Estimation of the production of a wind installation on a short-term horizon with a view to making the energy management easier for the electrical system operator. To achieve this, CENER has its own, continuously developing, software and methodology. Apart from providing information about real time prediction for a large number of wind farms on the Spanish market, CENER participates in different public financing projects (ANEMOS.plus, POW’WOW).
The LocalPred prediction model developed jointly by CIEMAT and CENER includes:
- Own weather prediction based on mesoscale models (SKIRON and MM5)
- Possibility of additionally using the prediction of the National Weather Institute
- MOS type statistical correction to adapt the prediction to the specific location of each farm
- Combination of the different available predictions (from the different meteorological models) for each wind farm. This methodology is known as “multi-model ensemble”
- Prediction for daily market
- Prediction for intradaily markets
- Prediction for fixed rate option
- Safety backup to guarantee delivery of predictions
Prediction of energy production for wind farms in agreement with the applicable regulation (Royal Decree 436).
Work is carried out in the following R&D lines, among others:
- Application of new meteorological models to wind prediction (MM5, WRF, SKIRON, HIRLAM)
- Development of advanced statistical models to detect and eliminate systematic prediction errors
- Prediction by combination adapted to the different atmospheric situations
- Application of fluid dynamics models (CFD) to the wind prediction in complex terrain
The research activities are backed up by collaborations with different national and European universities, as well as by the participation in EU-financed projects: ANEMOS, POWWOW.
The specialisation of the CENER work team, as well as the available calculation capability permit working with several meteorological models at the same time, thus generating complete information of specific predictions required by the wind farms.
This characteristic has permitted developing a combined prediction model, emphasising the precision of the best individual prediction and also offering a backup, which guarantees the delivery of a prediction even though there is a failure in any one of the individual models.