Suncast improves solar energy forecasting service by 57%

From the beginning, Suncast has focused on offering high-accuracy predictive models for energy generation forecasting in photovoltaic and wind parks. For this reason, to maintain and optimise the performance of the power plants, during 2022, the area in charge of Artificial Intelligence analysed the performance of the models used by Suncast in different power plants, calculating performance metrics and displaying graphs to show the performance of the forecasts.

In this regard, a complete analysis was performed on all plants, which was executed over several months and determined that, after applying the respective adjustments, Suncast currently has an average improvement of 57% for solar energy.

The quality of the forecasts can be seen in the attached figure, which displays the forecasts versus the actual generation of a photovoltaic park in August 2022. In this regard, it was found that the forecasts sent by Suncast to the National Electric Coordinator are adapted to the actual production of the wind farm, as well as anticipating days of low production.

In view of the advances and results in forecasting errors, Suncast's technical team believes that the work carried out allows guaranteeing a quality service, customized for each power plant so that customers can comply with the requirements of the Technical Standard, which will be increasingly demanding.

Forecasting sending and performance calculation methodology

In order to carry out the metrics, Suncast has followed the official indications of the document "Forecast Performance Index Calculation Methodology", established by the current regulation, where it is specified that the calculation is performed monthly by comparing the actual hourly generation data of the plants, and the forecasts correctly sent.

For this purpose, the performance of the medium-term forecasts is evaluated at each plant. The 24-hour power generation forecast is considered, which is extracted from the 48-hour forecast delivered at 18:00 hours of the previous day, both from wind and photovoltaic power plants.

According to the official document, the data obtained through this operation are used for the final version of the Short-Term Programming. The performance for different forecasting horizons is also considered, in order to have a view of how the quality of the forecasts changes as the forecasting horizon increases.