Lower energy estimates

Recently the Annual Net to grid energy estimates have been significantly lower than in previous analyses. Can you explain why? This is for residential detailed photovoltaic analyses.

For example a file created in SAM version 2015.1.30 (37.95 N, -120.35 E) estimated 10,210 kWh net to grid in year 1, but when updated to SAM version 2017.9.5 Revision 2, SSC Version 180, using a similar tmy file (lat 37.950_lon-120.260_TMY.csv downloaded 3/3/15) the estimated net to grid was only 9,860 kWh. When using a new tmy file (lat3795000_lon-120.34000_tmy.csv downloaded 1/9/18) the estimated net to grid decreased to 9,420 kWh.

This decrease of 7.7% appears to be a problem. The estimate made in 2015 was very close to the actual production measured by the inverter of this system. My most current SAM production estimate was approximately 10% lower than an estimate made by OnGrid for a project I am working on, which is causing a problem for the contract.

Is this a known occurrence? Is there a valid reason for it?

Hi Tom,

There are two factors that may be contributing to the difference in results you are seeing from the different versions:

1. The weather data is different. There are two factors contributing to that difference: 1) The data from SAM 2015.1.15 comes from a different database than 2017.9.5, and 2) the three latitude and longitudes, while similar, are for different map grid cells in the databases.

SAM 2015.1.30 downloaded files from the deprecated Solar Prospector database, so that weather file was for the 37.95 N, -120.35 E cell from Solar Prospector. The file you downloaded using SAM 2017.9.5 r2 on 3/3/2015 (did you mistype the year?) is from the National Solar Radiation Database (NSRDB) PSM dataset for the cell at 37.95 N, -120.26 E, and the file you downloaded on 1/19/2018 is for the cell at 37.95 N, -120.34 E. Different cells, even if they are neighboring cells, represent different locations have have different data.

The "Weather Data Example" section on the Weather Data page ( sam.nrel.gov/weather ) shows one strategy for using a set of weather files in your production estimates. The example uses out-of-date data, but, hopefully the concept is clear: If the results from different map cells around the project location vary a lot, then there is likely more uncertainty in your predictions.

Another thing to keep in mind is that the weather files you are downloading are TMY ("Typical Meteorological Year") data files that represent a long-term historical period. You can think of the TMY data as being a sort of average of the weather data over a long period of the past. (See the Weather Data page for a discussion of TMY data). The annual output prediction from a TMY weather file is good estimate for financial cash flow calculations where you want the annual value to be valid over the project life. (If you based the cash flow on a single year that happened to be especially sunny or cloudy, then your cash flow estimates would be off.) When you compare simulation results from a TMY file to measured data from a real system in a particular year, you are comparing the measured data for a specific year to an "average" long-term estimate, so you should expect the measured data from year to year to vary compared to the long-term estimate. It looks like your measured data happened to correspond well to the TMY-based results from the Solar Prospector database, but I would not expect that to always be the case.

2. The SAM models have changed between versions. You can see a list of changes between versions in the SAM Release Notes at sam.nrel.gov/sites/default/files/content/updates/releasenotes.html . How those changes might affect your results depend on which models you are using, and what parts of the models you are using.

When you compare results from different versions of SAM (or even from different files in the same version of SAM), you should be sure that the assumptions (or inputs), especially the weather data, in the different files are the same.

Finally, when you compare results from different models, like between SAM and OnGrid, you should also make sure the assumptions in both models are the same, or as close as possible.

I hope that helps shed some light on this.

Best regards,
Paul.