 The wind power production forecast comes in a tabular and graphical form revealing the amount of the predicted energy production along with the given probability.
It is updated every 12 hours and can provide an hourly data with 10 days horizon. The software is configured to operate autonomously and start as soon as it acquires the necessary information:
- Meteorological forecast for the respective area
- SCADA data
- Operational capacity of the facility
.jpg)
Weather forecast based on a Global Numerical Weather Prediction Model
In recent years the knowledge on the driving forces behind the formation of weather and movement of air masses has greatly improved. Furthermore, the growing capacity of computer hardware has increased the possibility to calculate vast amounts of data from various sources. The combination of those two factors has resulted in the development of weather prediction models such as the one we use. These models are characterized by the reliability of the meteorological data they supply and its low margin of error.
We can further decrease the statistical error from the global model by taking into account the local weather conditions and adjusting the forecast with data from our measurement masts in the area.
Online SCADA data from the park udner various weather conditions
This is used for the so-called adaptivity or self-calibration of the model. The reliability of the forecast can be significantly improved if the calculation software is fed with live data from the wind turbines in the park itself, preferably every hour. This way the model assesses the accuracy of its prediction and takes measures to adjust.
Information for the planned and already operational capacity in the park
This information is necessary for avoiding distortions of the forecast in the periods when scheduled maintenance must be conducted.
------------------------------------------------------------------------------------------------
Statistical Error
Root Mean Square Error (RMSE) is the most important indicator for the forecast accuracy.
In the context of the mathematical model used, the RMSE increases proportionally with time. In general, the increase is faster within the first few hours and slows down after the 4th hour.
The typical margin of error in a 24 hour forecast is between 12-15%, which is a good result. The error for the initial 4-5 hours is significantly lower - approximately 5-6%. High frequency of updating the forecast and feeding live SCADA data can maintain an acceptable margin of error. |
