world, from the free ones to the ones that cost thousands a year, walks through those four things in that order. So will this page.

Section 1Where the sun is

Start with the easy part. Based on where you are on the planet, the position of the sun in the sky at any moment is known. It is all orbital mechanics: the Earth spins once a day, goes around the sun once a year, and is tilted about 23.4 degrees. That tilt is the reason summer days are long and the sun is high, and winter days are short and the sun is low.

So for your address, a model can say exactly how high the sun is and which direction it is in at 2:37 PM on March 3rd of any year you like. There is nothing to argue about here.

Two things fall out of this that matter for your roof:

• In the northern hemisphere the sun lives in the southern half of the sky. It rises in the east, arcs through the south, and sets in the west. It is never north of you, unless you live in the tropics, where the midday sun passes north of you for part of the year. This is why south-facing roofs are ideal.
• The sun’s path is a low, short arc in winter and a high, long arc in summer. A roof that is great in July can be mediocre in January, and shade that does not exist in summer can appear in winter when the sun is low.

Section 2The atmosphere and the weather

Above the atmosphere, sunlight arrives at a steady 1,361 watts per square meter. There is not much going on between the surface of the sun and the top of the atmosphere. Then the light has to get through air, water vapor, dust, and whatever clouds are around, and what reaches your panels is split into pieces that solar people have names for:

• DNI, direct normal irradiance. The beam that comes straight from the sun’s disc. This is the part that casts sharp shadows.
• DHI, diffuse horizontal irradiance. Light scattered by the sky. On an overcast day this is all you get. It arrives from every direction, which is why there are no shadows and it is still bright.
• GHI, global horizontal irradiance. Everything landing on a flat, horizontal surface: the beam plus the diffuse.

The relationship between them is simple, and from any two you can derive the third.

Nobody measures any of this at your house. Instead, professionals use weather datasets built from decades of satellite images and ground stations. The usual move is to grab a typical meteorological year, or TMY: a stitched-together twelve months where each month is chosen to be typical of the long-term record for your location. That gives 8,760 hourly values of DNI, DHI, GHI, air temperature, and wind speed. Every serious performance model is run hour by hour against a file like that.

Two questions that will impress, or scare, your salesperson

• What source are you using for the weather file? A quick search of whatever they name will tell you a lot about its credibility.
• Is that a P50 or a P90? The key word is typical. A P50 means you would expect half of all years to come in higher and half lower. A P90 is the conservative version: you would expect 90 percent of years to beat it. Almost certainly you are being shown the P50.

Two honest caveats:

• A typical year is not next year. Real years swing about 5 to 10 percent either way around typical, and more in a few places. A quote is a long-run average, not a promise.
• Where you live matters more than almost anything else. The same panels make roughly 60 percent more energy in Phoenix than in Seattle. That is not the installer’s fault.

ReferenceNREL, “National Solar Radiation Database (NSRDB)” documentation; Sengupta et al. (2018), Renewable and Sustainable Energy Reviews.

Section 3Which way your panels face

The weather file tells you how much light lands on a flat surface. Your panels are not flat. They are tilted at your roof’s pitch and point in whatever direction your roof points. So the next step is to turn “light on a horizontal surface” into “light on the panel.” Solar people call this transposition, and it is where DNI, DHI, and GHI get recombined for a tilted surface.

The rules of thumb:

• South is best in the northern hemisphere, because that is where the sun spends most of its daylight hours. Southeast and southwest are close behind, typically within about 5 percent. East or west alone cost you something like 15 to 20 percent versus south, but they are far from useless — and a west-facing array makes its power in the late afternoon, when the grid and your air conditioner want it most.
• Tilt matters less than people think. Anywhere from about 15 to 40 degrees of pitch on a south-facing roof gets you within a few percent of the best case across most of the US. Steeper favors winter; flatter favors summer.
• North-facing roofs are usually a pass. Not zero, but you are living on diffuse light and the summer morning and evening beam.

ReferencePerez et al. (1990), Solar Energy — the sky model that almost every commercial tool uses for the diffuse part of transposition.

Section 4Dirt, glass, and shade

Now the light has reached the panel. Three things happen before it gets to the cells, and one of them is the thing homeowners should actually worry about.

Dirt

Module glass gets dirty. Dust, pollen, bird droppings, and everything else. This is soiling, and it is modeled as a flat percentage haircut on the light — usually 1 to 3 percent a year in places that get regular rain, and more in dusty, dry regions. If you are next to a farm or a gravel road, expect high soiling.

Snow usually gets bundled into soiling too. It is a kind of dust, I suppose? The effect is the same, and your proposal should carry large monthly loss percentages through the winter to account for modules being covered.

The glass

Light hitting the glass at a shallow angle partly reflects off instead of going through. Panels are built with anti-reflective glass to fight it, and the effect is small at noon and large near sunrise and sunset, when there is not much light anyway. Professionals call this the incidence angle modifier, or IAM, and it typically costs 1 to 3 percent over a year. You do not need to do anything about it. It is baked into any decent model.

Shade

Here is the one that matters. If something is between the sun and a panel, that panel is in shadow, and solar panels do not produce meaningful energy in shadow. A chimney, a vent pipe, a tree — any shadow is going to cut production. Remember the sun path from section 1: a tree that is nowhere near the sun’s summer arc can sit right on top of the winter arc.

This is typical of a residential system. I have a chimney that likes to cut my production for about an hour during the summer months. What you need to be careful of is: are these shading objects actually in your proposal’s performance model, and how are your panels wired?

Your ten panels are almost never ten independent generators. They are wired in strings — say two strings of five — and the panels in a string behave like batteries in series: the whole string is limited by its weakest member. In the simple picture, if one panel in a string of five goes into shade, all five drop toward the shaded one’s output. A chimney shadow on one panel at 4 PM can take out half your system at 4 PM.

Panel makers and inverter makers know this, and the market has answers:

• Bypass diodes. Every modern panel is divided into three sections with a diode across each, so a shadow on the bottom row of cells knocks out a third of that panel instead of the whole string. This is why “shade on the bottom edge” is much less bad than “shade over the entire panel.” Half-cut cell modules push this further.
• Microinverters. A small inverter bolted to each panel. Your effective string size becomes one. Each panel does what it can, independently.
• DC optimizers. A small box per panel that lets a string inverter treat each panel separately. Same idea, different hardware.

Which one is right for you depends entirely on your shading situation. On a clean, unshaded south roof, a plain string inverter is cheapest and works great. On a roof with a chimney, a dormer, and an oak tree, per-panel electronics often pay for themselves.

Here is the reassuring part. For most homeowners there is going to be a tree or two that shades part of the roof at some point in the year. That is fine. It costs you a few hours here and there. A good professional models the actual sun path against the actual obstructions and tells you the number. If your quote does not mention shade at all, run.

Section 5Inside the panel

How a solar panel turns photons into voltage is fascinating, but you do not need it. Here is the only model you need.

A panel is rated at a lab condition called STC: 1,000 watts per square meter of light, cell temperature of 25 °C. A “400-watt panel” makes 400 watts under exactly those conditions. Your roof is at those conditions approximately never.

Two things move the output away from the sticker:

• Light. Power is very nearly proportional to the light hitting the cells. Half the sunlight, half the power. A bright but cool spring day at 900 W/m² is a very good day.
• Temperature. Panels lose power as they heat up, roughly 0.3 to 0.4 percent per degree Celsius above 25. A panel in full sun on a still July afternoon can run 60 °C or hotter, which is 35 degrees over the rating and a 10 to 14 percent haircut right when the sun is strongest. Wind helps. Airflow under the panels helps. But most residential systems use black panels mounted very close to the roof surface, because it looks better. It is just hot.

This is why the best production days of the year are usually in April, May, and September — not July. Long days, high sun, cool air.

Go find this on your own quoteThe temperature coefficient is on every panel’s datasheet as the “Pmax temperature coefficient,” usually somewhere between −0.26 and −0.40 %/°C. Ask your installer for the datasheet and look it up.

A few percent more disappears in the DC wiring, and in the fact that no two panels are perfectly identical, so a string runs a hair below the sum of its parts. Models call these ohmic and mismatch losses. Together, a percent or two. Not worth your worry — and worth repeating that all of these losses are known quantities. A good solar professional models every one of them.

Section 6Batteries

If you have a battery, the DC from the panels can go straight into it. This is the most efficient path, because you skip a conversion. Later, when you want that energy in the house, it comes out of the battery through the inverter like any other DC.

There is a second design where the battery has its own inverter and sits on the AC side of the house. That is easier to add to an existing system and easier to charge from the grid, at the cost of an extra round trip through conversions. Either way, expect to get back roughly 85 to 92 percent of the energy you put into a battery. The rest is heat.

Whether a battery makes financial sense is a function of how much solar you are producing, how much of it you can use yourself, what the grid is willing to pay you for the excess, and how expensive the battery is. If your utility buys your excess power at $0.20/kWh and sells it back to you at $0.21/kWh, the $0.01/kWh you would capture by storing it yourself will never pay for a battery. On the other hand, many utilities pay you nothing, or very close to nothing, for excess generation. In those cases a battery is genuinely valuable and can pay for itself quickly. This is straightforward to model, and your installer should be able to show you that math.

Section 7The inverter and the DC:AC ratio

Your panels make DC. Your house and the grid run on AC. The inverter does the conversion, and it is where the last two important things happen.

First, it has an efficiency. Most of them are very good: 96 to 98 percent. You lose a couple of percent and that is that.

Second — and this is the one people find counterintuitive — it has a ceiling. An inverter rated at 8 kW will never put out more than 8 kW, no matter what the panels are doing.

Now think about what the panels do over a day. They ramp up from nothing at sunrise, peak around solar noon, and ramp back down. And from section 5, that peak is almost always below the sticker rating, because the panels are hot. So a “10 kW” array might touch 8 or 9 kW for an hour or two on the best days of the year and sit well below that the rest of the time.

That is why it is normal, and usually smart, to install more panel capacity than inverter capacity. The ratio of the two is the DC:AC ratio. Ten kilowatts of panels on an 8 kW inverter is 10:8, or 1.25. On the handful of hours a year when the panels could make more than 8 kW, the inverter “clips” the excess and you lose it. That lost energy is tiny, typically well under 2 percent of the year, and in exchange you got 25 percent more panels feeding an inverter you were paying for anyway. Residential systems commonly land between 1.1 and 1.3, and I would lean toward the higher side.

Section 8What comes out the other end

Add up all 8,760 hours and you get the annual number on your quote. There are two easy ways to check whether it is reasonable:

• Specific yield: kWh per year per kW of panels. In the US this runs from about 1,100 in the cloudy Pacific Northwest to about 1,800 in the desert Southwest, with most of the country between 1,300 and 1,600. A 10 kW system quoted at 11,400 kWh is a specific yield of 1,140 — plausible in Seattle, and deeply suspicious in Tucson.
• Performance ratio. The energy you actually get, divided by what you would get if every panel ran at its sticker rating for every hour of sun that hit it. Real rooftop systems land around 75 to 85 percent.

Professionals draw this as a loss diagram: start with the sunlight landing on the roof, subtract each effect in order, and end at the energy delivered to the house. It is the single most useful page in any solar report, because it shows you where your particular roof is losing energy, and therefore what — if anything — is worth doing about it.

ReferenceIEC 61724-1 defines performance ratio. NREL’s PVWatts documentation lists its default loss assumptions, which total about 14 percent. Both are public and citable.

In shortSo what should you actually do?

• Know where south is, and know where your sun arcs go.
• Get a quote that names its shading assumptions and its DC:AC ratio. If it does neither, it is a brochure, not a model.
• Run your own numbers in PVWatts. It is free, it uses the same weather data the professionals use, and it will land within about 10 percent of a professional model on an unshaded roof.
• Do not obsess over tilt, panel brand, or the last two percent of inverter efficiency. Obsess over orientation and shade. That is where the real money is.

Everything else is orbital mechanics and weather, and neither of those is negotiable.

AppendixCommon questions

AppendixSources and further reading

• NREL PVWatts Calculator — the free hourly production model, and the default loss assumptions behind most quick estimates.
• NREL National Solar Radiation Database (NSRDB) — the satellite-derived irradiance dataset behind most US models. See also Sengupta et al. (2018), “The National Solar Radiation Data Base (NSRDB),” Renewable and Sustainable Energy Reviews 89, 51–60.
• PVGIS — the European Commission’s equivalent, covering most of the world.
• Perez, R., et al. (1990), “Modeling daylight availability and irradiance components from direct and global irradiance,” Solar Energy 44(5), 271–289 — the diffuse sky model used by essentially every commercial tool.
• Reda, I. and Andreas, A. (2004), “Solar position algorithm for solar radiation applications,” Solar Energy 76(5), 577–589 — the sun-position algorithm behind the solar position tool on this site.
• IEC 61724-1, Photovoltaic system performance — Part 1: Monitoring — the standard that defines performance ratio.
• Google Project Sunroof — free shade-aware rooftop estimates from aerial imagery for many US addresses.