System Production: Calculating Yield

If you’ve been on the sales/installation/design side of the solar industry for even a few weeks, you’ve encountered questions related to how system production is calculated. This is a reasonable question. It’s a question that strikes at the heart of the solar value statement. As such, transparency is the appropriate posture for exploring this question.

Let’s start with a bit of groundwork information.

National Renewable Energy Laboratory

The United States Department of Energy National Renewable Energy Laboratory is the de-facto leader for the study of solar production, solar technicals, and solar industry dynamics for the US (and internationally to some degree). Their scientists have developed a well-vetted and trustworthy modeling tool called PVWatts. The latest version of PVWatts provides system production forecasts based on historical weather and irradiance data collected at 239 locations in the United States, over 30 years. THIRTY YEARS. And that includes data recorded during inclement weather months.

The PVWatts tooling allows for installers, or even lay-people to input details for an array and get back a yield forecast:

Location. This is simply input.

DC System Size. This is another simple input – its the number of panels x wattage of panels.

Tilt. This is a simple input – its the pitch of the array.

Azimuth. Another simple input – its the direction the array faces relative to solar South.

System Losses. This is a trickier input – its a calculated value based on the amount of shade an array receives and electrical details of the system.

With that information, PVWatts uses Typical Meteorological Year data and solid geometry to calculate how much sun irradiance will be collected, and how much electrical energy will be generated.

Installers, Designers, and Forecasts

Reputable installers do one of the following:
A. Calculate production for each array in a system by either using PVWatts directly
B. Use it indirectly via advanced software suites (which still rely on PVWatts at their core). This is more common.

PVWatts’ 30 years of cumulative weather data allows it to provide statistically derived stability for yield forecast in the face of day-to-day, month-to-month, and year-to-year weather variance.

Sure, its common that for some arbitrary stretch of time, the actual yield of an array may differ from PVWatts’ forecast. However, over a longer period, the system’s performance will converge to PVWatts’ forecast. A common refrain in the industry is that over 3 years, systems will produce within 10% of PVWatts’s forecast, and over 10 years, systems will produce within 3% of PVWatts’s forecast. As long as the inputs are correct, PVWatts’ results are reliable.*

Now, with all of this talk about the trustworthiness of PVWatts, its important to state clearly that its forecasts are predicated on accurate inputs by the person inputting the data.

Errors, Ignorance, and Fraud

Let’s take a closer look at ways humans may falter in using PVWatts to produce incorrect or even FRAUDULENT yield forecasts.

Back to those inputs…

Location. This is pretty hard to screw up, however if the PVWatts operator selected a location that was sunnier and had better weather, then sure, the forecast could be inflated relative to what would be observed by the system. The opposite is true as well; picking a less sunny location would provide an artificially low estimate for an array’s yield.

DC System Size. This is another hard one to screw up. Again, the DC system size is the number of panels times the wattage of those panels. A typo could swing forecasted production artificially higher or lower than would actually be observed by the system.

Tilt. This one is easier to get wrong. Because it relies on the actual pitch that an array will be set at. For roof mount systems, its possible that the roof pitch was estimated (its nearly impossible to measure the pitch of a home from a top-down image of it. If you don’t believe me, try it for yourself) or measured incorrectly. This will have a direct influence on production forecasts.

Azimuth. This is in the same boat as tilt, but its easier to confirm from aerial photography. Any designer/installer who knows what they are doing should have no problem deriving an array’s azimuth from aerial photography. For cases where aerial footage does not exist, a simple declination-adjusted compass measurement will provide an azimuth. Typos or incorrectly measured azimuths are the risk here, and have a direct influence on production forecast, and to a greater degree than would a Tilt error.

System Losses. This one is the big one. Its typically not the electrical characteristics that are the tricky part. The difficulty is primarily because the system losses input includes a measurement of the shade conditions for the array. Shade measurements determine how much sun an array will get over its surface over the course of a year, and they are easy to screw up:

– by people who don’t know how to measure properly.
– by people who don’t know how to interpret the results.
– by unscrupulous installers who fake shade reports/analysis to make a system’s forecast look better than it is, so they can sell the system, or maybe “squeak” a system past incentive thresholds (like NYSERDA’s) to make the system look more profitable.

If you ever see a solar system with a tree blocking most of its sunlight, then there’s a good chance fraud was at play. Also, I am aware of at least one WELL KNOWN residential solar leasing company (thats still in business) which consistently inflated production numbers to unachievable levels. We actually know some of their customers and how badly their systems are underproducing relative to what they were told at the time of sale.

How does EcoMen do it?

It depends on the scenario.

For cases such as new brand new construction, EcoMen uses the old-fashioned boots-on-the roof approach. We visit the site, photograph relevant roof surfaces and their surroundings, and then use a Solmetric Suneye tool to measure solar access at multiple array points. From there, we meticulously review and cleanse the Suneye data to accurately reflect the site conditions and then derive shade, azimuth, and tilt values. We then use PVWatts to generate array-by-array yield forecasts.

However, whenever possible, we use the Aurora solar modeling and design software suite. Aurora allows us to leverage LIDAR (LIght Detection And Ranging) data to model arrays and obstructions to accurate heights and dimensions, to derive array-wise azimuth and tilt measurements. We then use Aurora’s built in ray tracing engine to calculate solar access based on the 3D world we’ve constructed. From there we use its built-in PVWatts based production modeling to derive a system yield forecast**. The following images are from Aurora, and show the 3D model with LIDAR data points (the colored dots) and a heat map detailing the relative irradiance for various areas of the roof.