PVWatts vs the Google Solar API: two ways to estimate solar production

If you have run the same roof through NREL's PVWatts and Google's Solar API and gotten two different production numbers, you are not doing anything wrong. The two tools are built on different data and answer slightly different questions. Knowing what each one models, and why they disagree, tells you which number to trust for the decision in front of you.

What PVWatts actually does

PVWatts is a free calculator from the National Renewable Energy Laboratory (pvwatts.nrel.gov). You give it a location and a set of system specifications, and it models a year of production. Its weather engine uses Typical Meteorological Year (TMY) data: a synthesized "typical" year assembled from many years of real measurements at or near your location, capturing representative sun, temperature, and weather rather than any single actual year.

The defining trait of PVWatts is that you enter the system specs:

• DC system size (kW)
• Module type and array type (fixed roof mount, ground mount, tracking)
• Tilt and azimuth
• System losses, including a shading allowance you estimate
• DC-to-AC ratio and inverter efficiency

PVWatts then runs an hourly simulation and returns monthly and annual AC production. Because you control every input, it is only as good as the specs you type, garbage in, garbage out, but it gives you total flexibility to model any design, including one that doesn't exist yet. This is the same estimating logic in estimating solar production kWh, and the tilt and azimuth inputs it depends on are covered in solar panel tilt and azimuth.

What the Google Solar API does

The Google Solar API starts from the roof, not from your inputs. Using satellite and aerial imagery plus 3D building data, it reconstructs the actual building. It exposes two relevant products:

• Building Insights returns roof-segment geometry, each plane's area, pitch, and azimuth, along with a modeled annual production figure and how many panels of a given size fit.
• Data Layers returns per-pixel annual flux in kWh/kW/yr, effectively a sun and shade heat-map of the roof that accounts for surrounding trees, buildings, and terrain.

The strength is that it reflects your specific roof: its real shape, orientation, and shading, with no need to measure or type anything. The catch is that the imagery may predate roof changes, a new addition, a reroof, or a felled tree, and the modeled tree canopy is an approximation. It is a strong, roof-aware estimate to confirm, not a survey.

Side-by-side

Why the two disagree

When the numbers diverge, it is almost always because the inputs diverge, not because one tool is broken:

• Different weather. PVWatts uses a fixed TMY station; Google's model uses its own solar radiance modeling. Even for the same roof, the assumed annual sun differs.
• Different shade. In PVWatts you might type a 5% shading loss; Google's flux map may find a dormer and a tree that actually cost more, or less, on specific planes.
• Different geometry. If you enter a 20-degree tilt at 180-degree azimuth into PVWatts but the roof is really 25 degrees facing 160, PVWatts models a house that isn't there. Google reads the real plane.
• Different losses. PVWatts bundles soiling, wiring, mismatch, and inverter losses into a system-loss field; Google's annual figure makes its own assumptions.

Neither being "right" is the point. Use their disagreement as a prompt to check which assumption is off, usually shade or azimuth.

When to use which

Reach for Google Solar when you have only an address and want a fast, roof-specific screen: does this roof have good planes, where is the shade, roughly how many panels fit. Reach for PVWatts when you have a defined design and need to model it precisely, comparing a 20 versus 30 degree tilt, testing a west-facing plane for time-of-use rates, or sizing to a target. Many workflows use both: Google to find the roof's real geometry, then PVWatts to stress-test a specific build with those numbers.

Our free instant estimate uses a satellite roof-and-sun model to give you the roof-aware first pass instantly, then hands off cleanly to design and sizing. From there, how to size a residential solar system and how to calculate solar offset turn a production estimate into a system that meets your usage.

Both are estimates to confirm

Whichever tool you lean on, remember that both produce modeled numbers, not metered ones. TMY is a typical year, not next year; satellite geometry is a snapshot, not a survey. Treat either result as a well-grounded starting point to verify against a real site visit and, ultimately, against a year of actual meter data.

Educational reference, reviewed 2026-07. A design aid, not a substitute for a licensed electrician or PE. Confirm the enforced NEC edition and local amendments with your AHJ.