BACKGROUND:
The prediction of wind farm energy production is strategic in the current situation of the electricity market due to high saturation of wind power. The prediction applied to wind farms is a tool that has allowed it to manage the power system more effectively, providing predict ability of generation from wind power. The prediction model brings about offers of energy that will produce a wind farm, with the corresponding value added for the generation.
The CIEMAT (Centre for Energy, Environment and Technology) in 1999 conducted a review of the existing international state-of-the-art prediction models for wind farms. In addition, as part of this study, CIEMAT tested two prediction models, the Riso National Laboratory Prediktor and the WPPT from the Technical University of Denmark, in a wind farm with complex terrain in Spain. The findings of this study of situation prediction models indicated that:
- The models developed for flat land, did not give good results in complex terrain
- The weather forecast in Spain needed a specific adjustment for the wind farm that includes local effects
CIEMAT and CENER (National Renewable Energy Centre) established an agreement for technical and commercial development, technical and prediction wind programs and wind energy production.
Based on this agreement, Sotavento has made this project with CENER.
The Sotavento Experimental Wind Farm has shown from the beginning a great interest in evaluating predictive models in its fleet, preferring the LocalPred model.
OBJECTIVES:
The main objectives covered:
- To show an operative form, the operation of a prediction model in a wind farm with complex terrain
- Improve the integration of the energy produced in the electrical system, adapting to the requirements of the regulations at the time of its development
- Help plan daily electricity demand
METHODOLOGY(PHASES):
The different phases that took place throughout the project were:
- Acquisition of numerical predictions
Acquisition of the numerical prediction of a meteorological model, in this case we chose the National Meteorological Institute (NMI) by HIRLAM model.
We used historical predictions and forecasts in real time. The historical predictions were used to verify the prediction model and adjust it to the study area. Once optimised the model predictions were incorporated in real time. - Data collection of wind and production at the Wind Farm
The development of the prediction model needed to have wind data and wind farm energy production data. SOTAVENTO created a database with those measures, adapted to the needs of the prediction model.
- Tuning and Optimisation of wind prediction model
From historical predictions and measured speed data and production in the wind farm, the prediction model was adjusted for CENER-CIEMAT.
- Modelling of the wind farm operation against the wind. Obtaining the wind farm power curve
From the data on the functioning and operation of the wind farmsupplied for Sotavento, the operation of the wind farm by obtaining its power curve was modelled. The modelling of the wind farm takes into account the speed, wind direction and wind turbine availability, among other parameters.
- Calculation of model prediction errors
- Installing the prediction model in Sotavento equipment
- Evaluation of results
Tracking of the model performance and evaluation of the prediction model results.
- Final improvement phase from the obtained results
The results from part 7 served to identify possible improvements to the prediction model for the Sotavento Experimental Wind Farm. These improvements include:
- More advanced statistical models
- High resolution physical modeling with MM5 or FLUENT
PARTNER AGENCIES:
This collaborative project was conducted within the framework of contacts between SOTAVENTO, IDAE and the Centre for Energy, Environment and Technology (CIEMAT) in order to install a prediction model of energy production in the Sotavento Experimental Wind Farm.
CURRENT SITUATION:
The project was completed in 2003, at the time providing real-time data to estimate production for the web.
