We are constantly working to improve Solargis. Below are details of recent updates to Solargis model and services.
You can vote for features and submit new feature requests via our public product roadmap.
Solar radiation data for West Coast of USA and Canada, and Hawaii, for late afternoon hours on first and last day of each month in 2020 was incorrectly calculated, resulting in underestimation of cloud coverage. The issue was fixed and data for all affected days in 2020 was recalculated. Approximately 1% of all 2020 data in the west coast of continental US and Canada, and 2% of 2020 data for Hawaii was affected. Central and Eastern parts of Canada and US were not affected by this issue.
For Europe, forecasts for lead time from 2 hours up to 120 hours is now based on blending of forecasts from 2 NWP models - IFS (ECMWF) and ICON-EU (DWD). In general, IFS is considered as the best performing global NWP model. Yet, we've seen many instances when the IFS forecasts error was quite high. To limit the occurrence of large forecast errors, henceforth a blended forecasts of IFS and ICON-EU models will be used.
Overall, the new approach is expected to improve accuracy of both intra-day and day-ahead solar power forecasts.
Currently, forecasts from the ICON-EU model is being used only for Europe. We are working on implementation of forecasts of ICON model also for other regions.
For a more detailed explanation see this update on our Productboard.
Improved blending of satellite-based nowcasts with forecasts based on NWP models. Previously, nowcasts were used for forecast lead time of 0-4 hours, and NWP based forecasts were used for forecast lead times of 4+ hours. However, in some regions NWP forecasts already gives better results for forecast lead time of 3-4 hours. Thus a new blending method was implemented. Now, by default, full nowcasts is used up to hour 2, from hour 2 up to hour 4 the weight of nowcasts decreases from 1 to 0. From hour 4 only NWP forecasts are used. The default settings can be customized on per site basis where needed.
Improved aggregation of 10 minute time series (native resolution) to 15 minute time step. Previously implemented aggregation resulted in 'steppy' profile during cloudless situations. A new aggregation method was introduced to avoid these features.