Wind power forecast

from Wikipedia, the free encyclopedia
Typical wind power forecast

The aim of a wind power forecast is to predict the power supply of a single wind turbine , a wind farm or (most often) several wind farms in a region. The performance predictions are generally calculated for a time range from a few minutes up to around 10 days. In contrast to a weather forecast , the wind power forecast predicts the expected power in kW or energy in kWh and not the wind speed in m / s. For Germany, the expected Germany-wide or control area-wide feed-in from wind energy is particularly important.

Such predictions are used for energy trading and energy management; As a rule, short-term forecasts are useful for these purposes, i. H. Forecast periods from a few minutes up to approx. 48 hours. Medium-term forecasts, the forecast periods of which can be up to eight days, are more likely to be used for resource planning (e.g. maintenance).

Predictive models

Two different approaches have emerged for predicting the expected performance of a wind farm; a distinction is made between physical and statistical models. Methods from both approaches are used in many forecast systems.

Input variables

The most important input variable of wind power prediction models is a prediction of the local wind speed. Such forecasts can be obtained automatically from weather services, but mostly not exactly for the location of the wind farm and the height of the system. The weather services provide forecasts for points on a large-scale grid that covers a certain region (sometimes the whole earth) and has a mesh size between approx. 15 and 50 kilometers (see numerical weather forecast ). Wind speeds are generally predicted by the weather services for heights of 10 and 100 meters. In addition, some models also use the power actually provided by the wind farm in the last few hours as well as information about the wind farm itself (number, type and arrangement of the wind turbines and the surface properties of the surroundings).

Characteristic curve of a wind turbine

Physical models

A physical model first interpolates a wind forecast for the exact location of the wind farm from the available numerical weather forecast data for a specific point in time. Since the wind speed in this area of ​​the atmosphere increases with height, this wind speed is then extrapolated to the hub height of the system. Among other things, factors such as the stratification of the atmosphere and the roughness of the terrain are also included in the forecast. With the help of the expected wind speed at hub height determined in this way and a specific characteristic curve of the respective wind power plant, the expected power for this specific point in time is then calculated. In addition, such a prediction is usually compared with the currently actually generated power in order to better predict short-term changes.

Statistical models

Statistical approaches, on the other hand, do not use any system characteristics or details of the atmospheric processes. From the available wind and weather forecasts, they calculate the performance that is expected in a specific weather situation. Here the 'forecast history' serves as a source of information, which means that such models must be trained with historical data at the beginning.

reliability

To determine the accuracy of wind power forecasts, the difference between the predicted and the actually measured power is subsequently calculated (e.g. as mean average error or standard deviation ). Often the quality of a prediction is also measured using the so-called 'persistence' model, a trivial model that always predicts the current performance for the future. A forecasting model should by no means be worse than this 'pseudo-forecast'. The reliability of predictions can also be indicated by statistically determined performance values ​​which, with a certain probability, will not be undercut or exceeded.

Applications

A high level of accuracy in the wind power forecast helps electricity suppliers and transmission system operators to keep the power provided by the wind power plants and other power plants (in peak , medium and base load ) in balance with consumption. The accuracy of the forecast for the whole of Germany for the period of a 24h – 48h forecast is approx. 95 percent (normalized standard deviation approx. 5%). A high value is important because control energy and thus costs can be saved.

Further applications of wind power prediction can be found in the deployment planning of maintenance personnel for wind farms, in electricity trading and generally in energy trading.

Web links

literature