The Center for Renewable Energy from the Federal University of Pernambuco (CER-UFPE) has created and online forecasting tool with the funding support of the HPC4E project. CER-UFPE’s forecasting tool is being developed as an operational (online) short-term prediction tool oriented to the dispatch of renewable energy plants into the grid. In the first version, the focus is on the dispatch of wind farms with a 24-hours ahead prediction horizon and a time-step of 30 minutes as required by the Brazilian System Operator who is a special partner of HPC4E project.
The different models/checks contemplated in the forecasting tool by CER-UFPE are organized into four main groups:
- Quality assurance/control checks for flagging anomalous behavior to be neglected during models’ calibration/operation;
- Downscaling models to predict the behavior of very local surface wind (e.g., in the surroundings of an anemo mast) with a horizon of up to 24-hours ahead (note that power curve models are employed to convert the downscaling models output into forecasts of the wind farm output);
- Time series based models to predict fast dynamics within the wind power time series with a horizon of up to 6-hours ahead and a time-step of 30 minutes;
- Combination models to track the best aspects from downscaling and time series based models in order to produce final forecasts of the wind farm output.
The tool is still a demo and still being developed, thus it is not open for the public yet.
Note that CER-UFPE’s forecasting tool is being funded by two main R&D projects: EOLIPREV project, funded by CNPq; HPC4E project, funded by RNP. Undergraduate and graduate students with scholarships from both projects have been contributing for the tool. In addition, resources from both projects were employed for acquisition/upgrade of CER-UFPE’s computational cluster.
More information: https://cer.ufpe.br/forecast/