Home » FAQs

FAQs

How SolarAnywhere Data compares with other data sources

Is the Perez model used to generate SolarAnywhere Data new or different from the one used by the NREL NSRDB?

SolarAnywhere Data is generated using a more recent version of the Perez model than that used to generate the NSRDB 2005 and 2010 updated datasets. SolarAnywhere also provides more recent data through the last hour, forecasts and other features that are not available from the NSRDB.

How does SolarAnywhere Data compare with NREL TMY data?

Typical Meteorological Year irradiance data available from NREL is synthesized from historical sources to represent a “typical” year for a fixed number of sites in the U.S. SolarAnywhere represents actual hourly estimates of irradiance for each specific location based on satellite imagery and atmospheric conditions at the site. Visit Typical Year Files for more detail.

Learn more in this paper, “How Misuse of Solar Resource Datasets Is Reducing Solar Industry Profits.

How do satellite-derived irradiation sources compare with output from ground-based measurement instruments?

If properly calibrated and maintained, ground-based instruments can be very accurate for the immediate area around the equipment. As the area of interest is located further from the unit, the accuracy declines, increasing the usefulness of satellite-based observations. The cost of equipment, inaccuracies from calibration and long setup times can favor satellite over ground-based measurements. Satellite irradiance is also often used in conjunction with ground-based instruments, and is particularly useful for detecting ground device calibration drift and for filling in missing data.

 

SolarAnywhere methodology

What data sources are used when generating SolarAnywhere Data?

Cloud, albedo, elevation, temperature and wind speed data are used in conjunction with satellite imagery collected from geosynchronous satellite networks. Visit Historical Model for more detail.

How often is new SolarAnywhere Data made available?

SolarAnywhere real-time data is collected, processed and available for use within approximately one hour. When the forecast option is licensed, a continuous dataset extending from January 1, 1998 through the present, and up to 168 hours into the future is available.

Why are different models used when forecasting days-ahead and hours-ahead data periods?

Short-term SolarAnywhere forecasts utilize a vector-based cloud model. Longer-term forecasts rely on numerical analysis. The bifurcated approach utilizes the method expected to give the best results for the time interval requested. Visit the Forecast page for more detail.

What happens when the SolarAnywhere model changes?

SolarAnywhere modeling algorithms and input data sources are occasionally updated to improve accuracy, consistency and availability. Updates to the modeling algorithms are released as new SolarAnywhere versions. Within each new data version, historical data is reprocessed using the updated algorithm to take advantage of the improved accuracy. Data versioning is a key factor of solar resource data bankability as it allows investors, independent engineers, etc., to reproduce results. See the release notes for a summary of model improvements.

To provide continuity for users of older SolarAnywhere datasets, and allow comparisons between data versions, prior versions are also available. Support back to SolarAnywhere version 3.2 is available via data.solaranywhere.comContact us for information on prior datasets.

There are periods of missing measurements in my SolarAnywhere file. Why?

Missing data occur in the SolarAnywhere irradiance database due to missing satellite images. Missing images are normal and occur due to rare unplanned outages and regular maintenance performed by the National Oceanic and Atmospheric Administration (NOAA). Missing data also occur in the ancillary surface air temperature and wind speed data due to periods of missing measurement from ground-based sensor networks.

Visit the Missing Data page to learn more about how SolarAnywhere handles missing data.

What is the difference between TGY and P50 data?

We’re sometimes asked whether a TGY is a P50. In short, no. The typical year is constructed of closest-to-average months. In practice, the annual total of a typical year is very similar to the average of the annual totals of the timeseries. P50, on the other hand, represents the median year of the distribution. Half of future years are expected to fall above the value, and half are expected to fall below.

TGY and P50 are often similar but diverge for asymmetrical distributions. Visit Probability of Exceedance to learn more about SolarAnywhere PXX Data.

SolarAnywhere energy and loss modeling

Which snow loss model should I use? Has the model been validated?

SolarAnywhere supports both the Marion (NREL) and Townsend snow loss models.

The Marion model uses time-series weather data to estimate snow losses in each time step. Since snow losses are cumulative and time dependent, using the full, high-fidelity time-series meteorological data with a snow loss model can help make loss estimates more reliable and data-driven.

The Townsend model uses monthly average snowfall and snow event data to calculate losses. It also accounts for the effects of ground clearance on snow sliding, and has parameters that can represent single-axis tracking, both of which may be helpful for modeling commercial PV systems.

A comparison of different snow loss models and reported accuracy can be found in the PVPMC study: Dynamic Snow Loss Model and Validation.

Both of these models are validated and considered industry standard. The authors of the Marion snow loss model have validated it against ground measured data from fixed-tilt PV systems. Read more here. A case study comparing the two models can be found here

How can I account for expected increases in extreme weather events and irradiance variability in long-term PV generation uncertainty estimates?

In practice, the solar industry uses historical data to estimate future solar resource at a project location. Using a solar dataset that is up to date and has a long history of consistent measurements can help spot trends and quantify their potential impact on expected generation over the lifetime of the project. SolarAnywhere offers an accurate, consistent solar dataset with a period of record of more than twenty years and can be used to identify long-term trends in the solar resource.

In addition to identifying long-term trends, historical solar data can also help quantify the impact of extreme weather events such as wildfires or snowstorms on PV production. For example, a study from Clean Power Research used historical SolarAnywhere data to quantify the impact of wildfire smoke on PV production in western North America. The study found that wildfires can reduce PV output by up to 6% at some locations, rivaling cloud cover as a risk to PV production.

Additional research connecting extreme weather to long-term trends can be found in a recently published blog. You can also learn more about this topic in the on-demand webinar Wildfires and Extreme Weather: Quantifying Impacts and Risks with Improved Performance Benchmarking.

How can SolarAnywhere users access tools for quantifying soiling, wildfire smoke and snow related losses?

If you have a SolarAnywhere Typical Year license, you can include monthly average snow and soiling losses in your Average Year Summary file downloads. To include average year summary files with your typical year file downloads, select “Edit Settings” upon selecting a tile, then check the box next to “Avg. year summary” under “Typical year options.”

If you have a SolarAnywhere Sites license, you can access time-series DC snow and soiling losses through the Energy Modeling Services API. Contact us if you are interested in a demo of our API.

High-resolution aerosol optical depth data is used as an input to SolarAnywhere model versions 3.5 and later to capture smoke impact in SolarAnywhere historical irradiance data. Clear-sky irradiance data is helpful when using solar resource data to quantify aerosol impact on PV system performance. Since the clear-sky model is not influenced by cloud cover, it isolates the impact of aerosols on insolation.

More information is available in the IEEE PVSC paper: Quantifying the solar impacts of wildfire smoke in western North America.

Have you compared SolarAnywhere time-series GHI data to ground measured GHI data from September 2020 in wildfire areas such as California and Washington?

Yes. The IEEE publication, Quantifying the solar impacts of wildfire smoke in western North America shows that SolarAnywhere can accurately estimate aerosol impacts (see fig. 1).

How does the inclusion of an extreme-weather year such as 2020 impact SolarAnywhere typical-year data / the long-term average? How are you considering it going forward?

The impact of including 2020 in typical-year file generation will vary by location. Each year of weather is incorporated into the long-term averages with each version release. Visit our blog to stay up to date on our annual regional insolation deviation summaries:

Does pvlib have a tool that utilizes particulate matter data or manual washes?

Both the Kimber and the Humboldt State University (HSU) soiling models are supported and can be applied during pvlib simulations. The Kimber model uses daily soiling rates and manual wash input parameters, whereas the HSU model utilizes PM2.5 and PM10 data to estimate soiling losses. Visit the Soiling-loss Modeling page in the support center to learn more about the HSU and Kimber models.

Can SolarAnywhere data be used to quantify hail risk? It's becoming more and more of a hot-button issue in the industry.

SolarAnywhere currently offers outputs for solid and liquid precipitation. Unfortunately, these outputs do not provide insight into the frequency of large hail.

How do you estimate melt rate for a snow event?

The Marion and Townsend snow loss models do not account for the effect of snow melting. Snow sliding is considered to be the dominant process of snow removal.

The Marion model calculates the amount of snow sliding off the PV panels using:

 Snow Slide Amount = 1.97 * sin(tilt)

The Townsend model calculates the amount of snow sliding off the panels using monthly snowfall events, row slant height and ground clearance of the system.

You can learn more about the Marion snow model here and the Townsend model here.

Are the snow loss models adapted to single-axis tracking systems including specific leading-edge clearance and snow pile up?

Although the Marion model is not specifically adapted to single-axis tracking systems, the amount of snow removed is considered to be a function of the tilt angle. The Marion model does not consider specific leading-edge clearance and snow pile up effects when estimating snow losses.

The Townsend model is also not specifically adapted for single-axis tracking, however, it can be estimated using specific input parameters, see the pvlib documentation.

For more information about the validation of the snow loss models and a comparison of accuracy with other snow loss models, refer to the support center link here.

Can cleanings be considered with the snow loss models?

Cleanings cannot be considered by the Marion model or the Townsend model. For more information on model parameters, please see the terms and concepts defined in SolarAnywhere API documentation.

Do snow and soiling loss estimates consider accumulation on the modules in addition to particulates in the air? If so, can it account for accumulation on modules after the weather event has passed?

The HSU soiling loss model determines module accumulation based on the particulate matter concentration (PM10 and PM2.5) in the atmosphere at any given time step.

In comparison, the Kimber model uses a consistent soiling accumulation rate in combination with rainfall and manual washing data. The Kimber model accounts for a grace period where soiling does not accumulate after a weather event.

The NREL (Marion) snow-loss model uses difference in snow depth data during each time-step to estimate accumulated snowfall on PV panels.

The Townsend model uses monthly snowfall and total number of snowfall events.

You can learn more about the soiling and snow-loss models here.

When considering ground validation of the HSU model, some papers have shown lower measured soiling losses in desert areas compared to dryland agriculture. The sand may be in the air, but it doesn't necessarily stick as much.

The authors of the HSU model validated it against ground measurements from seven test sites in the southwestern United States, demonstrating the model’s ability to accurately predict soiling losses. You can read this validation here. Another study by Micheli et al. performed model validation using ground measured data from six measurement sites spread across the U.S. In this study, the HSU model consistently returned the lowest errors, with minimum mean absolute errors (MAE) and mean errors (ME) close to zero. With SolarAnywhere, users can easily access and more broadly validate the HSU model for diverse PV system types and geographical areas.

SolarAnywhere uses ambient particulate matter concentration with the HSU model to estimate soiling impact. Soiling on PV panels is a complex phenomenon. The actual amount of accumulated particulate matter can depend on a number of factors such as rebound and resuspension of particles from wind, cementation and dew formation, and PM larger than 10 μm in aerodynamic diameter. Research on soiling models is still evolving to more accurately account for these factors in soiling loss estimates.

Due to soiling rate variability, the Kimber model supports specifying expected daily loss in energy output due to soiling as opposed to using particulate matter data.

SolarAnywhere licensing options

How are geographic areas defined when licensing data?

The basic geographic unit is a satellite visible “tile.” Tile areas directly correspond to satellite image resolution. SolarAnywhere offers data in 10 km nominal (0.1 x 0.1 degree) resolution in all available regions, and up 1 km nominal (0.01 x 0.01 degree) resolution in select regions. For information on licensing, see purchase options or contact us.

What if I need more Typical Year sites in a year than my license allows?

You can upgrade to the unlimited license, purchase a fresh Professional license valid for 1 year or purchase single TGY credits. You can see all the licensing options here.

These purchases can be made online with a credit card by logging into SolarAnywhere and visiting the Purchase page, or by contacting us to receive an invoice.

What if I need more Sites in a year than my license allows?

You can purchase single Sites credits online with a credit card by logging into SolarAnywhere and visiting the Purchase page, or by contacting us to receive and invoice. You can also upgrade to a 10 or 50 pack of sites for a per-site pricing discount. If you anticipate needing more than 8 sites, it then makes sense to purchase a 10-pack of sites based on our Sites pricing structure. You can see all the licensing options here.

Can I access SolarAnywhere Data to support my research free-of-charge?

Yes, students, researchers and the solar industry have no-cost access to complete and current SolarAnywhere Typical Year and Time-series data at select locations around the world with a Public license. These data can be used for educational purposes, and to research new and improved solar PV models for risk assessment, operations and maintenance, intelligent energy dispatch and more.

For a limited time, data that was formerly available via NREL and CPR/SUNY’s Solar Prospector tool is available with a SolarAnywhere Academic license. This legacy dataset was generated using older SolarAnywhere data versions and is only available for locations in the U.S.

Commercial use of SolarAnywhere data obtained through a Public or Academic license is not supported. You can see all the licensing options here.