How accurate is your solar forecasting technology?
One of the most common user questions we receive is - “How accurate are your forecasts?”. We understand this question is very important to our customers, as accuracy is often tied to penalties, trading strategies, energy storage dispatch and system reliability.
But answering this question is not as straightforward as it may seem. Accuracy statistics vary widely by geography, climate, season and the forecast horizon of interest.
That’s why we’ve created this solar forecasting accuracy tool. Its aim is to allow you to learn more about Solcast’s forecast accuracy at your locations of interest and enable you to download some historical forecast data to inform your validation work.
Solar Forecasting Accuracy Tool
This tool applies to solar farms or large rooftop PV sites. After you choose your site’s location and climate type, we’ll match it to our nearest archived data point with the same climate type.
More than 120 days, more than 750mm
From 75 to 120 days, from 450mm to 750mm
Less than 75 days, less than 450mm
Enter a site location and select the site attributes to view accuracy data.
We've matched your site to actual sites in our archive. See below your expected accuracy, and a 12 month accuracy data sample.
Tabulated Forecast Error
Using the results above, you can review the expected accuracy statistics in tabulated form. We’ve presented these in terms of % of the actual AC capacity of the site. Simply multiply the result using a fractional percentage (32% = .32) times your AC capacity to determine the power or energy yield expected Mean Absolute Percentage Error (MAPE).
Historical Forecast Data File
We also provide access to a downloadable file of historical forecast data as compared to our ‘estimated actuals (‘Actuals’ in the data file). The estimated actuals data are our best guess of the PV site performance based on the observed cloud cover and our solar radiation modelling algorithms, which serve as the ground truth in the above solar forecasting accuracy tool.
Using our Solar Radiation Data
If you plan to use our solar radiation data to compute your own PV power, this tool will give you a good indication of the resulting accuracy. This is of course assuming that the PV power conversion model you choose is appropriate for your solar farm.
Forecast Accuracy and Grid Aggregations
It is important to note that the forecast accuracy for our Grid Aggregations is generally superior to the results shown below, due to the averaging effect of many sites being added together. If you need more accuracy information about Grid Aggregations, contact us.
Creating Actuals Data for Validation
While we can’t use actual PV generation data from our customers for the validation statistics shared publicly, we do produce statistically vigorous ‘Actuals’ data for this validation tool. To do this, we start with our global horizontal irradiance data, and convert them to hourly averages. We then add random error to these, based on the known statistical distribution of the errors as determined by our validation at BSRN sites, before creating an estimate of PV generation.
If this is not clear, feel free to reach out and ask us for more details!
We calculate the Mean Absolute Percentage Error (MAPE) using the method of Zhang et al, 2013 (read more about this approach). The MAPE is calculated as a percentage of plant AC capacity. For hourly average power, the MAPE values are normalised with the plant’s capacity. For daily energy MAPE values, the values are normalised with the plants capacity, multiplied by the number of daylight hours in the day.
Detailed Forecast Accuracy for Your Site:
We also automatically update forecast accuracy assessments exactly matched to your site of interest in our API Toolkit. Once you create your Utility Scale or Rooftop Solar site, you’ll be able to view recent forecast accuracy specific to that site (pictured).